1 2019-04-30 Jakub Jelinek <jakub@redhat.com>
3 Backported from mainline
4 2019-02-20 Jakub Jelinek <jakub@redhat.com>
7 * expr.c (expand_assignment): If result is a MEM, use change_address
8 instead of simplify_gen_subreg.
10 2019-02-20 Jakub Jelinek <jakub@redhat.com>
11 David Malcolm <dmalcolm@redhat.com>
14 * fold-const.c (decode_field_reference): Return NULL_TREE if
15 lang_hooks.types.type_for_size returns NULL. Check it before
16 overwriting *exp_. Use return NULL_TREE instead of return 0.
18 2019-02-20 Jakub Jelinek <jakub@redhat.com>
22 * toplev.c (do_compile): Double the emin/emax exponents to workaround
25 2019-02-19 Richard Biener <rguenther@suse.de>
28 * toplev.c (do_compile): Initialize mpfr's exponent range
29 based on available float modes.
31 2019-04-30 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
33 Backport from mainline.
34 2019-04-30 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
36 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
37 Define __ARM_FEATURE_ATOMICS.
39 2019-04-30 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
41 Backport from mainline
42 2018-12-11 Richard Earnshaw <Richard.Earnshaw@arm.com>
45 * config/aarch64/iterators.md (sizem1): Add sizes for
47 (Vbtype): Add SFmode mapping.
48 * config/aarch64/aarch64.md (copysigndf3, copysignsf3): Delete.
49 (copysign<GPF:mode>3): New expand pattern.
50 (copysign<GPF:mode>3_insn): New insn pattern.
52 2019-04-29 Richard Sandiford <richard.sandiford@arm.com>
54 Backport from mainline:
55 2019-01-25 Richard Sandiford <richard.sandiford@arm.com>
58 * varasm.c (output_constructor_bitfield): Use wi::extract_uhwi
59 instead of accessing TREE_INT_CST_ELT directly.
61 2019-04-29 Richard Sandiford <richard.sandiford@arm.com>
63 Backport from mainline:
64 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
67 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
68 aarch64_evpc_tbl guard.
70 2019-04-29 Richard Sandiford <richard.sandiford@arm.com>
72 Backport from mainline:
73 2019-03-08 Richard Sandiford <richard.sandiford@arm.com>
76 * dwarf2cfi.c (dwarf2out_frame_debug_expr): Use CONST_POLY_INT
77 instead of POLY_INT_CST.
79 2019-04-26 Roman Zhuykov <zhroma@ispras.ru>
81 Backport from mainline
82 2019-04-23 Roman Zhuykov <zhroma@ispras.ru>
84 PR rtl-optimization/87979
85 * modulo-sched.c (sms_schedule): Start ii value "mii" should
88 PR rtl-optimization/84032
89 * modulo-sched.c (ps_insn_find_column): Change condition so that
90 branch will always be the last insn in a row inside partial
93 2019-04-19 Pat Haugen <pthaugen@us.ibm.com>
95 Backport from mainline:
96 2019-04-16 Pat Haugen <pthaugen@us.ibm.com>
99 * config/rs6000/power9.md: Add store forwarding bypass.
101 2019-04-17 Martin Jambor <mjambor@suse.cz>
103 Backport from mainline
104 2019-03-10 Martin Jambor <mjambor@suse.cz>
106 PR tree-optimization/85762
107 PR tree-optimization/87008
108 PR tree-optimization/85459
109 * tree-sra.c (contains_vce_or_bfcref_p): New parameter, set the bool
110 it points to if there is a type changing MEM_REF. Adjust all callers.
111 (build_accesses_from_assign): Disable total scalarization if
112 contains_vce_or_bfcref_p returns true through the new parameter, for
115 2019-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
117 Backport from mainline
118 2019-03-15 Kelvin Nilsen <kelvin@gcc.gnu.org>
121 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
122 When handling vec_extract, use modular arithmetic to allow
123 constant selectors greater than vector length.
124 * config/rs6000/rs6000.c (rs6000_expand_vector_extract): Allow
125 V1TImode vectors to have constant selector values greater than 0.
126 Use modular arithmetic to compute vector index.
127 (rs6000_split_vec_extract_var): Use modular arithmetic to compute
128 index for in-memory vectors. Correct code generation for
129 in-register vectors. Use inner mode of vector rather than mode of
130 destination for move instruction.
131 (altivec_expand_vec_ext_builtin): Use modular arithmetic to
134 2019-04-12 Kelvin Nilsen <kelvin@gcc.gnu.org>
137 * config/rs6000/vsx.md (*vsx_extract_<mode>_<VS_scalar>mode_var):
138 Use QI inner mode with V16QI vector mode.
140 2019-04-15 Martin Jambor <mjambor@suse.cz>
142 Backport from mainline
143 2019-04-15 Martin Jambor <mjambor@suse.cz>
146 * cgraph.c (clone_of_p): Loop over clone chain for each step in
149 2019-04-15 Martin Liska <mliska@suse.cz>
151 Backport from mainline
152 2019-01-18 Martin Liska <mliska@suse.cz>
153 Richard Biener <rguenther@suse.de>
156 * cgraph.h (create_version_clone_with_body): Add new argument
158 * cgraphclones.c (cgraph_node::create_version_clone): Add
159 DECL_ATTRIBUTES to a newly created decl. And call
160 valid_attribute_p so that proper cl_target_optimization_node
161 is set for the newly created declaration.
162 * multiple_target.c (create_target_clone): Set DECL_ATTRIBUTES
164 (expand_target_clones): Do not call valid_attribute_p, it must
166 * tree-inline.c (copy_decl_for_dup_finish): Reset mode for
169 2019-04-11 Richard Biener <rguenther@suse.de>
171 PR tree-optimization/90018
172 * tree-vect-data-refs.c (vect_preserves_scalar_order_p):
173 Test both SLP and interleaving variants.
175 2019-04-10 Matthew Malcomson <matthew.malcomson@arm.com>
178 * config/arm/arm.c (neon_valid_immediate): Disallow VOIDmode parameter.
179 * config/arm/constraints.md (Dm, DN, Dn): Split previous Dn constraint
181 * config/arm/neon.md (*neon_mov<mode>): Account for TImode and DImode
182 differences directly.
183 (*smax<mode>3_neon, vashl<mode>3, vashr<mode>3_imm): Use Dm constraint.
185 2019-04-07 Uroš Bizjak <ubizjak@gmail.com>
188 * config/i386/i386.md (anddi_1 to andsi_1_zext splitter):
189 Avoid calling gen_lowpart with SYMBOL_REF and LABEL_REF operand.
191 2019-04-04 Martin Sebor <msebor@redhat.com>
194 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Bail
195 out if the number of arguments is less than expected.
197 2019-04-03 Richard Biener <rguenther@suse.de>
200 * lto-wrapper.c (run_gcc): Avoid implicit rules making
201 the all target phony.
203 2019-04-02 Uroš Bizjak <ubizjak@gmail.com>
207 * config/i386/i386.c (dimode_scalar_to_vector_candidate_p):
208 Return false for variable DImode shifts.
209 (dimode_scalar_chain::compute_convert_gain): Do not handle
210 register count operand in variable DImode shifts.
211 (dimode_scalar_chain::make_vector_copies): Remove support to copy
212 count argument of a variable shift instruction to a vector register.
213 (dimode_scalar_chain::convert_reg): Remove support to convert
214 count argument of a variable shift instruction.
216 2019-03-29 Kugan Vivekanandarajah <kuganv@linaro.org>
218 Backport from mainline
219 2019-03-29 Kugan Vivekanandarajah <kuganv@linaro.org>
220 Eric Botcazou <ebotcazou@adacore.com>
222 PR rtl-optimization/89862
223 * rtl.h (word_register_operation_p): Exclude CONST_INT from operations
224 that operates on the full registers for WORD_REGISTER_OPERATIONS
227 2019-03-28 Uroš Bizjak <ubizjak@gmail.com>
230 * config/i386/i386.c (dimode_scalar_chain::make_vector_copies):
231 Also process XEXP (src, 0) of a shift insn.
233 2019-03-28 Martin Liska <mliska@suse.cz>
235 Backport from mainline
236 2018-11-05 Martin Liska <mliska@suse.cz>
239 * doc/invoke.texi: Remove options that are
240 not disabled with -Os.
242 2019-03-26 Uroš Bizjak <ubizjak@gmail.com>
245 * config/i386/i386.c (dimode_scalar_chain::convert_reg):
246 Also process XEXP (src, 0) of a shift insn.
248 2019-03-25 Andreas Krebbel <krebbel@linux.ibm.com>
250 Backport from mainline
251 2019-03-20 Andreas Krebbel <krebbel@linux.ibm.com>
254 * config/s390/s390.c (global_not_special_regno_p): Move to make it
256 (s390_optimize_register_info): Use global_not_special_regno_p to
257 check for global regs.
259 2019-03-22 Bill Schmidt <wschmidt@linux.ibm.com>
261 Backport from mainline
262 2019-03-22 Bill Schmidt <wschmidt@linux.ibm.com>
264 * config/rs6000/mmintrin.h (_mm_sub_pi32): Fix typo.
266 2019-03-21 Bill Schmidt <wschmidt@linux.ibm.com>
268 Backport from mainline
269 2019-03-08 Bill Schmidt <wschmidt@linux.ibm.com>
271 * config/rs6000/rs6000-p8swap.c (rs6000_analyze_swaps): Rebuild
272 ud- and du-chains between phases.
274 2019-03-19 Richard Biener <rguenther@suse.de>
277 * opts.c (finish_options): Disable -gsplit-dwarf when doing LTO.
279 2019-03-19 Eric Botcazou <ebotcazou@adacore.com>
281 PR rtl-optimization/89753
282 * loop-unroll.c (decide_unroll_constant_iterations): Make guard for
283 explicit unrolling factor even more robust.
285 2019-03-18 Martin Sebor <msebor@redhat.com>
288 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range):
289 Handle anti-ranges the same as no range at all.
291 2019-03-18 Martin Jambor <mjambor@suse.cz>
293 Backport from mainline
294 2019-03-18 Martin Jambor <mjambor@suse.cz>
296 PR tree-optimization/89546
297 * tree-sra.c (propagate_subaccesses_across_link): Requeue new_acc if
298 any propagation to its children took place.
300 2019-03-17 H.J. Lu <hongjiu.lu@intel.com>
302 Backport from mainline
303 2019-03-14 H.J. Lu <hongjiu.lu@intel.com>
306 * config/i386/i386.c (ix86_print_operand): Handle 'M' to add
307 addr32 prefix to VSIB address for X32.
308 * config/i386/sse.md (*avx512pf_gatherpf<mode>sf_mask): Prepend
310 (*avx512pf_gatherpf<mode>df_mask): Likewise.
311 (*avx512pf_scatterpf<mode>sf_mask): Likewise.
312 (*avx512pf_scatterpf<mode>df_mask): Likewise.
313 (*avx2_gathersi<mode>): Prepend "%M3" to opcode.
314 (*avx2_gathersi<mode>_2): Prepend "%M2" to opcode.
315 (*avx2_gatherdi<mode>): Prepend "%M3" to opcode.
316 (*avx2_gatherdi<mode>_2): Prepend "%M2" to opcode.
317 (*avx2_gatherdi<mode>_3): Prepend "%M3" to opcode.
318 (*avx2_gatherdi<mode>_4): Prepend "%M2" to opcode.`
319 (*avx512f_gathersi<mode>): Prepend "%M4" to opcode.
320 (*avx512f_gathersi<mode>_2): Prepend "%M3" to opcode.
321 (*avx512f_gatherdi<mode>): Prepend "%M4" to opcode.
322 (*avx512f_gatherdi<mode>_2): Prepend "%M3" to opcode.
323 (*avx512f_scattersi<mode>): Prepend "%M0" to opcode.
324 (*avx512f_scatterdi<mode>): Likewise.
326 2019-03-15 Richard Biener <rguenther@suse.de>
328 Backport from mainline
329 2018-11-30 Jakub Jelinek <jakub@redhat.com>
332 * params.def (PARAM_LOGICAL_OP_NON_SHORT_CIRCUIT): New param.
333 * tree-ssa-ifcombine.c (ifcombine_ifandif): If
334 --param logical-op-non-short-circuit is present, override
335 LOGICAL_OP_NON_SHORT_CIRCUIT value from the param.
336 * fold-const.c (fold_range_test, fold_truth_andor): Likewise.
338 2019-03-15 Martin Liska <mliska@suse.cz>
341 * doc/invoke.texi: Remove -fdump-class-hierarchy option.
343 2019-03-14 Martin Jambor <mjambor@suse.cz>
345 Backport from mainline
346 2019-03-07 Martin Jambor <mjambor@suse.cz>
349 * ipa-cp.c (perform_estimation_of_a_value): Account zero time benefit
350 for extern inline functions.
352 2019-03-14 Richard Biener <rguenther@suse.de>
355 * tree-scalar-evolution.c (get_loop_exit_condition): Use
357 * tree-ssa-loop-ivcanon.c (canonicalize_loop_induction_variables):
358 Use gimple_location_safe.
360 2019-03-13 Vladimir Makarov <vmakarov@redhat.com>
363 * lra-constraints.c (inherit_in_ebb): Update
364 potential_reload_hard_regs along with live_hard_regs.
366 2019-03-14 Richard Biener <rguenther@suse.de>
368 Backport from mainline
369 2019-03-13 Richard Biener <rguenther@suse.de>
372 * tree-scalar-evolution.c (simplify_peeled_chrec): Do not
373 throw FP expressions at tree-affine.
375 2019-03-01 Richard Biener <rguenther@suse.de>
378 * tree-cfgcleanup.h (cleanup_tree_cfg): Add SSA update flags
379 argument, defaulted to zero.
380 * passes.c (execute_function_todo): Pass down SSA update flags
382 * tree-cfgcleanup.c: Include tree-into-ssa.h and tree-cfgcleanup.h.
383 (cleanup_tree_cfg_1): After cleanup_control_flow_pre update SSA
385 (cleanup_tree_cfg_noloop): Pass down SSA update flags.
386 (cleanup_tree_cfg): Get and pass down SSA update flags.
388 2019-02-18 Richard Biener <rguenther@suse.de>
390 PR tree-optimization/89296
391 * tree-ssa-loop-ch.c (ch_base::copy_headers): Restrict setting
392 of no-warning flag to cases that might emit the bogus warning.
394 2019-03-13 Andreas Krebbel <krebbel@linux.ibm.com>
396 Backport from mainline
397 2019-03-11 Andreas Krebbel <krebbel@linux.ibm.com>
399 * config/s390/s390.c (s390_expand_builtin): Do the copy_to_reg not
400 only on the else branch.
402 2019-03-13 Andreas Krebbel <krebbel@linux.ibm.com>
404 Backport from mainline
405 2019-02-07 Andreas Krebbel <krebbel@linux.ibm.com>
407 * config/s390/s390-builtin-types.def: Add new types.
408 * config/s390/s390-builtins.def: (s390_vec_xl, s390_vec_xld2)
409 (s390_vec_xlw4): Make the memory operand into a const pointer.
410 (s390_vec_xld2, s390_vec_xlw4): Add a variant for single precision
412 * config/s390/s390-c.c (s390_expand_overloaded_builtin): Generate
413 a new vector type with the alignment of the scalar memory operand.
415 2019-03-12 Richard Biener <rguenther@suse.de>
417 PR tree-optimization/89664
418 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Properly
419 free the occurance tree after the early out.
421 2019-03-12 Andre Vieira <andre.simoesdiasvieira@arm.com>
423 Backport from mainline
424 2019-03-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
426 * config/arm/arm.h (TARGET_FP16_TO_DOUBLE): Add TARGET_VFP_DOUBLE
429 2019-03-11 Eric Botcazou <ebotcazou@adacore.com>
431 PR rtl-optimization/89588
432 * loop-unroll.c (decide_unroll_constant_iterations): Make guard for
433 explicit unrolling factor more robust.
435 2019-03-11 Martin Liska <mliska@suse.cz>
437 Backport from mainline
438 2019-03-08 Martin Liska <mliska@suse.cz>
441 * config/i386/i386.c (ix86_option_override_internal): Disable
442 jump tables when retpolines are used.
444 2019-03-11 Martin Liska <mliska@suse.cz>
446 Backport from mainline
447 2019-02-11 David Malcolm <dmalcolm@redhat.com>
450 * input.c (selftest::test_line_offset_overflow): New selftest.
451 (selftest::input_c_tests): Call it.
453 2019-03-08 Martin Jambor <mjambor@suse.cz>
455 Backport from mainline
456 2019-03-07 Martin Jambor <mjambor@suse.cz>
459 * cgraph.h (cgraph_node): New inline method former_thunk_p.
460 * cgraph.c (cgraph_node::dump): Dump a note if node is a former thunk.
461 (clone_of_p): Treat expanded thunks like thunks, be optimistic if they
462 have multiple callees. At the end check if declarations match as
463 opposed to cgraph_nodes.
465 2019-03-06 Tamar Christina <tamar.christina@arm.com>
468 2019-02-25 Tamar Christina <tamar.christina@arm.com>
471 * common/config/aarch64/aarch64-common.c
472 (struct aarch64_option_extension): Add is_synthetic.
473 (all_extensions): Use it.
474 (TARGET_OPTION_INIT_STRUCT): Define hook.
475 (struct gcc_targetm_common): Moved to end.
476 (all_extensions_by_on): New.
477 (opt_ext_cmp, typedef opt_ext): New.
478 (aarch64_option_init_struct): New.
479 (aarch64_contains_opt): New.
480 (aarch64_get_extension_string_for_isa_flags): Output smallest set.
481 * config/aarch64/aarch64-option-extensions.def
482 (AARCH64_OPT_EXTENSION): Explicitly include AES and SHA2 in crypto.
483 (fp, simd, crc, lse, fp16, rcpc, rdma, dotprod, aes, sha2, sha3,
485 Set is_synthetic to false.
486 (crypto): Set is_synthetic to true.
487 * config/aarch64/driver-aarch64.c (AARCH64_OPT_EXTENSION): Add
490 2019-03-06 Tamar Christina <tamar.christina@arm.com>
493 2019-02-28 Tamar Christina <tamar.christina@arm.com>
496 * config/aarch64/aarch64-option-extensions.def: Document it.
497 * config/aarch64/driver-aarch64.c (host_detect_local_cpu): Skip feature
500 2019-03-06 Xiong Hu Luo <luoxhu@linux.ibm.com>
502 Backport of r268834 from mainline to gcc-8-branch.
503 2019-02-13 Xiong Hu Luo <luoxhu@linux.vnet.ibm.com>
505 * config/rs6000/altivec.h (vec_sbox_be, vec_cipher_be,
506 vec_cipherlast_be, vec_ncipher_be, vec_ncipherlast_be): New #defines.
507 * config/rs6000/crypto.md (CR_vqdi): New define_mode_iterator.
508 (crypto_vsbox_<mode>, crypto_<CR_insn>_<mode>): New define_insns.
509 * config/rs6000/rs6000-builtin.def (VSBOX_BE): New BU_CRYPTO_1.
510 (VCIPHER_BE, VCIPHERLAST_BE, VNCIPHER_BE, VNCIPHERLAST_BE):
512 * config/rs6000/rs6000.c (builtin_function_type)
513 <CRYPTO_BUILTIN_VSBOX_BE, CRYPTO_BUILTIN_VCIPHER_BE,
514 CRYPTO_BUILTIN_VCIPHERLAST_BE, CRYPTO_BUILTIN_VNCIPHER_BE,
515 CRYPTO_BUILTIN_VNCIPHERLAST_BE>: New switch options.
516 * doc/extend.texi (vec_sbox_be, vec_cipher_be, vec_cipherlast_be,
517 vec_ncipher_be, vec_ncipherlast_be): New builtin functions.
519 2019-03-05 Richard Biener <rguenther@suse.de>
521 Backport from mainline
522 2019-02-26 Richard Biener <rguenther@suse.de>
524 PR tree-optimization/89505
525 * tree-ssa-structalias.c (compute_dependence_clique): Make sure
526 to handle restrict pointed-to vars with multiple subvars
529 2019-03-01 Tamar Christina <tamar.christina@arm.com>
532 * config/aarch64/aarch64-option-extensions.def (fp, simd, crypto,
533 fp16): Collapse line.
535 2019-03-01 Richard Biener <rguenther@suse.de>
537 Backport from mainline
538 2019-03-01 Jakub Jelinek <jakub@redhat.com>
541 * dwarf2out.c (output_comdat_type_unit): Add ATTRIBUTE_UNUSED to
542 early_lto_debug argument.
544 2019-02-27 Richard Biener <rguenther@suse.de>
547 * dwarf2out.c (use_debug_types): Disable when in_lto_p.
549 2019-02-27 Richard Biener <rguenther@suse.de>
552 * dwarf2out.c (output_comdat_type_unit): Add early_lto_debug
553 parameter, prefix section name with .gnu.debuglto_ if true.
554 (dwarf2out_finish): Pass false to output_comdat_type_unit.
555 (dwarf2out_early_finish): Pass true to output_comdat_type_unit.
557 2019-02-27 Richard Biener <rguenther@suse.de>
560 * dwarf2out.c (size_of_die): Key on AT_ref (a)->comdat_type_p
561 rather than on use_debug_types, doing what output_die does.
562 (value_format): Likewise.
564 2019-02-28 Eric Botcazou <ebotcazou@adacore.com>
566 PR tree-optimization/89536
567 * tree-ssa-dom.c (edge_info::derive_equivalences) <BIT_NOT_EXPR>: Test
568 only whether bit #0 of the value is 0 instead of the entire value.
570 2019-02-28 Li Jia He <helijia@linux.ibm.com>
573 2019-02-20 Li Jia He <helijia@linux.ibm.com>
576 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin)
577 <case ALTIVEC_BUILTIN_VSPLTISB, ALTIVEC_BUILTIN_VSPLTISH,
578 ALTIVEC_BUILTIN_VSPLTISW>: Don't convert the operand before
581 2019-02-27 Uroš Bizjak <ubizjak@gmail.com>
584 * config/i386/i386.c (ix86_atomic_assign_expand_fenv): Check
585 TARGET_SSE in addition to TARGET_SSE_MATH.
587 (ix86_excess_precision): Ditto.
588 (ix86_float_exceptions_rounding_supported_p): Ditto.
589 (use_rsqrt_p): Ditto.
590 * config/i386/sse.md (rsqrt<mode>2): Ditto.
592 2019-02-27 Jakub Jelinek <jakub@redhat.com>
594 Backported from mainline
595 2019-02-18 Jakub Jelinek <jakub@redhat.com>
598 * config/s390/s390.c (s390_indirect_branch_attrvalue,
599 s390_indirect_branch_settings): Define unconditionally.
600 (s390_set_current_function): Likewise, but guard the whole body except
601 the s390_indirect_branch_settings call with
602 #if S390_USE_TARGET_ATTRIBUTE.
603 (TARGET_SET_CURRENT_FUNCTION): Redefine unconditionally.
605 2019-02-26 Eric Botcazou <ebotcazou@adacore.com>
607 Backport from mainline
608 2019-02-19 Eric Botcazou <ebotcazou@adacore.com>
610 * rtlanal.c (get_initial_register_offset): Fall back to the estimate
611 as long as the epilogue isn't completed.
613 2019-02-26 Eric Botcazou <ebotcazou@adacore.com>
615 * tree-ssa-dom.c (edge_info::derive_equivalences) <BIT_IOR_EXPR>: Fix
616 and move around comment.
617 <BIT_AND_EXPR>: Likewise.
618 <BIT_NOT_EXPR>: Add specific handling for boolean types.
620 2019-02-24 Alan Modra <amodra@gmail.com>
623 * config/rs6000/rs6000.md (<bd>_<mode> split): Check for an int
624 output reg on add insn.
625 (<bd>tf_<mode> split): Likewise. Match predicates with insn.
627 2019-02-22 Jakub Jelinek <jakub@redhat.com>
629 * BASE-VER: Set to 8.3.1.
631 2019-02-22 Release Manager
633 * GCC 8.3.0 released.
635 2019-02-15 Richard Biener <rguenther@suse.de>
636 Jakub Jelinek <jakub@redhat.com>
638 PR tree-optimization/89278
639 * tree-loop-distribution.c: Include tree-eh.h.
640 (generate_memset_builtin, generate_memcpy_builtin): Call
641 rewrite_to_non_trapping_overflow on builtin->size before passing it
642 to force_gimple_operand_gsi.
644 2019-02-15 Jakub Jelinek <jakub@redhat.com>
647 * optc-save-gen.awk: Handle optimize_fast like optimize_size or
649 * opth-gen.awk: Likewise.
651 2019-02-14 David Malcolm <dmalcolm@redhat.com>
653 Backport of r267957 from trunk.
654 2019-01-15 David Malcolm <dmalcolm@redhat.com>
657 * tree.c (build_function_type): Assert that arg_types is not
660 2019-02-14 Jakub Jelinek <jakub@redhat.com>
662 PR rtl-optimization/89354
663 * combine.c (make_extraction): Punt if extraction_mode is narrower
666 2019-02-14 David Malcolm <dmalcolm@redhat.com>
668 Backport of r263339 from trunk.
669 2018-08-06 David Malcolm <dmalcolm@redhat.com>
671 * function-tests.c (selftest::test_expansion_to_rtl): Call
672 free_after_compilation.
674 2019-02-14 David Malcolm <dmalcolm@redhat.com>
676 Backport of r263295 from trunk.
677 2018-08-03 David Malcolm <dmalcolm@redhat.com>
679 * doc/gcov.texi (-x): Remove duplicate "to".
680 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
681 (-Wif-not-aligned): Remove duplicate "is".
682 (-flto): Remove duplicate "the".
683 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
685 (MSP430 Options): Remove duplicate "and" from the description
686 of "-mgprel-sec=regexp".
687 (x86 Options): Remove duplicate copies of "vmldLog102" and
688 vmlsLog104 from description of "-mveclibabi=type".
690 2019-02-14 David Malcolm <dmalcolm@redhat.com>
692 Backport of r263275 from trunk.
693 2018-08-02 David Malcolm <dmalcolm@redhat.com>
695 * diagnostic-show-locus.c (diagnostic_show_locus): Use
696 pp_take_prefix when saving the existing prefix.
697 * diagnostic.c (diagnostic_append_note): Likewise.
698 * langhooks.c (lhd_print_error_function): Likewise.
699 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
700 param's type. Free the existing prefix.
701 (pp_take_prefix): New function.
702 (pretty_printer::pretty_printer): Drop the prefix parameter.
703 Rename the length parameter to match the comment.
704 (pretty_printer::~pretty_printer): Free the prefix.
705 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
707 (struct pretty_printer): Drop the "const" from "prefix" field's
708 type and clarify memory management.
709 (pp_set_prefix): Drop the "const" from the 2nd param.
710 (pp_take_prefix): New decl.
712 2019-02-14 Segher Boessenkool <segher@kernel.crashing.org>
715 2018-08-31 Segher Boessenkool <segher@kernel.crashing.org>
719 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
721 2019-02-14 Segher Boessenkool <segher@kernel.crashing.org>
724 2019-01-18 Segher Boessenkool <segher@kernel.crashing.org>
727 * config/rs6000/rs6000.md (*movsi_from_df): Allow only register
730 2019-02-14 Segher Boessenkool <segher@kernel.crashing.org>
733 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
735 PR rtl-optimization/85805
736 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
737 value for hard registers if that was written in the same mode.
739 2019-02-14 Martin Liska <mliska@suse.cz>
741 Backport from mainline
742 2019-02-14 Martin Liska <mliska@suse.cz>
744 PR rtl-optimization/89242
745 * dce.c (delete_unmarked_insns): Call free_dominance_info we
746 process a transformation.
748 2019-02-14 Martin Liska <mliska@suse.cz>
750 Backport from mainline
751 2019-02-13 Martin Liska <mliska@suse.cz>
754 * cfgrtl.c (remove_barriers_from_footer): New function.
755 (try_redirect_by_replacing_jump): Use it.
756 (cfg_layout_redirect_edge_and_branch): Likewise.
758 2019-02-14 Martin Liska <mliska@suse.cz>
760 Backport from mainline
761 2019-02-11 Martin Liska <mliska@suse.cz>
764 * ipa-cp.c (build_toporder_info): Remove usage of a param.
765 * ipa-inline.c (inline_small_functions): Likewise.
766 * ipa-pure-const.c (propagate_pure_const): Likewise.
767 (propagate_nothrow): Likewise.
768 * ipa-reference.c (propagate): Likewise.
769 * ipa-utils.c (struct searchc_env): Remove unused field.
770 (searchc): Always search across AVAIL_INTERPOSABLE.
771 (ipa_reduced_postorder): Always allow AVAIL_INTERPOSABLE as
772 the only called IPA pure const can properly not propagate
773 across interposable boundary.
774 * ipa-utils.h (ipa_reduced_postorder): Remove param.
776 2019-02-14 Jakub Jelinek <jakub@redhat.com>
778 PR tree-optimization/89314
779 * fold-const.c (fold_binary_loc): Cast strlen argument to
780 const char * before dereferencing it. Formatting fixes.
782 Backported from mainline
783 2019-02-13 Jakub Jelinek <jakub@redhat.com>
786 * tree-ssa-structalias.c (set_uids_in_ptset): Or in vi->is_heap_var
787 into pt->vars_contains_escaped_heap instead of setting
788 pt->vars_contains_escaped_heap to it.
791 * optabs.c (prepare_cmp_insn): Use UINTVAL (size) instead of
792 INTVAL (size), compare it to GET_MODE_MASK instead of
793 1 << GET_MODE_BITSIZE.
796 * config/i386/predicates.md (x86_64_immediate_operand): Allow
797 TLS UNSPECs offsetted by signed 32-bit CONST_INT even with
800 2019-02-09 Jakub Jelinek <jakub@redhat.com>
803 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
804 If !node->definition and TYPE_ARG_TYPES is non-NULL, use
805 TYPE_ARG_TYPES instead of DECL_ARGUMENTS.
807 2019-02-13 Bill Schmidt <wschmidt@linux.ibm.com>
809 Backport from mainline
810 2019-02-11 Bill Schmidt <wschmidt@linux.ibm.com>
812 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Shift-right
813 and shift-left vector built-ins need to include a TRUNC_MOD_EXPR
814 for correct semantics.
816 2019-02-13 Richard Biener <rguenther@suse.de>
818 Backport from mainline
819 2019-01-08 Richard Biener <rguenther@suse.de>
821 PR tree-optimization/86554
822 * tree-ssa-sccvn.c (visit_nary_op): When value-numbering to
823 expressions with different overflow behavior make sure there's an
824 available expression on the path.
826 2019-02-13 Martin Liska <mliska@suse.cz>
829 * lto-streamer.h (LTO_minor_version): Bump version due
832 2019-02-13 Richard Biener <rguenther@suse.de>
834 Backport from mainline
835 2019-02-12 Richard Biener <rguenther@suse.de>
837 PR tree-optimization/89253
838 * tree-ssa-loop-split.c (tree_ssa_split_loops): Check we can
841 2019-02-08 Richard Biener <rguenther@suse.de>
844 * tree-data-ref.c (initialize_matrix_A): Fail if constant
846 (analyze_subscript_affine_affine): Handle failure from
849 2019-01-28 Richard Biener <rguenther@suse.de>
851 PR tree-optimization/88739
852 * tree-ssa-sccvn.c (vn_reference_lookup_3): Avoid generating
853 BIT_FIELD_REFs of non-mode-precision integral operands.
855 2019-02-12 Jan Hubicka <hubicka@ucw.cz>
857 Backport from mainline:
858 2019-02-09 Jan Hubicka <hubicka@ucw.cz>
860 * ipa-visibility.c (localize_node): Also do not localize
861 LDPR_PREVAILING_DEF_IRONLY_EXP.
863 2019-02-11 Stefan Agner <stefan@agner.ch>
865 Backport from mainline.
866 2019-01-10 Stefan Agner <stefan@agner.ch>
869 * config/arm/arm.c (arm_option_override_internal): Force
870 opts->x_inline_asm_unified to true only if TARGET_THUMB2_P.
872 2019-02-09 Jan Hubicka <hubicka@ucw.cz>
874 Backport from mainline:
875 2019-01-05 Jan Hubicka <hubicka@ucw.cz>
879 2017-05-22 Jan Hubicka <hubicka@ucw.cz>
881 * ipa-inline.c (edge_badness): Use inlined_time instead of
882 inline_summaries->get.
884 2019-02-09 Aaron Sawdey <acsawdey@linux.ibm.com>
886 Backported from mainline
887 2019-02-05 Aaron Sawdey <acsawdey@linux.ibm.com>
890 * config/rs6000/rs6000.md (<bd>tf_<mode>): Generate a local label
891 for the long branch case.
893 2019-02-05 Aaron Sawdey <acsawdey@linux.ibm.com>
896 * config/rs6000/rs6000-string.c (do_ifelse, expand_cmp_vec_sequence,
897 expand_compare_loop, expand_block_compare_gpr,
898 expand_strncmp_align_check, expand_strncmp_gpr_sequence): Insert
899 REG_BR_PROB notes in inline expansion of memcmp/strncmp. Add
900 #include "profile-count.h" and "predict.h" for types and functions
901 needed to work with REG_BR_PROB notes.
903 2019-02-09 Aaron Sawdey <acsawdey@linux.ibm.com>
905 * config/rs6000/rs6000-string.c (expand_compare_loop,
906 expand_block_compare): Insert REG_BR_PROB notes in inline expansion of
909 2019-02-09 Jakub Jelinek <jakub@redhat.com>
911 Backported from mainline
912 2019-01-16 David Malcolm <dmalcolm@redhat.com>
915 * combine.c (delete_noop_moves): Convert to "bool" return,
916 returning true if any edges are eliminated.
917 (combine_instructions): Also return true if delete_noop_moves
920 2019-02-08 Jakub Jelinek <jakub@redhat.com>
922 PR rtl-optimization/89234
923 * except.c (copy_reg_eh_region_note_forward): Return if note_or_insn
924 is a NOTE, CODE_LABEL etc. - rtx_insn * other than INSN_P.
925 (copy_reg_eh_region_note_backward): Likewise.
927 2019-02-09 Alan Modra <amodra@gmail.com>
930 * config/rs6000/rs6000.c (save_reg_p): Match logic in
931 rs6000_emit_prologue emitting pic_offset_table setup.
933 2019-02-08 Richard Biener <rguenther@suse.de>
935 Backport from mainline
936 2018-11-13 Richard Biener <rguenther@suse.de>
938 PR tree-optimization/86991
939 * tree-vect-loop.c (vect_is_slp_reduction): Delay reduction
940 group building until we have successfully detected the SLP
942 (vect_is_simple_reduction): Remove fixup code here.
944 2019-02-08 Richard Biener <rguenther@suse.de>
946 Backport from mainline
947 2019-02-03 Richard Biener <rguenther@suse.de>
950 * dwarf2out.c (copy_ancestor_tree): Register non-stubs as
953 2019-01-29 Richard Biener <rguenther@suse.de>
956 * dwarf2out.c (collect_skeleton_dies): New helper.
957 (copy_decls_for_unworthy_types): Call it.
958 (build_abbrev_table): Do not try to replace
959 DW_AT_signature refs with local refs.
961 2019-01-17 Richard Biener <rguenther@suse.de>
964 * dwarf2out.c (want_pubnames): Never generate pubnames sections
965 and friends for the LTO part of debug info.
967 2019-01-24 Richard Biener <rguenther@suse.de>
970 * tree-streamer-out.c (write_ts_decl_common_tree_pointers):
971 When in "legacy" debug mode make sure to reset self-origins.
973 2019-01-15 Richard Biener <rguenther@suse.de>
976 * dwarf2out.c (gen_member_die): Do not generate inheritance
979 2018-11-22 Richard Biener <rguenther@suse.de>
983 * lto-streamer-out.c (lto_is_streamable): Allow CALL_EXPRs
984 which can appear in size expressions.
985 * tree-streamer-in.c (unpack_ts_base_value_fields): Stream
986 CALL_EXPR_BY_DESCRIPTOR.
987 (streamer_read_tree_bitfields): Stream CALL_EXPR_IFN.
988 * tree-streamer-out.c (pack_ts_base_value_fields): Stream
989 CALL_EXPR_BY_DESCRIPTOR.
990 (streamer_write_tree_bitfields): Stream CALL_EXPR_IFN.
992 2019-02-08 Christophe Lyon <christophe.lyon@linaro.org>
994 Backport from mainline
995 2019-01-07 Christophe Lyon <christophe.lyon@linaro.org>
998 * doc/install.texi (with-multilib-list): Document for aarch64.
1000 2019-02-08 Richard Biener <rguenther@suse.de>
1002 Backport from mainline
1003 2018-12-10 Richard Biener <rguenther@suse.de>
1005 PR tree-optimization/88427
1006 * vr-values.c (vr_values::extract_range_from_phi_node):
1007 Handle symbolic ranges conservatively when trying to drop
1010 2018-11-28 Richard Biener <rguenther@suse.de>
1012 PR tree-optimization/88217
1013 * vr-values.c (vr_values::extract_range_from_phi_node): Make
1014 sure to handle results > +INF and < -INF correctly when
1015 trying to drop down to +INF - 1 or -INF + 1.
1017 2018-11-23 Richard Biener <rguenther@suse.de>
1019 PR tree-optimization/88149
1020 * tree-vect-slp.c (vect_slp_analyze_node_operations): Detect
1021 the case where there are two different def types for the
1022 same operand at different operand position in the same stmt.
1024 2019-02-07 Jakub Jelinek <jakub@redhat.com>
1026 Backported from mainline
1027 2019-02-05 Jakub Jelinek <jakub@redhat.com>
1030 * dce.c (delete_unmarked_insns): Don't remove no-op moves if they
1031 can throw, non-call exceptions are enabled and we can't delete
1032 dead exceptions or alter cfg. Set must_clean if
1033 delete_insn_and_edges returns true, don't set it blindly for calls.
1035 PR rtl-optimization/89195
1036 * combine.c (make_extraction): For MEMs, don't extract bytes outside
1037 of the original MEM.
1040 * optabs.c (prepare_cmp_insn): Pass x and y to
1041 emit_block_comp_via_libcall rather than XEXP (x, 0) and XEXP (y, 0).
1043 2019-02-02 Jakub Jelinek <jakub@redhat.com>
1046 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
1047 Punt with warning on aggregate return or argument types. Ignore
1048 type/mode checking for uniform arguments.
1050 2019-02-01 Jakub Jelinek <jakub@redhat.com>
1052 PR tree-optimization/88107
1053 * tree-cfg.c (find_outermost_region_in_block): Add ALL argument,
1054 instead of assertion that eh_region_outermost is non-NULL, if it
1055 is NULL, set *ALL to true and return NULL.
1056 (move_sese_region_to_fn): Adjust caller, if all is set, call
1057 duplicate_eh_regions with NULL region.
1059 2019-01-31 Jakub Jelinek <jakub@redhat.com>
1062 * ipa-inline.c (sanitize_attrs_match_for_inline_p): Allow inlining
1063 always_inline callees into no_sanitize_address callers.
1065 2019-01-30 Jakub Jelinek <jakub@redhat.com>
1068 * config/i386/i386.c (ix86_warn_parameter_passing_abi): Don't warn
1069 for arguments to functions that are TU-local and shouldn't be
1070 referenced by assembly.
1072 2019-01-29 Jakub Jelinek <jakub@redhat.com>
1076 * omp-simd-clone.c (simd_clone_clauses_extract)
1077 <case OMP_CLAUSE_ALIGNED>: Ignore clauses with NULL
1078 OMP_CLAUSE_ALIGNED_ALIGNMENT.
1080 2019-01-28 Jakub Jelinek <jakub@redhat.com>
1083 * gimplify.c (gimplify_omp_for): When adding OMP_CLAUSE_*_GIMPLE_SEQ
1084 for lastprivate/linear IV, push gimplify context around gimplify_assign
1085 and, if it needed any temporaries, pop it into a gimple bind around the
1089 * doc/invoke.texi (-mclwb, -mprfchw, -mrdpid, -mrdseed, -msgx,
1090 -madx, -mhle, -mavx5124fmaps, -mavx512vnni, -mavx5124vnniw): Document
1092 (bmi2): Add missing @opindex.
1093 * doc/extend.texi (x86 target attribute): Move fma4, lwp, ssse3
1094 options alphabetically. Add missing 3dnow, 3dnowa, adx, avx, avx2,
1095 avx5124fmaps, avx5124vnniw, avx512bitalg, avx512bw, avx512cd,
1096 avx512dq, avx512er, avx512f, avx512ifma, avx512pf, avx512vbmi,
1097 avx512vbmi2, avx512vl, avx512vnni, avx512vpopcntdq, bmi, bmi2,
1098 clflushopt, clwb, clzero, crc32, cx16, f16c, fma, fsgsbase,
1099 fxsr, gfni, hle, lzcnt, movbe, movdir64b, movdiri, mwaitx, pconfig,
1100 pku, prefetchwt1, prfchw, rdpid, rdrnd, rdseed, rtm, sahf,
1101 sgx, sha, shstk, tbm, vaes, vpclmulqdq, wbnoinvd, xsave,
1102 xsavec, xsaveopt and xsaves options.
1104 2019-01-27 Jakub Jelinek <jakub@redhat.com>
1107 * config/i386/sse.md
1108 (<mask_codefor>avx512dq_shuf_<shuffletype>64x2_1<mask_name>,
1109 avx512f_shuf_<shuffletype>64x2_1<mask_name>): Ensure the
1110 first constants in pairs are multiples of 2. Formatting fixes.
1111 (avx512vl_shuf_<shuffletype>32x4_1<mask_name>,
1112 avx512vl_shuf_<shuffletype>32x4_1<mask_name>): Ensure the
1113 first constants in each quadruple are multiples of 4. Formatting fixes.
1115 2019-01-24 Jakub Jelinek <jakub@redhat.com>
1117 PR tree-optimization/88964
1118 * gimple-loop-interchange.cc (loop_cand::analyze_induction_var): Also
1119 punt if HONOR_SNANS (chrec).
1121 2019-01-23 Jakub Jelinek <jakub@redhat.com>
1123 PR tree-optimization/88964
1124 * gimple-loop-interchange.cc (loop_cand::analyze_induction_var): Use
1125 build_zero_cst instead of build_int_cst. Return false for loop
1126 invariants which honor signed zeros.
1128 2019-01-22 Jakub Jelinek <jakub@redhat.com>
1131 * config/rs6000/rs6000.c: Include tree-vrp.h and tree-ssanames.h.
1132 (rs6000_gimple_fold_builtin): If MEM_REF address doesn't satisfy
1133 is_gimple_mem_ref_addr predicate, force it into a SSA_NAME first.
1136 * gimplify.c (gimplify_omp_atomic): Handle bitfield atomics with
1137 non-integral DECL_BIT_FIELD_REPRESENTATIVEs.
1140 * optabs.c (add_equal_note): Add op0_mode argument, use it instead of
1142 (expand_binop_directly, expand_doubleword_clz,
1143 expand_doubleword_popcount, expand_ctz, expand_ffs,
1144 expand_unop_direct, maybe_emit_unop_insn): Adjust callers.
1146 PR rtl-optimization/49429
1148 PR rtl-optimization/86334
1150 * expr.c (emit_block_move_hints): Move marking of MEM_EXPRs
1151 addressable from here...
1152 (emit_block_op_via_libcall): ... to here.
1154 2019-01-17 Jakub Jelinek <jakub@redhat.com>
1157 * config/aarch64/arm_neon.h: Fix #pragma GCC target syntax - replace
1158 (("..."))) with ("..."). Use arch=armv8.2-a+sha3 instead of
1159 arch=armv8.2-a+crypto for vsha512hq_u64 etc. intrinsics.
1161 PR rtl-optimization/88870
1162 * dce.c (deletable_insn_p): Never delete const/pure calls that can
1163 throw if we can't alter the cfg or delete dead exceptions.
1164 (mark_insn): Don't call find_call_stack_args for such calls.
1166 2019-01-11 Jakub Jelinek <jakub@redhat.com>
1169 * passes.c (finish_optimization_passes): Call print_combine_total_stats
1170 inside of pass_combine_1 dump rather than pass_profile_1.
1172 2019-01-10 Jakub Jelinek <jakub@redhat.com>
1175 * attribs.c (handle_dll_attribute): Clear TREE_STATIC after setting
1178 2019-01-07 Jakub Jelinek <jakub@redhat.com>
1181 * dwarf2out.c (const_ok_for_output_1): Remove redundant call to
1182 const_not_ok_for_debug_p target hook.
1183 (mem_loc_descriptor) <case UNSPEC>: Only call const_ok_for_output_1
1184 on UNSPEC and subexpressions thereof if all subexpressions of the
1185 UNSPEC are CONSTANT_P.
1187 2019-01-05 Jakub Jelinek <jakub@redhat.com>
1190 * dwarf2out.c (const_ok_for_output_1): Reject MINUS that contains
1191 SYMBOL_REF, CODE_LABEL or UNSPEC in subexpressions of second argument.
1192 Reject PLUS that contains SYMBOL_REF, CODE_LABEL or UNSPEC in
1193 subexpressions of both operands.
1194 (mem_loc_descriptor): Handle UNSPEC if target hook acks it and all the
1195 subrtxes are CONSTANT_P.
1197 2019-02-07 Richard Biener <rguenther@suse.de>
1199 Backport from mainline
1200 2018-11-20 Richard Biener <rguenther@suse.de>
1202 PR tree-optimization/88105
1203 * tree-ssa-dom.c (pass_dominator::execute): Do not walk
1206 2018-11-28 Richard Biener <rguenther@suse.de>
1208 PR tree-optimization/88223
1209 * tree-ssa-sccvn.c (vn_reference_lookup_3): When skipping
1210 over a stored-same value may-alias store make sure to consider
1211 partial overlaps which are valid when TBAA reasonings do not
1212 apply and byte-granular overlaps are possible at all.
1214 2019-02-07 Richard Biener <rguenther@suse.de>
1216 Backport from mainline
1217 2018-12-11 Richard Biener <rguenther@suse.de>
1221 * tree-complex.c (update_complex_assignment): Properly transfer
1222 or clean EH info around gimple_assign_set_rhs_with_ops.
1224 2018-11-15 Richard Biener <rguenther@suse.de>
1226 PR tree-optimization/88030
1227 * tree-complex.c (need_eh_cleanup): New global.
1228 (update_complex_assignment): Mark blocks that need EH update.
1229 (expand_complex_comparison): Likewise.
1230 (tree_lower_complex): Allocate and deallocate need_eh_cleanup,
1231 perform EH cleanup and schedule CFG cleanup if that did anything.
1233 2018-11-08 Richard Biener <rguenther@suse.de>
1235 PR tree-optimization/87929
1236 * tree-complex.c (expand_complex_comparison): Clean EH.
1238 2019-02-06 Eric Botcazou <ebotcazou@adacore.com>
1240 * config/i386/i386.c (ix86_expand_prologue): Emit a memory blockage
1241 after restoring registers saved to allocate the frame on Windows.
1243 2019-02-06 Richard Biener <rguenther@suse.de>
1245 Backport from mainline
1246 2019-01-31 Richard Biener <rguenther@suse.de>
1248 PR tree-optimization/89135
1249 * tree-ssa-phiprop.c (pass_phiprop::execute): Skip blocks
1250 with abnormal preds.
1252 2019-01-18 Richard Biener <rguenther@suse.de>
1254 PR tree-optimization/88903
1255 * tree-vect-stmts.c (vectorizable_shift): Verify we see all
1256 scalar stmts a SLP shift amount is composed of when detecting
1259 2019-02-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
1261 Backport from mainline.
1262 2019-01-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
1263 * config/rs6000/rs6000-c.c (altivec-resolve_overloaded_builtin):
1264 Change handling of ALTIVEC_BUILTIN_VEC_EXTRACT. Coerce result to
1265 type of vector element when vec_extract is implemented by direct
1268 2019-02-05 Andreas Krebbel <krebbel@linux.ibm.com>
1270 Backport from mainline
1271 2019-02-05 Andreas Krebbel <krebbel@linux.ibm.com>
1274 * config/s390/s390.md: Remove load and test FP splitter.
1276 2019-02-04 Bill Schmidt <wshmidt@linux.ibm.com>
1279 Backport from mainline
1281 2019-01-30 Bill Schmidt <wschmidt@linux.ibm.com>
1284 * config/rs6000/vsx.md (*vsx_reduc_<VEC_reduc_name>_v4sf_scalar):
1285 Disable for little-endian.
1287 2019-01-22 Jakub Jelinek <jakub@redhat.com>
1290 * config/rs6000/vsx.md (*vsx_reduc_<VEC_reduc_name>_v2df_scalar):
1291 Disable for little endian.
1293 2019-02-04 Eric Botcazou <ebotcazou@adacore.com>
1295 * config/sparc/sparc.h: Remove superfluous blank lines.
1296 * config/sparc/sparc.c (global_offset_table_rtx): Rename into...
1297 (got_register_rtx): ...this.
1298 (sparc_got): Adjust to above renaming.
1299 (sparc_tls_got): Likewise.
1300 (sparc_output_mi_thunk): Likewise.
1301 (sparc_init_pic_reg): Likewise.
1302 (save_local_or_in_reg_p): Fix test on the GOT register.
1303 (USE_HIDDEN_LINKONCE): Move around.
1304 (get_pc_thunk_name): Likewise.
1305 (gen_load_pcrel_sym): Likewise.
1306 (load_got_register): Likewise.
1308 2019-02-01 Bin Cheng <bin.cheng@linux.alibaba.com>
1310 Backport from mainline
1311 2018-10-15 Bin Cheng <bin.cheng@linux.alibaba.com>
1313 PR tree-optimization/87022
1314 * tree-loop-distribution.c (pg_add_dependence_edges): Check all
1315 bits in dist vector rather than the first one.
1317 2019-02-01 Bin Cheng <bin.cheng@linux.alibaba.com>
1319 Backport from mainline
1320 2019-02-01 Bin Cheng <bin.cheng@linux.alibaba.com>
1322 PR tree-optimization/88932
1323 * tree-predcom.c (try_combine_chains): Get loop bbs in dom order.
1325 2018-01-31 Bill Schmidt <wschmidt@linux.ibm.com>
1327 Backport from mainline
1328 2018-01-31 Bill Schmidt <wschmidt@linux.ibm.com>
1330 PR tree-optimization/89008
1331 * gimple-ssa-strength-reduction.c (slsr_process_mul): Don't
1332 process anything of the form X * 0.
1334 2019-01-31 Richard Biener <rguenther@suse.de>
1336 Backport from mainline
1337 2019-01-31 Richard Biener <rguenther@suse.de>
1339 PR rtl-optimization/89115
1340 * lra.c (lra_rtx_hash): Properly hash CONST_INT values.
1342 2019-01-30 Richard Biener <rguenther@suse.de>
1344 PR rtl-optimization/89115
1345 * opts.c (default_options_optimization): Reduce
1346 PARAM_MAX_DSE_ACTIVE_LOCAL_STORES by a factor of 10 at -O1.
1347 Make PARAM_LOOP_INVARIANT_MAX_BBS_IN_LOOP reduction relative
1350 2019-01-30 Kewen Lin <linkw@gcc.gnu.org>
1352 Backport from mainline.
1353 2019-01-17 Kewen Lin <linkw@gcc.gnu.org>
1355 * doc/extend.texi: Add four new prototypes for vec_ld and seven new
1356 prototypes for vec_st.
1357 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add entries
1358 for scalar address type variants of altivec_vec_ld/altivec_vec_st,
1359 mainly on signed/unsigned long long and double.
1361 2019-01-29 Vineet Gupta <vgupta@synopsys.com>
1363 Backport from mainline.
1364 * gcc/config.gcc: Force .init_array for ARC.
1366 2019-01-29 Vineet Gupta <vgupta@synopsys.com>
1368 Backport from mainline.
1369 * config/arc/atomic.md: Add operand to DMB instruction.
1371 2019-01-27 Uroš Bizjak <ubizjak@gmail.com>
1374 * rtl.h (prepare_copy_insn): New prototype.
1375 * gcse.c (prepare_copy_insn): New function, split out from
1376 process_insert_insn.
1377 (process_insert_insn): Use prepare_copy_insn.
1378 * store-motion.c (replace_store_insn): Use prepare_copy_insn
1379 instead of gen_move_insn.
1381 2019-01-25 Richard Biener <rguenther@suse.de>
1383 PR tree-optimization/86865
1384 * graphite-scop-detection.c (scop_detection::can_represent_loop):
1385 Reject non-do-while loops.
1387 2019-01-24 Uroš Bizjak <ubizjak@gmail.com>
1390 * config/i386/sse.md (sse2_cvtpi2pd): Add SSE alternatives.
1391 Disparage MMX alternative.
1392 (sse2_cvtpd2pi): Ditto.
1393 (sse2_cvttpd2pi): Ditto.
1395 2019-01-24 Richard Earnshaw <rearnsha@arm.com>
1398 * profile-count.h (profile_count): On ARM systems using GCC 6/7/8
1399 force the alignment of m_val.
1401 2019-01-24 Richard Biener <rguenther@suse.de>
1403 Backport from mainline
1404 2019-01-23 Richard Biener <rguenther@suse.de>
1406 PR tree-optimization/89008
1407 * tree-ssa-reassoc.c (eliminate_using_constants): For * 0 do
1408 not leave another stray operand.
1410 2019-01-22 Uroš Bizjak <ubizjak@gmail.com>
1413 * config/i386/i386.c (ix86_expand_builtin) [case IX86_BUILTIN_BEXTRI32,
1414 case IX86_BUILTIN_BEXTRI64]: Sanitize operands.
1416 2019-01-18 Richard Earnshaw <rearnsha@arm.com>
1419 * config/arm/arm-cpus.in (mp): New feature.
1421 (fgroup ARMv7ve): Add mp and sec features.
1422 (arch armv7-a): Add options to allow mp and sec extensions.
1423 (cpu generic-armv7-a): Add options to allow mp and sec extensions.
1424 (cpu cortex-a5, cpu cortex-7, cpu cortex-a9): Add mp and sec
1425 extenstions to the base architecture.
1426 (cpu cortex-a8): Add sec extension to the base architecture.
1427 (cpu marvell-pj4): Add mp and sec extensions to the base architecture.
1428 * config/arm/t-aprofile (MULTILIB_MATCHES): Map all armv7-a arch
1429 variants down to the base v7-a varaint.
1430 * config/arm/t-multilib (v7_a_arch_variants): New variable.
1431 * doc/invoke.texi (ARM Options): Add +mp and +sec to the list
1432 of permitted extensions for -march=armv7-a and for
1433 -mcpu=generic-armv7-a.
1435 2019-01-17 Uroš Bizjak <ubizjak@gmail.com>
1437 * config/alpha/alpha.c (alpha_gimplify_va_arg):
1438 Handle split indirect COMPLEX_TYPE arguments.
1440 2019-01-17 Eric Botcazou <ebotcazou@adacore.com>
1442 * gimple-ssa-isolate-paths.c (stmt_uses_name_in_undefined_way): Replace
1443 flag_non_call_exceptions with cfun->can_throw_non_call_exceptions.
1444 (stmt_uses_0_or_null_in_undefined_way): Likewise.
1445 * tree-ssa-alias.c (same_addr_size_stores_p): Likewise.
1447 2019-01-16 Aaron Sawdey <acsawdey@linux.ibm.com>
1449 Backported from mainline
1450 2019-01-03 Aaron Sawdey <acsawdey@linux.ibm.com>
1451 * config/rs6000/rs6000-string.c (expand_block_move): Don't use
1452 unaligned vsx and avoid lxvd2x/stxvd2x.
1453 (gen_lvx_v4si_move): New function.
1455 2019-01-16 Martin Jambor <mjambor@suse.cz>
1457 Backported from mainline
1458 2018-12-10 Martin Jambor <mjambor@suse.cz>
1461 * ipa-prop.c (determine_locally_known_aggregate_parts): Make sure
1462 we check pointers against pointers.
1464 2019-01-10 Jan Hubicka <hubicka@ucw.cz>
1466 Backported from mainline
1467 2018-12-27 Jan Hubicka <hubicka@ucw.cz>
1469 * ipa-devirt.c (polymorphic_call_target_d): Add n_odr_types.
1470 (polymorphic_call_target_hasher::hash): Hash it.
1471 (polymorphic_call_target_hasher::equal): Compare it.
1472 (possible_polymorphic_call_targets): Set it.
1473 * tree.c (free_lang_data): Rebuild type inheritance graph even on
1476 2019-01-03 Jan Hubicka <hubicka@ucw.cz>
1478 Backported from mainline
1479 2019-01-02 Richard Biener <rguenther@suse.de>
1482 * ipa-icf.h (sem_item_optimizer::sort_congruence_split): Declare.
1483 * ipa-icf.c (sem_item_optimizer::sort_congruence_split): New
1485 (sem_item_optimizer::do_congruence_step_f): Sort the congruence
1486 set after UIDs before splitting them.
1488 2019-01-03 Jan Hubicka <hubicka@ucw.cz>
1490 Backported from mainline
1491 2019-01-03 Jan Hubicka <hubicka@ucw.cz>
1493 PR tree-optimization/85574
1494 * tree-ssa-uncprop.c (struct equiv_hash_elt): Remove unused
1496 (struct ssa_equip_hash_traits): Declare.
1497 (val_ssa_equiv): Use custom hash traits using operand_equal_p.
1499 2019-01-09 Eric Botcazou <ebotcazou@adacore.com>
1501 * config/sparc/sparc.md (*tablejump_sp32): Merge into...
1502 (*tablejump_sp64): Likewise.
1503 (*tablejump<P:mode>): ...this.
1504 (*call_address_sp32): Merge into...
1505 (*call_address_sp64): Likewise.
1506 (*call_address<P:mode>): ...this.
1507 (*call_symbolic_sp32): Merge into...
1508 (*call_symbolic_sp64): Likewise.
1509 (*call_symbolic<P:mode>): ...this.
1510 (call_value): Remove constraint and add predicate.
1511 (*call_value_address_sp32): Merge into...
1512 (*call_value_address_sp64): Likewise.
1513 (*call_value_address<P:mode>): ...this.
1514 (*call_value_symbolic_sp32): Merge into...
1515 (*call_value_symbolic_sp64): Likewise.
1516 (*call_value_symbolic<P:mode>): ...this.
1517 (*sibcall_symbolic_sp32): Merge into...
1518 (*sibcall_symbolic_sp64): Likewise.
1519 (*sibcall_symbolic<P:mode>): ...this.
1520 (sibcall_value): Remove constraint and add predicate.
1521 (*sibcall_value_symbolic_sp32): Merge into...
1522 (*sibcall_value_symbolic_sp64): Likewise.
1523 (*sibcall_value_symbolic<P:mode>): ...this.
1524 (window_save): Minor tweak.
1525 (*branch_sp32): Merge into...
1526 (*branch_sp64): Likewise.
1527 (*branch<P:mode>): ...this.
1529 2019-01-09 Eric Botcazou <ebotcazou@adacore.com>
1530 James Clarke <jrtc27@jrtc27.com>
1533 * config/sparc/sparc.c (sparc_legitimize_tls_address): Only use Pmode
1534 consistently in TLS address generation and adjust code to the renaming
1535 of patterns. Mark calls to __tls_get_addr as const.
1536 * config/sparc/sparc.md (tgd_hi22): Turn into...
1537 (tgd_hi22<P:mode>): ...this and use Pmode throughout.
1538 (tgd_lo10): Turn into...
1539 (tgd_lo10<P:mode>): ...this and use Pmode throughout.
1540 (tgd_add32): Merge into...
1541 (tgd_add64): Likewise.
1542 (tgd_add<P:mode>): ...this and use Pmode throughout.
1543 (tldm_hi22): Turn into...
1544 (tldm_hi22<P:mode>): ...this and use Pmode throughout.
1545 (tldm_lo10): Turn into...
1546 (tldm_lo10<P:mode>): ...this and use Pmode throughout.
1547 (tldm_add32): Merge into...
1548 (tldm_add64): Likewise.
1549 (tldm_add<P:mode>): ...this and use Pmode throughout.
1550 (tldm_call32): Merge into...
1551 (tldm_call64): Likewise.
1552 (tldm_call<P:mode>): ...this and use Pmode throughout.
1553 (tldo_hix22): Turn into...
1554 (tldo_hix22<P:mode>): ...this and use Pmode throughout.
1555 (tldo_lox10): Turn into...
1556 (tldo_lox10<P:mode>): ...this and use Pmode throughout.
1557 (tldo_add32): Merge into...
1558 (tldo_add64): Likewise.
1559 (tldo_add<P:mode>): ...this and use Pmode throughout.
1560 (tie_hi22): Turn into...
1561 (tie_hi22<P:mode>): ...this and use Pmode throughout.
1562 (tie_lo10): Turn into...
1563 (tie_lo10<P:mode>): ...this and use Pmode throughout.
1564 (tie_ld64): Use DImode throughout.
1565 (tie_add32): Merge into...
1566 (tie_add64): Likewise.
1567 (tie_add<P:mode>): ...this and use Pmode throughout.
1568 (tle_hix22_sp32): Merge into...
1569 (tle_hix22_sp64): Likewise.
1570 (tle_hix22<P:mode>): ...this and use Pmode throughout.
1571 (tle_lox22_sp32): Merge into...
1572 (tle_lox22_sp64): Likewise.
1573 (tle_lox22<P:mode>): ...this and use Pmode throughout.
1574 (*tldo_ldub_sp32): Merge into...
1575 (*tldo_ldub_sp64): Likewise.
1576 (*tldo_ldub<P:mode>): ...this and use Pmode throughout.
1577 (*tldo_ldub1_sp32): Merge into...
1578 (*tldo_ldub1_sp64): Likewise.
1579 (*tldo_ldub1<P:mode>): ...this and use Pmode throughout.
1580 (*tldo_ldub2_sp32): Merge into...
1581 (*tldo_ldub2_sp64): Likewise.
1582 (*tldo_ldub2<P:mode>): ...this and use Pmode throughout.
1583 (*tldo_ldsb1_sp32): Merge into...
1584 (*tldo_ldsb1_sp64): Likewise.
1585 (*tldo_ldsb1<P:mode>): ...this and use Pmode throughout.
1586 (*tldo_ldsb2_sp32): Merge into...
1587 (*tldo_ldsb2_sp64): Likewise.
1588 (*tldo_ldsb2<P:mode>): ...this and use Pmode throughout.
1589 (*tldo_ldub3_sp64): Use DImode throughout.
1590 (*tldo_ldsb3_sp64): Likewise.
1591 (*tldo_lduh_sp32): Merge into...
1592 (*tldo_lduh_sp64): Likewise.
1593 (*tldo_lduh<P:mode>): ...this and use Pmode throughout.
1594 (*tldo_lduh1_sp32): Merge into...
1595 (*tldo_lduh1_sp64): Likewise.
1596 (*tldo_lduh1<P:mode>): ...this and use Pmode throughout.
1597 (*tldo_ldsh1_sp32): Merge into...
1598 (*tldo_ldsh1_sp64): Likewise.
1599 (*tldo_ldsh1<P:mode>): ...this and use Pmode throughout.
1600 (*tldo_lduh2_sp64): Use DImode throughout.
1601 (*tldo_ldsh2_sp64): Likewise.
1602 (*tldo_lduw_sp32): Merge into...
1603 (*tldo_lduw_sp64): Likewise.
1604 (*tldo_lduw<P:mode>): ...this and use Pmode throughout.
1605 (*tldo_lduw1_sp64): Use DImode throughout.
1606 (*tldo_ldsw1_sp64): Likewise.
1607 (*tldo_ldx_sp64): Likewise.
1608 (*tldo_stb_sp32): Merge into...
1609 (*tldo_stb_sp64): Likewise.
1610 (*tldo_stb<P:mode>): ...this and use Pmode throughout.
1611 (*tldo_sth_sp32): Merge into...
1612 (*tldo_sth_sp64): Likewise.
1613 (*tldo_sth<P:mode>): ...this and use Pmode throughout.
1614 (*tldo_stw_sp32): Merge into...
1615 (*tldo_stw_sp64): Likewise.
1616 (*tldo_stw<P:mode>): ...this and use Pmode throughout.
1617 (*tldo_stx_sp64): Use DImode throughout.
1619 2019-01-09 Eric Botcazou <ebotcazou@adacore.com>
1621 * doc/invoke.texi (-Os): Add reference to -finline-functions.
1622 (-finline-functions): Remove reference to -O2, add references
1623 to -fprofile-use and -fauto-profile.
1625 2019-01-08 Jakub Jelinek <jakub@redhat.com>
1627 Backported from mainline
1628 2019-01-05 Jakub Jelinek <jakub@redhat.com>
1632 * expr.c (expand_assignment): For calls returning VLA structures
1633 if to_rtx is not a MEM, force it into a stack temporary.
1635 2019-01-04 Jakub Jelinek <jakub@redhat.com>
1638 * config/i386/i386.c (ix86_expand_divmod_libfunc): Use mode instead
1639 of GET_MODE (opN) as modes of the libcall arguments.
1641 2019-01-03 Jakub Jelinek <jakub@redhat.com>
1644 * dwarf2out.c (modified_type_die): If type is equal to sizetype,
1645 change it to qualified_type.
1647 2018-12-21 Jakub Jelinek <jakub@redhat.com>
1651 * omp-expand.c (extract_omp_for_update_vars): Regimplify the condition
1653 (expand_omp_for_generic): Don't clobber t temporary for ordered loops.
1656 * config/i386/sse.md (*avx512pf_gatherpf<mode>sf_mask,
1657 *avx512pf_gatherpf<mode>df_mask, *avx512pf_scatterpf<mode>sf_mask,
1658 *avx512pf_scatterpf<mode>df_mask): Use %X5 instead of %5 for
1660 (gatherq_mode): Remove mode iterator.
1661 (*avx512f_gathersi<mode>, *avx512f_gathersi<mode>_2): Use X instead
1663 (*avx512f_gatherdi<mode>): Use X instead of <gatherq_mode>.
1664 (*avx512f_gatherdi<mode>_2, *avx512f_scattersi<mode>,
1665 *avx512f_scatterdi<mode>): Use %X5 for -masm=intel.
1667 PR rtl-optimization/88563
1668 * expr.c (expand_expr_real_2) <case WIDEN_MULT_EXPR>: Swap innermode
1669 and mode arguments to convert_modes. Likewise swap mode and word_mode
1670 arguments. Handle both arguments with VOIDmode before convert_modes
1671 of one of them. Formatting fixes.
1673 2018-12-19 Jakub Jelinek <jakub@redhat.com>
1676 * config/i386/vpclmulqdqintrin.h (_mm256_clmulepi64_epi128): Enable
1677 for -mavx -mvpclmulqdq rather than just for -mavx512vl -mvpclmulqdq.
1679 2018-12-14 Jakub Jelinek <jakub@redhat.com>
1683 * expr.h (categorize_ctor_elements): Add p_unique_nz_elts argument.
1684 * expr.c (categorize_ctor_elements_1): Likewise. Compute it like
1685 p_nz_elts, except don't multiply it by mult. Adjust recursive call.
1686 Fix up COMPLEX_CST handling.
1687 (categorize_ctor_elements): Add p_unique_nz_elts argument, initialize
1688 it and pass it through to categorize_ctor_elements_1.
1689 (mostly_zeros_p, all_zeros_p): Adjust categorize_ctor_elements callers.
1690 * gimplify.c (gimplify_init_constructor): Likewise. Don't force
1691 ctor into readonly data section if num_unique_nonzero_elements is
1692 smaller or equal to 1/8 of num_nonzero_elements and size is >= 64
1695 2018-12-13 Jakub Jelinek <jakub@redhat.com>
1697 PR rtl-optimization/88470
1698 * cfgcleanup.c (outgoing_edges_match): If the function is
1699 shrink-wrapped and bb1 ends with a JUMP_INSN with a single fake
1700 edge to EXIT, return false.
1702 PR rtl-optimization/88416
1703 * valtrack.c (cleanup_auto_inc_dec): Handle pre/post-inc/dec/modify
1704 even if !AUTO_INC_DEC.
1706 2018-12-07 Jakub Jelinek <jakub@redhat.com>
1709 * final.c (rest_of_handle_final): Don't call collect_fn_hard_reg_usage
1710 for functions with naked attribute.
1712 2018-12-06 Jakub Jelinek <jakub@redhat.com>
1715 * config/aarch64/aarch64.c (aarch64_print_address_internal): Don't
1716 call output_operand_lossage on VOIDmode CONST_INTs. After
1717 output_operand_lossage do return false.
1719 2018-12-05 Jakub Jelinek <jakub@redhat.com>
1721 PR tree-optimization/87360
1722 * gimple-loop-jam.c (tree_loop_unroll_and_jam): On failure to analyze
1723 data dependencies, don't return false, just continue.
1725 2018-11-29 Jakub Jelinek <jakub@redhat.com>
1728 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): For
1729 vec_add and vec_sub builtins, perform PLUS_EXPR or MINUS_EXPR
1730 in unsigned_type_for instead of vector integral type where overflow
1733 2018-11-26 Jakub Jelinek <jakub@redhat.com>
1736 * dwarf2out.c (secname_for_decl): For functions with
1737 DECL_SECTION_NAME if in_cold_section_p, try to return
1738 current_function_section's name if it is a named section.
1740 2018-11-22 Jakub Jelinek <jakub@redhat.com>
1744 * config/i386/i386.c (ix86_option_override_internal): Default
1745 ix86_stack_protector_guard to SSP_TLS only if TARGET_THREAD_SSP_OFFSET
1747 * config/i386/i386.md (stack_protect_set, stack_protect_set_<mode>,
1748 stack_protect_test, stack_protect_test_<mode>): Use empty condition
1749 instead of TARGET_SSP_TLS_GUARD.
1751 2018-11-20 Jakub Jelinek <jakub@redhat.com>
1753 PR tree-optimization/87895
1754 * omp-simd-clone.c (ipa_simd_modify_function_body): When removing
1755 or replacing GIMPLE_RETURN, set EDGE_FALLTHRU on the edge to EXIT.
1756 (simd_clone_adjust): Don't set EDGE_FALLTHRU here. In a loop that
1757 redirects edges to EXIT to edges to incr_bb, iterate while EXIT
1758 has any preds and always use EDGE_PRED (, 0).
1760 2018-11-19 Jakub Jelinek <jakub@redhat.com>
1762 PR tree-optimization/88071
1763 * tree-vect-loop.c (vectorize_fold_left_reduction): Pass true instead
1764 of false as last argument to gsi_remove.
1765 * tree-vect-stmts.c (vect_finish_replace_stmt): Pass true instead of
1766 false as last argument to gsi_replace.
1769 * omp-expand.c: Don't include debug.h.
1770 (adjust_context_and_scope): Add REGION argument. Find DECL_CONTEXT
1771 from innermost outer parallel, task, or target that has a
1772 child_fn set, or, if there is no such outer region, use
1773 current_function_decl. Do the DECL_CONTEXT adjustment regardless of
1774 whether a suitable BLOCK is found or not.
1775 (expand_parallel_call): Don't call adjust_context_and_scope here.
1776 (grid_expand_target_grid_body): Revert 2017-01-25 changes.
1777 (expand_omp_taskreg, expand_omp_target): Likewise. Call
1778 adjust_context_and_scope.
1779 * dwarf2out.c (dwarf2out_early_global_decl): For
1780 decl_function_context recurse instead of calling dwarf2out_decl.
1782 2018-11-16 Jakub Jelinek <jakub@redhat.com>
1784 PR rtl-optimization/87475
1785 * cfgrtl.c (patch_jump_insn): Allow redirection failure for
1786 CROSSING_JUMP_P insns.
1787 (cfg_layout_redirect_edge_and_branch): Don't ICE if ret is NULL.
1789 2018-11-15 Jakub Jelinek <jakub@redhat.com>
1791 PR rtl-optimization/88018
1792 * cfgrtl.c (fixup_abnormal_edges): Guard moving insns to fallthru edge
1793 on the presence of fallthru edge, rather than if it is a USE or not.
1795 2018-11-13 Jakub Jelinek <jakub@redhat.com>
1797 PR tree-optimization/87898
1798 * omp-simd-clone.c (ipa_simd_modify_function_body): Remove debug stmts
1799 where the first argument was changed into a non-decl.
1801 2019-01-03 Jan Hubicka <hubicka@ucw.cz>
1803 Backport from mainline
1804 2019-01-02 Jan Hubicka <hubicka@ucw.cz>
1807 * varpool.c (varpool_node::ctor_useable_for_folding_p): Also return
1808 false at WPA time when body was removed.
1810 2019-01-03 Jan Hubicka <hubicka@ucw.cz>
1812 Backport from mainline
1813 2018-08-29 Jan Hubicka <jh@suse.cz>
1817 * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
1818 * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
1820 2019-01-04 Aaron Sawdey <acsawdey@linux.ibm.com>
1822 Backport from mainline
1823 2018-11-28 Aaron Sawdey <acsawdey@linux.ibm.com>
1825 * config/rs6000/rs6000-string.c (expand_block_clear): Change how
1826 we determine if unaligned vsx is ok.
1828 2019-01-03 Jan Hubicka <hubicka@ucw.cz>
1830 Backport from mainline
1831 2019-01-03 Jan Hubicka <hubicka@ucw.cz>
1833 * ipa-utils.c (scale_ipa_profile_for_fn): Break out from ...
1834 (ipa_merge_profiles): ... here; do not ICE on thunks and aliases.
1836 2019-01-03 Martin Liska <mliska@suse.cz>
1838 Backport from mainline
1839 2019-01-01 Jan Hubicka <hubicka@ucw.cz>
1841 * coverage.c (get_coverage_counts): Use current_function_decl.
1842 * profile.c (read_thunk_profile): New function.
1843 (branch_prob): Add THUNK parameter.
1844 * tree-profile.c (tree_profiling): Handle thunks.
1845 * value-prof.c (init_node_map): Handle thunks.
1846 * value-prof.h (branch_prob): Upate prototype.
1847 (read_thunk_profile): Declare.
1849 2019-01-03 Iain Sandoe <iain@sandoe.co.uk>
1852 2018-12-23 Iain Sandoe <iain@sandoe.co.uk>
1854 backport from mainline.
1855 2018-12-12 Segher Boessenkool <segher@kernel.crashing.org>
1856 Iain Sandoe <iain@sandoe.co.uk>
1859 * config/rs6000/rs6000.c (save_reg_p): Do not save the picbase reg
1860 unless it has been used.
1861 (first_reg_to_save): Remove dead code.
1863 2019-01-02 Segher Boessenkool <segher@kernel.crashing.org>
1866 2018-12-06 Segher Boessenkool <segher@kernel.crashing.org>
1869 * doc/extend.texi (Basic Asm): Update grammar.
1870 (Extended Asm): Update grammar.
1873 2018-12-06 Segher Boessenkool <segher@kernel.crashing.org>
1875 * doc/extend.texi (Using Assembly Language with C): Document asm inline.
1876 (Size of an asm): Fix typo. Document asm inline.
1877 * gimple-pretty-print.c (dump_gimple_asm): Handle asm inline.
1878 * gimple.h (enum gf_mask): Add GF_ASM_INLINE.
1879 (gimple_asm_set_volatile): Fix typo.
1880 (gimple_asm_inline_p): New.
1881 (gimple_asm_set_inline): New.
1882 * gimplify.c (gimplify_asm_expr): Propagate the asm inline flag from
1884 * ipa-icf-gimple.c (func_checker::compare_gimple_asm): Compare the
1885 gimple_asm_inline_p flag, too.
1886 * tree-core.h (tree_base): Document that protected_flag is ASM_INLINE_P
1888 * tree-inline.c (estimate_num_insns): If gimple_asm_inline_p return
1889 a minimum size for an asm.
1890 * tree.h (ASM_INLINE_P): New.
1892 2018-12-28 Thomas Schwinge <thomas@codesourcery.com>
1893 Julian Brown <julian@codesourcery.com>
1895 * gimplify.c (gimplify_scan_omp_clauses): Fix known_eq typo/bug.
1897 2018-12-27 Martin Liska <mliska@suse.cz>
1899 Backport from mainline
1900 2018-12-15 Jan Hubicka <hubicka@ucw.cz>
1903 * ipa-polymorphic-call.c
1904 (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Handle
1905 arguments of thunks correctly.
1906 (ipa_polymorphic_call_context::get_dynamic_context): Be ready for
1907 NULL instance pinter.
1908 * lto-cgraph.c (lto_output_node): Always stream thunk info.
1910 2018-12-23 Iain Sandoe <iain@sandoe.co.uk>
1912 backport from mainline.
1913 2018-12-12 Segher Boessenkool <segher@kernel.crashing.org>
1914 Iain Sandoe <iain@sandoe.co.uk>
1917 * config/rs6000/rs6000.c (save_reg_p): Do not save the picbase reg
1918 unless it has been used.
1919 (first_reg_to_save): Remove dead code.
1921 2018-12-23 Iain Sandoe <iain@sandoe.co.uk>
1923 Backport from mainline
1924 2018-12-06 Iain Sandoe <iain@sandoe.co.uk>
1927 * config/darwin.h (TARGET_WEAK_NOT_IN_ARCHIVE_TOC) Remove, use the
1929 * config/rs6000/darwin7.h (TARGET_WEAK_NOT_IN_ARCHIVE_TOC): New.
1931 2018-12-23 Iain Sandoe <iain@sandoe.co.uk>
1933 Backport from mainline
1934 2018-12-06 Iain Sandoe <iain@sandoe.co.uk>
1937 * config/i386/darwin.h (STACK_BOUNDARY): Remove macro.
1938 * config/i386/i386.c (ix86_compute_frame_layout): Ensure at least 128b
1939 stack alignment in non-leaf functions.
1941 2018-12-21 Uros Bizjak <ubizjak@gmail.com>
1943 Backport from mainline
1944 2018-12-10 Uros Bizjak <ubizjak@gmail.com>
1947 * config/i386/i386.c (ix86_expand_sse_cmp): For vector modes,
1948 check operand 1 with vector_operand predicate.
1949 (ix86_expand_sse_movcc): For vector modes, check op_true with
1950 vector_operand, not nonimmediate_operand.
1952 2018-12-19 Bill Schmidt <wschmidt@linux.ibm.com>
1954 Backport from mainline
1955 2018-12-18 Bill Schmidt <wschmidt@linux.ibm.com>
1957 * doc/extend.texi (PowerPC Altivec/VSX Built-in Functions):
1958 Describe when a typedef name can be used as the type specifier for
1959 a vector type, and when it cannot.
1961 2018-12-19 Segher Boessenkool <segher@kernel.crashing.org>
1964 2018-12-19 Segher Boessenkool <segher@kernel.crashing.org>
1967 * config/rs6000/vsx.md (*vsx_extract_<P:mode>_<VSX_D:mode>_load):
1968 Require TARGET_POWERPC64.
1970 2018-12-17 Senthil Kumar Selvaraj <senthilkumar.selvaraj@microchip.com>
1973 2018-12-17 Senthil Kumar Selvaraj <senthilkumar.selvaraj@microchip.com>
1975 PR rtl-optimization/88253
1976 * combine.c (combine_simplify_rtx): Test for side-effects before
1977 substituting by zero.
1979 2018-12-15 Segher Boessenkool <segher@kernel.crashing.org>
1982 2018-12-14 Segher Boessenkool <segher@kernel.crashing.org>
1984 PR rtl-optimization/88001
1985 * function.c (match_asm_constraints_1): Don't invalidly share RTL.
1987 2018-12-15 Jan Hubicka <jh@suse.cz>
1989 * ipa.c (cgraph_build_static_cdtor_1): Add OPTIMIZATION and TARGET
1991 (cgraph_build_static_cdtor): Update.
1992 (build_cdtor): Use OPTIMIZATION and TARGET of the first real cdtor
1995 2018-12-15 Jan Hubicka <jh@suse.cz>
1997 Backport from mainline
1998 2018-11-02 Richard Biener <rguenther@suse.de>
2000 * dwarf2out.c (build_abbrev_table): Guard lookup_external_ref call
2003 2018-12-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2005 Backport from mainline
2006 2018-12-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2008 * doc/invoke.texi (-msve-vector-bits): Clarify -msve-vector-bits=128
2011 2018-12-14 Andre Vieira <andre.simoesdiasvieira@arm.com>
2013 Backport from mainline
2015 * config/arm/arm-cpus.in (armv7-r): Add FP16conv configurations.
2016 (cortex-r7, cortex-r8): Update fpu and add new configuration.
2017 * doc/invoke.texi (armv7-r): Add two new vfp options.
2018 (nofp.dp): Add cortex-r7 and cortex-r8 to the list of targets that
2019 support this option.
2021 2018-12-13 H.J. Lu <hongjiu.lu@intel.com>
2023 Backport from mainline
2024 2018-11-05 Xuepeng Guo <xuepeng.guo@intel.com>
2027 * config/i386/emmintrin.h (__v16qs): New to cope with option
2029 (_mm_cmpeq_epi8): Replace __v16qi with __v16qs.
2030 (_mm_cmplt_epi8): Likewise.
2031 (_mm_cmpgt_epi8): Likewise.
2033 2018-12-13 Andreas Krebbel <krebbel@linux.ibm.com>
2035 Backport from mainline
2036 2018-12-13 Andreas Krebbel <krebbel@linux.ibm.com>
2038 * config/s390/s390-builtins.def (s390_vec_double_s64): Map to
2039 s390_vec_double_s64 instead of s390_vcdgb.
2040 (s390_vec_double_u64): Map to s390_vec_double_u64 instead of
2043 2018-12-13 Andreas Krebbel <krebbel@linux.ibm.com>
2045 Backport from mainline
2046 2018-12-13 Andreas Krebbel <krebbel@linux.ibm.com>
2048 * config/s390/vx-builtins.md ("vec_ctd_s64", "vec_ctd_u64")
2049 ("vec_ctsl", "vec_ctul"): Replace 0 with VEC_NOINEXACT.
2050 ("vec_double_s64", "vec_double_u64"): Replace 4 with VEC_INEXACT.
2052 2018-12-13 Eric Botcazou <ebotcazou@adacore.com>
2054 Backport from mainline
2055 2018-12-11 Eric Botcazou <ebotcazou@adacore.com>
2057 * config/rs6000/vxworks.h (RS6000_STARTING_FRAME_OFFSET): Define,
2058 accounting for STACK_BOUNDARY 128.
2059 (STACK_DYNAMIC_OFFSET): Likewise.
2061 2018-12-12 Peter Bergner <bergner@linux.ibm.com>
2063 Backport from mainline
2064 2018-12-07 Peter Bergner <bergner@linux.ibm.com>
2067 * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow
2068 -mabi=ieeelongdouble and -mabi=ibmlongdouble without -mlong-double-128.
2069 Do not error for -mabi=ibmlongdouble and no ISA 2.06 support.
2070 * doc/invoke.texi: Document -mabi=ibmlongdouble and -mabi=ieeelongdouble
2071 require -mlong-double-128.
2073 2018-12-05 Richard Biener <rguenther@suse.de>
2075 Backport from mainline
2076 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
2078 PR tree-optimization/87288
2079 * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
2080 into account when determining PEELING_FOR_NITERS.
2082 2018-12-05 Iain Sandoe <iain@sandoe.co.uk>
2084 Backport from mainline
2085 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
2087 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
2088 emit a diagnostic that it is not supported and reset the option.
2089 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
2090 supported and consume the option. (ASM_FINAL_SPEC): New.
2092 2018-12-04 Iain Sandoe <iain@sandoe.co.uk>
2094 Backport from mainline
2095 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
2098 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
2099 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
2100 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
2102 2018-11-29 Peter Bergner <bergner@linux.ibm.com>
2104 Backport from mainline
2105 2018-11-29 Peter Bergner <bergner@linux.ibm.com>
2108 * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow
2109 -mabi=ieeelongdouble without both -mpopcntd and -mvsx.
2111 2018-11-28 Aaron Sawdey <acsawdey@linux.ibm.com>
2113 Backport from mainline
2114 2018-10-25 Aaron Sawdey <acsawdey@linux.ibm.com>
2116 * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
2117 a shorter sequence with fewer branches.
2118 (emit_final_str_compare_gpr): Ditto.
2120 Backport from mainline to allow the above code to go in:
2121 2018-06-14 Aaron Sawdey <acsawdey@linux.ibm.com>
2123 * config/rs6000/rs6000-string.c (do_and3, do_and3_mask,
2124 do_cmpb3, do_rotl3): New functions.
2126 2018-11-28 Richard Biener <rguenther@suse.de>
2128 PR tree-optimization/79351
2129 * tree-ssa-sccvn.c (vn_reference_lookup_3): For assignments from
2130 empty CONSTRUCTORs ensure the store is at a constant position.
2132 2018-11-28 Richard Biener <rguenther@suse.de>
2134 PR tree-optimization/88229
2135 * tree-ssa.c (non_rewritable_mem_ref_base): Check DECL_SIZE_UNIT
2136 is an INTEGER_CST before accessing it so.
2138 2018-11-26 Andreas Krebbel <krebbel@linux.ibm.com>
2140 Backport from mainline
2141 2018-11-20 Andreas Krebbel <krebbel@linux.ibm.com>
2143 * config/s390/s390.md ("clztidi2"): Swap the RTX's written to the
2144 DImode parts of the target operand.
2146 2018-11-26 Andreas Krebbel <krebbel@linux.ibm.com>
2148 Backport from mainline
2149 2018-11-26 Andreas Krebbel <krebbel@linux.ibm.com>
2151 * doc/invoke.texi: Document z14/arch12 -march option.
2153 2018-11-23 Tamar Christina <tamar.christina@arm.com>
2155 Backported from mainline.
2156 2018-08-16 Tamar Christina <tamar.christina@arm.com>
2159 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
2160 (mov<mov>): ..this and enable unconditionally.
2162 2018-11-22 Uros Bizjak <ubizjak@gmail.com>
2164 Backport from mainline
2165 2018-11-16 Uros Bizjak <ubizjak@gmail.com>
2168 * config/i386/sse.md (UNSPEC_MOVDI_TO_SSE): New UNSPEC.
2169 (movdi_to_sse): Rewrite using UNSPEC_MOVDI_TO_SSE unspec.
2171 2018-11-20 Eric Botcazou <ebotcazou@adacore.com>
2173 PR rtl-optimization/85925
2174 * rtl.h (word_register_operation_p): New predicate.
2175 * combine.c (record_dead_and_set_regs_1): Only apply specific handling
2176 for WORD_REGISTER_OPERATIONS targets to word_register_operation_p RTX.
2177 * rtlanal.c (nonzero_bits1): Likewise. Adjust couple of comments.
2178 (num_sign_bit_copies1): Likewise.
2180 2018-11-19 Martin Liska <mliska@suse.cz>
2182 PR gcov-profile/88045
2183 * coverage.c (coverage_begin_function): One can't relly on the
2184 fact that function ends on the same line it starts or later.
2186 2018-11-18 Uros Bizjak <ubizjak@gmail.com>
2188 Backport from mainline
2189 2018-11-11 Uros Bizjak <ubizjak@gmail.com>
2192 * config/i386/i386.h (STACK_BOUNDARY): Use TARGET_64BIT_MS_ABI
2193 instead of (TARGET_64BIT && ix86_abi == MS_ABI).
2194 * config/i386/darwin.h (STACK_BOUNDARY): Ditto.
2195 * config/i386/cygming.h (STACK_BOUNDARY): Remove.
2197 2018-11-16 Eric Botcazou <ebotcazou@adacore.com>
2199 * tree-switch-conversion.c (emit_case_decision_tree): Pass location
2200 of switch statement to emit_case_nodes.
2201 (do_jump_if_equal): Add LOC parameter and set it on the newly built
2202 GIMPLE comparison statement.
2203 (emit_cmp_and_jump_insns): Likewise.
2204 (emit_case_nodes): Add LOC parameter and pass it to do_jump_if_equal
2205 as well as recursive calls.
2207 2018-11-15 Nathan Sidwell <nathan@acm.org>
2211 * dwarf2out.c (dwarf2out_finish): Apply resolve_addr to comdat
2214 2018-11-14 Paul A. Clarke <pc@us.ibm.com>
2216 Backport from mainline
2217 * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
2219 2018-11-14 Claudiu Zissulescu <claziss@synopsys.com>
2221 Backport from mainline
2223 * config/arc/arc.c (arc_return_address_register): Fix
2226 2018-11-14 Claudiu Zissulescu <claziss@synopsys.com>
2228 Backport from mainline
2229 * config/arc/arc.h (FUNCTION_PROFILER): Redefine to empty.
2230 * config/arc/elf.h (PROFILE_HOOK): Define.
2231 * config/arc/linux.h (PROFILE_HOOK): Likewise.
2233 2018-11-14 Claudiu Zissulescu <claziss@synopsys.com>
2235 Backport from mainline
2236 * config/arc/arc.c (hwloop_optimize): Bailout when detecting a
2237 jump table data in the text section.
2239 2018-11-14 Claudiu Zissulescu <claziss@synopsys.com>
2241 Backport from mainline
2242 * config/arc/arc.c (arc_eh_return_address_location): Repurpose it
2243 to fit the eh_return pattern.
2244 * config/arc/arc.md (eh_return): Define.
2245 (VUNSPEC_ARC_EH_RETURN): Likewise.
2246 * config/arc/arc-protos.h (arc_eh_return_address_location): Match
2248 * config/arc/arc.h (EH_RETURN_HANDLER_RTX): Remove it.
2250 2018-11-14 Claudiu Zissulescu <claziss@synopsys.com>
2252 Backport from mainline
2253 2018-09-17 Alexey Brodkin <abrodkin@synopsys.com>
2255 * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
2257 2018-11-14 Claudiu Zissulescu <claziss@synopsys.com>
2259 Backport from mainline
2260 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
2262 2018-11-11 Uros Bizjak <ubizjak@gmail.com>
2264 Backport from mainline
2265 2018-11-04 Uros Bizjak <ubizjak@gmail.com>
2268 * cfgexpand.c (pass_expand::execute): Move the call to
2269 finish_eh_generation in front of the call to expand_stack_alignment.
2271 2018-11-07 Max Filippov <jcmvbkbc@gmail.com>
2273 Backport from mainline
2274 2018-11-05 Max Filippov <jcmvbkbc@gmail.com>
2276 * config/xtensa/uclinux.h (XTENSA_ALWAYS_PIC): Change to 0.
2278 2018-11-05 Jakub Jelinek <jakub@redhat.com>
2280 PR tree-optimization/87859
2281 * gimple-ssa-store-merging.c (struct merged_store_group): Add
2282 first_nonmergeable_order member.
2283 (merged_store_group::merged_store_group): Initialize them.
2284 (imm_store_chain_info::coalesce_immediate_stores): Don't merge
2285 stores with order >= first_nonmergeable_order.
2286 Set merged_store->first_nonmergeable_order if we've skipped any
2287 stores. Attempt to merge overlapping INTEGER_CST stores that
2288 we would otherwise skip.
2291 * match.pd (X + Y < X): Don't optimize if TYPE_OVERFLOW_SANITIZED.
2293 Backported from mainline
2294 2018-10-20 Jakub Jelinek <jakub@redhat.com>
2297 * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
2299 2018-10-19 Jakub Jelinek <jakub@redhat.com>
2303 * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
2304 depend closely nested inside of loop with ordered clause with
2307 2018-10-29 Richard Biener <rguenther@suse.de>
2309 Backport from mainline
2310 2018-09-26 Richard Biener <rguenther@suse.de>
2314 * tree-inline.c (expand_call_inline): When the location
2315 of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
2316 or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
2317 the inserted BLOCK to make inlined_function_outer_scope_p
2319 * dwarf2out.c (add_call_src_coords_attributes): Do not add
2320 coords for reserved locations.
2322 2018-10-28 Iain Sandoe <iain@sandoe.co.uk>
2324 Backport from mainline
2325 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2330 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
2331 (dwarf2out_switch_text_section): Generate a local label for the second
2332 function sub-section and apply it as the second FDE start label.
2333 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
2334 second sub-section start.
2336 2018-10-28 Iain Sandoe <iain@sandoe.co.uk>
2338 Backport from mainline
2339 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
2342 (darwin_function_switched_text_sections): Delete.
2343 * gcc/config/darwin.h
2344 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
2346 2018-10-28 Iain Sandoe <iain@sandoe.co.uk>
2348 backport from mainline.
2349 2018-10-28 Iain Sandoe <iain@sandoe.co.uk>
2352 * config/rs6000/darwin.h (STACK_BOUNDARY): New.
2353 (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
2354 (STACK_DYNAMIC_OFFSET): Likewise.
2356 2018-10-26 Bill Schmidt <wschmidt@linux.ibm.com>
2358 Backport from mainline
2359 2018-10-19 Bill Schmidt <wschmidt@linux.ibm.com>
2361 PR tree-optimization/87473
2362 * gimple-ssa-strength-reduction.c (record_phi_increments_1): For
2363 phi arguments identical to the base expression of the phi
2364 candidate, record a phi-adjust increment of zero minus the index
2365 expression of the hidden basis.
2366 (phi_incr_cost_1): For phi arguments identical to the base
2367 expression of the phi candidate, the difference to compare against
2368 the increment is zero minus the index expression of the hidden
2369 basis, and there is no potential savings from replacing the (phi)
2371 (ncd_with_phi): For phi arguments identical to the base expression
2372 of the phi candidate, the difference to compare against the
2373 increment is zero minus the index expression of the hidden basis.
2374 (all_phi_incrs_profitable_1): For phi arguments identical to the
2375 base expression of the phi candidate, the increment to be checked
2376 for profitability is zero minus the index expression of the hidden
2379 2018-10-25 Richard Biener <rguenther@suse.de>
2381 PR tree-optimization/87665
2382 PR tree-optimization/87745
2383 * tree-vectorizer.h (get_earlier_stmt): Remove.
2384 (get_later_stmt): Pick up UID from the original non-pattern stmt.
2386 2018-10-24 Richard Biener <rguenther@suse.de>
2388 PR tree-optimization/87665
2389 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
2392 2018-10-23 Richard Biener <rguenther@suse.de>
2394 PR tree-optimization/87700
2395 * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
2397 2018-10-19 Richard Biener <rguenther@suse.de>
2400 Backport from mainline
2401 2018-07-12 Richard Biener <rguenther@suse.de>
2403 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
2404 (vn_lookup_simplify_result): Remove recursion limit applied
2406 (vn_nary_build_or_lookup_1): Adjust.
2407 (try_to_simplify): Likewise.
2408 * gimple-match-head.c (gimple_resimplify1): Instead apply one
2410 (gimple_resimplify2): Likewise.
2411 (gimple_resimplify3): Likewise.
2412 (gimple_resimplify4): Likewise.
2414 2018-10-19 Andreas Krebbel <krebbel@linux.ibm.com>
2416 Backport from mainline
2417 2018-10-15 Andreas Krebbel <krebbel@linux.ibm.com>
2419 * config/s390/s390.c (s390_expand_vec_init): Force vector element
2420 into reg if it isn't a general operand.
2422 2018-10-18 Richard Biener <rguenther@suse.de>
2426 2018-02-07 Richard Biener <rguenther@suse.de>
2428 PR tree-optimization/84204
2429 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2432 2018-10-17 Eric Botcazou <ebotcazou@adacore.com>
2435 * fold-const.c (fold_truth_andor_1): If the right side is not constant,
2436 bail out if both sides do not have the same storage order.
2438 2018-10-17 Richard Biener <rguenther@suse.de>
2440 Backport from mainline
2441 2018-10-08 Richard Sandiford <richard.sandiford@arm.com>
2444 * gimple-ssa-backprop.c (backprop::intersect_uses): Use
2445 FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
2447 2018-10-16 Richard Biener <rguenther@suse.de>
2449 Backport from mainline
2450 2018-10-08 Richard Biener <rguenther@suse.de>
2452 PR tree-optimization/63155
2453 * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
2454 (ssa_propagation_engine::ssa_propagate): Remove redundant
2455 bitmap bit clearing.
2457 2018-10-05 Richard Biener <rguenther@suse.de>
2459 PR tree-optimization/63155
2460 * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
2461 vertical space in dumpfiles.
2462 * tree-ssa-propagate.h
2463 (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
2464 * tree-ssa-propagate.c (cfg_blocks_back): New global.
2465 (ssa_edge_worklist_back): Likewise.
2466 (curr_order): Likewise.
2467 (cfg_blocks_get): Remove abstraction.
2468 (cfg_blocks_add): Likewise.
2469 (cfg_blocks_empty_p): Likewise.
2470 (add_ssa_edge): Add to current or next worklist based on
2472 (add_control_edge): Likewise.
2473 (ssa_propagation_engine::process_ssa_edge_worklist): Fold
2475 (ssa_propagation_engine::ssa_propagate): ... here. Unify
2476 iteration from CFG and SSA edge worklist so we process
2477 everything in RPO order, prioritizing forward progress
2479 (ssa_prop_init): Allocate new worklists, do not dump
2481 (ssa_prop_fini): Free new worklists.
2483 2018-09-24 Richard Biener <rguenther@suse.de>
2485 PR tree-optimization/63155
2486 * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
2487 the worklist when the edge of the respective argument isn't
2490 2018-10-16 Wilco Dijkstra <wdijkstr@arm.com>
2492 Backported from mainline
2494 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
2495 Use HOST_WIDE_INT_1U for shift.
2497 2018-10-16 Richard Biener <rguenther@suse.de>
2499 Backport from mainline
2500 2018-09-18 Richard Biener <rguenther@suse.de>
2503 * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
2504 (compute_samebase_partition_bases): Likewise.
2505 (coalesce_ssa_name): Always use compute_optimized_partition_bases.
2506 (gimple_can_coalesce_p): Simplify.
2508 2018-10-16 Richard Biener <rguenther@suse.de>
2510 Backport from mainline
2511 2018-10-15 Richard Biener <rguenther@suse.de>
2514 * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
2515 (visit_loadstore): When a used restrict tag escaped verify that
2516 the points-to solution of "other" pointers do not include
2518 (compute_dependence_clique): If a used restrict tag escaped
2519 communicated that down to visit_loadstore.
2521 2018-10-01 Richard Biener <rguenther@suse.de>
2523 PR tree-optimization/87465
2524 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
2525 causing branch miscounts.
2527 2018-10-11 Jakub Jelinek <jakub@redhat.com>
2529 Backported from mainline
2530 2018-10-10 Jakub Jelinek <jakub@redhat.com>
2533 * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
2534 to special_args set.
2536 2018-09-26 Jakub Jelinek <jakub@redhat.com>
2539 * config/i386/i386.c: Include debug.h and dwarf2out.h.
2540 (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
2543 2018-09-12 Jakub Jelinek <jakub@redhat.com>
2544 Andreas Krebbel <krebbel@linux.ibm.com>
2546 PR tree-optimization/86844
2547 * gimple-ssa-store-merging.c
2548 (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
2549 there are any overlapping stores in between them, make sure they are
2550 also coalesced or we give up completely.
2552 2018-10-09 Will Schmidt <will_schmidt@vnet.ibm.com>
2554 Backport from trunk.
2555 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
2558 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
2559 around folding of vec_sl to handle out of range shift values.
2561 2018-10-09 H.J. Lu <hongjiu.lu@intel.com>
2563 Backport from mainline
2564 2018-09-29 H.J. Lu <hongjiu.lu@intel.com>
2567 * config/i386/i386.c (construct_container): Use TImode for
2568 BLKmode values in 2 integer registers.
2570 2018-10-08 H.J. Lu <hongjiu.lu@intel.com>
2572 Backport from mainline
2573 2018-10-08 H.J. Lu <hongjiu.lu@intel.com>
2576 * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
2577 Defined with __builtin_ia32_vfmaddsubpd512_mask.
2579 2018-10-05 H.J. Lu <hongjiu.lu@intel.com>
2581 Backport from mainline
2582 2018-10-05 H.J. Lu <hongjiu.lu@intel.com>
2585 * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
2586 assembler for -mavx.
2587 * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
2589 2018-10-03 Uros Bizjak <ubizjak@gmail.com>
2591 Backport from mainline
2592 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
2594 * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
2595 (GET_SSE_REGNO): Rename from SSE_REGNO. Update all uses for rename.
2597 2018-10-03 Jonathan Wakely <jwakely@redhat.com>
2600 * doc/invoke.texi (Link Options): Fix formatting and grammar.
2602 2018-10-03 Martin Liska <mliska@suse.cz>
2604 Backport from mainline
2605 2018-10-03 Martin Liska <mliska@suse.cz>
2607 PR gcov-profile/86109
2608 * coverage.c (coverage_begin_function): Do not
2609 mark lambdas as artificial.
2610 * tree-core.h (struct GTY): Remove tm_clone_flag
2611 and introduce new lambda_function.
2612 * tree.h (DECL_LAMBDA_FUNCTION): New macro.
2614 2018-10-02 H.J. Lu <hongjiu.lu@intel.com>
2616 Backport from mainline
2617 2018-09-24 H.J. Lu <hongjiu.lu@intel.com>
2620 * config/i386/i386.c (rest_of_insert_endbranch): Set
2621 endbr_queued_at_entrance to true and don't insert ENDBR if
2622 x86_function_profiler will be called.
2623 (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
2625 * config/i386/i386.h (machine_function): Add
2626 endbr_queued_at_entrance.
2628 2018-10-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2630 Backport from mainline
2631 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2633 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
2634 if starting source register is not even.
2636 2018-09-29 Jakub Jelinek <jakub@redhat.com>
2639 * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
2640 __m512d type for __A argument rather than __m512.
2642 2018-09-27 Michael Meissner <meissner@linux.ibm.com>
2644 Backport from mainline
2645 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
2648 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
2649 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
2652 2018-09-25 Martin Liska <mliska@suse.cz>
2654 Backport from mainline
2655 2018-09-24 Martin Liska <mliska@suse.cz>
2658 * asan.c: Make asan_handled_variables extern.
2660 * cfgexpand.c (expand_stack_vars): Make sure
2661 a representative is unpoison if another
2662 variable in the partition is handled by
2663 use-after-scope sanitization.
2665 2018-09-24 Andrew Pinski <apinski@marvell.com>
2667 * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
2668 access prev before checking it for NULLness in the
2669 AARCH64_FUSE_CMP_BRANCH case.
2671 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
2673 * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
2674 if the call takes a static chain.
2676 2018-09-19 John David Anglin <danglin@gcc.gnu.org>
2678 * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
2679 (atomic_storehi): Likewise.
2680 (atomic_storesi): Likewise.
2681 (atomic_loaddi): Restore compare and swap exchange loop code.
2683 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
2686 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
2688 PR rtl-optimization/86882
2689 * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
2691 2018-09-18 Martin Liska <mliska@suse.cz>
2693 Backport from mainline
2694 2018-09-17 Martin Liska <mliska@suse.cz>
2696 PR gcov-profile/85871
2697 * gcov.c (output_intermediate_file): Fix out of bounds
2700 2018-09-16 Hans-Peter Nilsson <hp@bitrange.com>
2703 * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
2704 non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
2705 (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
2706 leaf_function_p, instead use has_hard_reg_initial_val.
2708 2018-09-14 John David Anglin <danglin@gcc.gnu.org>
2711 * dojump.c (do_compare_and_jump): Canonicalize function pointers
2712 when one operand is a function pointer. Use POINTER_TYPE_P and
2713 FUNC_OR_METHOD_TYPE_P.
2714 * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
2715 * fold-const.c (build_range_check): Likewise.
2716 * match.pd (simple_comparison): Likewise.
2718 2018-09-14 Segher Boessenkool <segher@kernel.crashing.org>
2721 2018-09-14 Segher Boessenkool <segher@kernel.crashing.org>
2724 * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
2727 2018-09-14 Carl Love <cel@us.ibm.com>
2730 2018-09-14 Carl Love <cel@us.ibm.com>
2732 * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
2733 * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
2735 2018-09-12 Segher Boessenkool <segher@kernel.crashing.org>
2738 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
2741 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
2744 2018-09-12 Segher Boessenkool <segher@kernel.crashing.org>
2747 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
2749 PR rtl-optimization/86771
2750 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
2751 of two SETs into those two SETs, one to be placed at i2, if that SETs
2752 destination is modified between i2 and i3.
2754 2018-09-12 Andreas Krebbel <krebbel@linux.ibm.com>
2756 Backport from mainline
2757 2018-09-12 Andreas Krebbel <krebbel@linux.ibm.com>
2759 * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
2761 ("trunc<BFP:mode><DFP_ALL:mode>2")
2762 ("trunc<DFP_ALL:mode><BFP:mode>2")
2763 ("extend<BFP:mode><DFP_ALL:mode>2")
2764 ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
2765 according to the target operand type.
2767 2018-09-12 Jakub Jelinek <jakub@redhat.com>
2770 * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
2771 BIT_AND_EXPR's second operand is a power of two. Formatting fix.
2773 2018-09-05 Nathan Sidwell <nathan@acm.org>
2776 * stor-layout.c (place_field): Scan forwards to check last
2777 bitfield when ms_bitfield_placement is in effect.
2779 2018-09-05 Richard Biener <rguenther@suse.de>
2782 * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
2785 2018-09-05 Jakub Jelinek <jakub@redhat.com>
2787 Backported from mainline
2788 2018-09-04 Jakub Jelinek <jakub@redhat.com>
2791 * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
2792 OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
2793 OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
2794 (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
2795 and OPTION_MASK_ISA_XSAVEC_UNSET.
2797 2018-08-31 Jakub Jelinek <jakub@redhat.com>
2800 * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
2801 gen_int_mode. Formatting fixes.
2803 2018-08-11 Jakub Jelinek <jakub@redhat.com>
2805 PR tree-optimization/86835
2806 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
2807 new_stmt after def_gsi, make sure to insert new_square_stmt after
2808 that stmt, not 2 stmts before it.
2810 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
2812 Backport from mainline
2813 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
2815 * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
2816 XOR operations in NAND case.
2818 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
2820 * doc/invoke.texi (Option Summary): Add -Waligned-new.
2822 2018-09-03 Tom de Vries <tdevries@suse.de>
2824 backport from trunk:
2825 2018-06-21 Tom de Vries <tdevries@suse.de>
2827 PR tree-optimization/85859
2828 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
2829 test with comment from bb_no_side_effects_p.
2831 2018-09-01 Michael Matz <matz@suse.de>
2833 Backport from mainline
2834 PR tree-optimization/87074
2835 * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
2836 PHIs for outer-loop uses.
2838 2018-08-31 Richard Biener <rguenther@suse.de>
2840 Backport from mainline
2841 2018-08-27 Richard Biener <rguenther@suse.de>
2843 PR tree-optimization/86927
2844 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
2845 use const cond reduction code.
2847 2018-08-23 Richard Biener <rguenther@suse.de>
2850 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
2853 2018-08-22 Richard Biener <rguenther@suse.de>
2855 PR tree-optimization/86945
2856 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
2858 2018-08-17 Richard Biener <rguenther@suse.de>
2861 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
2862 across a va-arg-pack using call adjust its return value accordingly.
2864 2018-08-02 Richard Biener <rguenther@suse.de>
2866 PR tree-optimization/86816
2867 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
2868 which checks for value availability before querying it.
2869 (gvn_uses_equal): Use it.
2870 (same_succ_hash): Likewise.
2871 (gimple_equal_p): Likewise.
2873 2018-07-17 Richard Biener <rguenther@suse.de>
2876 * dwarf2out.c (init_sections_and_labels): Always generate
2877 a debug_line_str_section for early LTO debug.
2878 (dwarf2out_finish): Reset debug_line_str_hash output early.
2879 Bump counter for extra dwarf5 .debug_loc labels to not conflict
2881 (dwarf2out_early_finish): Output debug_line_str.
2883 2018-08-29 Luis Machado <luis.machado@linaro.org>
2885 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static global.
2886 (qdf24xx_tunings): Set vector cost structure to qdf24xx_vector_cost.
2888 2018-08-29 Luis Machado <luis.machado@linaro.org>
2890 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
2891 <register_sextend>: Set to 3.
2893 2018-08-27 Martin Sebor <msebor@redhat.com>
2895 PR tree-optimization/86914
2896 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
2898 2018-08-28 Jakub Jelinek <jakub@redhat.com>
2901 * calls.c (maybe_warn_nonstring_arg): Punt early if
2902 warn_stringop_overflow is zero. Don't call get_range_strlen
2903 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
2904 Only use lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly
2905 increment lenrng[0].
2907 2018-08-27 Jakub Jelinek <jakub@redhat.com>
2909 PR rtl-optimization/87065
2910 * combine.c (simplify_if_then_else): Formatting fix.
2911 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
2913 (known_cond): Don't return const_true_rtx for vector modes. Use
2914 CONST0_RTX instead of const0_rtx. Formatting fixes.
2916 2018-08-25 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2918 Backport from mainline
2920 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
2921 with all enabled __intN types.
2923 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
2925 2018-08-21 H.J. Lu <hongjiu.lu@intel.com>
2927 Backport from mainline
2928 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
2931 * config/i386/i386.md (eh_return): Always update EH return
2932 address in word_mode.
2934 2018-08-17 John David Anglin <danglin@gcc.gnu.org>
2936 Backport from mainline
2937 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
2939 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
2940 Update comment for atomic instructions.
2941 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
2942 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
2944 (atomic_loaddi): Revise fence expansion to only emit fence prior to
2945 load for __ATOMIC_SEQ_CST model.
2946 (atomic_loaddi_1): Remove float register target.
2947 (atomic_storedi): Handle CONST_INT values.
2948 (atomic_storedi_1): Remove float register source. Add special case
2950 (memory_barrier): New expander and insn.
2952 2018-08-16 Tamar Christina <tamar.christina@arm.com>
2954 Backport from mainline
2955 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
2958 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
2961 2018-08-16 Martin Liska <mliska@suse.cz>
2963 Backport from mainline
2964 2018-08-02 Martin Liska <mliska@suse.cz>
2966 PR gcov-profile/86817
2967 * gcov.c (process_all_functions): New function.
2969 (process_file): Move functions processing to
2970 process_all_functions.
2972 2018-08-14 Siddhesh Poyarekar <siddhesh@sourceware.org>
2974 Backport from mainline
2975 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
2977 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
2979 (falkor_am_1_gtov_gtov): ... a new insn reservation.
2981 2018-08-13 Liu Hao <lh_mouse@126.com>
2983 Backport from mainline
2984 2018-08-13 Liu Hao <lh_mouse@126.com>
2986 * pretty-print.c (eat_esc_sequence): Swap the foreground and
2987 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
2988 and clear it thereafter, as it only works for DBCS.
2990 2018-08-13 Liu Hao <lh_mouse@126.com>
2992 Backport from mainline
2993 2018-08-13 Liu Hao <lh_mouse@126.com>
2995 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
2996 handle returned by _get_osf_handle().
2998 2018-08-14 Richard Sandiford <richard.sandiford@arm.com>
3000 Backport from mainline
3001 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
3003 PR tree-optimization/86871
3004 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
3005 instead of gimple_assign_lhs.
3007 2018-08-07 H.J. Lu <hongjiu.lu@intel.com>
3009 Backport from mainline
3010 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
3013 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
3014 cfun->machine->max_used_stack_alignment if needed.
3016 2018-08-02 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3018 Backport from mainline
3019 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3022 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
3023 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
3024 requested variable alignment.
3025 (expand_one_ssa_partition): Likewise.
3026 (expand_one_var): Likewise.
3028 2018-08-02 Andreas Schwab <schwab@linux-m68k.org>
3030 Backport from mainline
3031 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
3034 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
3035 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
3037 2018-08-01 Richard Biener <rguenther@suse.de>
3040 * graphite.h: Include isl/id.h and isl/space.h to allow build
3043 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
3046 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3047 mode for COFF targets.
3048 * defaults.h (TARGET_COFF): Define.
3049 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
3050 TARGET_COFF): Define.
3051 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
3052 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
3053 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
3055 2018-07-29 Bernd Edlinger <bernd.edlinger@hotmail.de>
3057 Backport from mainline
3058 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
3061 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
3063 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
3065 * config/pa/pa.c (pa_output_addr_vec): Align address table.
3066 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
3067 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
3069 2018-07-26 Uros Bizjak <ubizjak@gmail.com>
3071 Backport from mainline
3072 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
3075 * expmed.c (emit_store_flag): Do not emit setcc followed by a
3076 conditional move when trapping comparison was split to a
3077 non-trapping one (and vice versa).
3079 2018-07-26 Jakub Jelinek <jakub@redhat.com>
3081 Backported from mainline
3082 2018-07-24 Jakub Jelinek <jakub@redhat.com>
3085 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
3086 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
3087 and abs_d == d, do the power of two handling if profitable.
3089 2018-07-17 Jakub Jelinek <jakub@redhat.com>
3092 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
3093 to _looptemp_ clauses, other than the first two.
3096 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
3097 and cond temporaries don't have reference type if iterator has
3098 pointer type. For init use &for_pre_body instead of pre_p if
3099 for_pre_body is non-empty.
3101 2018-07-26 Jakub Jelinek <jakub@redhat.com>
3104 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
3105 declare target to variables if they have always,{to,from,tofrom} map
3108 * BASE-VER: Set to 8.2.1.
3110 2018-07-26 Release Manager
3112 * GCC 8.2.0 released.
3114 2018-07-18 Martin Sebor <msebor@redhat.com>
3117 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
3118 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
3119 Handle integer subtraction.
3120 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
3121 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
3122 * doc/invoke.texi (-Wstringop-truncation): Update.
3124 gcc/testsuite/ChangeLog:
3126 2018-07-16 Claudiu Zissulescu <claziss@synopsys.com>
3128 Backport from mainline
3129 2017-03-24 Claudiu Zissulescu <claziss@synopsys.com>
3131 * config/arc/arc-protos.h (arc_pad_return): Remove.
3132 * config/arc/arc.c (machine_function): Remove force_short_suffix
3134 (arc_print_operand): Adjust printing of '&'.
3135 (arc_verify_short): Remove conditional printing of short suffix.
3136 (arc_final_prescan_insn): Remove reference to size_reason.
3137 (pad_return): New function.
3138 (arc_reorg): Call pad_return.
3139 (arc_pad_return): Remove.
3140 (arc_init_machine_status): Remove reference to force_short_suffix.
3141 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
3142 (attr length): When attribute iscompact is true force to 2
3143 regardless; in the case of maybe check if we want to force the
3144 instruction to have 4 bytes length.
3145 (nopv): Change it to generate 4 byte long nop as well.
3146 (blockage): New pattern.
3147 (simple_return): Remove call to arc_pad_return.
3148 (p_return_i): Likewise.
3150 2018-07-19 Richard Biener <rguenther@suse.de>
3152 Backport from mainline
3153 2018-07-13 Richard Biener <rguenther@suse.de>
3156 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
3157 instead of get_context_die.
3159 2018-07-11 Richard Biener <rguenther@suse.de>
3162 * dwarf2out.c (init_sections_and_labels): Use
3163 output_asm_line_debug_info consistently.
3164 (dwarf2out_early_finish): Likewise.
3165 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
3168 2018-07-16 Richard Biener <rguenther@suse.de>
3170 Backport from mainline
3171 2018-07-13 Richard Biener <rguenther@suse.de>
3174 * match.pd (addr1 - addr2): Allow either of the operand to
3177 2018-06-15 Richard Biener <rguenther@suse.de>
3180 * tree-cfg.c (move_stmt_op): unshare invariant addresses
3181 before adjusting their block.
3183 2018-06-06 Richard Biener <rguenther@suse.de>
3185 PR tree-optimization/85935
3186 * graphite-scop-detection.c (find_params_in_bb): Analyze
3187 condition operands with respect to the correct loop. Assert
3188 the analysis doesn't fail.
3190 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
3192 PR tree-optimization/86514
3193 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
3194 conversion to a boolean type from a type with greater precision.
3196 2018-07-15 Bill Schmidt <wschmidt@linux.ibm.com>
3198 Backport from mainline
3199 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
3200 Steve Munroe <munroesj52@gmail.com>
3202 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
3203 (_mm_andnot_si128): Likewise.
3204 (_mm_or_si128): Likewise.
3205 (_mm_xor_si128): Likewise.
3207 2018-07-14 Martin Sebor <msebor@redhat.com>
3209 PR tree-optimization/86274
3210 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
3212 (format_floating): Correct handling of infinities and NaNs.
3214 2018-07-13 H.J. Lu <hongjiu.lu@intel.com>
3216 Backport from mainline
3217 2018-07-13 H.J. Lu <hongjiu.lu@intel.com>
3218 Sunil K Pandey <sunil.k.pandey@intel.com>
3221 * config/i386/i386.c (m_CORE_AVX512): New.
3222 (m_CORE_AVX2): Likewise.
3223 (m_CORE_ALL): Add m_CORE_AVX2.
3224 * config/i386/x86-tune.def: Replace m_HASWELL with m_CORE_AVX2.
3225 Replace m_SKYLAKE_AVX512 with m_CORE_AVX512 on avx256_optimal
3226 and remove the rest of m_SKYLAKE_AVX512.
3228 2018-07-13 Marek Polacek <polacek@redhat.com>
3231 2018-07-03 Marek Polacek <polacek@redhat.com>
3234 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
3236 2018-07-12 Richard Biener <rguenther@suse.de>
3239 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
3240 Remove -mieee-fp handling.
3242 2018-07-12 Jakub Jelinek <jakub@redhat.com>
3244 PR tree-optimization/86492
3245 * gimple-ssa-store-merging.c
3246 (imm_store_chain_info::coalesce_immediate_stores): Call
3247 check_no_overlap even for the merge_overlapping case.
3249 2018-07-11 Jakub Jelinek <jakub@redhat.com>
3251 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
3252 Use __mmask64 type instead of __mmask8 for __M argument.
3253 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
3254 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
3256 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
3257 __mmask16 for __M argument.
3258 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
3259 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
3260 to __mmask16 instead of __mmask8.
3261 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
3262 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
3263 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
3264 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
3265 instead of __mmask16 for __U argument.
3266 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
3267 __mmask16 instead of __mmask8 for __U argument.
3268 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
3270 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
3272 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
3274 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
3276 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
3278 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
3280 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
3282 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
3284 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
3286 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
3288 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
3289 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
3290 return type as well as __M argument type and all casts from __mmask8
3292 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
3293 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
3294 return type as well as __M argument type and all casts from __mmask8
3296 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
3297 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
3298 return type as well as __M argument type and all casts from __mmask8
3300 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
3301 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
3302 return type as well as __M argument type and all casts from __mmask8
3304 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
3305 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
3308 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
3310 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
3311 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
3314 2018-07-05 Martin Sebor <msebor@redhat.com>
3317 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
3318 than its domain to compute its the upper bound of a char array.
3320 2018-07-03 Carl Love <cel@us.ibm.com>
3323 2018-07-03 Carl Love <cel@us.ibm.com>
3325 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
3326 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
3327 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
3328 VSX_BUILTIN_DOUBLEL_V4SF.
3330 2018-07-04 Luis Machado <luis.machado@linaro.org>
3332 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
3334 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
3336 2018-07-02 Segher Boessenkool <segher@kernel.crashing.org>
3339 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
3342 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
3343 ieee128_float_type_node to long_double_type_node unless
3344 TARGET_LONG_DOUBLE_128 is set.
3346 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
3348 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
3349 -fno-omit-frame-pointer when not optimizing.
3351 2018-06-30 Alexandre Oliva <aoliva@redhat.com>
3354 * dwarf2out.c (loc_list_has_views): Adjust comments.
3355 (dw_loc_list): Split single cross-partition range with
3358 2018-06-29 Kelvin Nilsen <kelvin@gcc.gnu.org>
3360 Backport from mainline
3361 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3363 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
3364 behavior of vec_pack (vector double, vector double) to match
3365 behavior of vec_float2 (vector double, vector double).
3367 2018-06-29 Jakub Jelinek <jakub@redhat.com>
3369 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
3371 2018-06-28 Kelvin Nilsen <kelvin@gcc.gnu.org>
3373 Backport from mainline
3374 2017-05-23 Segher Boessenkool <segher@kernel.crashing.org>
3376 * doc/sourcebuild.texi (Endianness): New subsubsection.
3378 2018-06-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
3380 Backport from mainline
3381 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
3383 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
3384 behavior of vec_packsu (vector unsigned long long, vector unsigned
3385 long long) to match behavior of vec_packs with same signature.
3387 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
3390 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
3392 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
3393 unaligned vsx for 16B memset.
3395 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
3398 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
3401 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
3404 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
3407 2018-05-08 Segher Boessenkool <segher@kernel.crashing.org>
3409 PR rtl-optimization/85645
3410 * regrename.c (build_def_use): Also kill the chains that include the
3411 destination of a REG_CFA_REGISTER note.
3413 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
3416 2018-05-08 Segher Boessenkool <segher@kernel.crashing.org>
3418 PR rtl-optimization/85645
3419 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
3420 insn that has a REG_CFA_REGISTER note.
3422 2018-06-26 Robin Dapp <rdapp@linux.ibm.com>
3424 * config/s390/s390.h (enum processor_flags): Do not use
3425 default tune parameter when -march was specified.
3427 2018-06-26 Jakub Jelinek <jakub@redhat.com>
3430 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
3431 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
3434 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
3436 Back port from trunk
3437 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
3439 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
3440 added IEEE/IBM long double multilib support on PowerPC little
3441 endian Linux systems.
3442 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
3443 (MULTILIB_DEFAULTS): Likewise.
3444 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3446 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
3447 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
3448 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
3450 2018-06-25 Segher Boessenkool <segher@kernel.crashing.org>
3453 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
3455 * config/rs6000/rs6000.md (abs<mode>2): Handle IFmode.
3457 2018-05-02 Jan Hubicka <jh@suse.cz>
3459 Backport from mainline
3460 2018-04-30 Jan Hubicka <jh@suse.cz>
3462 * lto-wrapper.c (ltrans_priorities): New static var.
3463 (cmp_priority): New.
3464 (run_gcc): Read priorities and if doing parallel build order
3465 the Makefile by them.
3467 2018-06-25 Segher Boessenkool <segher@kernel.crashing.org>
3470 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
3473 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
3474 ieee128 argument takes up only one (vector) register, not two (floating
3477 2018-06-23 Richard Sandiford <richard.sandiford@linaro.org>
3479 PR tree-optimization/85989
3480 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
3482 (backprop::intersect_uses): Check it when deciding whether this
3483 is a backedge reference.
3484 (backprop::process_block): Add each phi to m_visited_phis
3485 after visiting it, then clear it at the end.
3487 2018-06-22 David Edelsohn <dje.gcc@gmail.com>
3489 Backport from mainline
3491 2018-06-19 Tony Reix <tony.reix@atos.com>
3492 Damien Bergamini <damien.bergamini@atos.com>
3493 David Edelsohn <dje.gcc@gmail.com>
3495 * collect2.c (static_obj): New variable.
3496 (static_libs): New variable.
3497 (is_in_list): Uncomment declaration.
3498 (main): Track AIX libraries linked statically.
3499 (is_in_list): Uncomment definition.
3500 (scan_prog_file): Don't add AIX shared libraries initializer
3501 to constructor list if linking statically.
3503 2018-06-22 Michael Meissner <meissner@linux.ibm.com>
3505 Back port from trunk
3506 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
3508 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
3509 conversion insn that shows up when pr85657-3.c is compiled using
3510 IEEE 128-bit long double.
3512 Back port from trunk
3513 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
3515 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
3516 multiply and divide external functions from being created more
3519 Back port from trunk
3520 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
3522 * config/rs6000/rs6000.md (neg<mode>2_internal): Use the correct
3523 mode to check whether the mode is IBM extended.
3525 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
3528 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
3529 floating point modes, so that IFmode is numerically greater than
3530 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
3531 to declare the ordering. This prevents IFmode from being
3532 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
3533 machine. Include rs6000-modes.h to share the fractional values
3534 between genmodes* and the rest of the compiler.
3538 * config/rs6000/rs6000-modes.h: New file.
3539 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
3540 meaning of rs6000_long_double_size so that 126..128 selects an
3541 appropriate 128-bit floating point type.
3542 (rs6000_option_override_internal): Likewise.
3543 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
3544 (TARGET_LONG_DOUBLE_128): Change the meaning of
3545 rs6000_long_double_size so that 126..128 selects an appropriate
3546 128-bit floating point type.
3547 (LONG_DOUBLE_TYPE_SIZE): Update comment.
3548 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
3549 source and destination to match the standard usage.
3550 (truncifkf2): Likewise.
3551 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
3552 ISA 2.07 to use an explicit clobber, instead of passing in a
3554 (copysign<mode>3_soft): Likewise.
3556 Back port from trunk
3557 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
3559 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
3562 Back port from trunk
3563 2018-06-04 Michael Meissner <meissner@linux.ibm.com>
3565 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
3566 track if we pass or return IEEE 128-bit floating point.
3567 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
3568 C++ mangling that is compatible with GCC 8.1.
3569 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
3570 (init_cumulative_args): Note if we pass or return IEEE 128-bit
3571 floating point types.
3572 (rs6000_function_arg_advance_1): Likewise.
3573 (rs6000_mangle_type): Optionally generate mangled names that match
3574 what GCC 8.1 generated for IEEE 128-bit floating point types.
3575 (rs6000_globalize_decl_name): If we have an external function that
3576 passes or returns IEEE 128-bit types, generate a weak reference
3577 from the mangled name used in GCC 8.1 to the current mangled
3579 (rs6000_init_builtins): Make __ibm128 use the long double type if
3580 long double is IBM extended double. Make __float128 use the long
3581 double type if long double is IEEE 128-bit.
3584 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
3585 macro for __ibm128 built-in functions.
3586 (PACK_IF): Add __ibm128 pack/unpack functions.
3587 (UNPACK_IF): Likewise.
3588 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
3589 enable long double built-in functions if long double is IEEE
3590 128-bit floating point.
3591 (rs6000_invalid_builtin): Update long double built-in function
3593 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
3594 functions, adjust the built-in function to use the long double
3595 built-in function if __ibm128 and long double are the same type.
3596 * doc/extend.texi (PowerPC builtins): Update documention for
3597 __builtin_{,un}pack_longdouble. Add documentation for
3598 __builtin_{,un}pack_ibm128.
3600 Back port from trunk
3601 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
3603 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
3604 the 128-bit floating point types. Fix function comment.
3606 2018-06-22 Michael Meissner <meissner@linux.ibm.com>
3608 Back port from trunk
3609 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
3612 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
3613 define __ibm128 as long double.
3614 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
3615 as a distinct type with IEEE 128-bit floating point is supported.
3616 (init_float128_ieee): Fix up conversions between IFmode and IEEE
3617 128-bit types to use the correct functions.
3618 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
3619 convert between 128-bit floating point types that have different
3620 modes but the same representation, instead of using gen_lowpart to
3622 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
3624 (IFKF_reg): New attributes to give the register constraints for
3626 (extend<mode>tf2_internal): New insns to mark an explicit
3627 conversion between 128-bit floating point types that have a
3628 different mode but share the same representation.
3630 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3633 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
3634 -x assembler-with-cpp.
3636 2018-06-21 Sebastian Huber <sebastian.huber@embedded-brains.de>
3638 Backported from mainline
3639 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
3641 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
3642 multilibs for *-*-rtems*.
3643 * config/riscv/t-rtems: New file.
3645 2018-06-20 Jakub Jelinek <jakub@redhat.com>
3647 Backported from mainline
3648 2018-06-16 Jakub Jelinek <jakub@redhat.com>
3650 PR rtl-optimization/86108
3651 * bb-reorder.c (create_forwarder_block): Renamed to ...
3652 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
3653 jump from new landing pad to the second part.
3654 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
3657 2018-06-20 Jakub Jelinek <jakub@redhat.com>
3660 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
3663 PR tree-optimization/86231
3664 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
3665 anti-range don't overwrite *vr0min before using it to compute *vr0max.
3667 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
3669 Backport from mainline
3670 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
3672 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
3674 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
3676 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
3678 * gimplify.c (gimplify_init_constructor): Really never clear for an
3679 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
3681 2018-06-18 Martin Sebor <msebor@redhat.com>
3684 * calls.c (alloc_max_size): Correct a logic error/typo.
3685 Treat excessive arguments as infinite. Warn for invalid arguments.
3686 * doc/invoke.texi (-Walloc-size-larger-than): Update.
3688 2018-06-16 Jakub Jelinek <jakub@redhat.com>
3691 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
3692 documented as preserved for backward compatibility only.
3693 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
3695 2018-06-15 Jakub Jelinek <jakub@redhat.com>
3698 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
3699 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
3700 Only call store_expr for halves if the mode is the same.
3703 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
3704 Fix up comment formatting.
3706 2018-06-14 Jakub Jelinek <jakub@redhat.com>
3709 * match.pd ((A +- CST1) +- CST2): Punt if last resort
3710 unsigned_type_for returns NULL.
3713 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
3714 subregs of multi-word pseudos unless the float mode has word size.
3716 2018-06-14 Richard Biener <rguenther@suse.de>
3719 * tree-vect-generic.c (build_word_mode_vector_type): Remove
3720 duplicate and harmful type_hash_canon.
3721 * tree.c (type_hash_canon): Assert we didn't find ourselves.
3723 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
3725 Backport from mainline
3726 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
3728 * config/rtems.h (STDINT_LONG32): Define.
3730 2018-05-05 Roland McGrath <mcgrathr@google.com>
3733 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
3734 any section for which we don't know a specific type it should have,
3735 regardless of name. Previously this was done only for the exact
3736 names ".init_array", ".fini_array", and ".preinit_array".
3737 (default_elf_asm_named_section): Add comment about
3738 relationship with default_section_type_flags and SECTION_NOTYPE.
3739 (get_section): Don't consider it a type conflict if one side has
3740 SECTION_NOTYPE and the other doesn't, as long as neither has the
3741 SECTION_BSS et al used in the default_section_type_flags logic.
3743 2018-06-13 Martin Sebor <msebor@redhat.com>
3745 PR tree-optimization/86114
3746 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
3748 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
3750 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
3753 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
3754 offsets for register save directives. Emit a second batch of save
3755 directives, if need be, when the function accesses prior frames.
3757 2018-06-11 Jason Merrill <jason@redhat.com>
3759 PR c++/80485 - inline function non-zero address.
3760 * symtab.c (nonzero_address): Check DECL_COMDAT.
3762 2018-06-11 Peter Bergner <bergner@vnet.ibm.com>
3764 Backport from mainline
3765 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
3768 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
3771 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
3774 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
3777 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
3778 on the correct operand.
3779 (*movdi_internal64): Ditto.
3781 2018-06-08 Carl Love <cel@us.ibm.com>
3783 Backport from mainline
3784 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
3785 vextract_fp_from_shortl): Add BE support.
3787 2018-06-08 Carl Love <cel@us.ibm.com>
3789 Backport from mainline
3790 * gcc/config/rs6000/vsx.md (first_match_index_<mode>):
3791 Calculate index using natureal element order.
3792 (first_match_or_eos_index_<mode>):
3793 Calculate index using natural element order.
3794 (first_match_index_<mode>):
3795 Calculate index using natural element order.
3796 (first_match_or_eos_index_<mode>):
3797 Calculate index using natural order.
3798 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
3799 for BE and LE modes.
3800 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
3801 P9V_BUILTIN_VCLZLSBB_V16QI.
3802 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
3805 2018-06-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3807 Backport from mainline
3808 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3811 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
3812 qualifier_void_pointer and qualifier_const_void_pointer.
3813 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
3814 (arm_init_builtins): Handle the above.
3815 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
3816 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
3817 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
3820 2018-06-07 Uros Bizjak <ubizjak@gmail.com>
3823 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
3824 Generate SImode target register for null target.
3825 <case IX86_BUILTIN_XGETBV>: Ditto.
3826 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
3827 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
3829 2018-06-07 Peter Bergner <bergner@vnet.ibm.com>
3831 Backport from mainline
3832 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
3835 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
3836 Don't handle -mcpu=power8 if -mpower9-vector is also used.
3838 2018-06-06 Richard Biener <rguenther@suse.de>
3840 Backport from mainline
3841 2018-06-04 Richard Biener <rguenther@suse.de>
3843 PR tree-optimization/86038
3844 * tracer.c (find_best_successor): Check probability for
3845 being initialized, bail out if not.
3847 2018-05-30 Richard Biener <rguenther@suse.de>
3849 PR tree-optimization/85964
3850 * tracer.c (better_p): Drop initialized count check, we only
3851 call the function with initialized counts now.
3852 (find_best_successor): Do find a best edge if one
3853 has uninitialized count.
3854 (find_best_predecessor): Likewise. Do BB frequency check only
3855 if count is initialized.
3857 2018-05-28 Richard Biener <rguenther@suse.de>
3859 PR tree-optimization/85934
3860 * tree-vect-generic.c (expand_vector_operations_1): Hoist
3861 vector boolean check before scalar optimization.
3863 2018-05-22 Richard Biener <rguenther@suse.de>
3865 PR tree-optimization/85863
3866 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
3867 comparisons when vectype is specified.
3868 (vectorizable_condition): Do not specify vectype for
3869 vect_is_simple_cond when SLP vectorizing.
3871 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
3873 Backport from mainline
3874 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
3876 * config/s390/s390-builtin-types.def: Add void function type.
3877 * config/s390/s390-builtins.def: Use the function type for the
3880 2018-06-04 Martin Sebor <msebor@redhat.com>
3883 * calls.c (maybe_warn_nonstring_arg): Use string length to set
3884 or ajust the presumed bound on an operation to avoid unnecessary
3887 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
3890 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
3891 of bits to ignore when comparing architectures.
3893 2018-06-04 Jakub Jelinek <jakub@redhat.com>
3896 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
3898 2018-06-01 Bill Schmidt <wschmidt@linux.ibm.com>
3900 PR tree-optimization/85712
3901 Backport from mainline:
3902 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
3904 PR tree-optimization/85712
3905 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
3907 (alloc_cand_and_find_basis): Initialize first_interp field.
3908 (slsr_process_mul): Modify first_interp field.
3909 (slsr_process_add): Likewise.
3910 (slsr_process_cast): Modify first_interp field for each new
3912 (slsr_process_copy): Likewise.
3913 (dump_candidate): Dump first_interp field.
3914 (replace_mult_candidate): Process all interpretations, not just
3916 (replace_rhs_if_not_dup): Likewise.
3917 (replace_one_candidate): Likewise.
3919 Backport from mainline:
3920 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
3922 PR tree-optimization/85712
3923 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
3924 this candidate has already been replaced in-situ by a copy.
3926 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
3929 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
3930 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
3932 (sse4_1_round<mode>2): Use nonimmediate_operand
3933 for operand 1 predicate.
3935 2018-05-31 Jakub Jelinek <jakub@redhat.com>
3938 * bb-reorder.c (pass_partition_blocks::gate): Return false for
3939 functions with naked attribute.
3941 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
3944 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
3945 and movx for Haswell.
3947 2018-05-30 Jakub Jelinek <jakub@redhat.com>
3949 Backported from mainline
3950 2018-05-14 Jakub Jelinek <jakub@redhat.com>
3953 * config/i386/i386.md: Disallow non-commutative arithmetics in
3954 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
3955 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
3956 in the peephole2 before it.
3958 2018-05-08 Jakub Jelinek <jakub@redhat.com>
3961 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
3962 after cmpelim optimization.
3964 2018-05-28 H.J. Lu <hongjiu.lu@intel.com>
3966 Backport from mainline
3967 2018-05-26 H.J. Lu <hongjiu.lu@intel.com>
3971 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
3973 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3977 * varasm.c (assemble_alias): Check ifunc_resolver only on
3980 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
3983 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
3985 (cgraph_node::create_alias): Likewise.
3986 (cgraph_node::get_availability): Check ifunc_resolver instead
3987 of looking up ifunc attribute.
3988 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
3989 * varasm.c (do_assemble_alias): Likewise.
3990 (assemble_alias): Likewise.
3991 (default_binds_local_p_3): Likewise.
3992 * cgraph.h (cgraph_node): Add ifunc_resolver.
3993 (cgraph_node::only_called_directly_or_aliased_p): Return false
3995 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
3997 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
3998 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
3999 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
4000 instead of looking up ifunc attribute.
4002 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
4004 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
4006 2018-05-25 Richard Biener <rguenther@suse.de>
4009 * tree-dump.c (dequeue_and_dump): Remove access to removed
4010 operand 2 of a SWITCH_EXPR.
4012 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
4014 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
4015 Add {q} suffix to insn mnemonic.
4017 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
4020 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
4021 when memory input operand is handled.
4023 2018-05-23 Bin Cheng <bin.cheng@arm.com>
4025 Backport from mainline
4026 2018-05-17 Bin Cheng <bin.cheng@arm.com>
4027 Richard Biener <rguenther@suse.de>
4029 PR tree-optimization/85793
4030 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
4031 for VMAT_ELEMENTWISE.
4033 2018-05-23 Richard Biener <rguenther@suse.de>
4036 * tree-data-ref.c (create_runtime_alias_checks): Defer
4037 and ignore overflow warnings.
4039 2018-05-23 Martin Liska <mliska@suse.cz>
4041 Backport from mainline
4042 2018-05-23 Yury Gribov <tetra2005@gmail.com>
4044 PR tree-optimization/85822
4045 * tree-vrp.c (is_masked_range_test): Fix handling of negative
4048 2018-05-23 Martin Liska <mliska@suse.cz>
4050 Backport from mainline
4051 2018-05-11 Martin Liska <mliska@suse.cz>
4054 * doc/extend.texi: Document LLVM style format for no_sanitize
4057 2018-05-22 Martin Sebor <msebor@redhat.com>
4060 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
4062 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
4064 PR tree-optimization/85814
4065 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
4066 a null return from get_strinfo when unsharing the next
4067 strinfo in the chain.
4069 2018-05-21 Pat Haugen <pthaugen@us.ibm.com>
4071 Backport from mainline
4072 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
4073 Segher Boessenkool <segher@kernel.crashing.org>
4076 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest operand.
4078 2018-05-18 Toon Moene <toon@moene.org>
4080 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
4081 directly after that of -floop-interchange. Indicate that both
4082 options are enabled by default when specifying -O3.
4084 2018-05-11 Martin Jambor <mjambor@suse.cz>
4087 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
4090 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4093 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
4095 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4098 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
4100 (cortex-m0): Use armv6s-m isa.
4101 (cortex-m0plus): Likewise.
4102 (cortex-m1): Likewise.
4103 (cortex-m0.small-multiply): Likewise.
4104 (cortex-m0plus.small-multiply): Likewise.
4105 (cortex-m1.small-multiply): Likewise.
4107 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
4109 PR rtl-optimization/85638
4110 * bb-reorder.c: Include common/common-target.h.
4111 (create_forwarder_block): New function extracted from...
4112 (fix_up_crossing_landing_pad): ...here. Rename into...
4113 (dw2_fix_up_crossing_landing_pad): ...this.
4114 (sjlj_fix_up_crossing_landing_pad): New function.
4115 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
4116 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
4117 from both partitions and exit the loop after one iteration.
4119 2018-05-08 Richard Biener <rguenther@suse.de>
4121 Backport from mainline
4122 2018-05-04 Richard Biener <rguenther@suse.de>
4125 * fold-const.c (negate_expr_p): Restrict negation of operand
4126 zero of a division to when we know that can happen without
4128 (fold_negate_expr_1): Likewise.
4130 2018-05-03 Richard Biener <rguenther@suse.de>
4132 PR tree-optimization/85615
4133 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
4134 to loops not nested in BBs loop father to avoid creating multi-entry
4137 2018-05-02 Richard Biener <rguenther@suse.de>
4140 * gimplify.c (gimplify_save_expr): When in SSA form allow
4141 SAVE_EXPRs to compute to SSA vars.
4143 2018-05-02 Richard Biener <rguenther@suse.de>
4145 PR tree-optimization/85597
4146 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
4147 do not use split vect_get_vec_defs call but call vect_get_slp_defs
4150 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
4153 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
4154 (check_arch): Likewise.
4155 (check_fpu): Return the result rather than printing it.
4156 (end arch): Fix operator precedence.
4157 (end cpu): Likewise.
4158 (END): Print the result from check_fpu.
4160 2018-05-08 Richard Biener <rguenther@suse.de>
4162 Backport from mainline
4163 2018-05-04 Richard Biener <rguenther@suse.de>
4165 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
4166 * gimple-ssa-store-merging.c
4167 (imm_store_chain_info::output_merged_store): Remove redundant create,
4168 release split_store vector contents on failure.
4169 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
4170 scalar stmt vector on cache hit.
4172 2018-05-04 Richard Biener <rguenther@suse.de>
4175 * tree-complex.c (tree_lower_complex): Handle removed blocks.
4177 2018-05-07 Chung-Ju Wu <jasonwucj@gmail.com>
4179 Backport from mainline.
4180 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4182 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
4183 GET_MODE_MASK before any checking.
4184 (nds32_can_use_bset_p): Likewise.
4185 (nds32_can_use_btgl_p): Likewise.
4187 2018-05-07 Chung-Ju Wu <jasonwucj@gmail.com>
4189 Backport from mainline.
4190 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4192 * config/nds32/nds32-doubleword.md: New define_split pattern for
4193 illegal register number.
4195 2018-05-07 Chung-Ju Wu <jasonwucj@gmail.com>
4197 Backport from mainline.
4198 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4200 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
4202 2018-05-07 Chung-Ju Wu <jasonwucj@gmail.com>
4204 Backport from mainline.
4205 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4207 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
4209 2018-05-06 Jakub Jelinek <jakub@redhat.com>
4212 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
4213 the type is addressable. Don't force op into register if it has
4216 2018-05-04 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4218 Backport from trunk.
4219 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4222 * config/aarch64/constraints.md (Usg): Limit to 31.
4225 2018-05-02 Jakub Jelinek <jakub@redhat.com>
4227 * BASE-VER: Set to 8.1.1.
4229 2018-05-02 Release Manager
4231 * GCC 8.1.0 released.
4233 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
4235 PR tree-optimization/85586
4236 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
4237 exit early for statements in the same group if the accesses are
4240 2018-05-01 Jakub Jelinek <jakub@redhat.com>
4243 * doc/install.texi2html: Replace _002d with - and _002a with * in
4244 generated html files using sed.
4246 2018-04-27 Martin Jambor <mjambor@suse.cz>
4249 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
4250 the jump function allows for passing through aggregate values.
4252 2018-04-27 Jeff Law <law@redhat.com>
4254 * config.gcc: Mark tile* targets as deprecated/obsolete.
4256 2018-04-27 Jakub Jelinek <jakub@redhat.com>
4258 PR tree-optimization/85529
4259 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
4260 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
4261 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
4262 zero extension or masking of the MSB bit.
4263 (optimize_range_tests): Add FIRST_BB argument, pass it through
4264 to optimize_range_tests_var_bound.
4265 (maybe_optimize_range_tests, reassociate_bb): Adjust
4266 optimize_range_tests callers.
4268 2018-04-26 Richard Biener <rguenther@suse.de>
4269 Jakub Jelinek <jakub@redhat.com>
4271 * cgraph.h (symbol_table): Just declare debug method here.
4272 * symtab.c (symbol_table::debug): Define.
4274 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
4276 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
4278 2018-04-25 H.J. Lu <hongjiu.lu@intel.com>
4280 Backport from mainline
4281 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
4284 * config/i386/i386.c (ix86_expand_builtin): Change memory
4285 operand to XI, extend p0 to Pmode.
4286 * config/i386/i386.md: Change unspec volatile and operand
4287 1 mode to XI, change operand 0 mode to P.
4289 2018-04-25 Jakub Jelinek <jakub@redhat.com>
4291 * DEV-PHASE: Set to prerelease.
4293 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
4295 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
4296 in __abskf2 and __powikf2.
4298 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4301 * config/aarch64/constraints.md (Usg, Usj): New constraints.
4302 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
4303 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
4304 Use the above on operand 2. Reindent.
4305 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
4307 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
4310 * common/config/i386/i386-common.c (ix86_handle_option): Don't
4312 * config/i386/i386.opt (mcet): Removed.
4313 * doc/install.texi: Remove -mcet documentation.
4314 * doc/invoke.texi: Likewise.
4316 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
4319 * doc/install.texi: Remove -mcet from bootstrap-cet.
4321 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4324 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
4325 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
4329 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
4330 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
4331 containing a CONST_VECTOR.
4333 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
4335 * doc/install.texi: Update newlib dependency for nvptx.
4337 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4340 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
4341 instead of INTVAL when shifting x left.
4343 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
4345 PR tree-optimization/85478
4346 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
4347 vect_grouped_store_supported for single element vectors.
4349 2018-04-24 Richard Biener <rguenther@suse.de>
4352 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
4353 load cost increase to the case of non-constant step.
4355 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4358 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
4359 destination if any_malformed_asm.
4361 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
4364 * expr.c (store_field): In the bitfield case, if the value comes from
4365 a function call and is returned in registers by means of a PARALLEL,
4366 do not change the mode of the temporary unless BLKmode and VOIDmode.
4368 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
4370 PR rtl-optimization/85423
4371 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
4372 dependencies to debug insns when the previous insn is non-debug.
4374 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
4376 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
4377 enums into a single definition.
4378 (fls): Fix predicates and printing.
4381 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
4383 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
4384 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
4385 and short u6 immediate.
4386 (check_if_valid_sleep_operand): Remove.
4387 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
4389 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4391 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
4392 flag_always_save_lp condition.
4393 * config/nds32/nds32.opt (malways-save-lp): New option.
4395 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4397 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
4398 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
4399 * config/nds32/nds32.h
4400 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
4401 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
4403 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4405 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
4406 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
4408 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4409 Chung-Ju Wu <jasonwucj@gmail.com>
4411 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
4413 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
4414 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
4416 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4418 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
4420 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4422 * config/nds32/nds32-protos.h (nds32_data_alignment,
4423 nds32_local_alignment): Declare.
4424 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
4425 nds32_local_alignment): New functions.
4426 (TARGET_CONSTANT_ALIGNMENT): Define.
4427 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
4429 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4431 * config/nds32/nds32.c
4432 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
4433 (TARGET_MODES_TIEABLE_P): Likewise.
4435 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4437 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
4440 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
4441 Chung-Ju Wu <jasonwucj@gmail.com>
4443 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4444 for unaligned access.
4445 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4447 * config/nds32/nds32-intrinsic.md: Likewise.
4448 * config/nds32/nds32_intrinsic.h: Likewise.
4449 * config/nds32/nds32.h (nds32_builtins): Likewise.
4450 * config/nds32/nds32.opt (munaligned-access): New option.
4451 * config/nds32/nds32.c (nds32_asm_file_start): Display
4452 flag_unaligned_access status.
4454 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4456 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
4457 -mno-relax is present.
4458 * config/riscv/linux.h (LINK_SPEC): Ditto.
4460 2018-04-20 Martin Sebor <msebor@redhat.com>
4463 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
4465 (gimple_fold_builtin_stxcpy_chk): Same.
4466 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
4468 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
4471 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
4472 __powikf2 when long double is IEEE 128-bit.
4474 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4476 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
4477 step to make sure stack always aligned.
4479 2018-04-20 Carl Love <cel@us.ibm.com>
4482 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
4483 size check for arg0.
4485 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
4486 Tom de Vries <tom@codesourcery.com>
4489 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
4490 Emit insns for calls too.
4491 (nvptx_find_par): Always look for worker-level predecessor insn.
4492 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
4494 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
4495 (nvptx_process_pars): Propagate frames for calls.
4497 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
4500 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
4502 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
4503 (ix86_handle_option): Don't handle OPT_mibt.
4504 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
4506 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
4508 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
4509 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
4510 (ix86_target_macros): Define __CET__ with flag_cf_protection
4511 for -fcf-protection.
4512 * config/i386/i386.c (isa2_opts): Remove -mibt.
4513 * config/i386/i386.h (TARGET_IBT): Removed.
4514 (TARGET_IBT_P): Likewise.
4515 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
4516 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
4517 * config/i386/i386.opt (mcet): Update help message.
4520 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
4521 -mcet as an alias for -mshstk.
4523 2018-04-20 Richard Biener <rguenther@suse.de>
4526 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
4527 complexity by forcing a single use of the multiply operand.
4529 2018-04-20 Martin Jambor <mjambor@suse.cz>
4532 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
4533 recursion dependency to only apply to non-clones.
4535 2018-04-20 Martin Jambor <mjambor@suse.cz>
4538 * ipa-cp.c (create_specialized_node): Check that clones of
4539 self-recursive edges exist during IPA-CP.
4541 2018-04-19 Toon Moene <toon@moene.org>
4543 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
4546 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4548 PR tree-optimization/85467
4549 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
4550 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
4551 VECTOR_CST element to type.
4553 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4556 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
4557 * config/i386/i386.md (builtin_setjmp_setup): Removed.
4558 (builtin_longjmp): Likewise.
4559 (save_stack_nonlocal): New pattern.
4560 (restore_stack_nonlocal): Likewise.
4562 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4565 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4566 Replace ASM_OUTPUT_LABEL with fprintf.
4568 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4571 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4572 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
4573 * config/i386/i386-c.c (ix86_target_macros_internal): Also
4574 define __IBT__ and __SHSTK__ for -fcf-protection.
4575 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
4577 (ix86_trampoline_init): Likewise.
4578 (x86_output_mi_thunk): Likewise.
4579 (ix86_notrack_prefixed_insn_p): Likewise.
4580 (ix86_option_override_internal): Don't disallow -fcf-protection.
4581 * config/i386/i386.md (rdssp<mode>): Also enable for
4583 (incssp<mode>): Likewise.
4584 (nop_endbr): Likewise.
4585 * config/i386/i386.opt (mcet): Change help message to built-in
4589 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
4590 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
4591 enable CET built-in functions.
4593 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
4595 * common/config/i386/i386-common.c
4596 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
4597 OPTION_MASK_ISA_MOVDIRI_UNSET,
4598 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
4599 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
4600 * config.gcc (movdirintrin.h): New header.
4601 * config/i386/cpuid.h (bit_MOVDIRI,
4602 bit_MOVDIR64B): New bits.
4603 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
4605 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
4606 (VOID, PVOID, PCVOID)): New function types.
4607 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
4608 __builtin_ia32_directstoreu_u64,
4609 __builtin_ia32_movdir64b): New builtins.
4610 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
4611 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
4613 (ix86_valid_target_attribute_inner_p): Ditto.
4614 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
4615 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
4616 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
4617 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
4618 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
4619 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
4620 (movdiri<mode>, movdir64b_<mode>): New.
4621 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
4622 * config/i386/immintrin.h: Include movdirintrin.h.
4623 * config/i386/movdirintrin.h: New file.
4624 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
4626 2018-04-19 Richard Biener <rguenther@suse.de>
4629 * cfg.c (clear_bb_flags): When loop state says we have
4630 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
4632 2018-04-19 Richard Biener <rguenther@suse.de>
4634 PR tree-optimization/84737
4635 * tree-vect-data-refs.c (vect_copy_ref_info): New function
4636 copying restrict info.
4637 (vect_setup_realignment): Use it.
4638 * tree-vectorizer.h (vect_copy_ref_info): Declare.
4639 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
4640 the first DR to all generated stores.
4641 (vectorizable_load): Likewise for loads.
4643 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4645 PR tree-optimization/85446
4646 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
4647 the integral and pointer types to have the same precision.
4649 * doc/install.texi: Document --disable-cet being the default and
4652 2018-04-18 Martin Liska <mliska@suse.cz>
4654 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
4657 2018-04-18 Martin Liska <mliska@suse.cz>
4660 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4663 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4664 arguments if they are comparable.
4666 2018-04-18 Martin Liska <mliska@suse.cz>
4669 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4672 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4675 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
4678 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
4679 ENDBR after calling __morestack.
4681 2018-04-18 David Malcolm <dmalcolm@redhat.com>
4684 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
4685 by using gcc_base_ver to generate a gcc_driver_version, and use
4686 it when generating GCC_DRIVER_NAME.
4687 * configure: Regenerate.
4689 2018-04-18 Jakub Jelinek <jakub@redhat.com>
4692 * config.gcc: Obsolete powerpc*-*-*spe*.
4694 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4697 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
4698 (stabstr_D): Change type of unum from unsigned int to
4699 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
4702 2018-04-17 Jim Wilson <jimw@sifive.com>
4705 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
4706 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
4707 Set arg_pointer_offset after using pretend_args_size.
4709 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4711 PR rtl-optimization/85431
4712 * dse.c (record_store): Ignore zero width stores.
4715 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
4716 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
4717 __builtin_stack_restore rather than after it.
4718 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
4719 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
4720 argument instead of virtual_dynamic_stack_rtx.
4722 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
4724 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
4726 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4727 Add note to error message to explain internal mapping of overloaded
4728 built-in function name to non-overloaded built-in function name.
4729 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
4732 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4735 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
4736 where the inputs overlap with the output.
4738 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4740 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
4741 (=v, v) alternative and explicit "memory" attribute.
4742 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
4743 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4745 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
4746 "sselog1" type instead of "sselog".
4747 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
4748 "sselog". Remove explicit "memory" attribute.
4749 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
4750 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4752 (vec_extract_hi_v32hi): Merge all alternatives into one, use
4753 "sselog1" type instead of "sselog". Remove explicit "memory"
4755 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
4756 use "sselog1" type instead of "sselog". Remove explicit "memory"
4758 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
4759 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4761 (vec_extract_hi_v64qi): Merge all alternatives into one, use
4762 "sselog1" type instead of "sselog". Remove explicit "memory"
4764 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
4765 use "sselog1" type instead of "sselog". Remove explicit "memory"
4769 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
4772 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
4775 2018-04-17 Martin Jambor <mjambor@suse.cz>
4778 * ipa-cp.c (create_specialized_node): Call
4779 expand_all_artificial_thunks if necessary.
4781 2018-04-17 Martin Liska <mliska@suse.cz>
4784 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
4785 in message, remote space in between '_G' and '('.
4787 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4790 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
4791 avx512f_vmcmp<mode>3<round_saeonly_name>,
4792 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
4793 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
4794 avx512f_rndscale<mode><round_saeonly_name>,
4795 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
4796 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
4797 Use %<iptr>2 instead of %2 for -masm=intel.
4798 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
4799 avx512f_vcvttss2usi<round_saeonly_name>,
4800 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
4802 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
4803 avx512f_vcvttsd2usi<round_saeonly_name>,
4804 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
4805 Use %q1 instead of %1 for -masm=intel.
4806 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
4807 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
4808 of %3 for -masm=intel.
4809 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
4811 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
4813 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
4815 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
4816 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
4818 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
4820 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
4821 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
4822 %g1 and one with %0 and %1.
4823 (avx512er_vmrcp28<mode><round_saeonly_name>,
4824 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
4826 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
4827 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
4828 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
4829 of %0 and %{%4%} for -masm=intel.
4830 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
4831 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
4832 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
4833 order of %0 and %{%5%}%{z%} for -masm=intel.
4835 2018-04-17 Jan Hubicka <jh@suse.cz>
4838 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
4840 2018-04-17 Martin Liska <mliska@suse.cz>
4843 * multiple_target.c (create_dispatcher_calls): Set apostrophes
4844 for target_clone error message. Make default implementation
4845 clone to be a local declaration.
4846 (separate_attrs): Add new argument and check for an empty
4848 (expand_target_clones): Handle it.
4849 (ipa_target_clone): Make redirection just for target_clones
4852 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
4853 Tom de Vries <tom@codesourcery.com>
4856 * omp-expand.c (expand_oacc_for): Add dummy false branch for
4857 tiled basic blocks without omp continue statements.
4859 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
4862 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
4863 vec_extract expression as having side effects to make sure it gets
4866 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
4869 * config/i386/i386.c (get_builtin_code_for_version): Check
4872 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
4875 * gcc/config.gcc: Support "skylake".
4876 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
4878 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
4879 (processor_target_table): Add "skylake".
4880 (ix86_option_override_internal): Add "skylake".
4881 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
4882 PROCESSOR_CANNONLAKE.
4883 (get_builtin_code_for_version): Fix priority for
4884 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
4885 PROCESSOR_SKYLAKE-AVX512.
4886 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
4887 (processor_type): Add PROCESSOR_SKYLAKE.
4889 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
4890 Jason Merrill <jason@redhat.com>
4893 * convert.c (convert_to_integer_1): Use direct recursion for
4894 enumeral types and types with a precision less than the number
4895 of bits in their mode.
4897 2018-04-16 Julia Koval <julia.koval@intel.com>
4900 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
4901 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
4903 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
4906 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
4907 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
4908 and -mno-direct-move.
4910 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
4913 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
4914 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
4915 Ensure negative shifts result in {0}.
4917 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
4919 PR rtl-optimization/79916
4920 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
4921 regs (if any) to define how to gnerate SD moves when LRA is in
4924 2018-04-13 Jakub Jelinek <jakub@redhat.com>
4926 PR rtl-optimization/85393
4927 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
4928 * except.c (expand_dw2_landing_pad_for_region): Make static.
4929 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
4930 a label and unconditional jump to old_bb, rather than
4931 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
4934 PR rtl-optimization/85376
4935 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
4936 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
4937 instead of a specific value.
4939 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4940 Bin Cheng <bin.cheng@arm.com>
4942 PR tree-optimization/82965
4943 PR tree-optimization/83991
4944 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
4945 by_profile_only parameter.
4946 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
4947 information if the loop was predicted to iterate too many times.
4948 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
4950 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4953 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
4956 2018-04-13 Martin Liska <mliska@suse.cz>
4957 Jakub Jelinek <jakub@redhat.com>
4960 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
4961 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
4962 * builtins.c (expand_builtin_memory_copy_args): Use
4963 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
4964 handle dest_addr == pc_rtx.
4966 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
4969 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
4970 asked to not generate direct moves.
4971 (fix_trunc<mode>si2_stfiwx): Similar.
4972 (fix_trunc<mode>si2_internal): Similar.
4974 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4977 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
4978 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
4979 lookup if dest in some wider mode is known to be const0_rtx and
4980 if so, record permanent equivalence for it to be ZERO_EXTEND of
4981 the narrower mode destination.
4983 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4985 * lto-streamer-out.c (output_function): Revert 259346.
4986 * omp-expand.c (expand_oacc_for): Likewise.
4988 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
4990 PR rtl-optimization/85354
4991 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
4992 * sel-sched.c (sel_global_init): ... here.
4994 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
4997 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
4998 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4999 mode for PE-COFF targets.
5000 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
5001 (i386_pe_asm_lto_end): Likewise.
5002 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
5003 (TARGET_ASM_LTO_END): Likewise.
5004 * config/i386/winnt.c (saved_debug_info_level): New static variable.
5005 (i386_pe_asm_lto_start): New function.
5006 (i386_pe_asm_lto_end): Likewise.
5008 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
5009 Richard Biener <rguenther@suse.de>
5012 * lto-streamer-out.c (output_function): Fix CFG loop state before
5014 * omp-expand.c (expand_oacc_for): Handle calls to internal
5015 functions like regular functions.
5017 2018-04-12 Richard Biener <rguenther@suse.de>
5020 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
5021 for the early LTO debug to properly generate references to it
5022 during DIE emission. Do not re-use that for the skeleton for
5024 (dwarf2out_early_finish): Likewise.
5026 2018-04-12 Jakub Jelinek <jakub@redhat.com>
5029 * config/i386/sse.md
5030 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
5031 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
5032 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
5033 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
5034 and output is a reg, avoid creating invalid lowpart subreg, but
5035 instead split into a 512-bit move. Don't split if not AVX512VL,
5036 input is xmm16+ reg and output is a mem.
5037 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
5038 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
5039 xmm16+ reg and output is a mem.
5041 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5043 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
5044 also for flag_dwarf2_cfi_asm.
5046 2018-04-12 Jakub Jelinek <jakub@redhat.com>
5048 PR rtl-optimization/85342
5049 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
5050 a bool scalar var inside of the loop instead. Don't try to update
5051 recog_data.operand after failed apply_change_group.
5053 2018-04-12 Tom de Vries <tom@codesourcery.com>
5056 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
5057 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
5058 array with flexible array member as array without given dimension.
5059 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
5060 argument for undefined param to true.
5062 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
5065 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
5066 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
5067 from PowerPC section.
5068 * config/rs6000/sysv4.opt (mcall-): Improve help text.
5069 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
5070 help text that is too long.
5071 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
5072 help text that is too long.
5073 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
5074 help text that is too long.
5076 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
5078 * config/alpha/alpha.md (stack_probe_internal): Rename
5079 from "probe_stack". Update all callers.
5081 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
5083 PR rtl-optimization/84566
5084 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
5085 sched_macro_fuse_insns.
5087 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
5090 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
5091 (compute_block_dependences): ... from here.
5093 2018-04-11 Jakub Jelinek <jakub@redhat.com>
5095 PR tree-optimization/85331
5096 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
5097 from int to HOST_WIDE_INT.
5099 2018-04-11 Martin Jambor <mjambor@suse.cz>
5102 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
5103 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
5104 not the same as the source val.
5105 (cgraph_edge_brings_value_p): New parameter.
5106 (gather_edges_for_value): Pass destination value to
5107 cgraph_edge_brings_value_p.
5108 (perhaps_add_new_callers): Likewise.
5109 (get_info_about_necessary_edges): Likewise and exclude values brought
5110 only by self-recursive edges.
5111 (create_specialized_node): Redirect only clones of self-calling edges.
5112 (+self_recursive_pass_through_p): New function.
5113 (find_more_scalar_values_for_callers_subset): Use it.
5114 (find_aggregate_values_for_callers_subset): Likewise.
5115 (known_aggs_to_agg_replacement_list): Removed.
5116 (decide_whether_version_node): Re-calculate known constants for all
5117 remaining context clones.
5119 2018-04-11 Richard Biener <rguenther@suse.de>
5122 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
5123 from early DWARF output.
5124 (dwarf2out_early_finish): Output line info unconditionally into
5125 early DWARF and add reference to it.
5127 2018-04-11 Jakub Jelinek <jakub@redhat.com>
5130 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
5131 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
5132 other than V2DFmode using iptr mode attribute.
5133 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
5135 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
5137 PR rtl-optimization/84659
5138 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
5140 2018-04-11 Jakub Jelinek <jakub@redhat.com>
5143 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
5145 (output_loc_list): Pass address of a dummy size variable even in the
5146 locview handling loop.
5147 (index_location_lists): Add comment on why skip_loc_list_entry can't
5150 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
5153 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
5156 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
5159 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
5160 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
5161 and -mstring-compare-inline-limit.
5163 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
5166 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
5167 for stack clash protection in a register whenever we need it to be in
5170 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
5172 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
5173 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
5175 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
5178 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
5180 (mlong-double-): Ditto.
5181 * config/rs6000/sysv4.opt (msdata=): Ditto.
5182 (mtls-size=): Ditto.
5184 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
5186 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
5187 erroneous entries for
5188 "vector int vec_ldl (int, long int *)", and
5189 "vector unsigned int vec_ldl (int, unsigned long int *)".
5190 Add comments and entries for
5191 "vector bool char vec_ldl (int, bool char *)",
5192 "vector bool short vec_ldl (int, bool short *)",
5193 "vector bool int vec_ldl (int, bool int *)",
5194 "vector bool long long vec_ldl (int, bool long long *)",
5195 "vector pixel vec_ldl (int, pixel *)",
5196 "vector long long vec_ldl (int, long long *)",
5197 "vector unsigned long long vec_ldl (int, unsigned long long *)".
5198 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
5199 type tree bool_long_long_type_node and correct definition of
5200 bool_V2DI_type_node to make reference to this new type tree.
5201 (rs6000_mangle_type): Replace erroneous reference to
5202 bool_long_type_node with bool_long_long_type_node.
5203 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
5204 comments to emphasize sign distinctions for char and int types and
5205 replace RS6000_BTI_bool_long constant with
5206 RS6000_BTI_bool_long_long constant. Also add comment to restrict
5207 use of RS6000_BTI_pixel.
5208 (bool_long_type_node): Remove this macro definition.
5209 (bool_long_long_type_node): New macro definition
5211 2018-04-10 Jakub Jelinek <jakub@redhat.com>
5213 PR rtl-optimization/85300
5214 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
5215 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
5216 simplify_unary_operation fails.
5218 2018-04-10 Martin Liska <mliska@suse.cz>
5220 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
5221 cgraph_edge and ipa_ref.
5223 2018-04-10 Jakub Jelinek <jakub@redhat.com>
5227 * config/i386/sse.md
5228 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
5229 computation of the VEC_MERGE selector from mask.
5230 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
5231 Fix decoding of the VEC_MERGE selector into mask.
5233 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
5235 PR tree-optimization/85286
5236 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
5238 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
5240 * final.c (final_1): Set insn_last_address as well as
5241 insn_current_address.
5243 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5246 * explow.c (emit_stack_probe): Call validize_mem on memory location
5247 before passing it to gen_probe_stack. Create address operand and
5248 legitimize it for the probe_stack_address case.
5250 2018-04-09 Jan Hubicka <jh@suse.cz>
5253 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
5254 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
5255 * tree.c (free_lang_data_in_type): Fix handling of binfos;
5257 (free_lang_data): Rebuild type inheritance graph.
5259 2018-04-09 Martin Sebor <msebor@redhat.com>
5261 * invoke.texi (-finline-small-functions): Mention other optimization
5263 (-findirect-inlining, -fpartial-inlining): Same.
5264 (-finline-functions-called-once): Same.
5265 (-freorder-blocks-and-partition): Same.
5267 2018-04-09 Jan Hubicka <jh@suse.cz>
5270 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
5271 jumps; choose last target that matches the criteria (i.e.
5272 no partition changes for non-crossing jumps).
5273 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
5274 support for redirecting crossing jumps to non-crossing.
5276 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
5278 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
5279 also for naked functions.
5281 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
5283 * config/arc/arc.md (add_shift): New pattern.
5284 (add_shift2): Likewise.
5285 (sub_shift): Likewise.
5286 (sub_shift_cmp0_noout): Likewise.
5287 (compare_si_ashiftsi): Likewise.
5288 (xbfu_cmp0_noout): New combine pattern.
5289 (xbfu_cmp0"): Likewise.
5290 (movsi_set_cc_insn): Place the predicable variant first.
5291 (commutative_binary_cmp0_noout): Remove clobber.
5292 (commutative_binary_cmp0): New pattern.
5293 (noncommutative_binary_cmp0): Likewise.
5294 (noncommutative_binary_cmp0_noout): Likewise.
5295 (noncommutative_binary_comparison_result_used): Removed.
5296 (rsub_cmp0): New pattern.
5297 (rsub_cmp0_noout): Likewise.
5298 (extzvsi): Changed, keep only meaningful variants.
5299 (SQH, SEZ): New iterators.
5300 (SQH_postfix): New mode attribute.
5301 (SEZ_prefix): New code attribute.
5302 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
5303 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
5304 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
5306 (noncommutative_operator): Check the availability of barrel
5309 2018-04-09 Richard Biener <rguenther@suse.de>
5311 PR tree-optimization/85284
5312 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
5313 Only use the niter constraining form of simple_iv when the exit
5316 2018-04-09 Tom de Vries <tom@codesourcery.com>
5319 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
5320 (define_expand "*memory_barrier"): New define_expand.
5321 (define_insn "memory_barrier"): New insn.
5323 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5325 PR rtl-optimization/80463
5326 PR rtl-optimization/83972
5327 PR rtl-optimization/83480
5329 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
5330 correct producer for the insn.
5331 (tidy_control_flow): Fixup seqnos in case of debug insns.
5333 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5335 PR rtl-optimization/83913
5337 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
5338 different sched-times when merging exprs.
5340 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5342 PR rtl-optimization/83962
5344 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
5345 tidy_fallthru_edge and tidy_control_flow.
5347 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5349 PR rtl-optimization/83530
5351 * sel-sched.c (force_next_insn): New global variable.
5352 (remove_insn_for_debug): When force_next_insn is true, also leave only
5353 next insn in the ready list.
5354 (sel_sched_region): When the region wasn't scheduled, make another pass
5355 over it with force_next_insn set to 1.
5357 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
5359 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
5361 * config/nds32/constants.md (unspec_volatile_element): Add enum values
5362 for interrupt control.
5363 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5364 functions for interrupt control.
5365 * config/nds32/nds32-intrinsic.md: Likewise.
5366 * config/nds32/nds32_intrinsic.h: Likewise.
5367 * config/nds32/nds32.h (nds32_builtins): Likewise.
5369 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
5371 * config/nds32/nds32.c (nds32_init_machine_status,
5372 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
5373 strict_aligned_p field.
5374 (nds32_expand_to_rtl_hook): New function.
5375 (TARGET_EXPAND_TO_RTL_HOOK): Define.
5376 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
5378 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5379 Chung-Ju Wu <jasonwucj@gmail.com>
5381 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
5382 * config/nds32/nds32-n7.md: New file.
5383 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
5384 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
5386 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
5387 * config/nds32/nds32.md (pipeline_model): Add n7.
5388 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
5389 * config/nds32/pipelines.md: Include n7 settings.
5391 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5392 Chung-Ju Wu <jasonwucj@gmail.com>
5394 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
5395 * config/nds32/nds32-e8.md: New file.
5396 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
5397 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
5399 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
5400 * config/nds32/nds32.md (pipeline_model): Add e8.
5401 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
5402 * config/nds32/pipelines.md: Include e8 settings.
5404 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5405 Chung-Ju Wu <jasonwucj@gmail.com>
5407 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
5408 * config/nds32/nds32-n8.md: New file.
5409 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
5410 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
5412 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
5413 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
5414 * config/nds32/nds32.md (pipeline_model): Add n8.
5415 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
5416 * config/nds32/pipelines.md: Include n8 settings.
5418 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5419 Chung-Ju Wu <jasonwucj@gmail.com>
5421 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
5422 * config/nds32/nds32-n9-2r1w.md: New file.
5423 * config/nds32/nds32-n9-3r2w.md: New file.
5424 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
5425 nds32_register_ports): New or modify for cpu n9.
5426 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
5428 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
5429 * config/nds32/nds32-utils.c: New file.
5430 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
5431 TARGET_MUL_SLOW): Define.
5432 * config/nds32/nds32.md (pipeline_model): New attribute.
5433 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
5434 New options that support cpu n9.
5435 * config/nds32/pipelines.md: Include n9 settings.
5436 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
5438 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
5440 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
5441 information if necessary.
5442 (output_cond_branch_compare_zero): Likewise.
5443 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
5444 (nds32_target_alignment): Refine for alignment.
5445 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
5446 (FUNCTION_BOUNDARY): Modify.
5447 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
5449 * config/nds32/nds32.opt (malways-align, malign-functions): New.
5451 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
5453 * config/nds32/constants.md (unspec_volatile_element): Add values for
5454 TLB operation and data prefetch.
5455 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5456 functions for TLB operation and data prefetch.
5457 * config/nds32/nds32-intrinsic.md: Likewise.
5458 * config/nds32/nds32_intrinsic.h: Likewise.
5459 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
5460 (nds32_print_operand): Likewise.
5461 * config/nds32/nds32.h (nds32_builtins): Likewise.
5463 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
5464 Andrew Pinski <pinsika@gcc.gnu.org>
5467 * match.pd: Use constant_boolean_node of correct type instead of
5468 boolean_true_node or boolean_false_node for simplifying
5469 pointer comparisons to zero.
5471 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5473 PR tree-optimization/80021
5474 * tree.c (verify_type_variant): Make error call in verify_variant_match
5475 translatable and remove final full stop.
5477 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5479 * config/nds32/constants.md (unspec_volatile_element): Add
5480 UNSPEC_VOLATILE_EH_RETURN.
5481 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
5482 nds32_output_stack_pop): Support dwarf exception handling process.
5483 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
5484 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
5485 exception handling process.
5486 (nds32_compute_stack_frame): Likewise.
5487 (nds32_return_addr_rtx): Likewise.
5488 (nds32_initial_elimination_offset): Likewise.
5489 (nds32_expand_prologue): Likewise.
5490 (nds32_expand_epilogue): Likewise.
5491 (nds32_dynamic_chain_address): New function.
5492 * config/nds32/nds32.h (machine_function): Add fields for dwarf
5494 (DYNAMIC_CHAIN_ADDRESS): Define.
5495 (EH_RETURN_DATA_REGNO): Define.
5496 (EH_RETURN_STACKADJ_RTX): Define.
5497 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
5498 patterns for dwarf exception handling.
5500 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5502 * config/nds32/nds32.h: Clean up obsolete macros.
5504 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5506 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5507 Add enum values for particular instructions.
5508 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5509 particular intrinsic functions.
5510 * config/nds32/nds32-intrinsic.md: Likewise.
5511 * config/nds32/nds32_intrinsic.h: Likewise.
5512 * config/nds32/nds32.h (nds32_builtins): Likewise.
5513 * config/nds32/nds32.md (type): Add pbsad and pbsada.
5514 (btst, ave): New patterns for particular instructions.
5516 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5518 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5519 Add enum values for atomic load/store and memory sync.
5520 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
5522 * config/nds32/nds32-intrinsic.md: Likewise.
5523 * config/nds32/nds32_intrinsic.h: Likewise.
5524 * config/nds32/nds32.h (nds32_builtins): Likewise.
5526 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5528 PR tree-optimization/85257
5529 * fold-const.c (native_encode_vector): If not all elts could fit
5530 and off is -1, return 0 rather than offset.
5531 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
5532 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
5533 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
5534 adjust buffer in native_interpret_expr call.
5536 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5538 * config/nds32/constants.md (unspec_volatile_element): Add cache
5539 control enum values.
5540 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
5541 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
5542 * config/nds32/nds32.c (nds32_cctl_names): New.
5543 (nds32_print_operand): Handle cache control register names.
5544 * config/nds32/nds32.h (nds32_builtins): New enum values.
5545 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
5547 * config/nds32/nds32.md (type): Add mmu.
5548 * config/nds32/pipelines.md (simple_insn): Add mmu.
5550 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5552 * config/nds32/nds32.md (type): Remove call.
5553 * config/nds32/pipelines.md (simple_insn): Likewise.
5555 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5557 * config/nds32/constants.md (unspec_volatile_element): Add
5558 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
5559 UNSPEC_VOLATILE_FMFCFG.
5560 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
5561 description for fmfcfg and fmfcsr.
5562 (bdesc_1arg): Add fmtcsr.
5563 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
5564 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
5565 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
5566 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
5567 unspec_fmfcfg): New patterns.
5568 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
5569 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
5570 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
5571 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
5572 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
5573 __nds32__fmfcfg): Define.
5575 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5577 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
5578 intrinsic register names.
5579 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
5580 intrinsic register enum values and macros.
5582 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5584 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
5585 for load/store addressing form.
5586 (nds32_print_operand_address): Likewise.
5588 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
5591 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
5592 based on LABEL_REF. Remove useless assertion.
5593 (pic_address_needs_scratch): Fix formatting.
5594 (sparc_legitimize_pic_address): Minor tweaks.
5595 (sparc_delegitimize_address): Adjust assertion accordingly.
5596 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
5597 into symbolic_operand.
5598 (movsi_high_pic_label_ref): Likewise.
5599 (movsi_lo_sum_pic_label_ref): Likewise.
5600 (movdi_pic_label_ref): Likewise.
5601 (movdi_high_pic_label_ref): Likewise.
5602 (movdi_lo_sum_pic_label_ref): Likewise.
5604 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
5606 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
5607 custom LIB_SPEC setup.
5609 2018-04-06 Ruslan Bukin <br@bsdpad.com>
5610 Kito Cheng <kito.cheng@gmail.com>
5612 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
5613 * config/riscv/freebsd.h: New.
5615 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5617 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
5618 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
5621 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5622 Kito Cheng <kito.cheng@gmail.com>
5624 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
5625 nds32_output_call, nds32_symbol_binds_local_p): New functions.
5626 * config/nds32/nds32-protos.h (nds32_output_call,
5627 nds32_output_return): Declare.
5628 * config/nds32/nds32.md: Refine all the call and return patterns.
5630 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5633 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
5634 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
5636 PR rtl-optimization/84872
5637 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
5638 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
5641 2018-04-06 Tamar Christina <tamar.christina@arm.com>
5643 * expr.c (copy_blkmode_to_reg): Revert 254862.
5644 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
5646 2018-04-06 Richard Biener <rguenther@suse.de>
5649 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
5650 after seeing a component reference with an adjacent field. Treat
5651 refs to arrays at struct end of external decls similar to
5652 refs to unconstrained commons.
5654 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5657 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
5658 look through SAVE_EXPRs with non-side-effects argument. Adjust
5660 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
5663 2018-04-06 Richard Biener <rguenther@suse.de>
5666 * alias.c (find_base_term): New wrapper around find_base_term
5667 unwinding CSELIB_VAL_PTR changes.
5668 (find_base_term): Do not restore CSELIB_VAL_PTR during the
5671 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5673 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
5675 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
5676 constant definitions.
5677 ("nop"): lr 0,0 -> nopr r0
5678 ("nop_lr0", "nop_lr1"): New insn definitions.
5680 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5682 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
5683 NDS32_V3PUSH_AVAILABLE_P macro.
5685 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
5686 Chung-Ju Wu <jasonwucj@gmail.com>
5688 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
5689 (nds32*-*-*): Add float and fpu_config into supported_defaults.
5690 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
5691 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
5692 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
5693 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
5694 * config/nds32/constraints.md: New constraints and checking for hard
5695 float configuration.
5696 * config/nds32/iterators.md: New mode iterator and attribute for hard
5697 float configuration.
5698 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
5700 * config/nds32/nds32-fpu.md: New file.
5701 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
5702 deal with hard float code generation.
5703 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
5705 (abi_type, float_reg_number): New enum type.
5706 * config/nds32/nds32-predicates.c: New predicates for hard float.
5707 * config/nds32/nds32-protos.h: Declare functions for hard float.
5708 * config/nds32/nds32.c: Implementation for hard float configuration.
5709 * config/nds32/nds32.h: Definitions for hard float configuration.
5710 * config/nds32/nds32.md: Include hard float machine description and
5711 modify patterns for hard float configuration.
5712 * config/nds32/nds32.opt: New options for hard float configuration.
5713 * config/nds32/predicates.md: New predicates for hard float
5716 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5718 * common/config/nds32/nds32-common.c
5719 (nds32_option_optimization_table): Enable -mreleax-hint by default.
5721 2018-04-05 Jakub Jelinek <jakub@redhat.com>
5724 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
5725 CONSTRUCTOR_ELT (ctor, ...)->value.
5727 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
5730 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
5732 2018-04-05 Tom de Vries <tom@codesourcery.com>
5735 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
5738 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5739 Kito Cheng <kito.cheng@gmail.com>
5741 * config/nds32/constraints.md (U33): Fine-tune checking condition.
5742 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
5743 * config/nds32/nds32.h (nds32_16bit_address_type): Add
5744 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
5746 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5747 Kito Cheng <kito.cheng@gmail.com>
5749 * config/nds32/constraints.md (Ufe): New memory constraint.
5750 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
5751 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
5752 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
5754 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
5755 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
5757 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5759 * config/nds32/nds32.md: Use optimize_size in the condition for
5760 alu-shift instructions.
5762 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5764 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
5766 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5768 * config/nds32/nds32.md (negsi2): Refine pattern.
5770 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5771 Chung-Ju Wu <jasonwucj@gmail.com>
5773 * config/nds32/iterators.md (shift_rotate): New code iterator.
5774 (shift): New code attribute.
5775 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
5776 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
5777 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
5778 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
5779 bit-wise operations.
5780 (andsi3, *andsi3): Ditto.
5781 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
5782 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
5783 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
5784 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
5785 nds32_ior_operand, nds32_xor_operand): New predicates.
5787 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5789 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
5790 (addsi3, subsi3): ... this.
5792 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5794 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
5796 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5798 * config/nds32/nds32.md: Adjust indention.
5800 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5802 * config/nds32/nds32.md (feature): New attribute.
5804 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5806 * config/nds32/nds32.md (subtype): New attribute.
5808 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
5811 * config/arm/arm-builtins.c (arm_expand_builtin): Change
5812 expansion to perform a bitwise AND of the argument followed by a
5813 boolean negation of the result.
5815 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
5817 PR rtl-optimization/84878
5818 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
5819 the basic block. Assert the use reference is not artificial and that
5820 it has an associated insn.
5822 2018-04-04 Michael Matz <matz@suse.de>
5824 * builtins.c (compute_objsize): Pass correct operand
5825 to array_at_struct_end_p.
5827 2018-04-04 Richard Biener <rguenther@suse.de>
5830 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
5831 from contexts for DINFO_LEVEL_TERSE and below.
5833 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5835 * config/nds32/nds32-doubleword.md (move_<mode>): Require
5836 resiter_operand condition.
5837 * config/nds32/nds32.md (*move<mode>): Ditto.
5839 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5840 Monk Chiang <sh.chiang04@gmail.com>
5842 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
5844 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5846 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
5848 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5849 Kito Cheng <kito.cheng@gmail.com>
5851 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
5852 nds32_cond_code_str, output_cond_branch,
5853 output_cond_branch_compare_zero, nds32_expand_cbranch,
5854 nds32_expand_cstore, nds32_expand_movcc,
5855 nds32_output_cbranchsi4_equality_zero,
5856 nds32_output_cbranchsi4_equality_reg,
5857 nds32_output_cbranchsi4_equality_reg_or_const_int,
5858 nds32_output_cbranchsi4_greater_less_zero: New functions.
5859 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
5860 nds32_expand_cstore, nds32_expand_movcc,
5861 nds32_output_cbranchsi4_equality_zero,
5862 nds32_output_cbranchsi4_equality_reg,
5863 nds32_output_cbranchsi4_equality_reg_or_const_int,
5864 nds32_output_cbranchsi4_greater_less_zero): Declare.
5865 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
5866 nds32_rimm11s_operand): New predicates.
5867 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
5868 * config/nds32/nds32.md: Rewrite all the branch and conditional move
5871 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5873 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
5874 * config/nds32/nds32.md: Ditto.
5875 * config/nds32/pipelines.md: Ditto.
5877 2018-04-04 Richard Biener <rguenther@suse.de>
5879 PR tree-optimization/85168
5880 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
5881 propagating abnormals.
5883 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5885 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
5887 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5888 Kito Cheng <kito.cheng@gmail.com>
5890 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
5891 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
5892 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
5893 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
5894 * config/nds32/nds32.md (sibcall_internal): New.
5895 (sibcall_register): Remove.
5896 (sibcall_immediate): Remove.
5897 (sibcall_value_internal): New.
5898 (sibcall_value_register): Remove.
5899 (sibcall_value_immediate): Remove.
5900 * config/nds32/predicates.md (nds32_general_register_operand): New.
5901 (nds32_call_address_operand): New.
5903 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5905 PR rtl-optimization/85167
5906 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
5907 bb_defs if *split_p, instead preinitialize it to NULL.
5909 PR tree-optimization/85156
5910 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
5911 evaluating the argument multiple times.
5913 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
5915 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
5917 (_mm_cvtpd_ps): Likewise.
5918 (_mm_cvttpd_epi32): Likewise.
5919 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
5920 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
5921 vector, pixel, and bool following altivec.h include.
5923 2018-04-03 Martin Sebor <msebor@redhat.com>
5925 * doc/extend.texi (Common Function Attributes): Clarify.
5926 (const attribute): Likewise.
5927 (pure attribute): Likewise.
5929 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5932 * config/i386/i386.c (ix86_expand_vector_set): Use
5933 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
5935 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
5937 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
5938 instructions when changing rounding bits to preserve precision bits
5939 in the x87 control word.
5941 2018-04-03 Martin Liska <mliska@suse.cz>
5943 PR tree-optimization/82491
5944 * rtl.h (strip_offset_and_add): Replace += suboffset with
5945 poly_uint64 () + suboffset.
5947 2018-03-29 Martin Liska <mliska@suse.cz>
5948 Martin Jambor <mjambor@suse.cz>
5951 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
5952 param_type is not an integral or pointer type.
5954 2018-04-03 Richard Biener <rguenther@suse.de>
5956 * sese.h (recompute_all_dominators): Remove.
5958 2018-04-02 Martin Sebor <msebor@redhat.com>
5960 * doc/invoke.texi (-Wrestrict): Fix typos.
5962 2018-04-02 Jim Wilson <jimw@sifive.com>
5964 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
5965 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
5966 (<optab>di3, <optab>si3_extend): Likewise.
5967 (<optab>si3_mask, <optab>si3_mask_1): New.
5968 (<optab>di3_mask, <optab>di3_mask_1): New.
5969 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
5970 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
5971 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
5973 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
5975 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
5978 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
5980 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
5981 (nds32_canonicalize_comparison): New function.
5983 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5984 Kito Cheng <kito.cheng@gmail.com>
5985 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5987 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
5988 * config/nds32/constants.md (unspec_volatile_element): Add
5989 UNSPEC_VOLATILE_RELAX_GROUP.
5990 * config/nds32/nds32-relax-opt.c: New file.
5991 * config/nds32/nds32-predicates.c
5992 (nds32_symbol_load_store_p): New function.
5993 * config/nds32/nds32-protos.h
5994 (nds32_symbol_load_store_p): Declare function.
5995 (make_pass_nds32_relax_opt): Declare new rtl pass function.
5996 * config/nds32/nds32.c
5997 (nds32_register_pass): New function to register pass.
5998 (nds32_register_passes): New function to register passes.
5999 * config/nds32/nds32.md (relax_group): New pattern.
6000 * config/nds32/nds32.opt (mrelax-hint): New option.
6001 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
6003 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
6005 * config/nds32/t-nds32: Modify files dependency.
6007 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
6009 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
6010 (PROFILE_HOOK): Define its implementation.
6012 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
6014 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
6015 type and 32-bit size.
6017 2018-04-01 Jakub Jelinek <jakub@redhat.com>
6020 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
6021 (V_128_256): New mode iterator.
6022 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
6023 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
6024 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
6026 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
6027 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
6029 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
6032 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
6033 NaN inputs correctly.
6035 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
6038 * config/rs6000/vsx.md (??r): New mode attribute.
6039 (*vsx_mov<mode>_64bit): Use it.
6040 (*vsx_mov<mode>_32bit): Likewise.
6042 2018-03-30 Martin Sebor <msebor@redhat.com>
6044 PR tree-optimization/84818
6045 * builtins.c (check_access): Use warning_n.
6047 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6050 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
6052 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
6055 2018-03-30 Julia Koval <julia.koval@intel.com>
6058 * x86-tune.def (movx, partial_reg_dependency): Enable for
6061 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
6064 * lra-constraints.c (process_alt_operands): Move setting
6065 this_alternative_matches below.
6067 2018-03-29 Martin Liska <mliska@suse.cz>
6070 * doc/invoke.texi: Document how LTO works with debug info.
6071 Describe auto-load support of binutils. Mention 'x86-64'
6072 as valid option value of -march option.
6074 2018-03-29 Jakub Jelinek <jakub@redhat.com>
6076 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
6079 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
6080 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
6081 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
6084 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
6087 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
6088 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
6089 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
6090 for RS6000_BTM_POWERPC64.
6091 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
6092 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
6093 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
6098 2018-03-28 Carl Love <cel@us.ibm.com>
6101 2017-09-27 Carl Love <cel@us.ibm.com>
6103 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
6104 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
6105 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
6108 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6110 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
6111 instead of __vector bool.
6112 (_mm_max_pu8): Likewise.
6113 (_mm_min_pi16): Likewise.
6115 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
6118 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
6119 (DIVWEUO): Likewise.
6121 (DIVDEUO): Likewise.
6122 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
6123 DIVWEUO and DIVDEUO.
6124 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
6125 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
6126 (div_extend): Likewise.
6127 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
6129 (__builtin_divweuo): Likewise.
6130 (__builtin_divdeo): Likewise.
6131 (__builtin_divdeuo): Likewise.
6133 2018-03-28 Jakub Jelinek <jakub@redhat.com>
6136 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
6137 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
6139 PR tree-optimization/82004
6140 * gimple-match-head.c (optimize_pow_to_exp): New function.
6141 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
6142 Don't fold to exp if optimize_pow_to_exp is false.
6144 2018-03-28 Martin Liska <mliska@suse.cz>
6147 * calls.c (initialize_argument_information): Fix trailing space.
6148 * common.opt: Fix typo and provide better explanation for
6149 -fsanitize-coverage option.
6150 * config/i386/i386.opt: Fix typo.
6152 2018-03-28 Jakub Jelinek <jakub@redhat.com>
6153 Martin Liska <mliska@suse.cz>
6156 * gimplify.c (asan_poison_variable): Don't do the check for
6157 gimplify_omp_ctxp here.
6158 (gimplify_decl_expr): Do it here.
6159 (gimplify_target_expr): Likewise.
6161 2018-03-28 Martin Liska <mliska@suse.cz>
6164 * config/i386/i386.c (ix86_function_arg_advance): Do not call
6165 chkp_type_bounds_count if MPX is not enabled.
6167 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
6169 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
6171 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
6174 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
6175 function to create the function decl for complex long double
6176 multiply and divide for -mabi=ieeelongdouble.
6177 (init_float128_ieee): Call it.
6179 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
6182 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
6183 -fcf-protection=branch -mibt.
6184 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
6186 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6189 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
6191 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
6194 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
6195 extern array declarations.
6197 2018-03-27 Richard Biener <rguenther@suse.de>
6200 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
6201 explicit single_use checks.
6203 2018-03-27 Richard Biener <rguenther@suse.de>
6205 PR tree-optimization/85082
6206 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
6209 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6211 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
6212 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
6213 Turn on fasynchronous-unwind-tables and funwind-tables.
6215 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
6218 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
6219 (*bmi_blsr_<mode>_ccz): Ditto.
6221 2018-03-26 Tom de Vries <tom@codesourcery.com>
6223 PR tree-optimization/85063
6224 * omp-general.c (offloading_function_p): New function. Factor out
6226 * omp-offload.c (pass_omp_target_link::gate): ... here.
6227 * omp-general.h (offloading_function_p): Declare.
6228 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
6229 with attribute omp declare target for offloading functions.
6231 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
6233 PR tree-optimization/84005
6234 * tree-data-ref.h (get_base_for_alignment): Declare.
6235 * tree-data-ref.c (get_base_for_alignment_1): New function.
6236 (get_base_for_alignment): Likewise.
6237 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
6238 get_base_for_alignment to find a suitable base object, instead
6239 of always using drb->base_address.
6241 2018-03-23 Jakub Jelinek <jakub@redhat.com>
6244 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
6245 known size by default.
6247 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
6250 * lra-constraints.c (process_alt_operands): Don't match BLKmode
6251 and non BLKmode operands.
6253 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6256 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
6257 Clean up attributes.
6259 2018-03-23 Richard Biener <rguenther@suse.de>
6262 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
6263 we are going to emit early debug for LTO.
6265 2018-03-23 Jakub Jelinek <jakub@redhat.com>
6268 * function.c (match_asm_constraints_1): Don't optimize if input
6269 doesn't satisfy general_operand predicate for output's mode.
6272 * alias.c (write_dependence_p): Don't require for x_canonicalized
6273 non-VOIDmode if x has VOIDmode.
6276 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
6277 just don't try to optimize it rather than assert it never happens.
6279 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
6281 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
6282 macro expansions for definition of ST_INTERNAL_<mode> and
6283 LD_INTERNAL_<mode> builtins.
6284 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
6286 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
6288 (altivec_expand_st_builtin): Likewise.
6289 (altivec_expand_builtin): Remove calls to deleted functions.
6290 (rs6000_address_for_altivec): Delete this function.
6291 * config/rs6000/vector.md: Remove expands for
6292 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
6294 2018-03-22 Sudakshina Das <sudi.das@arm.com>
6297 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
6298 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
6299 re-computing once computed.
6300 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
6301 (arm_init_machine_status): Initialize
6302 machine->static_chain_stack_bytes.
6304 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
6307 * doc/extend.texi: Add four new prototypes for vec_ld.
6308 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
6309 definitions for more logical presentation.
6310 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
6311 entries for V1TI variants of __builtin_altivec_ld builtin.
6312 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
6313 handling of V1TI variant of LVX icode pattern.
6314 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
6315 (rs6000_gimple_fold_builtin): Likewise.
6316 (altivec_init_builtins): Add code to define
6317 __builtin_altivec_lvx_v1ti function.
6319 2018-03-22 Jakub Jelinek <jakub@redhat.com>
6322 * function.c (match_asm_constraints_1): Don't do the optimization
6323 if input isn't a REG, SUBREG, MEM or constant.
6325 2018-03-22 Tom de Vries <tom@codesourcery.com>
6327 PR tree-optimization/84956
6328 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
6329 bb_has_abnormal_pred.
6331 2018-03-22 Jakub Jelinek <jakub@redhat.com>
6334 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
6335 DECL_INITIAL (decl) to decl at the end.
6336 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
6339 2018-03-21 Joseph Myers <joseph@codesourcery.com>
6341 * doc/extend.texi (__builtin_tgmath): Document when complex
6342 integer types are treated as _Complex _Float64.
6344 2018-03-21 Tom de Vries <tom@codesourcery.com>
6346 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
6348 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6350 PR tree-optimization/84960
6351 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
6352 if it is ENTRY block, move them into single succ of ENTRY in that case.
6354 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
6356 PR tree-optimization/84811
6357 * poly-int.h (poly_span_traits): Remove the T3 parameter and
6358 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
6359 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
6360 (known_subrange_p): Update accordingly. Cast each value involved
6361 in the size comparison, rather than casting the result of the
6364 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6366 PR tree-optimization/84982
6367 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
6368 by flipping the least significant bit rather than all bits from
6369 bitpos to bitpos + bitsize - 1.
6371 2018-03-21 Nathan Sidwell <nathan@acm.org>
6373 * doc/extend.texi (Deprecated Features): Remove mention of
6374 long-deleted deprecations.
6376 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6379 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
6380 * configure: Regenerate.
6382 2018-03-21 Tom de Vries <tom@codesourcery.com>
6384 PR tree-optimization/83126
6385 * tree-parloops.c (num_phis): New function.
6386 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
6388 2018-03-21 Nathan Sidwell <nathan@acm.org>
6390 * doc/extend.texi (Deprecated Features): Update deprecated flags,
6391 mention anon-struct/union members and trailing attributes.
6393 2018-03-21 Bin Cheng <bin.cheng@arm.com>
6395 PR tree-optimization/84969
6396 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
6397 builtin memset partitions if they set different rhs values.
6399 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6401 PR rtl-optimization/84989
6402 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
6403 VEC_DUPLICATE with scalar result mode.
6405 2018-03-21 Martin Liska <mliska@suse.cz>
6408 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
6409 not intended return statement.
6411 2018-03-21 Martin Liska <mliska@suse.cz>
6414 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
6415 (chkp_find_bound_slots_1): Limit number of iterations.
6417 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
6420 * Minor grammar fixes for x86 options.
6422 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6425 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
6426 holding REG_CFA_RESTORE notes, instead turn them into a USE.
6428 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
6431 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
6432 (altivec_lvx_<mode>_1op): Likewise.
6433 (altivec_stvx_<mode>_2op): Likewise.
6434 (altivec_stvx_<mode>_1op): Likewise.
6435 (altivec_lvx_<VM2:mode>): New define_expand.
6436 (altivec_stvx_<VM2:mode>): Likewise.
6437 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
6438 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6439 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
6440 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6441 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
6442 (rs6000_gen_lvx): Likewise.
6443 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
6444 (altivec_expand_stv_builtin): Likewise.
6445 (altivec_expand_builtin): Likewise.
6446 * config/rs6000/vector.md: Likewise.
6448 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6451 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
6454 2018-03-20 Richard Biener <rguenther@suse.de>
6457 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
6458 sign-conversions as zero, fall back to standard scalar_stmt
6461 2018-03-20 Martin Liska <mliska@suse.cz>
6464 * predict.c (rebuild_frequencies): Handle case when we have
6465 PROFILE_ABSENT, but flag_guess_branch_prob is false.
6467 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6470 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
6471 flag_section_anchors.
6472 * varasm.c (use_blocks_for_decl_p): Remove hack for
6473 dw2_force_const_mem.
6476 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
6478 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
6479 be created, use lowpart_subreg of operands[0] rather than operands[0]
6481 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
6482 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
6483 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
6484 and n constraint instead of aarch64_shift_imm_di and Usd.
6485 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
6486 (*aarch64_<optab>_reg_minus<mode>3): ... this.
6488 2018-03-20 Sudakshina Das <sudi.das@arm.com>
6491 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
6492 to favor GPR over NEON registers.
6493 (<shift>di3_neon): Likewise.
6495 2018-03-20 Tom de Vries <tom@codesourcery.com>
6498 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
6499 (nvptx_process_pars): Emit bar.sync asap and alap.
6501 2018-03-20 Tom de Vries <tom@codesourcery.com>
6504 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
6505 seen_label if seen_label is already set.
6507 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6510 * config/i386/i386.c (fold_builtin_cpu): For features above 31
6511 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
6512 Use 1U instead of 1. Formatting fixes.
6515 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
6516 instead of TREE_TYPE (s1) for the return value.
6518 2018-03-19 Jakub Jelinek <jakub@redhat.com>
6520 PR tree-optimization/84946
6521 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
6522 bitsize + bitsize in poly_uint64 rather than poly_int64.
6525 * dwarf2asm.c: Include fold-const.c.
6526 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
6527 of decl rather than decl itself.
6529 PR rtl-optimization/84643
6530 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
6532 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
6535 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
6536 calling assemble_variable.
6538 2018-03-19 Sudakshina Das <sudi.das@arm.com>
6541 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
6542 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
6544 2018-03-19 Jim Wilson <jimw@sifive.com>
6547 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
6548 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
6549 (riscv_first_stack_step): Likewise.
6550 (riscv_option_override): Use STACK_BOUNDARY instead of
6552 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
6554 (BIGGEST_ALIGNMENT): Set to 128.
6555 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
6556 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
6559 2018-03-19 Richard Biener <rguenther@suse.de>
6561 PR tree-optimization/84933
6562 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
6563 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
6565 2018-03-19 Richard Biener <rguenther@suse.de>
6567 PR tree-optimization/84859
6568 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
6569 (cond_if_else_store_replacement): Perform sinking operation on
6570 single-store BBs regardless of MAX_STORES_TO_SINK setting.
6571 Generalize what a BB with a single eligible store is.
6573 2018-03-19 Richard Biener <rguenther@suse.de>
6575 PR tree-optimization/84929
6576 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
6577 chrec_is_positive against non-chrec arg.
6579 2018-03-19 Tamar Christina <tamar.christina@arm.com>
6582 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
6584 2018-03-18 Martin Liska <mliska@suse.cz>
6586 PR rtl-optimization/84635
6587 * regrename.c (build_def_use): Use matches_mode only when
6590 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
6592 PR tree-optimization/84913
6593 * tree-vect-loop.c (vectorizable_reduction): Don't try to
6594 vectorize chains of COND_EXPRs.
6596 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6598 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
6600 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6602 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
6604 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6606 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
6608 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6609 Kito Cheng <kito.cheng@gmail.com>
6611 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
6612 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
6613 (nds32_adjust_reg_alloc_order): New function.
6614 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
6616 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
6618 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
6619 nds32_print_operand, nds32_print_operand_address): Use
6620 HOST_WIDE_INT_PRINT_DEC instead.
6622 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6624 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
6626 2018-03-17 Jakub Jelinek <jakub@redhat.com>
6629 * config/i386/i386.c (initial_ix86_tune_features,
6630 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
6632 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
6633 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
6634 rather than 1u << ix86_tune. Formatting fix.
6635 (ix86_option_override_internal): Change ix86_arch_mask from
6636 unsigned int to unsigned HOST_WIDE_INT, initialize to
6637 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
6638 (ix86_function_specific_restore): Likewise.
6640 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6643 * postreload.c (reload_combine_recognize_pattern): Perform
6644 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
6645 truncate_int_for_mode the result for the destination's mode.
6648 * hsa-gen.c (mem_type_for_type): Fix comment typo.
6649 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
6651 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6654 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
6657 * lra-assigns.c (lra_split_hard_reg_for): Don't use
6658 regno_allocno_class_array and sorted_pseudos.
6659 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
6660 insns where regno is used.
6662 2018-03-16 Martin Liska <mliska@suse.cz>
6665 * multiple_target.c (create_dispatcher_calls): Redirect
6666 reference in the symbol table.
6668 2018-03-16 Martin Liska <mliska@suse.cz>
6671 * multiple_target.c (create_dispatcher_calls): Redirect also
6674 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6678 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
6679 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
6680 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
6682 2018-03-16 Julia Koval <julia.koval@intel.com>
6684 * doc/invoke.texi (Skylake Server): Add CLWB.
6685 Cannonlake): Remove CLWB.
6687 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6689 PR tree-optimization/84841
6690 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
6692 (FLOAT_ONE_CONST_TYPE): Define.
6693 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
6694 (sort_by_operand_rank): Put entries with higher constant_type last
6695 rather than first to match comments.
6697 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
6699 * config/nios2/nios2.md (movsi_internal): Fix thinko in
6702 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6705 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
6706 check and use address of target always.
6708 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
6711 * config/i386/i386.c (indirect_thunk_needed): Update comments.
6712 (indirect_thunk_bnd_needed): Likewise.
6713 (indirect_thunks_used): Likewise.
6714 (indirect_thunks_bnd_used): Likewise.
6715 (indirect_return_needed): New.
6716 (indirect_return_bnd_needed): Likewise.
6717 (output_indirect_thunk_function): Add a bool argument for
6719 (output_indirect_thunk_function): Don't generate alias for
6720 function return thunk.
6721 (ix86_code_end): Call output_indirect_thunk_function to generate
6722 function return thunks.
6723 (ix86_output_function_return): Set indirect_return_bnd_needed
6724 and indirect_return_needed instead of indirect_thunk_bnd_needed
6725 and indirect_thunk_needed.
6727 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
6729 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
6730 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
6731 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
6733 2018-03-15 David Malcolm <dmalcolm@redhat.com>
6734 Paul Hua <paul.hua.gm@gmail.com>
6737 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
6739 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
6741 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
6742 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
6745 2018-03-15 Tamar Christina <tamar.christina@arm.com>
6748 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
6749 instead of GET_MODE_SIZE when comparing Units.
6751 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6754 * varasm.c (hash_section): Return an unchangeble hash value
6755 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
6756 Return !aarch64_can_use_per_function_literal_pools_p ().
6758 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6761 * optabs.c (emit_conditional_move): Pass address of cmode's copy
6762 rather than address of cmode as last argument to prepare_cmp_insn.
6764 2018-03-15 Julia Koval <julia.koval@intel.com>
6766 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
6767 F_AVX512VNNI, F_AVX512BITALG): New.
6769 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
6772 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
6773 insn for floating-point loads and stores.
6775 2018-03-14 Carl Love <cel@us.ibm.com>
6777 * config/rs6000/rs6000-c.c: Add macro definitions for
6778 ALTIVEC_BUILTIN_VEC_PERMXOR.
6779 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
6780 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
6781 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
6782 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
6784 * config/doc/extend.texi: Add prototypes for vec_permxor.
6786 2018-03-14 David Malcolm <dmalcolm@redhat.com>
6789 * diagnostic-show-locus.c (class layout_point): Convert m_line
6790 from int to linenum_type.
6791 (line_span::comparator): Use linenum "compare" function when
6792 comparing line numbers.
6793 (test_line_span): New function.
6794 (layout_range::contains_point): Convert param "row" from int to
6796 (layout_range::intersects_line_p): Likewise.
6797 (layout::will_show_line_p): Likewise.
6798 (layout::print_source_line): Likewise.
6799 (layout::should_print_annotation_line_p): Likewise.
6800 (layout::print_annotation_line): Likewise.
6801 (layout::print_leading_fixits): Likewise.
6802 (layout::annotation_line_showed_range_p): Likewise.
6803 (struct line_corrections): Likewise for field m_row.
6804 (line_corrections::line_corrections): Likewise for param "row".
6805 (layout::print_trailing_fixits): Likewise.
6806 (layout::get_state_at_point): Likewise.
6807 (layout::get_x_bound_for_row): Likewise.
6808 (layout::print_line): Likewise.
6809 (diagnostic_show_locus): Likewise for locals "last_line" and
6811 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
6812 * input.c (selftest::test_linenum_comparisons): New function.
6813 (selftest::input_c_tests): Call it.
6814 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
6815 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
6816 * selftest.h (ASSERT_GT): New macro.
6817 (ASSERT_GT_AT): New macro.
6818 (ASSERT_LT): New macro.
6819 (ASSERT_LT_AT): New macro.
6821 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
6823 PR rtl-optimization/84780
6824 * combine.c (distribute_links): Don't make a link based on pc_rtx.
6826 2018-03-14 Martin Liska <mliska@suse.cz>
6828 * tree.c (record_node_allocation_statistics): Use
6829 get_stats_node_kind.
6830 (get_stats_node_kind): New function extracted from
6831 record_node_allocation_statistics.
6832 (free_node): Use get_stats_node_kind.
6834 2018-03-14 Richard Biener <rguenther@suse.de>
6836 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
6837 that the value-set of ANTIC_IN doesn't grow.
6840 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6842 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6843 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6845 2018-03-14 Julia Koval <julia.koval@intel.com>
6847 * config.gcc (icelake-client, icelake-server): New.
6849 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
6850 (initial_ix86_arch_features): Ditto.
6851 (PTA_SKYLAKE): Add SGX.
6852 (PTA_ICELAKE): Remove.
6853 (PTA_ICELAKE_CLIENT): New.
6854 (PTA_ICELAKE_SERVER): New.
6855 (ix86_option_override_internal): Split up icelake on icelake client and
6857 (get_builtin_code_for_version): Ditto.
6858 (fold_builtin_cpu): Ditto.
6859 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
6860 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
6861 * config/i386/i386.h (processor_type): Ditto.
6862 * doc/invoke.texi: Ditto.
6864 2018-03-14 Jakub Jelinek <jakub@redhat.com>
6867 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
6868 INTEGER_CST offset, add it together with bitpos / 8 and
6869 sign extend based on POINTER_SIZE.
6873 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
6876 * config/i386/constraints.md (Yc): New register constraint.
6877 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
6878 Use Yc constraint for alternative 2 of operand 0. Remove
6879 preferred_for_speed attribute.
6881 2018-03-14 Richard Biener <rguenther@suse.de>
6883 PR tree-optimization/84830
6884 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
6885 with the old one to avoid oscillations.
6887 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
6890 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
6892 (assign_by_spills): Return a flag of reload assignment failure.
6893 Do not process the reload assignment failures. Do not spill other
6894 reload pseudos if they has the same reg class. Update n if
6896 (lra_assign): Add a return arg. Set up from the result of
6897 assign_by_spills call.
6898 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6899 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6900 usage_insns if it is not NULL.
6901 (spill_hard_reg_in_range): New function.
6902 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6903 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6904 function prototypes.
6905 (lra_assign): Change prototype.
6906 * lra.c (lra): Add code to deal with fails by splitting hard reg
6909 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
6911 * config/riscv/riscv.opt (mrelax): New option.
6912 * config/riscv/riscv.c (riscv_file_start): Emit ".option
6913 "norelax" when riscv_mrelax is disabled.
6914 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
6916 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6919 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
6920 reassociation for int modes.
6922 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6924 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
6925 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
6927 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
6928 * config/aarch64/aarch64-sve.md
6929 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
6930 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
6931 (*extend<mode><Vwide>2): Rename to...
6932 (aarch64_sve_extend<mode><Vwide>2): ...this.
6933 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
6934 renaming the old pattern to...
6935 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
6937 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
6938 define_expand, renaming the old pattern to...
6939 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
6940 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
6941 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
6942 account when deciding which SVE instruction the optab should use.
6943 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
6945 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6947 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
6948 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
6949 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
6950 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
6951 (tlsdesc_small_<mode>): Turn a define_expand and use
6952 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
6953 (tlsdesc_small_advsimd_<mode>): ...this.
6954 (tlsdesc_small_sve_<mode>): New pattern.
6956 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6958 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
6959 (UNSPEC_UMUL_HIGHPART): New constants.
6960 (MUL_HIGHPART): New int iteraor.
6961 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
6962 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
6964 (*<su>mul<mode>3_highpart): New define_insn.
6966 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6969 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6972 2018-03-13 Martin Liska <mliska@suse.cz>
6975 * (sem_item_optimizer::sem_item_optimizer): Initialize new
6977 (sem_item_optimizer::~sem_item_optimizer): Release it.
6978 (sem_item_optimizer::merge_classes): Register variable aliases.
6979 (sem_item_optimizer::fixup_pt_set): New function.
6980 (sem_item_optimizer::fixup_points_to_sets): Likewise.
6981 * ipa-icf.h: Declare new variables and functions.
6983 2018-03-13 Jakub Jelinek <jakub@redhat.com>
6986 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
6987 integer_pow2p@2 and test integer_pow2p in condition.
6988 (A < 0 ? C : 0): Similarly for @1.
6991 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
6992 characters starting at p contain '\0' character, don't look beyond
6996 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
6997 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
7000 * reg-stack.c (change_stack): Change update_end var from int to
7001 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
7002 also call set_block_for_insn on the newly added insns and rescan.
7005 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
7006 on the last operand.
7009 * tree.c (stabilize_reference_1): Return save_expr (e) for
7010 STATEMENT_LIST even if it doesn't have side-effects.
7012 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
7014 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
7016 2018-03-12 Renlin Li <renlin.li@arm.com>
7018 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
7019 aarch64_output_scalar_simd_mov_immediate.
7021 2018-03-12 Martin Sebor <msebor@redhat.com>
7023 PR tree-optimization/83456
7024 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
7025 for perfectly overlapping calls to memcpy.
7026 (gimple_fold_builtin_memory_chk): Same.
7027 (gimple_fold_builtin_strcpy): Handle no-warning.
7028 (gimple_fold_builtin_stxcpy_chk): Same.
7029 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
7031 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
7033 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
7034 parameter. Use it for SFmode.
7035 (rs6000_function_arg_advance_1): Adjust.
7036 (rs6000_function_arg): Adjust.
7037 (rs6000_gimplify_va_arg): Pass false for that new parameter.
7039 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
7041 PR rtl-optimization/84169
7042 PR rtl-optimization/84780
7043 * combine.c (can_combine_p): Check for a 2-insn combination whether
7044 the destination register is used between the two insns, too.
7046 2018-03-12 Richard Biener <rguenther@suse.de>
7048 PR tree-optimization/84803
7049 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
7050 for refs DR analysis didn't process.
7052 2018-03-12 Richard Biener <rguenther@suse.de>
7054 PR tree-optimization/84777
7055 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
7056 force-vectorize loops ignore whether we are optimizing for size.
7058 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
7060 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
7061 (TARGET_MD_ASM_ADJUST): Define.
7063 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
7064 Kito Cheng <kito.cheng@gmail.com>
7065 Chung-Ju Wu <jasonwucj@gmail.com>
7067 * config/nds32/nds32.c (nds32_compute_stack_frame,
7068 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
7069 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
7070 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
7071 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
7072 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
7073 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
7074 * config/nds32/nds32.md (prologue, epilogue): Use macro
7075 NDS32_V3PUSH_AVAILABLE_P to do checking.
7077 2018-03-11 Jakub Jelinek <jakub@redhat.com>
7080 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
7081 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
7082 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
7083 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
7084 addition of most attributes on !orig_type_die or the attribute not
7085 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
7087 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
7088 Chung-Ju Wu <jasonwucj@gmail.com>
7090 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
7092 * config/nds32/nds32.opt (mvh): New option.
7094 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
7095 Chung-Ju Wu <jasonwucj@gmail.com>
7097 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
7099 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
7100 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
7103 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
7104 Chung-Ju Wu <jasonwucj@gmail.com>
7106 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
7108 * config/nds32/nds32-multiple.md (strlensi): New pattern.
7109 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
7111 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
7112 Kito Cheng <kito.cheng@gmail.com>
7113 Chung-Ju Wu <jasonwucj@gmail.com>
7115 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
7116 UNSPEC_FFMISM and UNSPEC_FLMISM.
7117 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
7118 for ffb, ffmism and flmism.
7119 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
7120 (unspec_ffmism): Ditto.
7121 (unspec_flmism): Ditto.
7122 (nds32_expand_builtin_impl): Check if string extension is available.
7123 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
7124 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
7126 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
7129 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
7132 * lra-assigns.c (assign_by_spills): Return a flag of reload
7133 assignment failure. Do not process the reload assignment
7134 failures. Do not spill other reload pseudos if they has the same
7136 (lra_assign): Add a return arg. Set up from the result of
7137 assign_by_spills call.
7138 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
7139 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
7140 usage_insns if it is not NULL.
7141 (spill_hard_reg_in_range): New function.
7142 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
7143 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
7144 function prototypes.
7145 (lra_assign): Change prototype.
7146 * lra.c (lra): Add code to deal with fails by splitting hard reg
7149 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
7152 * config/i386/i386.opt: Replace Enforcment with Enforcement.
7154 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
7157 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
7158 (dw_val_node): Add val_symbolic_view.
7159 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
7160 (symview_upper_bound): New.
7161 (new_line_info_table): Initialize symviews_since_reset.
7162 (dwarf2out_source_line): Count symviews_since_reset and set
7163 symview_upper_bound.
7164 (dw_val_equal_p): Handle symview.
7165 (add_AT_symview): New.
7166 (print_dw_val): Handle symview.
7167 (attr_checksum, attr_checksum_ordered): Likewise.
7168 (same_dw_val_p, size_of_die): Likewise.
7169 (value_format, output_die): Likewise.
7170 (add_high_low_attributes): Use add_AT_symview for entry_view.
7171 (dwarf2out_finish): Reset symview_upper_bound, clear
7174 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
7177 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
7178 Add strict argument and use it.
7179 (rs6000_split_multireg_move): Update for new strict argument.
7180 (mem_operand_gpr): Disallow all non-offsettable addresses.
7181 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
7183 2018-03-09 Jakub Jelinek <jakub@redhat.com>
7186 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
7187 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
7188 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
7191 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
7192 decl, use remap_type if we want to use the type.
7194 2018-03-09 Martin Sebor <msebor@redhat.com>
7196 PR tree-optimization/84526
7197 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
7199 (builtin_access::generic_overlap): Be prepared to handle non-array
7202 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
7204 PR rtl-optimization/84682
7205 * lra-constraints.c (process_address_1): Check is_address flag
7206 for address constraints.
7207 (process_alt_operands): Likewise.
7208 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
7209 preprocess_constraints.
7210 * recog.h (preprocess_constraints): Add oploc parameter.
7212 * recog.c (preprocess_constraints): Test address_operand for
7213 CT_ADDRESS constraints.
7215 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
7218 * lra-assigns.c (assign_by_spills): Return a flag of reload
7219 assignment failure. Do not process the reload assignment
7220 failures. Do not spill other reload pseudos if they has the same
7222 (lra_assign): Add a return arg. Set up from the result of
7223 assign_by_spills call.
7224 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
7225 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
7226 usage_insns if it is not NULL.
7227 (spill_hard_reg_in_range): New function.
7228 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
7229 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
7230 function prototypes.
7231 (lra_assign): Change prototype.
7232 * lra.c (lra): Add code to deal with fails by splitting hard reg
7235 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7238 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
7239 Accept complain bool parameter. Only emit errors if it is true.
7240 (arm_parse_cpu_option_name): Likewise.
7241 (arm_target_thumb_only): Adjust callers of the above.
7242 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
7243 prototype to take a default true bool parameter.
7244 (arm_parse_arch_option_name): Likewise.
7246 2018-03-09 David Malcolm <dmalcolm@redhat.com>
7247 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
7251 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
7252 * configure: Regenerate.
7253 * configure.ac ("linker --version-script option"): New.
7254 ("linker soname option"): New.
7256 2018-03-09 Richard Biener <rguenther@suse.de>
7258 PR tree-optimization/84775
7259 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
7260 immediate uses of predicate stmts and mark them modified.
7263 PR tree-optimization/84178
7264 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
7266 (version_loop_for_if_conversion): Delay update_ssa call.
7267 (tree_if_conversion): Delay update_ssa until after predicate
7270 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
7273 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
7274 when the function accesses prior frames.
7276 2018-03-08 Jakub Jelinek <jakub@redhat.com>
7279 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
7280 gen_llsym, otherwise call maybe_gen_llsym.
7283 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
7284 has ',' character inside of it.
7286 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7289 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
7290 as clobbering CC_REGNUM.
7292 2018-03-08 Richard Biener <rguenther@suse.de>
7295 * tree-scalar-evolution.c: Include tree-into-ssa.h.
7296 (follow_copies_to_constant): Do not follow SSA names registered
7299 2018-03-08 Richard Biener <rguenther@suse.de>
7301 PR tree-optimization/84178
7302 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
7304 (version_loop_for_if_conversion): Delay update_ssa call.
7305 (tree_if_conversion): Delay update_ssa until after predicate
7308 2018-03-08 David Malcolm <dmalcolm@redhat.com>
7310 PR tree-optimization/84178
7311 * tree-if-conv.c (release_bb_predicate): Remove the
7312 the assertion that the stmts have NULL use_ops.
7313 Discard the statements, asserting that they haven't
7314 yet been added to a BB.
7316 2018-03-08 Richard Biener <rguenther@suse.de>
7318 PR tree-optimization/84746
7319 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
7320 (phi_translate): Pass in destination ANTIC_OUT set.
7321 (phi_translate_1): Likewise. For a simplified result lookup
7322 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
7323 (phi_translate_set): Adjust.
7324 (do_pre_regular_insertion): Likewise.
7325 (do_pre_partial_partial_insertion): Likewise.
7327 2018-03-08 Martin Liska <mliska@suse.cz>
7329 PR gcov-profile/84735
7330 * doc/gcov.texi: Document usage of profile files.
7331 * gcov-io.h: Document changes in the format.
7333 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
7337 * dwarf2out.c (struct dw_line_info_table): Update comments for
7339 (FORCE_RESET_NEXT_VIEW): New.
7340 (FORCE_RESETTING_VIEW_P): New.
7341 (RESETTING_VIEW_P): Check for -1 too.
7342 (ZERO_VIEW_P): Likewise.
7343 (new_line_info_table): Force-reset next view.
7344 (dwarf2out_begin_function): Likewise.
7345 (dwarf2out_source_line): Simplify zero_view_p initialization.
7346 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
7347 view directly. Omit view when omitting .loc at line 0.
7349 2018-03-08 Jakub Jelinek <jakub@redhat.com>
7351 PR tree-optimization/84740
7352 * tree-switch-conversion.c (process_switch): Call build_constructors
7353 only if info.phi_count is non-zero.
7355 PR tree-optimization/84739
7356 * tree-tailcall.c (find_tail_calls): Check call arguments against
7357 DECL_ARGUMENTS (current_function_decl) rather than
7358 DECL_ARGUMENTS (func) when checking for tail recursion.
7360 2018-03-07 Jakub Jelinek <jakub@redhat.com>
7362 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
7363 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
7364 Volker Reichelt's entry and add entries for people that perform
7365 GCC fuzzy testing and report numerous bugs.
7367 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
7370 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
7371 readonly data in sdata, if that is disabled.
7372 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
7373 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
7374 -mreadonly-in-sdata option.
7376 2018-03-07 Martin Sebor <msebor@redhat.com>
7378 PR tree-optimization/84468
7379 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
7380 basic block when looking for nul assignment.
7382 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
7385 * except.h (output_function_exception_table): Adjust prototype.
7386 * except.c (output_function_exception_table): Remove FNNAME parameter
7387 and add SECTION parameter. Ouput one part of the table at a time.
7388 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
7389 the first part of the exception table and emit unwind directives.
7390 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
7391 (i386_pe_seh_cold_init): Likewise.
7392 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
7393 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
7394 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
7395 (ix86_output_call_insn): Emit a nop in one more case for SEH.
7396 * config/i386/winnt.c: Include except.h.
7397 (struct seh_frame_state): Add reg_offset, after_prologue and
7398 in_cold_section fields.
7399 (i386_pe_seh_end_prologue): Set seh->after_prologue.
7400 (i386_pe_seh_cold_init): New function.
7401 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
7402 to seh->in_cold_section.
7403 (seh_emit_push): Record the offset of the push.
7404 (seh_emit_save): Record the offet of the save.
7405 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
7406 Test seh->after_prologue to disregard the epilogue.
7407 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
7408 (i386_pe_end_cold_function): New function.
7410 2018-03-07 Jakub Jelinek <jakub@redhat.com>
7413 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
7414 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
7417 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
7419 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
7420 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
7423 * multiple_target.c: Include tree-inline.h and intl.h.
7424 (expand_target_clones): Diagnose and fail if node->definition and
7425 !tree_versionable_function_p (node->decl).
7427 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
7429 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
7431 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
7432 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
7433 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
7435 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7438 * combine.c (try_combine): Use reg_or_subregno instead of handling
7439 just paradoxical SUBREGs and REGs.
7441 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
7443 * config/arc/arc.c (arc_finalize_pic): Remove function.
7444 (arc_must_save_register): We use single base PIC register, remove
7445 checks to save/restore the PIC register.
7446 (arc_expand_prologue): Likewise.
7447 * config/arc/arc-protos.h (arc_set_default_type_attributes):
7449 (arc_verify_short): Likewise.
7450 (arc_attr_type): Likewise.
7451 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
7452 (walk_stores): Likewise.
7453 (arc_address_cost): Make it static.
7454 (arc_verify_short): Likewise.
7455 (branch_dest): Likewise.
7456 (arc_attr_type): Likewise.
7457 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
7458 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
7459 (arc_final_prescan_insn): Remove inserting the nops due to
7460 hardware hazards. It is done in reorg step.
7461 (insn_length_variant_t): Remove.
7462 (insn_length_parameters_t): Likewise.
7463 (arc_insn_length_parameters): Likewise.
7464 (arc_get_insn_variants): Likewise.
7465 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
7467 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7470 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
7473 PR tree-optimization/84687
7474 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
7476 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
7478 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7480 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
7481 Rename to ppc_speculation_barrier.
7482 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
7483 __builtin_ppc_speculation_barrier.
7485 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7488 * combine.c (combine_simplify_rtx): Don't try to simplify if
7489 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
7492 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
7494 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
7495 to 32 bytes when compiling for POWER9.
7497 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7500 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
7501 regparm >= 3 with no arg reg available also for calls with
7502 flag_force_indirect_call. Pass decl to ix86_function_regparm.
7505 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
7507 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
7509 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
7512 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
7514 2018-03-05 Richard Biener <rguenther@suse.de>
7516 PR tree-optimization/84486
7517 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
7518 When inserting a __builtin_assume_aligned call set the LHS
7519 SSA name alignment info accordingly.
7521 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
7523 PR tree-optimization/84114
7524 * config/aarch64/aarch64.c (aarch64_reassociation_width)
7525 Avoid reassociation of FLOAT_MODE addition.
7527 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
7529 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
7530 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
7531 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
7532 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
7533 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
7534 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
7535 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
7537 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
7538 __builtin_ia32_wbinvd): New builtins.
7539 (SPECIAL_ARGS2): New.
7540 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
7541 (SPECIAL_ARGS2): New.
7542 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
7543 (ix86_valid_target_attribute_inner_p): Ditto.
7544 (ix86_init_mmx_sse_builtins): Add special_args2.
7545 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
7546 TARGET_WBNOINVD_P): New.
7547 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
7548 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
7549 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
7550 * config/i386/immintrin.h (_wbinvd): New intrinsic.
7551 * config/i386/pconfigintrin.h: New file.
7552 * config/i386/wbnoinvdintrin.h: Ditto.
7553 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
7554 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
7556 2018-03-05 Richard Biener <rguenther@suse.de>
7558 PR tree-optimization/84670
7559 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7561 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7562 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7563 (compute_antic_aux): Only assert the number of values in ANTIC_IN
7564 doesn't grow if all successors (recursively) were visited at least
7567 2018-03-05 Richard Biener <rguenther@suse.de>
7569 PR tree-optimization/84650
7570 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
7571 if executed in the loop pipeline.
7573 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
7575 * doc/configfiles.texi (Configuration Files): Move info about
7576 conditionalizing $target-protos.h to...
7577 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
7578 differs from $target-protos.h.
7580 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
7581 Chung-Ju Wu <jasonwucj@gmail.com>
7583 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
7584 * config/nds32/nds32-multiple.md (setmemsi): Define.
7585 * config/nds32/nds32-memory-manipulation.c
7586 (nds32_gen_dup_4_byte_to_word_value): New.
7587 (emit_setmem_word_loop): New.
7588 (emit_setmem_byte_loop): New.
7589 (nds32_expand_setmem_loop): New.
7590 (nds32_expand_setmem_loop_v3m): New.
7591 (nds32_expand_setmem_unroll): New.
7592 (nds32_expand_setmem): New.
7594 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7595 Chung-Ju Wu <jasonwucj@gmail.com>
7597 * config/nds32/nds32-memory-manipulation.c
7598 (nds32_emit_load_store): New.
7599 (nds32_emit_post_inc_load_store): New.
7600 (nds32_emit_mem_move): New.
7601 (nds32_emit_mem_move_block): New.
7602 (nds32_expand_movmemsi_loop_unknown_size): New.
7603 (nds32_expand_movmemsi_loop_known_size): New.
7604 (nds32_expand_movmemsi_loop): New.
7605 (nds32_expand_movmemsi_unroll): New.
7606 (nds32_expand_movmemqi): Rename ...
7607 (nds32_expand_movmemsi): ... to this.
7608 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
7609 (movmemsi): ... to this.
7610 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
7611 (nds32_expand_movmemsi): ... to this.
7613 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7614 Monk Chiang <sh.chiang04@gmail.com>
7615 Chung-Ju Wu <jasonwucj@gmail.com>
7617 * config/nds32/nds32-protos.h
7618 (nds32_expand_load_multiple): New arguments.
7619 (nds32_expand_store_multiple): Ditto.
7620 (nds32_valid_multiple_load_store): Rename ...
7621 (nds32_valid_multiple_load_store_p): ... to this.
7622 * config/nds32/nds32-memory-manipulation.c
7623 (nds32_expand_load_multiple): Refine implementation.
7624 (nds32_expand_store_multiple): Ditto.
7625 * config/nds32/nds32-multiple.md
7626 (load_multiple): Update nds32_expand_load_multiple interface.
7627 (store_multiple): Update nds32_expand_store_multiple interface.
7628 * config/nds32/nds32-predicates.c
7629 (nds32_valid_multiple_load_store): Rename ...
7630 (nds32_valid_multiple_load_store_p): ... to this and refine
7632 * config/nds32/predicates.md
7633 (nds32_load_multiple_and_update_address_operation): New predicate.
7634 (nds32_store_multiple_and_update_address_operation): New predicate.
7636 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7637 Chung-Ju Wu <jasonwucj@gmail.com>
7639 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
7640 (combo): New attribute.
7641 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
7643 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
7645 * config/nds32/nds32.opt: Change -mcmodel= default value.
7647 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
7648 Monk Chiang <sh.chiang04@gmail.com>
7649 Chung-Ju Wu <jasonwucj@gmail.com>
7651 * config/nds32/constants.md (unspec_element): New enum.
7652 * config/nds32/constraints.md (Umw): New constraint.
7653 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
7654 * config/nds32/nds32-intrinsic.md: Likewise.
7655 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
7656 (nds32_valid_smw_lwm_base_p): New.
7657 (nds32_output_smw_single_word): New.
7658 (nds32_output_lmw_single_word): New.
7659 (nds32_expand_unaligned_load): New.
7660 (nds32_expand_unaligned_store): New.
7661 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
7662 (nds32_output_smw_single_word): Declare.
7663 (nds32_output_lmw_single_word): Declare.
7664 (nds32_expand_unaligned_load): Declare.
7665 (nds32_expand_unaligned_store): Declare.
7666 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
7667 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
7668 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
7669 NDS32_BUILTIN_UASTORE_DW.
7670 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
7673 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
7674 Kito Cheng <kito.cheng@gmail.com>
7675 Chung-Ju Wu <jasonwucj@gmail.com>
7677 * config/nds32/nds32-intrinsic.c
7678 (nds32_expand_builtin_null_ftype_reg): Delete.
7679 (nds32_expand_builtin_reg_ftype_imm): Ditto.
7680 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
7681 (nds32_read_argument): New.
7682 (nds32_legitimize_target): Ditto.
7683 (nds32_legitimize_argument): Ditto.
7684 (nds32_check_constant_argument): Ditto.
7685 (nds32_expand_unop_builtin): Ditto.
7686 (nds32_expand_unopimm_builtin): Ditto.
7687 (nds32_expand_binop_builtin): Ditto.
7688 (nds32_builtin_decl_impl): Ditto.
7689 (builtin_description): Ditto.
7690 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
7691 (nds32_init_builtins_impl): Ditto.
7692 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
7693 (nds32_builtin_decl): New.
7694 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
7695 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
7697 2018-03-02 Jeff Law <law@redhat.com>
7699 * reorg.c (stop_search_p): Handle DEBUG_INSN.
7700 (redundant_insn, fill_simple_delay_slots): Likewise.
7701 (fill_slots_from_thread): Likewise.
7702 * resource.c (mark_referenced_resources): Likewise.
7703 (mark_set_resources, find_dead_or_set_registers): Likewise.
7705 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7707 * substring-locations.h (format_warning_va): Formatting fix for
7709 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
7711 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
7712 * substring-locations.c: Include intl.h.
7713 (format_warning_va): Turned into small wrapper around
7714 format_warning_n_va, renamed to ...
7715 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
7716 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
7718 (format_warning_at_substring_n): New function.
7719 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
7720 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
7721 format_warning_at_substring with just a shorter name instead of
7722 const function pointer.
7723 (fmtwarn_n): New function.
7724 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
7725 appropriate, get rid of all the fmtstr temporaries, move conditionals
7726 with G_() wrapped string literals directly into fmtwarn arguments,
7727 cast dir.len to (int), formatting fixes.
7729 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
7731 * doc/invoke.texi: Remove "Cilk Plus" references.
7733 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7734 Richard Biener <rguenther@suse.de>
7737 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
7738 for error or warning attributes if CALL_FROM_THUNK_P is set.
7741 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7744 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
7745 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
7748 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
7749 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
7751 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
7752 instead of -1U in last predictors element's probability member.
7754 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7757 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7758 arguments if they are comparable.
7760 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7762 PR tree-optimization/84634
7763 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
7764 masks and masked_loop_p with a single loop_masks, making sure it's
7765 null for bb vectorization.
7767 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7769 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7770 (vect_analyze_data_ref_access): Use loop->safe_len rather than
7771 loop->force_vectorize to check whether there is no alias.
7773 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7776 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
7778 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
7780 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
7781 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
7782 instead of a loop around prev_real_insn.
7783 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
7787 * config/i386/i386.c (ix86_print_operand): Use conditional
7788 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
7791 2018-03-02 Richard Biener <rguenther@suse.de>
7793 PR tree-optimization/84427
7794 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
7795 (bitmap_set_subtract_values): Rewrite to handle multiple
7798 (prune_clobbered_mems): Likewise.
7799 (phi_translate): Take edge instead of pred/phiblock.
7800 (phi_translate_1): Likewise.
7801 (phi_translate_set): Likewise. Insert all translated
7802 exprs for a value into the set, keeping possibly multiple
7803 expressions per value.
7804 (compute_antic_aux): Adjust for phi_translate changes.
7805 When intersecting union the expressions and prune those
7806 not in the final value set, keeping possibly multiple
7807 expressions per value. Do not use value-insertion
7808 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
7809 all expressions. Add verification that the value-sets
7810 only shrink during iteration.
7811 (compute_partial_antic_aux): Adjust for the phi_translate changes.
7812 (do_pre_regular_insertion): Likewise.
7813 (do_pre_partial_partial_insertion): Likewise.
7815 2018-03-02 Richard Biener <rguenther@suse.de>
7818 * config/darwin.c (saved_debug_info_level): New static global.
7819 (darwin_asm_lto_start): Disable debug info generation for LTO out.
7820 (darwin_asm_lto_end): Restore debug info generation settings.
7822 2018-03-01 Martin Liska <mliska@suse.cz>
7825 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
7828 2018-03-01 Richard Biener <rguenther@suse.de>
7831 * dwarf2out.c (gen_variable_die): Properly handle late VLA
7832 type annotation with LTO when debug was disabled at compile-time.
7834 2018-03-01 Matthew Fortune <mfortune@gmail.com>
7836 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
7838 (mips_final_postscan_insn): Likewise.
7840 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
7842 PR rtl-optimization/84528
7843 * alias.c (init_alias_target): Add commentary.
7844 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
7845 a unique base value if the frame pointer is not eliminated
7846 to the stack pointer.
7848 2018-03-01 Tom de Vries <tom@codesourcery.com>
7850 PR rtl-optimization/83327
7851 * lra-int.h (hard_regs_spilled_into): Declare.
7852 * lra.c (hard_regs_spilled_into): Define.
7853 (init_reg_info): Init hard_regs_spilled_into.
7854 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
7855 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
7856 (process_bb_lives): Handle hard_regs_spilled_into.
7857 (lra_create_live_ranges_1): Before doing liveness propagation, clear
7858 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
7860 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
7862 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
7863 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
7864 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
7865 * config/rs6000/aix72.h: New file.
7867 2018-02-28 Jakub Jelinek <jakub@redhat.com>
7869 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
7870 instead of warning_at with conditional singular and plural messages
7874 * stor-layout.c (update_alignment_for_field): For
7875 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
7876 && !DECL_PACKED (field), do the alignment update, just use
7877 only desired_align instead of MAX (type_align, desired_align)
7879 (place_field): Don't do known_align < desired_align handling
7880 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
7881 is non-NULL, instead do it after rli->prev_field handling and
7882 only if not within a bitfield word. For DECL_PACKED (field)
7883 use type_align of BITS_PER_UNIT.
7885 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
7887 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
7888 superfluous parentheses and trailing spaces.
7890 2018-02-28 Richard Biener <rguenther@suse.de>
7892 PR tree-optimization/84584
7893 * graphite-scop-detection.c (scop_detection::add_scop): Discard
7894 SCoPs with fake exit edge.
7896 2018-02-28 Martin Liska <mliska@suse.cz>
7899 * timevar.c (timer::print): Fix format to properly print 100%
7902 2018-02-28 Richard Biener <rguenther@suse.de>
7905 * genmatch.c (capture_info::walk_match): Do not mark
7906 captured expressions without operands as expr_p given
7907 they act more like predicates and should be subject to
7908 "lost tail" side-effect preserving.
7910 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
7912 PR rtl-optimization/81611
7913 * auto-inc-dec.c (attempt_change): Move dead note from
7914 mem_insn if it's the next use of regno
7915 (find_address): Take address use of reg holding
7916 non-incremented value. Add parm to limit search to the named
7918 (merge_in_block): Attempt to use a mem insn that is the next
7919 use of the original regno.
7921 2018-02-27 Martin Sebor <msebor@redhat.com>
7924 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
7925 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
7927 2018-02-27 Martin Sebor <msebor@redhat.com>
7929 PR translation/84207
7930 * diagnostic-core.h (warning_n, error_n, inform_n): Change
7931 n argument to unsigned HOST_WIDE_INT.
7932 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
7933 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
7934 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
7935 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
7937 2018-02-27 Richard Biener <rguenther@suse.de>
7939 PR tree-optimization/84512
7940 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7941 Do not use the estimate returned from record_stmt_cost for
7942 the scalar iteration cost but sum properly using add_stmt_cost.
7944 2018-02-27 Richard Biener <rguenther@suse.de>
7946 PR tree-optimization/84466
7947 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7948 Adjust last change to less strictly validate use operands.
7950 2018-02-27 Martin Liska <mliska@suse.cz>
7952 PR gcov-profile/84548
7953 * gcov.c (process_file): Allow partial overlap and consider it
7954 also as group functions.
7955 (output_lines): Properly calculate range of lines for a group.
7957 2018-02-27 Martin Liska <mliska@suse.cz>
7959 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
7960 'ggc' suffixes. Change first column width.
7961 (timer::print): Fix formatting of the column.
7963 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
7965 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
7966 preserve inline entry blocks for the sake of debug inline
7967 entry point markers alone.
7968 (remove_unused_locals): Suggest in comments a better place to
7969 force the preservation of inline entry blocks that are
7970 otherwise unused, but do not preserve them.
7972 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7974 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
7976 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7979 * config/i386/constraints.md (Bs): Replace
7980 ix86_indirect_branch_register with
7981 TARGET_INDIRECT_BRANCH_REGISTER.
7983 * config/i386/i386.md (indirect_jump): Likewise.
7984 (tablejump): Likewise.
7985 (*sibcall_memory): Likewise.
7986 (*sibcall_value_memory): Likewise.
7987 Peepholes of indirect call and jump via memory: Likewise.
7988 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
7989 (*sibcall_value_GOT_32): Likewise.
7990 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7991 (GOT_memory_operand): Likewise.
7992 (call_insn_operand): Likewise.
7993 (sibcall_insn_operand): Likewise.
7994 (GOT32_symbol_operand): Likewise.
7995 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
7997 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7999 PR rtl-optimization/83496
8000 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
8001 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
8002 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
8003 redundant insn, if any.
8004 (relax_delay_slots): Likewise.
8005 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
8007 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
8009 PR tree-optimization/83965
8010 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
8011 that grouped statements are part of a reduction chain. Return
8012 true if the statement is not marked as a reduction itself but
8014 (vect_recog_dot_prod_pattern): Don't check whether the statement
8015 is part of a group here.
8016 (vect_recog_sad_pattern): Likewise.
8017 (vect_recog_widen_sum_pattern): Likewise.
8019 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
8022 * final.c (rest_of_clean_state): Also look for calls inside sequences.
8024 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
8027 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
8029 (ix86_output_indirect_function_return): New prototype.
8030 (ix86_split_simple_return_pop_internal): Likewise.
8031 * config/i386/i386.c (indirect_return_via_cx): New.
8032 (indirect_return_via_cx_bnd): Likewise.
8033 (indirect_thunk_name): Handle return va CX_REG.
8034 (output_indirect_thunk_function): Create alias for
8035 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
8036 (ix86_output_indirect_jmp): Remove the bool argument.
8037 (ix86_output_indirect_function_return): New function.
8038 (ix86_split_simple_return_pop_internal): Likewise.
8039 * config/i386/i386.md (*indirect_jump): Don't pass false
8040 to ix86_output_indirect_jmp.
8041 (*tablejump_1): Likewise.
8042 (simple_return_pop_internal): Change it to define_insn_and_split.
8043 Call ix86_split_simple_return_pop_internal to split it for
8045 (simple_return_indirect_internal): Call
8046 ix86_output_indirect_function_return instead of
8047 ix86_output_indirect_jmp.
8049 2018-02-26 Jakub Jelinek <jakub@redhat.com>
8052 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
8053 memset and value initialization afterwards.
8055 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
8057 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
8059 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8062 * common/config/aarch64/aarch64-common.c
8063 (aarch_option_optimization_table[]): Switch
8064 off fomit-frame-pointer
8066 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
8067 Chung-Ju Wu <jasonwucj@gmail.com>
8069 * config/nds32/nds32-multiple.md (load_multiple): Disallow
8071 (store_multiple): Ditto.
8073 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
8075 * config.gcc: Add --with-cpu support for nds32 target.
8076 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
8077 * config/nds32/nds32.opt: Add -mcpu= option.
8079 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
8081 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
8082 isel=yes): Warn for these deprecated options.
8084 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
8086 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
8087 ISA_2_5_MASKS_EMBEDDED.
8089 2018-02-23 Jakub Jelinek <jakub@redhat.com>
8091 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
8092 p->max as pointers rather than using iterative_hash_expr.
8094 2018-02-23 Carl Love <cel@us.ibm.com>
8096 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
8097 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
8099 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
8100 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
8101 P8V_BUILTIN_VEC_VUNSIGNED2.
8103 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
8106 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
8107 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
8109 * lra-constraints.c (curr_insn_transform): Set up
8110 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
8112 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
8113 * lra-eliminations.c (spill_pseudos): Ditto.
8114 (process_insn_for_elimination): Ditto.
8115 * lra-lives.c (reg_early_clobber_p): Use the new macros.
8116 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
8117 LRA_NON_CLOBBERED_ALT.
8119 2018-02-22 Martin Sebor <msebor@redhat.com>
8121 PR tree-optimization/84480
8122 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
8123 to maybe_diag_stxncpy_trunc. Call it.
8124 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
8125 from gimple_fold_builtin_strcpy. Print inlining stack.
8126 (handle_builtin_stxncpy): Print inlining stack.
8127 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
8129 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
8132 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
8133 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
8134 and -fcheck-pointer-bounds are used together.
8135 (indirect_thunk_prefix): New enum.
8136 (indirect_thunk_need_prefix): New function.
8137 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
8138 "_nt" instead of "_bnd" for NOTRACK prefix.
8139 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
8140 (output_indirect_thunk_function): Likewise.
8142 (ix86_code_end): Update output_indirect_thunk_function calls.
8143 (ix86_output_indirect_branch_via_reg): Replace
8144 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
8145 (ix86_output_indirect_branch_via_push): Likewise.
8146 (ix86_output_function_return): Likewise.
8147 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
8148 incompatible with -fcf-protection=branch and
8149 -fcheck-pointer-bounds.
8151 2018-02-22 Steve Ellcey <sellcey@cavium.com>
8154 * config/aarch64/aarch64.c (aarch64_print_address_internal):
8155 Change gcc_assert call to output_operand_lossage.
8157 2018-02-22 Steve Ellcey <sellcey@cavium.com>
8159 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
8161 2018-02-22 DJ Delorie <dj@redhat.com>
8162 Sebastian Perta <sebastian.perta@renesas.com>
8163 Oleg Endo <olegendo@gcc.gnu.org>
8165 * config/rx/rx.c (rx_rtx_costs): New function.
8166 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
8168 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
8170 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
8172 2018-02-22 Martin Liska <mliska@suse.cz>
8175 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
8176 Add "native" as a possible value.
8178 2018-02-22 Martin Liska <mliska@suse.cz>
8181 * config/i386/i386.c (ix86_option_override_internal):
8182 Add "native" as a possible value for -march and -mtune.
8184 2018-02-22 Jakub Jelinek <jakub@redhat.com>
8187 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
8188 to all type variants.
8190 PR tree-optimization/84503
8191 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
8192 width as info->bitpos + info->bitsize - start.
8193 (merged_store_group::merge_overlapping): Simplify width computation.
8194 (check_no_overlap): New function.
8195 (imm_store_chain_info::try_coalesce_bswap): Compute expected
8196 start + width and last_order of the group, fail if check_no_overlap
8198 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
8199 to group if check_no_overlap fails.
8201 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
8203 * config/rs6000/altivec.md: Delete contraint arguments to
8204 define_expand, define_split, and define_peephole2, and in
8205 define_insn_and_split if always unused.
8206 * config/rs6000/darwin.md: Ditto.
8207 * config/rs6000/dfp.md: Ditto.
8208 * config/rs6000/rs6000.md: Ditto.
8209 * config/rs6000/sync.md: Ditto.
8210 * config/rs6000/vector.md: Ditto.
8211 * config/rs6000/vsx.md: Ditto.
8213 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
8215 * config/rs6000/altivec.md: Write output control strings as braced
8216 blocks instead of double-quoted strings.
8217 * config/rs6000/darwin.md: Ditto.
8218 * config/rs6000/rs6000.md: Ditto.
8219 * config/rs6000/vector.md: Ditto.
8220 * config/rs6000/vsx.md: Ditto.
8222 2018-02-21 Jason Merrill <jason@redhat.com>
8224 PR c++/84314 - ICE with templates and fastcall attribute.
8225 * attribs.c (build_type_attribute_qual_variant): Remove assert.
8227 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
8229 * ipa-cp.c (determine_versionability): Fix comment typos.
8231 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
8234 * ipa-cp.c (determine_versionability): Do not version functions caling
8237 2018-02-21 Martin Liska <mliska@suse.cz>
8240 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
8241 Add "native" as a possible value.
8242 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
8243 the macro when native cpu detection is available.
8245 2018-02-21 Martin Liska <mliska@suse.cz>
8248 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
8249 Add "native" as a possible value.
8250 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
8251 when native cpu detection is available.
8253 2018-02-21 Jakub Jelinek <jakub@redhat.com>
8254 Martin Sebor <msebor@redhat.com>
8256 PR tree-optimization/84478
8257 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
8259 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
8260 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
8261 support which is conservatively correct, for 2 only stay conservative
8262 for maxlen. Formatting and comment capitalization fixes. Add STRICT
8263 argument to the 2 argument get_range_strlen, adjust 6 arg
8264 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
8266 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
8267 (gimple_fold_builtin_strlen): Pass true as last argument to
8270 2018-02-20 Martin Sebor <msebor@redhat.com>
8273 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
8274 (builtin_memref::set_base_and_offset): Same. Handle inner references.
8275 (builtin_memref::builtin_memref): Factor out parts into
8276 set_base_and_offset and call it.
8278 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
8281 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
8282 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
8283 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
8284 search at the associated MODE_INT.
8286 2018-02-20 Jeff Law <law@redhat.com>
8289 PR tree-optimization/81592
8291 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
8292 for range data rather than using global data.
8293 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
8294 range data rather than using global data.
8295 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
8296 pass it to children as needed.
8297 (struct directive::fmtresult): Similarly.
8298 (struct directive::set_width): Similarly.
8299 (struct directive::set_precision): Similarly.
8300 (format_integer, format_directive, parse_directive): Similarly.
8301 (format_none): Accept unnamed vr_values parameter.
8302 (format_percent, format_floating, format_character): Similarly.
8303 (format_string, format_plain): Similarly.
8304 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
8305 the EVRP range analyzer for range data rather than using global data.
8306 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
8307 gimple-ssa-evrp-analyze.h
8308 (class sprintf_dom_walker): Add after_dom_children member function.
8309 Add evrp_range_analyzer member.
8310 (sprintf_dom_walker::before_dom_children): Call into the EVRP
8311 range analyzer as needed.
8312 (sprintf_dom_walker::after_dom_children): New member function.
8313 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
8315 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
8316 (evrp_range_analyzer::pop_to_marker): Likewise.
8318 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
8320 PR tree-optimization/84419
8321 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
8322 with the required type if its current type is compatible but
8325 2018-02-20 Jakub Jelinek <jakub@redhat.com>
8328 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
8329 after vectorization.
8331 2018-02-20 Martin Liska <mliska@suse.cz>
8334 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
8335 possible values if we don't have a hint.
8337 2018-02-20 Martin Liska <mliska@suse.cz>
8341 * final.c (shorten_branches): Build align_tab array with one
8343 * opts.c (finish_options): Add alignment option limit check.
8344 (MAX_CODE_ALIGN): Likewise.
8345 (MAX_CODE_ALIGN_VALUE): Likewise.
8346 * doc/invoke.texi: Document maximum allowed option value for
8347 all -falign-* options.
8349 2018-02-19 Jakub Jelinek <jakub@redhat.com>
8352 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
8353 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
8354 * var-tracking.c (emit_note_insn_var_location): Remove all references
8355 to NOTE_INSN_CALL_ARG_LOCATION.
8356 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
8357 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
8358 Use copy_rtx_if_shared.
8359 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
8360 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
8361 (dwarf2out_var_location): Remove handling of
8362 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
8364 * final.c (final_scan_insn): Remove all references to
8365 NOTE_INSN_CALL_ARG_LOCATION.
8366 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
8367 before dumping final insns.
8368 * except.c (emit_note_eh_region_end): Remove all references to
8369 NOTE_INSN_CALL_ARG_LOCATION.
8370 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
8371 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
8372 * config/arc/arc.c (hwloop_optimize): Likewise.
8373 * config/arm/arm.c (create_fix_barrier): Likewise.
8374 * config/s390/s390.c (s390_chunkify_start): Likewise.
8375 * config/sh/sh.c (find_barrier): Likewise.
8376 * config/i386/i386.c (rest_of_insert_endbranch,
8377 ix86_seh_fixup_eh_fallthru): Likewise.
8378 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
8379 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
8380 * config/frv/frv.c (frv_function_prologue): Likewise.
8381 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
8383 (note_outside_basic_block_p): Remove all references to
8384 NOTE_INSN_CALL_ARG_LOCATION.
8385 * gengtype.c (adjust_field_rtx_def): Likewise.
8386 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
8388 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
8389 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
8392 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
8395 PR tree-optimization/84452
8396 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
8397 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
8400 2018-02-19 Martin Liska <mliska@suse.cz>
8403 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
8405 2018-02-19 Martin Liska <mliska@suse.cz>
8406 Richard Sandiford <richard.sandiford@linaro.org>
8408 PR tree-optimization/82491
8409 * gimple-fold.c (get_base_constructor): Make earlier bail out
8412 2018-02-19 Carl Love <cel@us.ibm.com>
8414 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
8415 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
8417 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
8418 P8V_BUILTIN_VEC_NEG.
8420 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
8422 * config/rl78/rl78.md (movdf): New define expand.
8424 2018-02-19 Martin Liska <mliska@suse.cz>
8427 * doc/invoke.texi: Fix typo.
8428 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
8430 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
8432 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8433 handle rs6000_single_float and rs6000_double_float specially for
8436 2018-02-16 Jeff Law <law@redhat.com>
8438 * config/rx/rx.c (add_pop_cfi_notes): New function.;
8441 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8444 * ipa-inline.c (inline_small_functions): Fix a typo.
8446 2018-02-16 Nathan Sidwell <nathan@acm.org>
8448 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
8450 2018-02-16 Carl Love <cel@us.ibm.com>
8452 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
8453 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
8454 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
8455 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
8456 expansion to P8V_BUILTIN_VEC_FLOAT2.
8458 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
8460 PR rtl-optimization/70023
8461 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
8462 src_regno into account.
8464 2018-02-16 Carl Love <cel@us.ibm.com>
8466 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
8467 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
8468 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
8469 * config/rs6000/rs6000.c: Remove case statements for
8470 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
8471 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
8472 and P9V_BUILTIN_VEC_VINSERT4B.
8473 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
8474 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
8475 * config/rs6000/vsx.md:
8476 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
8479 2018-02-16 Carl Love <cel@us.ibm.com>
8481 * config/rs6000/altivec.h: Add builtin names vec_extract4b
8483 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
8485 * config/rs6000/rs6000-c.c: Add the definitions for
8486 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
8487 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
8488 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
8489 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
8490 definition for insert4b and define insn *insert3b_internal.
8491 * doc/extend.texi: Add documentation for vec_extract4b.
8493 2018-02-16 Nathan Sidwell <nathan@acm.org>
8495 * doc/extend.texi (Backwards Compatibility): Mention friend
8496 injection. Note for-scope is deprecated.
8497 * doc/invoke.texi (-ffriend-injection): Deprecate.
8499 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
8501 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
8502 that moved to I2, also allow destinations that are a paradoxical
8503 subreg (instead of a normal reg).
8505 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
8508 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
8511 2018-02-16 Richard Biener <rguenther@suse.de>
8513 PR tree-optimization/84037
8514 PR tree-optimization/84016
8516 * config/i386/i386.c (ix86_builtin_vectorization_cost):
8517 Adjust vec_construct for the fact we need additional higher latency
8518 128bit inserts for AVX256 and AVX512 vector builds.
8519 (ix86_add_stmt_cost): Scale vector construction cost for
8522 2018-02-16 Richard Biener <rguenther@suse.de>
8524 PR tree-optimization/84417
8525 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
8526 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
8527 (non_rewritable_lvalue_p): Likewise, use poly-ints.
8529 2018-02-16 Martin Liska <mliska@suse.cz>
8532 * internal-fn.def (ASAN_CHECK): Set proper flags.
8533 (ASAN_MARK): Likewise.
8535 2018-02-16 Julia Koval <julia.koval@intel.com>
8537 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
8538 from PTA_CANNONLAKE.
8540 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8543 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
8544 Use ++iter rather than iter++ for std::list iterators.
8545 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
8546 defer deleting them until all nodes in the forest are processed. Do
8547 free even leaf nodes. Change to_process into auto_vec.
8550 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
8551 * vec.h (vec_default_construct): Use memset instead of placement new
8552 if BROKEN_VALUE_INITIALIZATION is defined.
8553 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
8554 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
8557 PR rtl-optimization/83723
8558 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
8559 * lra.c (lra_substitute_pseudo): Likewise. If true, use
8560 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
8562 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
8564 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
8566 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
8568 PR rtl-optimization/81443
8569 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
8570 from inner REGs to paradoxical SUBREGs.
8572 2018-02-16 Richard Biener <rguenther@suse.de>
8574 PR tree-optimization/84399
8575 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8576 For operands we can analyze at their definition make sure we can
8577 analyze them at each use as well.
8579 2018-02-16 Richard Biener <rguenther@suse.de>
8581 PR tree-optimization/84190
8582 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
8583 volatile accesses if the decl isn't volatile.
8585 2018-02-15 Jason Merrill <jason@redhat.com>
8587 PR c++/84314 - ICE with templates and fastcall attribute.
8588 * attribs.c (build_type_attribute_qual_variant): Don't clobber
8589 TYPE_CANONICAL on an existing type.
8591 2018-02-15 Jakub Jelinek <jakub@redhat.com>
8593 PR tree-optimization/84383
8594 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
8595 dstoff nor call operand_equal_p if dstbase is NULL.
8597 PR tree-optimization/84334
8598 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
8599 also a CONSTANT_CLASS_P, punt.
8601 2018-02-14 Jim Wilson <jimw@sifive.com>
8603 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
8604 first SMALL_OPERAND check. New local min_second_step. Move assert
8605 to where locals are set. Add TARGET_RVC support.
8606 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
8608 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
8610 * doc/invoke.texi: Correct -Wformat-overflow code sample.
8612 2018-02-14 Martin Sebor <msebor@redhat.com>
8614 PR tree-optimization/83698
8615 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
8616 arrays constrain the offset range to their bounds.
8617 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
8618 (builtin_access::overlap): Avoid setting the size of overlap if it's
8620 (maybe_diag_overlap): Also consider arrays when deciding what values
8621 of offsets to include in diagnostics.
8623 2018-02-14 Martin Sebor <msebor@redhat.com>
8626 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
8627 that correspond to the kind of a declaration.
8629 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
8632 * config/pa/pa.md: Load address of PIC label using the linkage table
8633 if the label is nonlocal.
8635 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
8637 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
8638 warning message if user requests -maltivec=be.
8639 * doc/invoke.texi: Document deprecation of -maltivec=be.
8641 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
8644 * config/rs6000/rs6000-c.c: Update definitions for
8645 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
8646 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
8648 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8651 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
8652 add _get_ssp intrinsics. Remove argument from
8653 __builtin_ia32_rdssp[d|q].
8654 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
8655 * config/i386/i386-builtin.def: Remove argument from
8656 __builtin_ia32_rdssp[d|q].
8657 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
8658 ix86_expand_special_args_builtin for _rdssp[d|q].
8659 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
8660 Clear register before usage.
8661 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
8662 Add documentation for new _get_ssp and _inc_ssp intrinsics.
8664 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
8666 PR tree-optimization/84357
8667 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
8668 operand 1 of an ARRAY_REF too.
8670 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
8673 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
8674 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8676 (set_of_reg): New struct.
8677 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
8678 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
8679 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8681 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
8682 Split into bitclr, bitset, bitinvert patterns if appropriate.
8683 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
8684 use rx_fuse_in_memory_bitop.
8685 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
8686 to named insn, correct maximum insn length.
8688 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
8691 * machmode.def: Define a complex mode for PARTIAL_INT.
8692 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
8694 * doc/rtl.texi: Document CSPImode.
8695 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
8697 (msp430_hard_regno_nregs_with_padding): Likewise.
8699 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
8702 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
8704 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
8706 PR rtl-optimization/84169
8707 * combine.c (try_combine): New variable split_i2i3. Set it to true if
8708 we generated a parallel as new i3 and we split that to new i2 and i3
8709 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
8710 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
8711 those to i2, not i1. Partially rewrite this scan code.
8713 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8716 * stor-layout.c (place_field): For variable length fields, adjust
8717 offset_align afterwards not just based on the field's alignment,
8718 but also on the size.
8721 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
8722 of exps and logs in the use_exp2 case.
8724 2018-02-13 Jeff Law <law@redhat.com>
8726 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
8729 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
8732 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
8736 * common.opt (gas-loc-support, gas-locview-support): New.
8737 (ginline-points, ginternal-reset-location-views): New.
8738 * doc/invoke.texi: Document them. Use @itemx where intended.
8739 (gvariable-location-views): Adjust.
8740 * target.def (reset_location_view): New.
8741 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
8742 (TARGET_RESET_LOCATION_VIEW): New.
8743 * doc/tm.texi: Rebuilt.
8744 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
8745 (dwarf2out_default_as_locview_support): New.
8746 (output_asm_line_debug_info): Use option variables.
8747 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
8748 (output_loc_list): Likewise.
8749 (add_high_low_attributes): Check option variables.
8750 Don't output entry view attribute in strict mode.
8751 (gen_inlined_subroutine_die): Check option variables.
8752 (dwarf2out_inline_entry): Likewise.
8753 (init_sections_and_labels): Likewise.
8754 (dwarf2out_early_finish): Likewise.
8755 (maybe_reset_location_view): New, from...
8756 (dwarf2out_var_location): ... here. Call it.
8757 * debug.h (dwarf2out_default_as_loc_support): Declare.
8758 (dwarf2out_default_as_locview_support): Declare.
8759 * hooks.c (hook_int_rtx_insn_0): New.
8760 * hooks.h (hook_int_rtx_insn_0): Declare.
8761 * toplev.c (process_options): Take -gas-loc-support and
8762 -gas-locview-support from dwarf2out. Enable
8763 -gvariable-location-views by default only with locview
8764 assembler support. Enable -ginternal-reset-location-views by
8765 default only if the target defines the corresponding hook.
8766 Enable -ginline-points by default if location views are
8767 enabled; force it disabled if statement frontiers are
8769 * tree-inline.c (expand_call_inline): Check option variables.
8770 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
8772 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
8774 PR tree-optimization/84321
8775 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
8776 handling. Also check whether the anti-range contains any values
8777 that satisfy the mask; switch to a VR_RANGE if not.
8779 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
8782 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
8784 2018-02-13 Martin Jambor <mjambor@suse.cz>
8787 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
8788 of call statements, also set location of a load to a temporary.
8790 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
8792 * config/rl78/rl78.c (add_vector_labels): New function.
8793 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
8794 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
8795 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
8796 which checks that no arguments are passed.
8797 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
8798 * doc/extend.texi: Documentation for the new attribute.
8800 2018-02-13 Andreas Schwab <schwab@suse.de>
8802 * config/riscv/linux.h (CPP_SPEC): Define.
8804 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8807 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
8808 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
8809 OPTION_MASK_ISA_AES as first argument to def_builtin_const
8810 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
8811 instead of OPTION_MASK_ISA_PCLMUL as first argument to
8812 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
8813 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
8814 temporarily for AES and PCLMUL builtins.
8816 PR tree-optimization/84339
8817 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
8818 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
8822 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
8823 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
8824 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
8826 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
8828 * omp-simd-clone.h: New file.
8829 * omp-simd-clone.c: Include omp-simd-clone.h.
8830 (expand_simd_clones): No longer static.
8831 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
8832 cgraph.h and omp-simd-clone.h.
8833 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
8834 (vect_recog_widen_shift_pattern): Formatting fix.
8835 (vect_pattern_recog_1): Don't check optab for calls.
8838 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
8839 operands[2] into a REG before using gen_lowpart on it.
8841 2018-02-12 Jeff Law <law@redhat.com>
8844 * config/sh/sh.c (find_barrier): Consider a sibling call
8847 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
8848 successfully back substituting a reg.
8850 2018-02-12 Richard Biener <rguenther@suse.de>
8852 PR tree-optimization/84037
8853 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
8854 parameter, move visited init to caller.
8855 (vect_slp_analyze_operations): Separate cost from validity
8856 check, initialize visited once for all instances.
8857 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
8859 * tree-vect-stmts.c (vect_model_simple_cost): Make early
8861 (vect_model_promotion_demotion_cost): Likewise.
8862 (vectorizable_bswap): Guard cost modeling with !slp_node
8863 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
8865 (vectorizable_call): Likewise.
8866 (vectorizable_conversion): Likewise.
8867 (vectorizable_assignment): Likewise.
8868 (vectorizable_shift): Likewise.
8869 (vectorizable_operation): Likewise.
8870 (vectorizable_store): Likewise.
8871 (vectorizable_load): Likewise.
8872 (vectorizable_condition): Likewise.
8873 (vectorizable_comparison): Likewise.
8875 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
8878 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
8879 (ASAN_MARK): Fix fnspec to account for return value, change pointer
8880 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
8882 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8885 * params.def (inline-min-speedup): Increase from 8 to 15.
8886 (max-inline-insns-auto): Decrease from 40 to 30.
8887 * ipa-split.c (consider_split): Add some buffer for function to
8888 be considered inlining candidate.
8889 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
8892 2018-02-12 Richard Biener <rguenther@suse.de>
8894 PR tree-optimization/84037
8895 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
8896 matched stmts if we cannot swap the non-matched ones.
8898 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8900 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
8901 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
8902 _mm_maskz_scalef_round_ss): New intrinsics.
8903 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
8904 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
8905 __builtin_ia32_scalefss_round): Remove.
8906 (__builtin_ia32_scalefsd_mask_round,
8907 __builtin_ia32_scalefss_mask_round): New intrinsics.
8908 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
8909 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
8910 ((match_operand:VF_128 2 "<round_nimm_predicate>"
8911 "<round_constraint>")): Changed to ...
8912 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
8913 "<round_scalar_constraint>")): ... this.
8914 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
8915 %0, %1, %2<round_op3>}"): Changed to ...
8916 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
8917 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
8918 %2<round_scalar_mask_op3>}"): ... this.
8919 * config/i386/subst.md (round_scalar_nimm_predicate): New.
8921 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8923 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
8924 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
8925 (_mm_maskz_sqrt_round_ss): New intrinsics.
8926 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
8927 (__builtin_ia32_sqrtsd_mask_round)
8928 (__builtin_ia32_sqrtss_mask_round): New builtins.
8929 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
8930 (__builtin_ia32_sqrtss_round): Remove.
8931 (__builtin_ia32_sqrtsd_mask_round)
8932 (__builtin_ia32_sqrtss_mask_round): New builtins.
8933 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
8934 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
8935 ((match_operand:VF_128 1 "vector_operand"
8936 "xBm,<round_constraint>")): Changed to ...
8937 ((match_operand:VF_128 1 "vector_operand"
8938 "xBm,<round_scalar_constraint>")): ... this.
8939 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
8940 %0, %2, %<iptr>1<round_op3>}): Changed to ...
8941 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
8942 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
8943 %<iptr>1<round_scalar_mask_op3>}): ... this.
8944 ((set_attr "prefix" "<round_prefix>")): Changed to ...
8945 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
8947 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
8950 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
8951 Cast vec_cmpeq result to correct type.
8952 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
8953 Cast vec_cmpgt result to correct type.
8955 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
8957 * final.c (final_scan_insn_1): Renamed from...
8958 (final_scan_insn): ... this. New wrapper, to recover
8959 seen from the outermost call in recursive ones.
8960 * config/sparc/sparc.c (output_return): Drop seen from call.
8961 (output_sibcall): Likewise.
8962 * config/visium/visium.c (output_branch): Likewise.
8964 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
8966 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
8969 2018-02-10 Alan Modra <amodra@gmail.com>
8972 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
8973 Specify LR as an input.
8975 2018-02-10 Jakub Jelinek <jakub@redhat.com>
8978 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
8979 remove_member_access_dummy_vars): New functions.
8980 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
8981 lower_omp_1, execute_lower_omp): Use them.
8983 PR rtl-optimization/84308
8984 * shrink-wrap.c (spread_components): Release todo vector.
8986 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
8988 PR rtl-optimization/57193
8989 * ira-color.c (struct allocno_color_data): Add member
8990 conflict_allocno_hard_prefs.
8991 (update_conflict_allocno_hard_prefs): New.
8992 (bucket_allocno_compare_func): Add a preference based on
8993 conflict_allocno_hard_prefs.
8994 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
8995 (color_allocnos): Remove a dead code. Initiate
8996 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
8998 2018-02-09 Jakub Jelinek <jakub@redhat.com>
9001 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
9002 constraint from =wa to wa. Avoid a subreg on the output operand,
9003 instead use a pseudo and subreg it in a move.
9004 (p9_xxbrd_<mode>): Changed to ...
9005 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
9006 (p9_xxbrd_v2df): New expander.
9007 (p9_xxbrw_<mode>): Changed to ...
9008 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
9009 (p9_xxbrw_v4sf): New expander.
9011 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
9013 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
9015 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
9018 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
9019 multiply in 32-bit mode.
9020 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
9021 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
9024 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
9026 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
9027 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
9028 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
9029 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
9031 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
9034 * dwarf2out.c (is_trivial_indirect_ref): New function.
9035 (dwarf2out_late_global_decl): Do not generate a location
9036 attribute for variables that have a non-trivial DECL_VALUE_EXPR
9037 and that are not defined in the current unit.
9039 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
9041 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
9042 instead of a libcall for UNORDERED.
9044 2018-02-09 Tamar Christina <tamar.christina@arm.com>
9047 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
9048 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
9050 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9053 * config/s390/s390.c (s390_set_current_function): Invoke
9054 s390_indirect_branch_settings also if fndecl didn't change.
9056 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
9058 * config/rs6000/rs6000.md (blockage): Set length to zero.
9060 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
9062 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
9064 2018-02-09 Jakub Jelinek <jakub@redhat.com>
9067 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
9068 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
9072 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
9073 PARALLEL incoming that failed vt_get_decl_and_offset check.
9076 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
9077 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
9079 (get_variable_section): For decls in named .bss* sections pass true as
9080 second argument to bss_initializer_p.
9082 2018-02-09 Marek Polacek <polacek@redhat.com>
9083 Jakub Jelinek <jakub@redhat.com>
9086 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
9087 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
9088 Sync some changes from cxx_fold_indirect_ref.
9090 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
9092 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
9094 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
9095 (BLOCK_INLINE_ENTRY_LABEL): New.
9096 (dwarf2out_var_location): Disregard inline entry markers.
9097 (inline_entry_data): New struct.
9098 (inline_entry_data_hasher): New hashtable type.
9099 (inline_entry_data_hasher::hash): New.
9100 (inline_entry_data_hasher::equal): New.
9101 (inline_entry_data_table): New variable.
9102 (add_high_low_attributes): Add DW_AT_entry_pc and
9103 DW_AT_GNU_entry_view attributes if a pending entry is found
9104 in inline_entry_data_table. Add old entry_pc attribute only
9105 if debug nonbinding markers are disabled.
9106 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
9107 markers are enabled.
9108 (block_within_block_p, dwarf2out_inline_entry): New.
9109 (dwarf2out_finish): Check that no entries remained in
9110 inline_entry_data_table.
9111 * final.c (reemit_insn_block_notes): Handle inline entry notes.
9112 (final_scan_insn, notice_source_line): Likewise.
9113 (rest_of_clean_state): Skip inline entry markers.
9114 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
9116 * gimple.c (gimple_build_debug_inline_entry): New.
9117 * gimple.h (enum gimple_debug_subcode): Add
9118 GIMPLE_DEBUG_INLINE_ENTRY.
9119 (gimple_build_debug_inline_entry): Declare.
9120 (gimple_debug_inline_entry_p): New.
9121 (gimple_debug_nonbind_marker_p): Adjust.
9122 * insn-notes.def (INLINE_ENTRY): New.
9123 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
9124 inline entry marker notes.
9125 (print_insn): Likewise.
9126 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
9127 (INSN_DEBUG_MARKER_KIND): Likewise.
9128 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
9129 * tree-inline.c (expand_call_inline): Build and insert
9130 debug_inline_entry stmt.
9131 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
9132 inline entry blocks early, if nonbind markers are enabled.
9133 (dump_scope_block): Dump fragment info.
9134 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
9135 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
9136 (gimple_build_debug_inline_entry): New.
9137 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
9138 Enable/disable inline entry points too.
9139 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
9140 (DEBUG_INSN): Describe inline entry markers.
9142 * common.opt (gvariable-location-views): New.
9143 (gvariable-location-views=incompat5): New.
9144 * config.in: Rebuilt.
9145 * configure: Rebuilt.
9146 * configure.ac: Test assembler for view support.
9147 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
9148 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
9149 * dwarf2out.c (var_loc_view): New typedef.
9150 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
9151 (dwarf2out_locviews_in_attribute): New.
9152 (dwarf2out_locviews_in_loclist): New.
9153 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
9154 (enum dw_line_info_opcode): Add LI_adv_address.
9155 (struct dw_line_info_table): Add view.
9156 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
9157 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
9158 (zero_view_p): New variable.
9159 (ZERO_VIEW_P): New macro.
9160 (output_asm_line_debug_info): New.
9161 (struct var_loc_node): Add view.
9162 (add_AT_view_list, AT_loc_list): New.
9163 (add_var_loc_to_decl): Add view param. Test it against last.
9164 (new_loc_list): Add view params. Record them.
9165 (AT_loc_list_ptr): Handle loc and view lists.
9166 (view_list_to_loc_list_val_node): New.
9167 (print_dw_val): Handle dw_val_class_view_list.
9168 (size_of_die): Likewise.
9169 (value_format): Likewise.
9170 (loc_list_has_views): New.
9171 (gen_llsym): Set vl_symbol too.
9172 (maybe_gen_llsym, skip_loc_list_entry): New.
9173 (dwarf2out_maybe_output_loclist_view_pair): New.
9174 (output_loc_list): Output view list or entries too.
9175 (output_view_list_offset): New.
9176 (output_die): Handle dw_val_class_view_list.
9177 (output_dwarf_version): New.
9178 (output_compilation_unit_header): Use it.
9179 (output_skeleton_debug_sections): Likewise.
9180 (output_rnglists, output_line_info): Likewise.
9181 (output_pubnames, output_aranges): Update version comments.
9182 (output_one_line_info_table): Output view numbers in asm comments.
9183 (dw_loc_list): Determine current endview, pass it to new_loc_list.
9184 Call maybe_gen_llsym.
9185 (loc_list_from_tree_1): Adjust.
9186 (add_AT_location_description): Create view list attribute if
9187 needed, check it's absent otherwise.
9188 (convert_cfa_to_fb_loc_list): Adjust.
9189 (maybe_emit_file): Call output_asm_line_debug_info for test.
9190 (dwarf2out_var_location): Reset views as needed. Precompute
9191 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
9192 attribute. Set view.
9193 (new_line_info_table): Reset next view.
9194 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
9195 (dwarf2out_source_line): Likewise. Output view resets and labels to
9196 the assembler, or select appropriate line info opcodes.
9197 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
9198 (optimize_string_length): Catch it. Adjust.
9199 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
9200 dw_val_class_view_list, and remove it if no longer needed.
9201 (hash_loc_list): Hash view numbers.
9202 (loc_list_hasher::equal): Compare them.
9203 (optimize_location_lists): Check whether a view list symbol is
9204 needed, and whether the locview attribute is present, and
9205 whether they match. Remove the locview attribute if no longer
9207 (index_location_lists): Call skip_loc_list_entry for test.
9208 (dwarf2out_finish): Call output_asm_line_debug_info for test.
9209 Use output_dwarf_version.
9210 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
9211 (struct dw_val_node): Add val_view_list.
9212 * final.c (SEEN_NEXT_VIEW): New.
9213 (set_next_view_needed): New.
9214 (clear_next_view_needed): New.
9215 (maybe_output_next_view): New.
9216 (final_start_function): Rename to...
9217 (final_start_function_1): ... this. Take pointer to FIRST,
9218 add SEEN parameter. Emit param bindings in the initial view.
9219 (final_start_function): Reintroduce SEEN-less interface.
9220 (final): Rename to...
9221 (final_1): ... this. Take SEEN parameter. Output final pending
9222 next view at the end.
9223 (final): Reintroduce seen-less interface.
9224 (final_scan_insn): Output pending next view before switching
9225 sections or ending a block. Mark the next view as needed when
9226 outputting variable locations. Notify debug backend of section
9227 changes, and of location view changes.
9228 (rest_of_handle_final): Adjust.
9229 * toplev.c (process_options): Autodetect value for debug variable
9230 location views option. Warn on incompat5 without -gdwarf-5.
9231 * doc/invoke.texi (gvariable-location-views): New.
9232 (gvariable-location-views=incompat5): New.
9233 (gno-variable-location-views): New.
9235 2018-02-08 David Malcolm <dmalcolm@redhat.com>
9237 PR tree-optimization/84136
9238 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
9239 that the result of find_edge is non-NULL.
9241 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
9244 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
9245 storing integer register in SImode. Fix cost of 256 and 512
9246 byte aligned SSE register store.
9248 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
9250 * config/i386/i386.c (ix86_multiplication_cost): Fix
9251 multiplication cost for TARGET_AVX512DQ.
9253 2018-02-08 Marek Polacek <polacek@redhat.com>
9255 PR tree-optimization/84238
9256 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
9259 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9261 PR tree-optimization/84265
9262 * tree-vect-stmts.c (vectorizable_store): Don't treat
9263 VMAT_CONTIGUOUS accesses as grouped.
9264 (vectorizable_load): Likewise.
9266 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9268 PR tree-optimization/81635
9269 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
9270 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
9271 (test_round_for_mask): New functions.
9272 (wide_int_cc_tests): Call test_round_for_mask.
9273 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
9274 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
9275 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
9276 range returned by get_range_info.
9278 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
9281 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
9282 * symtab.c: Include builtins.h
9283 (symtab_node::output_to_lto_symbol_table_p): Move here
9284 from lto-streamer-out.c:output_symbol_p.
9285 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
9286 (output_symbol_p): Move all logic to symtab.c
9287 (produce_symtab): Update.
9289 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9291 * config/s390/s390-opts.h (enum indirect_branch): Define.
9292 * config/s390/s390-protos.h (s390_return_addr_from_memory)
9293 (s390_indirect_branch_via_thunk)
9294 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
9295 (enum s390_indirect_branch_type): Define.
9296 * config/s390/s390.c (struct s390_frame_layout, struct
9297 machine_function): Remove.
9298 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
9299 (indirect_branch_table_label_no, indirect_branch_table_name):
9301 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
9302 (enum s390_indirect_branch_option): Define.
9303 (s390_return_addr_from_memory): New function.
9304 (s390_handle_string_attribute): New function.
9305 (s390_attribute_table): Add new attribute handler.
9306 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
9307 (s390_indirect_branch_via_thunk): New function.
9308 (s390_indirect_branch_via_inline_thunk): New function.
9309 (s390_function_ok_for_sibcall): When jumping via thunk disallow
9310 sibling call optimization for non z10 compiles.
9311 (s390_emit_call): Force indirect branch target to be a single
9312 register. Add r1 clobber for non-z10 compiles.
9313 (s390_emit_epilogue): Emit return jump via return_use expander.
9314 (s390_reorg): Handle JUMP_INSNs as execute targets.
9315 (s390_option_override_internal): Perform validity checks for the
9316 new command line options.
9317 (s390_indirect_branch_attrvalue): New function.
9318 (s390_indirect_branch_settings): New function.
9319 (s390_set_current_function): Invoke s390_indirect_branch_settings.
9320 (s390_output_indirect_thunk_function): New function.
9321 (s390_code_end): Implement target hook.
9322 (s390_case_values_threshold): Implement target hook.
9323 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
9325 * config/s390/s390.h (struct s390_frame_layout)
9326 (struct machine_function): Move here from s390.c.
9327 (TARGET_INDIRECT_BRANCH_NOBP_RET)
9328 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
9329 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
9330 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
9331 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
9332 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
9333 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
9334 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
9335 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
9336 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
9337 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
9338 (mnemonic attribute): Add values which aren't recognized
9340 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
9341 pattern for branch conversion. Fix mnemonic attribute.
9342 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
9343 indirect branch via thunk if requested.
9344 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
9345 ("*indirect_jump"): Disable for branch conversion using out of
9347 ("indirect_jump_via_thunk<mode>_z10")
9348 ("indirect_jump_via_thunk<mode>")
9349 ("indirect_jump_via_inlinethunk<mode>_z10")
9350 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
9351 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
9352 ("casesi_jump_via_inlinethunk<mode>_z10")
9353 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
9354 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
9355 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
9356 ("*indirect2_jump"): Disable for branch conversion.
9357 ("casesi_jump"): Turn into expander and expand patterns for branch
9359 ("return_use"): New expander.
9360 ("*return"): Emit return via thunk and rename it to ...
9361 ("*return<mode>"): ... this one.
9362 * config/s390/s390.opt: Add new options and and enum for the
9365 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9367 * lra-constraints.c (match_reload): Unconditionally use
9368 gen_lowpart_SUBREG, rather than selecting between that
9369 and equivalent gen_rtx_SUBREG code.
9371 2018-02-08 Richard Biener <rguenther@suse.de>
9373 PR tree-optimization/84233
9374 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
9375 changed flag instead of boguously re-using phi_inserted.
9377 2018-02-08 Martin Jambor <mjambor@suse.cz>
9379 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
9380 static local variables.
9382 2018-02-08 Richard Biener <rguenther@suse.de>
9384 PR tree-optimization/84278
9385 * tree-vect-stmts.c (vectorizable_store): When looking for
9386 smaller vector types to perform grouped strided loads/stores
9387 make sure the mode is supported by the target.
9388 (vectorizable_load): Likewise.
9390 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
9392 * config/aarch64/aarch64.c (aarch64_components_for_bb):
9393 Increase LDP/STP opportunities by adding adjacent callee-saves.
9395 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
9397 PR rtl-optimization/84068
9398 PR rtl-optimization/83459
9399 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
9401 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
9403 PR tree-optimization/84224
9404 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
9405 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
9408 2018-02-07 Iain Sandoe <iain@codesourcery.com>
9411 * config/rs6000/altivec.md (*restore_world): Remove LR use.
9412 * config/rs6000/predicates.md (restore_world_operation): Adjust op
9413 count, remove one USE.
9415 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
9417 * doc/install.texi (Configuration): Document the
9418 --with-long-double-format={ibm,ieee} PowerPC configuration
9422 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
9423 Convert from define_expand to be define_insn_and_split. Rework
9424 float/double/_Float128 conversions to QI/HI/SImode to work with
9425 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
9426 conversions to QI/HImode types did a store and then a load to
9427 truncate the value. For conversions to VSX registers, don't split
9428 the insn, instead emit the code directly. Use the code iterator
9429 any_fix to combine signed and unsigned conversions.
9430 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
9431 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
9432 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9433 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
9434 (fix_<mode>di2_hw): Likewise.
9435 (fixuns_<mode>di2_hw): Likewise.
9436 (fix_<mode>si2_hw): Likewise.
9437 (fixuns_<mode>si2_hw): Likewise.
9438 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
9439 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9440 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
9441 fix<uns>_trunc<SFDF:mode>si2_p8.
9442 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
9444 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
9445 (fix<uns>_<mode>_mem): Likewise.
9446 (fctiw<u>z_<mode>_mem): Likewise.
9447 (fix<uns>_<mode>_mem): Likewise.
9448 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
9449 the register allocator from doing a direct move to the GPRs to do
9450 a store, and instead use the ISA 3.0 store byte/half-word from
9451 vector register instruction. For IEEE 128-bit floating point,
9452 also optimize stores of 32-bit ints.
9453 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
9455 2018-02-07 Alan Hayward <alan.hayward@arm.com>
9457 * genextract.c (push_pathstr_operand): New function to support
9459 (walk_rtx): Call push_pathstr_operand.
9460 (print_path): Support [a-zA-Z].
9462 2018-02-07 Richard Biener <rguenther@suse.de>
9464 PR tree-optimization/84037
9465 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
9466 (cse_and_gimplify_to_preheader): Declare.
9467 (vect_get_place_in_interleaving_chain): Likewise.
9468 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9470 (_loop_vec_info::~_loop_vec_info): Delete it.
9471 (cse_and_gimplify_to_preheader): New function.
9472 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
9473 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
9474 (vectorizable_load): Likewise. For grouped stores always base
9475 the IV on the first element.
9476 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
9477 condition before gimplifying.
9479 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9481 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
9482 *DIV_EXPR and *MOD_EXPR.
9484 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
9487 * config/i386/i386.c (ix86_option_override_internal): Mask out
9488 the CF_SET bit when checking -fcf-protection.
9490 2018-02-07 Tom de Vries <tom@codesourcery.com>
9493 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
9496 2018-02-07 Richard Biener <rguenther@suse.de>
9498 PR tree-optimization/84204
9499 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
9502 PR tree-optimization/84205
9503 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
9504 special-case isl_ast_op_zdiv_r.
9506 PR tree-optimization/84223
9507 * graphite-scop-detection.c (gather_bbs::before_dom_children):
9508 Only add conditions from within the region.
9509 (gather_bbs::after_dom_children): Adjust.
9511 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
9514 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
9515 * config/avr/avr.md: Only post-reload split REG-REG moves if
9516 either register is GENERAL_REG_P.
9518 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9520 PR tree-optimization/84235
9521 * tree-ssa-scopedtables.c
9522 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
9523 if the subtraction is performed in floating point type where NaNs are
9524 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
9525 build 1. Formatting fix.
9527 2018-02-06 Jakub Jelinek <jakub@redhat.com>
9530 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
9531 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
9532 and skip it regardless of bb boundaries. Use CALL_P macro,
9533 don't test INSN_P (insn) together with CALL_P or JUMP_P check
9534 unnecessarily, formatting fix.
9536 2018-02-06 Michael Collison <michael.collison@arm.com>
9538 * config/arm/thumb2.md:
9539 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
9540 (*thumb_mov_notscc): Ditto.
9542 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
9545 * config/rs6000/rs6000.md (su code attribute): Use "u" for
9546 unsigned_fix, not "s".
9548 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9550 * configure.ac (gcc_fn_eh_frame_ro): New function.
9551 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
9552 correct .eh_frame permissions.
9553 * configure: Regenerate.
9555 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
9557 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
9560 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9562 * config/rs6000/rs6000.c (rs6000_option_override_internal):
9563 Display warning message for -mno-speculate-indirect-jumps.
9565 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
9567 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
9569 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
9571 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
9573 PR tree-optimization/84225
9574 * tree-eh.c (find_trapping_overflow): Only call
9575 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
9577 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9580 * config/i386/i386.c: Reimplement the check of possible options
9581 -mibt/-mshstk conbination. Change error messages.
9582 * doc/invoke.texi: Fix a typo: remove extra '='.
9584 2018-02-06 Marek Polacek <polacek@redhat.com>
9586 PR tree-optimization/84228
9587 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
9589 2018-02-06 Tamar Christina <tamar.christina@arm.com>
9592 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
9593 emitted arch directives.
9594 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
9595 __ARM_FEATURE_COPROC before changing architectures.
9597 2018-02-06 Richard Biener <rguenther@suse.de>
9599 * config/i386/i386.c (print_reg): Fix typo.
9600 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
9602 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
9604 * configure: Regenerate.
9606 2018-02-05 Martin Sebor <msebor@redhat.com>
9608 PR tree-optimization/83369
9609 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
9612 2018-02-05 Martin Liska <mliska@suse.cz>
9614 * doc/invoke.texi: Cherry-pick upstream r323995.
9616 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
9618 * ira.c (ira_init_register_move_cost): Adjust comment.
9620 2018-02-05 Martin Liska <mliska@suse.cz>
9622 PR gcov-profile/84137
9623 * doc/gcov.texi: Fix typo in documentation.
9625 2018-02-05 Martin Liska <mliska@suse.cz>
9627 PR gcov-profile/83879
9628 * doc/gcov.texi: Document necessity of --dynamic-list-data when
9629 using dlopen functionality.
9631 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
9633 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
9634 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
9635 _mm_maskz_range_ss, _mm_mask_range_round_ss,
9636 _mm_maskz_range_round_ss): New intrinsics.
9637 (__builtin_ia32_rangesd128_round)
9638 (__builtin_ia32_rangess128_round): Remove.
9639 (__builtin_ia32_rangesd128_mask_round,
9640 __builtin_ia32_rangess128_mask_round): New builtins.
9641 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
9642 __builtin_ia32_rangess128_round): Remove.
9643 (__builtin_ia32_rangesd128_mask_round,
9644 __builtin_ia32_rangess128_mask_round): New builtins.
9645 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
9646 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
9647 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
9648 "<round_saeonly_constraint>")): Changed to ...
9649 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
9650 "<round_saeonly_scalar_constraint>")): ... this.
9651 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
9652 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
9653 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
9654 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
9655 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
9657 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
9659 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
9661 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
9662 Remove all values except native, 8540 and 8548.
9664 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
9666 * config/i386/i386.c (ix86_output_function_return): Pass
9667 INVALID_REGNUM, instead of -1, as invalid register number to
9668 indirect_thunk_name and output_indirect_thunk.
9670 2018-02-02 Julia Koval <julia.koval@intel.com>
9672 * config.gcc: Add -march=icelake.
9673 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
9674 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
9675 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
9676 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
9677 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
9678 (processor_target_table): Add icelake.
9679 (ix86_option_override_internal): Handle new PTAs.
9680 (get_builtin_code_for_version): Handle icelake.
9681 (M_INTEL_COREI7_ICELAKE): New.
9682 (fold_builtin_cpu): Handle icelake.
9683 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
9684 * doc/invoke.texi: Add -march=icelake.
9686 2018-02-02 Julia Koval <julia.koval@intel.com>
9688 * config/i386/i386.c (ix86_option_override_internal): Change flags type
9689 to wide_int_bitmask.
9690 * wide-int-bitmask.h: New.
9692 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9695 * config/i386/i386.md: Replace Pmode with word_mode in
9696 builtin_setjmp_setup and builtin_longjmp to support x32.
9698 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
9702 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
9704 (rs6000_supported_cpu_names): New static variable.
9705 (linux_cpu_translation_table): Likewise.
9706 (elf_platform) <cpu>: Define new static variable and use it.
9707 Translate kernel AT_PLATFORM name to canonical name if needed.
9708 Error if platform name is unknown.
9710 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
9713 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
9715 2018-02-01 Jeff Law <law@redhat.com>
9718 * config/i386/i386.c (release_scratch_register_on_entry): Add new
9719 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
9720 the scratch if RELEASE_VIA_POP is false.
9721 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
9722 If we have to save a temporary register, decrement SIZE appropriately.
9723 Pass new arguments to release_scratch_register_on_entry.
9724 (ix86_adjust_stack_and_probe): Likewise.
9725 (ix86_emit_probe_stack_range): Pass new arguments to
9726 release_scratch_register_on_entry.
9728 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
9730 PR rtl-optimization/84157
9731 * combine.c (change_zero_ext): Use REG_P predicate in
9732 front of HARD_REGISTER_P predicate.
9734 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
9736 * config/avr/avr.c (avr_option_override): Move disabling of
9737 -fdelete-null-pointer-checks to...
9738 * common/config/avr/avr-common.c (avr_option_optimization_table):
9741 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9743 PR tree-optimization/81635
9744 * tree-data-ref.c (split_constant_offset_1): For types that
9745 wrap on overflow, try to use range info to prove that wrapping
9748 2018-02-01 Renlin Li <renlin.li@arm.com>
9751 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
9753 (aarch64_register_move_cost): Likewise.
9754 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
9756 (REG_CLASS_NAMES): Likewise.
9757 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
9758 TAILCALL_ADDR_REGS. Remove IP registers.
9759 * config/aarch64/aarch64.md (Ucs): Update register constraint.
9761 2018-02-01 Richard Biener <rguenther@suse.de>
9763 * domwalk.h (dom_walker::dom_walker): Add additional constructor
9764 for specifying RPO order and allow NULL for that.
9765 * domwalk.c (dom_walker::dom_walker): Likewise.
9766 (dom_walker::walk): Handle NULL RPO order.
9767 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
9769 (rewrite_update_dom_walker): Likewise.
9770 (mark_def_dom_walker): Likewise.
9772 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9774 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
9775 (aarch64_maybe_expand_sve_subreg_move): Declare.
9776 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
9777 * config/aarch64/predicates.md (aarch64_any_register_operand): New
9779 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
9780 that are semantically a reverse operation.
9781 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
9782 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
9783 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
9785 (aarch64_can_change_mode_class): For big-endian, forbid changes
9786 between two SVE modes if they have different element sizes.
9788 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9790 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
9791 the TImode handling for big-endian targets.
9793 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9795 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
9796 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
9798 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
9799 Remove BSWAP handing for big-endian targets and use the form of
9800 LD1RQ appropariate for the mode.
9802 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9804 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
9805 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
9808 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9811 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
9812 check for operands that need to go through aarch64_sve_reload_be.
9814 2018-02-01 Jakub Jelinek <jakub@redhat.com>
9816 PR tree-optimization/81661
9817 PR tree-optimization/84117
9818 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
9819 * tree-eh.c: Include gimplify.h.
9820 (find_trapping_overflow, replace_trapping_overflow,
9821 rewrite_to_non_trapping_overflow): New functions.
9822 * tree-vect-loop.c: Include tree-eh.h.
9823 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
9824 * tree-data-ref.c: Include tree-eh.h.
9825 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
9827 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
9829 PR rtl-optimization/84123
9830 * combine.c (change_zero_ext): Check if hard register satisfies
9831 can_change_dest_mode before calling gen_lowpart_SUBREG.
9833 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
9836 * ira.c (ira_init_register_move_cost): Remove assert.
9838 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9840 PR rtl-optimization/84071
9841 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
9842 * doc/tm.texi: Regenerate.
9844 2018-01-31 Richard Biener <rguenther@suse.de>
9846 PR tree-optimization/84132
9847 * tree-data-ref.c (analyze_miv_subscript): Properly
9848 check whether evolution_function_is_affine_multivariate_p
9849 before calling gcd_of_steps_may_divide_p.
9851 2018-01-31 Julia Koval <julia.koval@intel.com>
9854 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
9855 * config/i386/i386.md (rdpid_rex64) New.
9856 (rdpid): Make 32bit only.
9858 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
9861 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
9862 an IDENTIFIER_NODE for FUNCTION_TYPE's.
9864 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9867 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9869 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9871 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9873 PR rtl-optimization/84071
9874 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
9875 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
9877 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9879 * config/arc/arc.c (arc_handle_aux_attribute): New function.
9880 (arc_attribute_table): Add 'aux' attribute.
9881 (arc_in_small_data_p): Consider aux like variables.
9882 (arc_is_aux_reg_p): New function.
9883 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
9884 (arc_get_aux_arg): New function.
9885 (prepare_move_operands): Handle aux-register access.
9886 (arc_handle_aux_attribute): New function.
9887 * doc/extend.texi (ARC Variable attributes): Add subsection.
9889 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9891 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
9892 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
9893 (arc_attribute_table): Add 'uncached' attribute.
9894 (arc_print_operand): Print '.di' flag for uncached memory
9896 (arc_in_small_data_p): Do not consider for small data the uncached
9898 (arc_is_uncached_mem_p): New function.
9899 * config/arc/predicates.md (compact_store_memory_operand): Check
9900 for uncached memory accesses.
9901 (nonvol_nonimm_operand): Likewise.
9902 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
9904 2018-01-31 Jakub Jelinek <jakub@redhat.com>
9907 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
9908 falign-loops=): Add Optimization flag.
9910 2018-01-30 Jeff Law <law@redhat.com>
9913 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
9914 INT_REGISTERS_SAVED. Check it prior to calling
9915 get_scratch_register_on_entry.
9916 (ix86_adjust_stack_and_probe): Similarly.
9917 (ix86_emit_probe_stack_range): Similarly.
9918 (ix86_expand_prologue): Corresponding changes.
9920 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9923 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
9924 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
9926 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
9929 * lra-constraints.c (curr_insn_transform): Process AND in the
9932 2018-01-30 Jakub Jelinek <jakub@redhat.com>
9934 PR rtl-optimization/83986
9935 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
9936 dependence against last_pending_memory_flush in addition to
9939 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
9941 PR tree-optimization/81611
9942 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
9945 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9948 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
9951 2018-01-30 Richard Biener <rguenther@suse.de>
9952 Jakub Jelinek <jakub@redhat.com>
9954 PR tree-optimization/84111
9955 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
9956 inner loops added during recursion, as they don't have up-to-date
9959 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9962 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
9963 (can_inline_edge_by_limits_p): ... here.
9964 (can_early_inline_edge_p, check_callers,
9965 update_caller_keys, update_callee_keys, recursive_inlining,
9966 add_new_edges_to_heap, speculation_useful_p,
9967 inline_small_functions,
9968 inline_small_functions, flatten_function,
9969 inline_to_all_callers_1): Update.
9971 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9973 * profile-count.c (profile_count::combine_with_ipa_count): Handle
9976 2018-01-30 Richard Biener <rguenther@suse.de>
9978 PR tree-optimization/83008
9979 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
9980 invariant and constant vector uses in stmts when they need
9983 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9986 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
9987 * configure: Regenerate.
9989 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9991 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
9993 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
9994 Use gen_rtx_REG rather than gen_lowpart.
9996 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9998 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9999 rather than 0 when creating partial subregs.
10001 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
10003 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
10006 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
10009 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
10010 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
10011 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
10012 flags. This restores the settings used before the 2017-07-24.
10013 Turning off pre increment/decrement/modify allows IVOPTS to
10014 optimize DF/SF loops where the index is an int.
10016 2018-01-29 Richard Biener <rguenther@suse.de>
10017 Kelvin Nilsen <kelvin@gcc.gnu.org>
10020 * tree-vect-stmts.c (vectorizable_call): Don't call
10021 targetm.vectorize_builtin_md_vectorized_function if callee is
10024 2018-01-22 Carl Love <cel@us.ibm.com>
10026 * doc/extend.tex: Fix typo in second arg in
10027 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
10029 2018-01-29 Richard Biener <rguenther@suse.de>
10031 PR tree-optimization/84086
10032 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
10033 (flush_ssaname_freelist): When SSA names were released reset
10034 the SCEV hash table.
10036 2018-01-29 Richard Biener <rguenther@suse.de>
10038 PR tree-optimization/84057
10039 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
10040 removed paths when removing edges.
10042 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
10044 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
10045 -mfunction-return=@var{choice}.
10047 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
10049 PR diagnostic/84034
10050 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
10051 Handle CR like TAB.
10052 (layout::print_source_line): Likewise.
10053 (test_get_line_width_without_trailing_whitespace): Add test cases.
10055 2018-01-27 Jakub Jelinek <jakub@redhat.com>
10057 PR middle-end/84040
10058 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
10061 2018-01-26 Jim Wilson <jimw@sifive.com>
10063 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
10065 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
10068 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10070 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
10071 and CMP + SUB-immediate -> SUBS.
10073 2018-01-26 Martin Sebor <msebor@redhat.com>
10075 PR tree-optimization/83896
10076 * tree-ssa-strlen.c (get_string_len): Rename...
10077 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
10078 Avoid assuming length is constant.
10079 (handle_char_store): Use HOST_WIDE_INT for string length.
10081 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
10084 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
10085 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
10087 2018-01-26 Richard Biener <rguenther@suse.de>
10089 PR rtl-optimization/84003
10090 * dse.c (record_store): Only record redundant stores when
10091 the earlier store aliases at least all accesses the later one does.
10093 2018-01-26 Jakub Jelinek <jakub@redhat.com>
10095 PR rtl-optimization/83985
10096 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
10097 REG_CFA_RESTORE insns.
10098 (delete_unmarked_insns): Don't ignore separate shrink wrapping
10099 REG_CFA_RESTORE insns here.
10102 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
10103 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
10105 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10107 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
10108 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
10109 (arc_init): Likewise.
10110 (arc_override_options): Likewise.
10111 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
10113 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
10115 * config/arc/arc.h (TARGET_DBNZ): Define.
10116 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
10117 properly set the tune attribute.
10118 (dbnz): Use TARGET_DBNZ guard.
10119 * config/arc/arc.opt (mtune): Add core3 option.
10121 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10123 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
10124 recognize new pic like addresses.
10125 (arc_delegitimize_address): Clean up.
10127 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10129 * config/arc/arc-arches.def: Option mrf16 valid for all
10131 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
10132 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
10133 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
10134 * config/arc/arc-tables.opt: Regenerate.
10135 * config/arc/arc.c (arc_conditional_register_usage): Handle
10136 reduced register file case.
10137 (arc_file_start): Set must have build attributes.
10138 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
10139 mrf16 option value.
10140 * config/arc/arc.opt (mrf16): Add new option.
10141 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
10142 * config/arc/genmultilib.awk: Handle new mrf16 option.
10143 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
10144 * config/arc/t-multilib: Regenerate.
10145 * doc/invoke.texi (ARC Options): Document mrf16 option.
10147 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10149 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
10150 * config/arc/arc.c (arc_handle_secure_attribute): New function.
10151 (arc_attribute_table): Add 'secure_call' attribute.
10152 (arc_print_operand): Print secure call operand.
10153 (arc_function_ok_for_sibcall): Don't optimize tail calls when
10155 (arc_is_secure_call_p): New function. * config/arc/arc.md
10156 (call_i): Add support for sjli instruction.
10157 (call_value_i): Likewise.
10158 * config/arc/constraints.md (Csc): New constraint.
10160 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10161 John Eric Martin <John.Martin@emmicro-us.com>
10163 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
10164 * config/arc/arc.c (_arc_jli_section): New struct.
10165 (arc_jli_section): New type.
10166 (rc_jli_sections): New static variable.
10167 (arc_handle_jli_attribute): New function.
10168 (arc_attribute_table): Add jli_always and jli_fixed attribute.
10169 (arc_file_end): New function.
10170 (TARGET_ASM_FILE_END): Define.
10171 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
10172 (arc_add_jli_section): New function.
10173 (jli_call_scan): Likewise.
10174 (arc_reorg): Call jli_call_scan.
10175 (arc_output_addsi): Remove 'S' from printing asm operand.
10176 (arc_is_jli_call_p): New function.
10177 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
10179 (movhi_insn): Likewise.
10180 (movsi_insn): Likewise.
10181 (movsi_set_cc_insn): Likewise.
10182 (loadqi_update): Likewise.
10183 (load_zeroextendqisi_update): Likewise.
10184 (load_signextendqisi_update): Likewise.
10185 (loadhi_update): Likewise.
10186 (load_zeroextendhisi_update): Likewise.
10187 (load_signextendhisi_update): Likewise.
10188 (loadsi_update): Likewise.
10189 (loadsf_update): Likewise.
10190 (movsicc_insn): Likewise.
10191 (bset_insn): Likewise.
10192 (bxor_insn): Likewise.
10193 (bclr_insn): Likewise.
10194 (bmsk_insn): Likewise.
10195 (bicsi3_insn): Likewise.
10196 (cmpsi_cc_c_insn): Likewise.
10197 (movsi_ne): Likewise.
10198 (movsi_cond_exec): Likewise.
10199 (clrsbsi2): Likewise.
10200 (norm_f): Likewise.
10210 (call_i): Remove 'S' asm letter, add jli instruction.
10211 (call_value_i): Likewise.
10212 * config/arc/arc.op (mjli-always): New option.
10213 * config/arc/constraints.md (Cji): New constraint.
10214 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
10216 (subsf3_fpx): Likewise.
10217 (mulsf3_fpx): Likewise.
10218 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
10220 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
10221 function attrbutes.
10222 * doc/invoke.texi (ARC): Document mjli-always option.
10224 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
10226 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
10227 avoid addition with 0 and use incw and decw where possible.
10229 2018-01-26 Richard Biener <rguenther@suse.de>
10231 PR tree-optimization/81082
10232 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
10233 association if it requires casting to unsigned.
10234 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
10235 from fold_plusminus_mult_expr to catch important cases late when
10236 range info is available.
10238 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10240 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
10241 * configure.ac (hidden_linkonce): New test.
10242 * configure: Regenerate.
10243 * config.in: Regenerate.
10245 2018-01-26 Julia Koval <julia.koval@intel.com>
10247 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
10248 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
10249 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
10250 _mm_mask_bitshuffle_epi64_mask): Fix type.
10251 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
10252 USI_FTYPE_V4DI_V4DI_USI): Remove.
10253 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
10254 __builtin_ia32_vpshufbitqmb256_mask,
10255 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
10256 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
10257 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
10259 2018-01-26 Alan Modra <amodra@gmail.com>
10262 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
10263 UNSPEC_VBPERMQ. Sort other unspecs.
10265 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
10267 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
10269 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
10271 PR middle-end/83055
10272 * predict.c (drop_profile): Do not push/pop cfun; update also
10274 (handle_missing_profiles): Fix logic looking for zero profiles.
10276 2018-01-25 Jakub Jelinek <jakub@redhat.com>
10278 PR middle-end/83977
10279 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
10280 on functions with #pragma omp declare simd or functions with simd
10282 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
10283 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
10284 Remove trailing \n from warning_at calls.
10286 2018-01-25 Tom de Vries <tom@codesourcery.com>
10289 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
10290 for neutered workers.
10292 2018-01-24 Joseph Myers <joseph@codesourcery.com>
10295 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
10296 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
10298 2018-01-24 Jeff Law <law@redhat.com>
10301 * i386.c (get_probe_interval): Move to earlier point.
10302 (ix86_compute_frame_layout): If -fstack-clash-protection and
10303 the frame is larger than the probe interval, then use pushes
10304 to save registers rather than reg->mem moves.
10305 (ix86_expand_prologue): Remove conditional for int_registers_saved
10308 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
10311 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
10312 min/max for never referenced object.
10314 2018-01-24 Jakub Jelinek <jakub@redhat.com>
10316 PR middle-end/83977
10317 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
10319 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
10320 attributes from DECL_ATTRIBUTES (decl) without affecting
10321 DECL_ATTRIBUTES (current_function_decl).
10322 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
10323 functions with non-NULL DECL_ABSTRACT_ORIGIN.
10325 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
10327 PR tree-optimization/83979
10328 * fold-const.c (fold_comparison): Use constant_boolean_node
10329 instead of boolean_{true,false}_node.
10331 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
10333 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
10336 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10338 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10339 Simplify the clause that sets the length attribute.
10340 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10341 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
10342 clause that sets the length attribute.
10343 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10345 2018-01-24 Tom de Vries <tom@codesourcery.com>
10348 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
10349 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
10350 Add strict parameter.
10351 (prevent_branch_around_nothing): Insert dummy insn between branch to
10352 label and label with no ptx insn inbetween.
10353 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
10355 2018-01-24 Tom de Vries <tom@codesourcery.com>
10358 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
10359 for neutered threads in warp.
10360 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
10362 2018-01-24 Richard Biener <rguenther@suse.de>
10364 PR tree-optimization/83176
10365 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
10368 2018-01-24 Richard Biener <rguenther@suse.de>
10370 PR tree-optimization/82819
10371 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
10372 code generating pluses that are no-ops in the target precision.
10374 2018-01-24 Richard Biener <rguenther@suse.de>
10376 PR middle-end/84000
10377 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
10379 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10381 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
10382 to merge probabilities.
10383 * predict.c (probably_never_executed): Also mark as cold functions
10384 with global 0 profile and guessed local profile.
10385 * profile-count.c (profile_probability::combine_with_count): New
10387 * profile-count.h (profile_probability::operator*,
10388 profile_probability::operator*=, profile_probability::operator/,
10389 profile_probability::operator/=): Reduce precision to adjusted
10390 and set value to guessed on contradictory divisions.
10391 (profile_probability::combine_with_freq): Remove.
10392 (profile_probability::combine_wiht_count): Declare.
10393 (profile_count::force_nonzero):: Set to adjusted.
10394 (profile_count::probability_in):: Set quality to adjusted.
10395 * tree-ssa-tail-merge.c (replace_block_by): Use
10396 combine_with_count.
10398 2018-01-23 Andrew Waterman <andrew@sifive.com>
10399 Jim Wilson <jimw@sifive.com>
10401 * config/riscv/riscv.c (riscv_stack_boundary): New.
10402 (riscv_option_override): Set riscv_stack_boundary. Handle
10403 riscv_preferred_stack_boundary_arg.
10404 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
10405 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
10406 (STACK_BOUNDARY): Set to riscv_stack_boundary.
10407 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
10408 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
10409 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
10411 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
10414 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
10415 of struct ix86_frame.
10416 (ix86_expand_epilogue): Likewise. Add a local variable for
10417 the reg_save_offset field in struct ix86_frame.
10419 2018-01-23 Bin Cheng <bin.cheng@arm.com>
10421 PR tree-optimization/82604
10422 * tree-loop-distribution.c (enum partition_kind): New enum item
10423 PKIND_PARTIAL_MEMSET.
10424 (partition_builtin_p): Support above new enum item.
10425 (generate_code_for_partition): Ditto.
10426 (compute_access_range): Differentiate cases that equality can be
10427 proven at all loops, the innermost loops or no loops.
10428 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
10429 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
10430 (finalize_partitions, distribute_loop): Don't fuse partition of
10431 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
10432 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
10433 parloop is enabled.
10435 2018-01-23 Martin Liska <mliska@suse.cz>
10437 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
10438 order to ignore the predictor.
10439 (PRED_POLYMORPHIC_CALL): Likewise.
10440 (PRED_RECURSIVE_CALL): Likewise.
10442 2018-01-23 Martin Liska <mliska@suse.cz>
10444 * tree-profile.c (tree_profiling): Print function header to
10445 aware reader which function we are working on.
10446 * value-prof.c (gimple_find_values_to_profile): Do not print
10447 not interesting value histograms.
10449 2018-01-23 Martin Liska <mliska@suse.cz>
10451 * profile-count.h (enum profile_quality): Add
10452 profile_uninitialized as the first value. Do not number values
10453 as they are zero based.
10454 (profile_count::verify): Update sanity check.
10455 (profile_probability::verify): Likewise.
10457 2018-01-23 Nathan Sidwell <nathan@acm.org>
10459 * doc/invoke.texi (ffor-scope): Deprecate.
10461 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10463 PR tree-optimization/83510
10464 * domwalk.c (set_all_edges_as_executable): New function.
10465 (dom_walker::dom_walker): Convert bool param
10466 "skip_unreachable_blocks" to enum reachability. Move setup of
10467 edge flags to set_all_edges_as_executable and only do it when
10468 reachability is REACHABLE_BLOCKS.
10469 * domwalk.h (enum dom_walker::reachability): New enum.
10470 (dom_walker::dom_walker): Convert bool param
10471 "skip_unreachable_blocks" to enum reachability.
10472 (set_all_edges_as_executable): New decl.
10473 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
10474 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
10476 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
10477 but converting true to REACHABLE_BLOCKS.
10478 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
10480 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
10481 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
10482 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
10484 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
10485 if check_all_array_refs will be called.
10487 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10489 * tree.c (selftest::test_location_wrappers): Add more test
10492 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10494 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
10495 (selftest::test_bit_in_range): Likewise.
10497 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10500 * doc/sourcebuild.texi (vect_float): Say that the selector
10501 only describes the situation when -funsafe-math-optimizations is on.
10502 (vect_float_strict): Document.
10504 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10506 PR tree-optimization/83965
10507 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
10508 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
10509 instead of checking only for a reduction.
10510 (vect_recog_widen_sum_pattern): Likewise.
10512 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10514 * predict.c (probably_never_executed): Only use precise profile info.
10515 (compute_function_frequency): Skip after inlining hack since we now
10516 have quality checking.
10518 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10520 * profile-count.h (profile_probability::very_unlikely,
10521 profile_probability::unlikely, profile_probability::even): Set
10522 precision to guessed.
10524 2018-01-23 Richard Biener <rguenther@suse.de>
10526 PR tree-optimization/83963
10527 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
10528 Properly terminate dominator walk when crossing the exit edge not
10529 when visiting its source block.
10531 2018-01-23 Jakub Jelinek <jakub@redhat.com>
10534 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
10535 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
10537 2018-01-22 Jakub Jelinek <jakub@redhat.com>
10539 PR tree-optimization/83957
10540 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
10541 semicolon after for body surrounded by braces.
10543 PR tree-optimization/83081
10544 * profile-count.h (profile_probability::split): New method.
10545 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
10546 Use profile_probability::split.
10547 (do_compare_rtx_and_jump): Fix adjustment of probabilities
10548 when splitting a single conditional jump into 2.
10550 2018-01-22 David Malcolm <dmalcolm@redhat.com>
10552 PR tree-optimization/69452
10553 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
10556 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10558 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
10559 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
10560 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
10562 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10564 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
10565 * config/rl78/rl78.md (movdi): New define_expand.
10566 * config/rl78/rl78.c (rl78_split_movdi): New function.
10568 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
10571 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
10573 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
10574 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
10575 128-bit to produce an UNSPEC move to get the double word with the
10576 signbit and then a shift directly to do signbit.
10577 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
10578 implementation with a new version that just does either a direct
10579 move or a regular move. Move memory interface to separate insns.
10580 Move insns so they are next to the expander.
10581 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
10582 with signbit move. Split big and little endian case.
10583 (signbit<mode>2_dm_mem_le): Likewise.
10584 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
10585 (signbit<mode>2_dm2): Likewise.
10587 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10589 * config/rl78/rl78.md (anddi3): New define_expand.
10591 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10593 * config/rl78/rl78.md (umindi3): New define_expand.
10595 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10597 * config/rl78/rl78.md (smindi3): New define_expand.
10599 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10601 * config/rl78/rl78.md (smaxdi3): New define_expand.
10603 2018-01-22 Carl Love <cel@us.ibm.com>
10605 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
10606 LVX_V1TI): Add macro expansion.
10607 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
10608 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
10609 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
10610 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
10611 Change check to determine if the instruction is a byte reversing
10612 entry. Fix typo in comment.
10613 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
10614 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
10615 Add def_builtin calls for new builtins.
10616 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
10617 Add define_insn expansion.
10619 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10621 * config/rl78/rl78.md (umaxdi3): New define_expand.
10623 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10625 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
10626 for non-QImode registers.
10628 2018-01-22 Richard Biener <rguenther@suse.de>
10630 PR tree-optimization/83963
10631 * graphite-scop-detection.c (scop_detection::get_sese): Delay
10632 including the loop exit block.
10633 (scop_detection::merge_sese): Likewise.
10634 (scop_detection::add_scop): Do it here instead.
10636 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10638 * doc/sourcebuild.texi (arm_softfloat): Document.
10640 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
10643 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
10644 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
10645 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
10647 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10648 David Edelsohn <dje.gcc@gmail.com>
10651 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10652 Change "crset eq" to "crset 2".
10653 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10654 (*call_indirect_aix<mode>_nospec): Likewise.
10655 (*call_value_indirect_aix<mode>_nospec): Likewise.
10656 (*call_indirect_elfv2<mode>_nospec): Likewise.
10657 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
10658 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
10659 change assembly output from . to $.
10660 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10661 (indirect_jump<mode>_nospec): Change assembly output from . to $.
10662 (*tablejump<mode>_internal1_nospec): Likewise.
10664 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
10667 * config/sh/sh_optimize_sett_clrt.cc:
10668 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
10670 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
10672 PR tree-optimization/83940
10673 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
10674 offset_dt to vect_constant_def rather than vect_unknown_def_type.
10675 (vect_check_load_store_mask): Add a mask_dt_out parameter and
10676 use it to pass back the definition type.
10677 (vect_check_store_rhs): Likewise rhs_dt_out.
10678 (vect_build_gather_load_calls): Add a mask_dt argument and use
10679 it instead of a call to vect_is_simple_use.
10680 (vectorizable_store): Update calls to vect_check_load_store_mask
10681 and vect_check_store_rhs. Use the dt returned by the latter instead
10682 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
10683 instead of calls to vect_is_simple_use. Pass the scalar rather
10684 than the vector operand to vect_is_simple_use when handling
10685 second and subsequent copies of an rhs value.
10686 (vectorizable_load): Update calls to vect_check_load_store_mask
10687 and vect_build_gather_load_calls. Use the cached mask_dt and
10688 gs_info.offset_dt instead of calls to vect_is_simple_use.
10690 2018-01-20 Jakub Jelinek <jakub@redhat.com>
10692 PR middle-end/83945
10693 * tree-emutls.c: Include gimplify.h.
10694 (lower_emutls_2): New function.
10695 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
10696 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
10697 it before further processing.
10700 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
10701 UINTVAL (trueop1) instead of INTVAL (op1).
10703 2018-01-19 Jakub Jelinek <jakub@redhat.com>
10707 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
10708 INCOMING_FRAME_SP_OFFSET if not defined.
10709 (scan_trace): Add ENTRY argument. If true and
10710 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
10711 emit a note to adjust the CFA offset.
10712 (create_cfi_notes): Adjust scan_trace callers.
10713 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
10714 INCOMING_FRAME_SP_OFFSET in the CIE.
10715 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
10716 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
10718 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
10719 * doc/tm.texi: Regenerated.
10721 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10723 PR rtl-optimization/83147
10724 * lra-constraints.c (remove_inheritance_pseudos): Use
10725 lra_substitute_pseudo_within_insn.
10727 2018-01-19 Tom de Vries <tom@codesourcery.com>
10728 Cesar Philippidis <cesar@codesourcery.com>
10731 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
10733 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
10736 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
10737 spaces for function labels.
10739 2018-01-19 Martin Liska <mliska@suse.cz>
10741 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
10742 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
10743 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
10744 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
10745 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
10746 (PRED_CONST_RETURN): Change from 69 to 65.
10747 (PRED_NULL_RETURN): Change from 91 to 71.
10748 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
10749 (PRED_LOOP_GUARD): Change from 66 to 73.
10751 2018-01-19 Martin Liska <mliska@suse.cz>
10753 * predict.c (predict_insn_def): Add new assert.
10754 (struct branch_predictor): Change type to signed integer.
10755 (test_prediction_value_range): Amend test to cover
10756 PROB_UNINITIALIZED.
10757 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
10758 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
10759 (PRED_LOOP_ITERATIONS_MAX): Likewise.
10760 (PRED_LOOP_IV_COMPARE): Likewise.
10761 * predict.h (PROB_UNINITIALIZED): Define new constant.
10763 2018-01-19 Martin Liska <mliska@suse.cz>
10765 * predict.c (dump_prediction): Add new format for
10766 analyze_brprob.py script which is enabled with -details
10768 * profile-count.h (precise_p): New function.
10770 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10772 PR tree-optimization/83922
10773 * tree-vect-loop.c (vect_verify_full_masking): Return false if
10774 there are no statements that need masking.
10775 (vect_active_double_reduction_p): New function.
10776 (vect_analyze_loop_operations): Use it when handling phis that
10777 are not in the loop header.
10779 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10781 PR tree-optimization/83914
10782 * tree-vect-loop.c (vectorizable_induction): Don't convert
10783 init_expr or apply the peeling adjustment for inductions
10784 that are nested within the vectorized loop.
10786 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10788 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
10791 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10795 * function.h (gimplify_parameters): Add gimple_seq * argument.
10796 * function.c: Include gimple.h and options.h.
10797 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
10798 for the added local temporaries if needed.
10799 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
10800 if there are any parameter cleanups, wrap whole body into a
10801 try/finally with the cleanups.
10803 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
10806 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
10807 Use GET_MODE_CLASS for scalar floating point.
10809 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10813 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
10814 Fix call of call_cgraph_insertion_hooks.
10816 2018-01-18 Martin Sebor <msebor@redhat.com>
10818 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
10820 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10823 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
10826 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
10829 * common.opt: (-ffile-prefix-map): New option.
10830 * opts.c (common_handle_option): Defer it.
10831 * opts-global.c (handle_common_deferred_options): Handle it.
10832 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
10833 * file-prefix-map.h: New file.
10834 (remap_debug_filename, add_debug_prefix_map): ...here.
10835 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
10836 * final.c (debug_prefix_map, add_debug_prefix_map
10837 remap_debug_filename): Move to...
10838 * file-prefix-map.c: New file.
10839 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
10840 generalize, get rid of alloca(), use strrchr() instead of strchr().
10841 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
10842 Implement in terms of add_prefix_map().
10843 (remap_macro_filename, remap_debug_filename): Implement in term of
10845 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
10846 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
10847 * dbxout.c: Include file-prefix-map.h.
10848 * varasm.c: Likewise.
10849 * vmsdbgout.c: Likewise.
10850 * xcoffout.c: Likewise.
10851 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
10852 * doc/cppopts.texi (-fmacro-prefix-map): Document.
10853 * doc/invoke.texi (-ffile-prefix-map): Document.
10854 (-fdebug-prefix-map): Update description.
10856 2018-01-18 Martin Liska <mliska@suse.cz>
10858 * config/i386/i386.c (indirect_thunk_name): Document that also
10860 (output_indirect_thunk): Document why both instructions
10861 (pause and lfence) are generated.
10863 2018-01-18 Richard Biener <rguenther@suse.de>
10865 PR tree-optimization/83887
10866 * graphite-scop-detection.c
10867 (scop_detection::get_nearest_dom_with_single_entry): Remove.
10868 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
10869 (scop_detection::merge_sese): Re-implement with a flood-fill
10870 algorithm that properly finds a SESE region if it exists.
10872 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10875 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
10876 pointer_diff optimizations use view_convert instead of convert.
10878 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10880 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10881 Generate different code for -mno-speculate-indirect-jumps.
10882 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10883 (*call_indirect_aix<mode>): Disable for
10884 -mno-speculate-indirect-jumps.
10885 (*call_indirect_aix<mode>_nospec): New define_insn.
10886 (*call_value_indirect_aix<mode>): Disable for
10887 -mno-speculate-indirect-jumps.
10888 (*call_value_indirect_aix<mode>_nospec): New define_insn.
10889 (*sibcall_nonlocal_sysv<mode>): Generate different code for
10890 -mno-speculate-indirect-jumps.
10891 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10893 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
10895 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
10896 long double type, set the flags for noting the default long double
10897 type, even if we don't pass or return a long double type.
10899 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
10902 * ipa-inline.c (flatten_function): Do not overwrite final inlining
10905 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
10907 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
10908 support for merge[hl].
10909 (fold_mergehl_helper): New helper function.
10910 (tree-vector-builder.h): New #include for tree_vector_builder usage.
10911 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
10912 (altivec_vmrglw_direct): Add xxmrglw insn.
10914 2018-01-17 Andrew Waterman <andrew@sifive.com>
10916 * config/riscv/riscv.c (riscv_conditional_register_usage): If
10917 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
10919 2018-01-17 David Malcolm <dmalcolm@redhat.com>
10922 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
10923 call the lto_location_cache before reading the
10924 DECL_SOURCE_LOCATION of the types.
10926 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
10927 Richard Sandiford <richard.sandiford@linaro.org>
10929 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
10930 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
10931 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
10932 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
10933 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
10935 * config/aarch64/constraints.md (aarch64_movti_operand):
10937 * config/aarch64/predicates.md (Uti): Add new constraint.
10939 2018-01-17 Carl Love <cel@us.ibm.com>
10941 * config/rs6000/vsx.md (define_expand xl_len_r,
10942 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
10943 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
10945 (define_expand, define_insn): Move the shift left from the
10946 define_insn to the define_expand for lxvl and stxvl instructions.
10947 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
10948 and XL_LEN_R definitions to PURE.
10950 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
10952 * config/i386/i386.c (indirect_thunk_name): Declare regno
10953 as unsigned int. Compare regno with INVALID_REGNUM.
10954 (output_indirect_thunk): Ditto.
10955 (output_indirect_thunk_function): Ditto.
10956 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
10957 in the call to output_indirect_thunk_function.
10959 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
10961 PR middle-end/83884
10962 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
10963 rather than the size of inner_type to determine the stack slot size
10964 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
10966 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
10969 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
10972 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
10974 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
10975 endian Linux systems to optionally enable multilibs for selecting
10976 the long double type if the user configured an explicit type.
10977 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
10978 have no long double multilibs if not defined.
10979 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
10980 warn if the user used -mabi={ieee,ibm}longdouble and we built
10981 multilibs for long double.
10982 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
10983 appropriate multilib option.
10984 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
10986 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
10987 for building long double multilibs.
10988 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
10990 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
10992 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
10995 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
10997 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
11000 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
11003 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
11006 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
11008 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
11009 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
11011 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
11013 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
11014 different rtl trees depending on TARGET_64BIT.
11015 (rs6000_gen_lvx): Likewise.
11017 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
11019 * config/visium/visium.md (nop): Tweak comment.
11020 (hazard_nop): Likewise.
11022 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11024 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
11025 -mspeculate-indirect-jumps.
11026 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
11027 for -mno-speculate-indirect-jumps.
11028 (*call_indirect_elfv2<mode>_nospec): New define_insn.
11029 (*call_value_indirect_elfv2<mode>): Disable for
11030 -mno-speculate-indirect-jumps.
11031 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
11032 (indirect_jump): Emit different RTL for
11033 -mno-speculate-indirect-jumps.
11034 (*indirect_jump<mode>): Disable for
11035 -mno-speculate-indirect-jumps.
11036 (*indirect_jump<mode>_nospec): New define_insn.
11037 (tablejump): Emit different RTL for
11038 -mno-speculate-indirect-jumps.
11039 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
11040 (tablejumpsi_nospec): New define_expand.
11041 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
11042 (tablejumpdi_nospec): New define_expand.
11043 (*tablejump<mode>_internal1): Disable for
11044 -mno-speculate-indirect-jumps.
11045 (*tablejump<mode>_internal1_nospec): New define_insn.
11046 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
11049 2018-01-16 Artyom Skrobov tyomitch@gmail.com
11051 * caller-save.c (insert_save): Drop unnecessary parameter. All
11054 2018-01-16 Jakub Jelinek <jakub@redhat.com>
11055 Richard Biener <rguenth@suse.de>
11058 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
11059 return early, inline manually is_gimple_sizepos. Make sure if we
11060 call gimplify_expr we don't end up with a gimple constant.
11061 * tree.c (variably_modified_type_p): Don't return true for
11062 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
11063 * gimplify.h (is_gimple_sizepos): Remove.
11065 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11067 PR tree-optimization/83857
11068 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
11069 vectorizable_live_operation for pure SLP statements.
11070 (vectorizable_live_operation): Handle PHIs.
11072 2018-01-16 Richard Biener <rguenther@suse.de>
11074 PR tree-optimization/83867
11075 * tree-vect-stmts.c (vect_transform_stmt): Precompute
11076 nested_in_vect_loop_p since the scalar stmt may get invalidated.
11078 2018-01-16 Jakub Jelinek <jakub@redhat.com>
11081 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
11082 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
11083 If off is not INTEGER_CST, issue a may not be aligned warning
11084 rather than isn't aligned. Use isn%'t rather than isn't.
11085 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
11087 <case MULT_EXPR>: Improve the case when bottom and one of the
11088 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
11089 operand, in that case check if the other operand is multiple of
11090 bottom divided by the INTEGER_CST operand.
11092 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11095 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
11096 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
11097 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
11098 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
11099 * config/pa/pa.c (pa_function_arg_advance): Likewise.
11100 (pa_function_arg, pa_arg_partial_bytes): Likewise.
11101 (pa_function_arg_size): New function.
11103 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11105 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
11106 in a separate statement.
11108 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11110 PR tree-optimization/83847
11111 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
11112 group gathers and scatters.
11114 2018-01-16 Jakub Jelinek <jakub@redhat.com>
11116 PR rtl-optimization/86620
11117 * params.def (max-sched-ready-insns): Bump minimum value to 1.
11119 PR rtl-optimization/83213
11120 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
11121 to last if both are JUMP_INSNs.
11123 PR tree-optimization/83843
11124 * gimple-ssa-store-merging.c
11125 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
11126 store_immediate_info for bswap/nop orig_stores.
11128 2018-01-15 Andrew Waterman <andrew@sifive.com>
11130 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
11132 <UDIV>: Increase cost if !TARGET_DIV.
11134 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
11136 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
11137 (define_attr "cr_logical_3op"): New.
11138 (cceq_ior_compare): Adjust.
11139 (cceq_ior_compare_complement): Adjust.
11140 (*cceq_rev_compare): Adjust.
11141 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
11142 (is_cracked_insn): Adjust.
11143 (insn_must_be_first_in_group): Adjust.
11144 * config/rs6000/40x.md: Adjust.
11145 * config/rs6000/440.md: Adjust.
11146 * config/rs6000/476.md: Adjust.
11147 * config/rs6000/601.md: Adjust.
11148 * config/rs6000/603.md: Adjust.
11149 * config/rs6000/6xx.md: Adjust.
11150 * config/rs6000/7450.md: Adjust.
11151 * config/rs6000/7xx.md: Adjust.
11152 * config/rs6000/8540.md: Adjust.
11153 * config/rs6000/cell.md: Adjust.
11154 * config/rs6000/e300c2c3.md: Adjust.
11155 * config/rs6000/e500mc.md: Adjust.
11156 * config/rs6000/e500mc64.md: Adjust.
11157 * config/rs6000/e5500.md: Adjust.
11158 * config/rs6000/e6500.md: Adjust.
11159 * config/rs6000/mpc.md: Adjust.
11160 * config/rs6000/power4.md: Adjust.
11161 * config/rs6000/power5.md: Adjust.
11162 * config/rs6000/power6.md: Adjust.
11163 * config/rs6000/power7.md: Adjust.
11164 * config/rs6000/power8.md: Adjust.
11165 * config/rs6000/power9.md: Adjust.
11166 * config/rs6000/rs64.md: Adjust.
11167 * config/rs6000/titan.md: Adjust.
11169 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11171 * config/i386/predicates.md (indirect_branch_operand): Rewrite
11172 ix86_indirect_branch_register logic.
11174 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11176 * config/i386/constraints.md (Bs): Update
11177 ix86_indirect_branch_register check. Don't check
11178 ix86_indirect_branch_register with GOT_memory_operand.
11180 * config/i386/predicates.md (GOT_memory_operand): Don't check
11181 ix86_indirect_branch_register here.
11182 (GOT32_symbol_operand): Likewise.
11184 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11186 * config/i386/predicates.md (constant_call_address_operand):
11187 Rewrite ix86_indirect_branch_register logic.
11188 (sibcall_insn_operand): Likewise.
11190 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11192 * config/i386/constraints.md (Bs): Replace
11193 ix86_indirect_branch_thunk_register with
11194 ix86_indirect_branch_register.
11196 * config/i386/i386.md (indirect_jump): Likewise.
11197 (tablejump): Likewise.
11198 (*sibcall_memory): Likewise.
11199 (*sibcall_value_memory): Likewise.
11200 Peepholes of indirect call and jump via memory: Likewise.
11201 * config/i386/i386.opt: Likewise.
11202 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11203 (GOT_memory_operand): Likewise.
11204 (call_insn_operand): Likewise.
11205 (sibcall_insn_operand): Likewise.
11206 (GOT32_symbol_operand): Likewise.
11208 2018-01-15 Jakub Jelinek <jakub@redhat.com>
11210 PR middle-end/83837
11211 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
11212 type rather than type addr's type points to.
11213 (expand_omp_atomic_mutex): Likewise.
11214 (expand_omp_atomic): Likewise.
11216 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11219 * config/i386/i386.c (output_indirect_thunk_function): Use
11220 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
11221 for __x86_return_thunk.
11223 2018-01-15 Richard Biener <rguenther@suse.de>
11225 PR middle-end/83850
11226 * expmed.c (extract_bit_field_1): Fix typo.
11228 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11231 * config/arm/iterators.md (VF): New mode iterator.
11232 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
11233 Remove integer-related logic from pattern.
11234 (neon_vabd<mode>_3): Likewise.
11236 2018-01-15 Jakub Jelinek <jakub@redhat.com>
11238 PR middle-end/82694
11239 * common.opt (fstrict-overflow): No longer an alias.
11240 (fwrapv-pointer): New option.
11241 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
11242 also for pointer types based on flag_wrapv_pointer.
11243 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
11244 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
11245 opts->x_flag_wrapv got set.
11246 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
11247 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
11248 POINTER_TYPE_OVERFLOW_UNDEFINED.
11249 * match.pd: Likewise in address comparison pattern.
11250 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
11252 2018-01-15 Richard Biener <rguenther@suse.de>
11255 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
11256 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
11257 Reset type names to their identifier if their TYPE_DECL doesn't
11258 have linkage (and thus is used for ODR and devirt).
11259 (save_debug_info_for_decl): Remove.
11260 (save_debug_info_for_type): Likewise.
11261 (add_tree_to_fld_list): Adjust.
11262 * tree-pretty-print.c (dump_generic_node): Make dumping of
11263 type names more robust.
11265 2018-01-15 Richard Biener <rguenther@suse.de>
11267 * BASE-VER: Bump to 8.0.1.
11269 2018-01-14 Martin Sebor <msebor@redhat.com>
11272 * builtins.c (check_access): Avoid warning when the no-warning bit
11275 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
11277 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
11278 * ira-color (allocno_hard_regs_compare): Likewise.
11280 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
11283 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
11284 Use .pushsection/.popsection.
11286 2018-01-14 Martin Sebor <msebor@redhat.com>
11289 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
11291 2018-01-14 Jakub Jelinek <jakub@redhat.com>
11293 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
11294 entry from extra_headers.
11295 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
11296 extra_headers, make the list bitwise identical to the i?86-*-* one.
11298 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11300 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
11301 -mcmodel=large with -mindirect-branch=thunk,
11302 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
11303 -mfunction-return=thunk-extern.
11304 * doc/invoke.texi: Document -mcmodel=large is incompatible with
11305 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
11306 -mfunction-return=thunk and -mfunction-return=thunk-extern.
11308 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11310 * config/i386/i386.c (print_reg): Print the name of the full
11311 integer register without '%'.
11312 (ix86_print_operand): Handle 'V'.
11313 * doc/extend.texi: Document 'V' modifier.
11315 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11317 * config/i386/constraints.md (Bs): Disallow memory operand for
11318 -mindirect-branch-register.
11320 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11321 (GOT_memory_operand): Likewise.
11322 (call_insn_operand): Likewise.
11323 (sibcall_insn_operand): Likewise.
11324 (GOT32_symbol_operand): Likewise.
11325 * config/i386/i386.md (indirect_jump): Call convert_memory_address
11326 for -mindirect-branch-register.
11327 (tablejump): Likewise.
11328 (*sibcall_memory): Likewise.
11329 (*sibcall_value_memory): Likewise.
11330 Disallow peepholes of indirect call and jump via memory for
11331 -mindirect-branch-register.
11332 (*call_pop): Replace m with Bw.
11333 (*call_value_pop): Likewise.
11334 (*sibcall_pop_memory): Replace m with Bs.
11335 * config/i386/i386.opt (mindirect-branch-register): New option.
11336 * doc/invoke.texi: Document -mindirect-branch-register option.
11338 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11340 * config/i386/i386-protos.h (ix86_output_function_return): New.
11341 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
11342 set function_return_type.
11343 (indirect_thunk_name): Add ret_p to indicate thunk for function
11345 (output_indirect_thunk_function): Pass false to
11346 indirect_thunk_name.
11347 (ix86_output_indirect_branch_via_reg): Likewise.
11348 (ix86_output_indirect_branch_via_push): Likewise.
11349 (output_indirect_thunk_function): Create alias for function
11350 return thunk if regno < 0.
11351 (ix86_output_function_return): New function.
11352 (ix86_handle_fndecl_attribute): Handle function_return.
11353 (ix86_attribute_table): Add function_return.
11354 * config/i386/i386.h (machine_function): Add
11355 function_return_type.
11356 * config/i386/i386.md (simple_return_internal): Use
11357 ix86_output_function_return.
11358 (simple_return_internal_long): Likewise.
11359 * config/i386/i386.opt (mfunction-return=): New option.
11360 (indirect_branch): Mention -mfunction-return=.
11361 * doc/extend.texi: Document function_return function attribute.
11362 * doc/invoke.texi: Document -mfunction-return= option.
11364 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11366 * config/i386/i386-opts.h (indirect_branch): New.
11367 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
11368 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
11369 with local indirect jump when converting indirect call and jump.
11370 (ix86_set_indirect_branch_type): New.
11371 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
11372 (indirectlabelno): New.
11373 (indirect_thunk_needed): Likewise.
11374 (indirect_thunk_bnd_needed): Likewise.
11375 (indirect_thunks_used): Likewise.
11376 (indirect_thunks_bnd_used): Likewise.
11377 (INDIRECT_LABEL): Likewise.
11378 (indirect_thunk_name): Likewise.
11379 (output_indirect_thunk): Likewise.
11380 (output_indirect_thunk_function): Likewise.
11381 (ix86_output_indirect_branch_via_reg): Likewise.
11382 (ix86_output_indirect_branch_via_push): Likewise.
11383 (ix86_output_indirect_branch): Likewise.
11384 (ix86_output_indirect_jmp): Likewise.
11385 (ix86_code_end): Call output_indirect_thunk_function if needed.
11386 (ix86_output_call_insn): Call ix86_output_indirect_branch if
11388 (ix86_handle_fndecl_attribute): Handle indirect_branch.
11389 (ix86_attribute_table): Add indirect_branch.
11390 * config/i386/i386.h (machine_function): Add indirect_branch_type
11391 and has_local_indirect_jump.
11392 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
11394 (tablejump): Likewise.
11395 (*indirect_jump): Use ix86_output_indirect_jmp.
11396 (*tablejump_1): Likewise.
11397 (simple_return_indirect_internal): Likewise.
11398 * config/i386/i386.opt (mindirect-branch=): New option.
11399 (indirect_branch): New.
11402 (thunk-inline): Likewise.
11403 (thunk-extern): Likewise.
11404 * doc/extend.texi: Document indirect_branch function attribute.
11405 * doc/invoke.texi: Document -mindirect-branch= option.
11407 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
11410 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
11412 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
11414 * ipa-inline.c (want_inline_small_function_p): Return false if
11415 inlining has already failed with CIF_FINAL_ERROR.
11416 (update_caller_keys): Call want_inline_small_function_p before
11418 (update_callee_keys): Likewise.
11420 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
11422 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
11424 (rs6000_quadword_masked_address_p): Likewise.
11425 (quad_aligned_load_p): Likewise.
11426 (quad_aligned_store_p): Likewise.
11427 (const_load_sequence_p): Add comment to describe the outer-most loop.
11428 (mimic_memory_attributes_and_flags): New function.
11429 (rs6000_gen_stvx): Likewise.
11430 (replace_swapped_aligned_store): Likewise.
11431 (rs6000_gen_lvx): Likewise.
11432 (replace_swapped_aligned_load): Likewise.
11433 (replace_swapped_load_constant): Capitalize argument name in
11434 comment describing this function.
11435 (rs6000_analyze_swaps): Add a third pass to search for vector loads
11436 and stores that access quad-word aligned addresses and replace
11437 with stvx or lvx instructions when appropriate.
11438 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
11439 New function prototype.
11440 (rs6000_quadword_masked_address_p): Likewise.
11441 (rs6000_gen_lvx): Likewise.
11442 (rs6000_gen_stvx): Likewise.
11443 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
11444 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
11445 when memory address is aligned.
11446 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
11447 this split to select lvx instruction when memory address is aligned.
11448 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
11449 instruction when memory address is aligned.
11450 (*vsx_le_perm_load_v16qi): Likewise.
11451 (four unnamed splitters): Modify to select the stvx instruction
11452 when memory is aligned.
11454 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
11456 * predict.c (determine_unlikely_bbs): Handle correctly BBs
11457 which appears in the queue multiple times.
11459 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11460 Alan Hayward <alan.hayward@arm.com>
11461 David Sherwood <david.sherwood@arm.com>
11463 * tree-vectorizer.h (vec_lower_bound): New structure.
11464 (_loop_vec_info): Add check_nonzero and lower_bounds.
11465 (LOOP_VINFO_CHECK_NONZERO): New macro.
11466 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
11467 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
11468 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
11469 fields. Make seg_len the distance travelled, not including the
11471 (dr_direction_indicator): Declare.
11472 (dr_zero_step_indicator): Likewise.
11473 (dr_known_forward_stride_p): Likewise.
11474 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
11476 (runtime_alias_check_p): Allow runtime alias checks with
11478 (operator ==): Compare access_size and align.
11479 (prune_runtime_alias_test_list): Rework for new distinction between
11480 the access_size and seg_len.
11481 (create_intersect_range_checks_index): Likewise. Cope with polynomial
11483 (get_segment_min_max): New function.
11484 (create_intersect_range_checks): Use it.
11485 (dr_step_indicator): New function.
11486 (dr_direction_indicator): Likewise.
11487 (dr_zero_step_indicator): Likewise.
11488 (dr_known_forward_stride_p): Likewise.
11489 * tree-loop-distribution.c (data_ref_segment_size): Return
11490 DR_STEP * (niters - 1).
11491 (compute_alias_check_pairs): Update call to the dr_with_seg_len
11493 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
11494 (vect_preserves_scalar_order_p): New function, split out from...
11495 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
11496 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
11497 (vect_vfa_access_size): New function.
11498 (vect_vfa_align): Likewise.
11499 (vect_compile_time_alias): Take access_size_a and access_b arguments.
11500 (dump_lower_bound): New function.
11501 (vect_check_lower_bound): Likewise.
11502 (vect_small_gap_p): Likewise.
11503 (vectorizable_with_step_bound_p): Likewise.
11504 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
11505 depencies if the vectorization factor is 1. Convert the checks
11506 for nonzero steps into checks on the bounds of DR_STEP. Try using
11507 a bunds check for variable steps if the minimum required step is
11508 relatively small. Update calls to the dr_with_seg_len
11509 constructor and to vect_compile_time_alias.
11510 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
11512 (vect_loop_versioning): Call it.
11513 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
11515 (vect_estimate_min_profitable_iters): Account for any bounds checks.
11517 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11518 Alan Hayward <alan.hayward@arm.com>
11519 David Sherwood <david.sherwood@arm.com>
11521 * doc/sourcebuild.texi (vect_scatter_store): Document.
11522 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
11524 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
11526 * genopinit.c (main): Add supports_vec_scatter_store and
11527 supports_vec_scatter_store_cached to target_optabs.
11528 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
11529 IFN_MASK_SCATTER_STORE.
11530 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
11532 * internal-fn.h (internal_store_fn_p): Declare.
11533 (internal_fn_stored_value_index): Likewise.
11534 * internal-fn.c (scatter_store_direct): New macro.
11535 (expand_scatter_store_optab_fn): New function.
11536 (direct_scatter_store_optab_supported_p): New macro.
11537 (internal_store_fn_p): New function.
11538 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
11539 IFN_MASK_SCATTER_STORE.
11540 (internal_fn_mask_index): Likewise.
11541 (internal_fn_stored_value_index): New function.
11542 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
11543 for scatter stores.
11544 * optabs-query.h (supports_vec_scatter_store_p): Declare.
11545 * optabs-query.c (supports_vec_scatter_store_p): New function.
11546 * tree-vectorizer.h (vect_get_store_rhs): Declare.
11547 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
11548 true for scatter stores.
11549 (vect_gather_scatter_fn_p): Handle scatter stores too.
11550 (vect_check_gather_scatter): Consider using scatter stores if
11551 supports_vec_scatter_store_p.
11552 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
11553 scatter stores too.
11554 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11555 internal_fn_stored_value_index.
11556 (check_load_store_masking): Handle scatter stores too.
11557 (vect_get_store_rhs): Make public.
11558 (vectorizable_call): Use internal_store_fn_p.
11559 (vectorizable_store): Handle scatter store internal functions.
11560 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
11561 when deciding whether the end of the group has been reached.
11562 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
11563 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
11564 (mask_scatter_store<mode>): New insns.
11566 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11567 Alan Hayward <alan.hayward@arm.com>
11568 David Sherwood <david.sherwood@arm.com>
11570 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
11571 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
11572 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
11574 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
11575 Use vect_truncate_gather_scatter_offset if we can't treat the
11576 operation as a normal gather load or scatter store.
11577 (get_group_load_store_type): Take the gather_scatter_info
11578 as argument. Try using a gather load or scatter store for
11579 single-element groups.
11580 (get_load_store_type): Update calls to get_group_load_store_type
11581 and vect_use_strided_gather_scatters_p.
11583 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11584 Alan Hayward <alan.hayward@arm.com>
11585 David Sherwood <david.sherwood@arm.com>
11587 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
11588 optional tree argument.
11589 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
11591 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
11592 but continue to use the current value as a fallback.
11593 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
11594 to compare the updates.
11595 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
11596 (get_load_store_type): Use it when handling a strided access.
11597 (vect_get_strided_load_store_ops): New function.
11598 (vect_get_data_ptr_increment): Likewise.
11599 (vectorizable_load): Handle strided gather loads. Always pass
11600 a step to vect_create_data_ref_ptr and bump_vector_ptr.
11602 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11603 Alan Hayward <alan.hayward@arm.com>
11604 David Sherwood <david.sherwood@arm.com>
11606 * doc/md.texi (gather_load@var{m}): Document.
11607 (mask_gather_load@var{m}): Likewise.
11608 * genopinit.c (main): Add supports_vec_gather_load and
11609 supports_vec_gather_load_cached to target_optabs.
11610 * optabs-tree.c (init_tree_optimization_optabs): Use
11611 ggc_cleared_alloc to allocate target_optabs.
11612 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
11613 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
11615 * internal-fn.h (internal_load_fn_p): Declare.
11616 (internal_gather_scatter_fn_p): Likewise.
11617 (internal_fn_mask_index): Likewise.
11618 (internal_gather_scatter_fn_supported_p): Likewise.
11619 * internal-fn.c (gather_load_direct): New macro.
11620 (expand_gather_load_optab_fn): New function.
11621 (direct_gather_load_optab_supported_p): New macro.
11622 (direct_internal_fn_optab): New function.
11623 (internal_load_fn_p): Likewise.
11624 (internal_gather_scatter_fn_p): Likewise.
11625 (internal_fn_mask_index): Likewise.
11626 (internal_gather_scatter_fn_supported_p): Likewise.
11627 * optabs-query.c (supports_at_least_one_mode_p): New function.
11628 (supports_vec_gather_load_p): Likewise.
11629 * optabs-query.h (supports_vec_gather_load_p): Declare.
11630 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
11631 and memory_type field.
11632 (NUM_PATTERNS): Bump to 15.
11633 * tree-vect-data-refs.c: Include internal-fn.h.
11634 (vect_gather_scatter_fn_p): New function.
11635 (vect_describe_gather_scatter_call): Likewise.
11636 (vect_check_gather_scatter): Try using internal functions for
11637 gather loads. Recognize existing calls to a gather load function.
11638 (vect_analyze_data_refs): Consider using gather loads if
11639 supports_vec_gather_load_p.
11640 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
11641 (vect_get_gather_scatter_offset_type): Likewise.
11642 (vect_convert_mask_for_vectype): Likewise.
11643 (vect_add_conversion_to_patterm): Likewise.
11644 (vect_try_gather_scatter_pattern): Likewise.
11645 (vect_recog_gather_scatter_pattern): New pattern recognizer.
11646 (vect_vect_recog_func_ptrs): Add it.
11647 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11648 internal_fn_mask_index and internal_gather_scatter_fn_p.
11649 (check_load_store_masking): Take the gather_scatter_info as an
11650 argument and handle gather loads.
11651 (vect_get_gather_scatter_ops): New function.
11652 (vectorizable_call): Check internal_load_fn_p.
11653 (vectorizable_load): Likewise. Handle gather load internal
11655 (vectorizable_store): Update call to check_load_store_masking.
11656 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
11657 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
11658 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
11659 (aarch64_gather_scale_operand_d): New predicates.
11660 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
11661 (mask_gather_load<mode>): New insns.
11663 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11664 Alan Hayward <alan.hayward@arm.com>
11665 David Sherwood <david.sherwood@arm.com>
11667 * optabs.def (fold_left_plus_optab): New optab.
11668 * doc/md.texi (fold_left_plus_@var{m}): Document.
11669 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
11670 * internal-fn.c (fold_left_direct): Define.
11671 (expand_fold_left_optab_fn): Likewise.
11672 (direct_fold_left_optab_supported_p): Likewise.
11673 * fold-const-call.c (fold_const_fold_left): New function.
11674 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
11675 * tree-parloops.c (valid_reduction_p): New function.
11676 (gather_scalar_reductions): Use it.
11677 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
11678 (vect_finish_replace_stmt): Declare.
11679 * tree-vect-loop.c (fold_left_reduction_fn): New function.
11680 (needs_fold_left_reduction_p): New function, split out from...
11681 (vect_is_simple_reduction): ...here. Accept reductions that
11682 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
11683 (vect_force_simple_reduction): Also store the reduction type in
11684 the assignment's STMT_VINFO_REDUC_TYPE.
11685 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
11686 (merge_with_identity): New function.
11687 (vect_expand_fold_left): Likewise.
11688 (vectorize_fold_left_reduction): Likewise.
11689 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
11690 scalar phi in place for it. Check for target support and reject
11691 cases that would reassociate the operation. Defer the transform
11692 phase to vectorize_fold_left_reduction.
11693 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
11694 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
11695 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
11697 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11699 * tree-if-conv.c (predicate_mem_writes): Remove redundant
11700 call to ifc_temp_var.
11702 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11703 Alan Hayward <alan.hayward@arm.com>
11704 David Sherwood <david.sherwood@arm.com>
11706 * target.def (legitimize_address_displacement): Take the original
11707 offset as a poly_int.
11708 * targhooks.h (default_legitimize_address_displacement): Update
11710 * targhooks.c (default_legitimize_address_displacement): Likewise.
11711 * doc/tm.texi: Regenerate.
11712 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
11713 as an argument, moving assert of ad->disp == ad->disp_term to...
11714 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
11715 Try calling targetm.legitimize_address_displacement before expanding
11716 the address rather than afterwards, and adjust for the new interface.
11717 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
11718 Match the new hook interface. Handle SVE addresses.
11719 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
11720 new hook interface.
11722 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11724 * Makefile.in (OBJS): Add early-remat.o.
11725 * target.def (select_early_remat_modes): New hook.
11726 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
11727 * doc/tm.texi: Regenerate.
11728 * targhooks.h (default_select_early_remat_modes): Declare.
11729 * targhooks.c (default_select_early_remat_modes): New function.
11730 * timevar.def (TV_EARLY_REMAT): New timevar.
11731 * passes.def (pass_early_remat): New pass.
11732 * tree-pass.h (make_pass_early_remat): Declare.
11733 * early-remat.c: New file.
11734 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
11736 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
11738 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11739 Alan Hayward <alan.hayward@arm.com>
11740 David Sherwood <david.sherwood@arm.com>
11742 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
11743 vfm1 with a bound_epilog parameter.
11744 (vect_do_peeling): Update calls accordingly, and move the prologue
11745 call earlier in the function. Treat the base bound_epilog as 0 for
11746 fully-masked loops and retain vf - 1 for other loops. Add 1 to
11747 this base when peeling for gaps.
11748 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
11749 with fully-masked loops.
11750 (vect_estimate_min_profitable_iters): Handle the single peeled
11751 iteration in that case.
11753 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11754 Alan Hayward <alan.hayward@arm.com>
11755 David Sherwood <david.sherwood@arm.com>
11757 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
11758 single-element interleaving even if the size is not a power of 2.
11759 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
11760 accesses for single-element interleaving if the group size is
11763 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11764 Alan Hayward <alan.hayward@arm.com>
11765 David Sherwood <david.sherwood@arm.com>
11767 * doc/md.texi (fold_extract_last_@var{m}): Document.
11768 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
11769 * optabs.def (fold_extract_last_optab): New optab.
11770 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
11771 * internal-fn.c (fold_extract_direct): New macro.
11772 (expand_fold_extract_optab_fn): Likewise.
11773 (direct_fold_extract_optab_supported_p): Likewise.
11774 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
11775 * tree-vect-loop.c (vect_model_reduction_cost): Handle
11776 EXTRACT_LAST_REDUCTION.
11777 (get_initial_def_for_reduction): Do not create an initial vector
11778 for EXTRACT_LAST_REDUCTION reductions.
11779 (vectorizable_reduction): Leave the scalar phi in place for
11780 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
11781 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
11782 epilogue code for EXTRACT_LAST_REDUCTION and defer the
11783 transform phase to vectorizable_condition.
11784 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
11786 (vect_finish_stmt_generation): ...here.
11787 (vect_finish_replace_stmt): New function.
11788 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
11789 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
11791 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
11793 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11794 Alan Hayward <alan.hayward@arm.com>
11795 David Sherwood <david.sherwood@arm.com>
11797 * doc/md.texi (extract_last_@var{m}): Document.
11798 * optabs.def (extract_last_optab): New optab.
11799 * internal-fn.def (EXTRACT_LAST): New internal function.
11800 * internal-fn.c (cond_unary_direct): New macro.
11801 (expand_cond_unary_optab_fn): Likewise.
11802 (direct_cond_unary_optab_supported_p): Likewise.
11803 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
11804 loops using EXTRACT_LAST.
11805 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
11806 (extract_last_<mode>): ...this optab.
11807 (vec_extract<mode><Vel>): Update accordingly.
11809 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11810 Alan Hayward <alan.hayward@arm.com>
11811 David Sherwood <david.sherwood@arm.com>
11813 * target.def (empty_mask_is_expensive): New hook.
11814 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
11815 * doc/tm.texi: Regenerate.
11816 * targhooks.h (default_empty_mask_is_expensive): Declare.
11817 * targhooks.c (default_empty_mask_is_expensive): New function.
11818 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
11819 if the target says that empty masks are expensive.
11820 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
11822 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
11824 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11825 Alan Hayward <alan.hayward@arm.com>
11826 David Sherwood <david.sherwood@arm.com>
11828 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
11829 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
11830 (vect_use_loop_mask_for_alignment_p): New function.
11831 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
11832 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
11833 niters_skip argument. Make sure that the first niters_skip elements
11834 of the first iteration are inactive.
11835 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
11836 Update call to vect_set_loop_masks_directly.
11837 (get_misalign_in_elems): New function, split out from...
11838 (vect_gen_prolog_loop_niters): ...here.
11839 (vect_update_init_of_dr): Take a code argument that specifies whether
11840 the adjustment should be added or subtracted.
11841 (vect_update_init_of_drs): Likewise.
11842 (vect_prepare_for_masked_peels): New function.
11843 (vect_do_peeling): Skip prologue peeling if we're using a mask
11844 instead. Update call to vect_update_inits_of_drs.
11845 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11847 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
11848 alignment. Do not include the number of peeled iterations in
11849 the minimum threshold in that case.
11850 (vectorizable_induction): Adjust the start value down by
11851 LOOP_VINFO_MASK_SKIP_NITERS iterations.
11852 (vect_transform_loop): Call vect_prepare_for_masked_peels.
11853 Take the number of skipped iterations into account when calculating
11855 * tree-vect-stmts.c (vect_gen_while_not): New function.
11857 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11858 Alan Hayward <alan.hayward@arm.com>
11859 David Sherwood <david.sherwood@arm.com>
11861 * doc/sourcebuild.texi (vect_fully_masked): Document.
11862 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
11863 default value to 0.
11864 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
11866 (vect_analyze_loop_2): ...here. Don't check the vectorization
11867 factor against the number of loop iterations if the loop is
11870 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11871 Alan Hayward <alan.hayward@arm.com>
11872 David Sherwood <david.sherwood@arm.com>
11874 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
11875 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
11876 (dump_groups): Update accordingly.
11877 (iv_use::mem_type): New member variable.
11878 (address_p): New function.
11879 (record_use): Add a mem_type argument and initialize the new
11881 (record_group_use): Add a mem_type argument. Use address_p.
11882 Remove obsolete null checks of base_object. Update call to record_use.
11883 (find_interesting_uses_op): Update call to record_group_use.
11884 (find_interesting_uses_cond): Likewise.
11885 (find_interesting_uses_address): Likewise.
11886 (get_mem_type_for_internal_fn): New function.
11887 (find_address_like_use): Likewise.
11888 (find_interesting_uses_stmt): Try find_address_like_use before
11889 calling find_interesting_uses_op.
11890 (addr_offset_valid_p): Use the iv mem_type field as the type
11891 of the addressed memory.
11892 (add_autoinc_candidates): Likewise.
11893 (get_address_cost): Likewise.
11894 (split_small_address_groups_p): Use address_p.
11895 (split_address_groups): Likewise.
11896 (add_iv_candidate_for_use): Likewise.
11897 (autoinc_possible_for_pair): Likewise.
11898 (rewrite_groups): Likewise.
11899 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
11900 (determine_group_iv_cost): Update after split of USE_ADDRESS.
11901 (get_alias_ptr_type_for_ptr_address): New function.
11902 (rewrite_use_address): Rewrite address uses in calls that were
11903 identified by find_address_like_use.
11905 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11906 Alan Hayward <alan.hayward@arm.com>
11907 David Sherwood <david.sherwood@arm.com>
11909 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
11911 * gimple-expr.h (is_gimple_addressable: Likewise.
11912 * gimple-expr.c (is_gimple_address): Likewise.
11913 * internal-fn.c (expand_call_mem_ref): New function.
11914 (expand_mask_load_optab_fn): Use it.
11915 (expand_mask_store_optab_fn): Likewise.
11917 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11918 Alan Hayward <alan.hayward@arm.com>
11919 David Sherwood <david.sherwood@arm.com>
11921 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
11922 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
11923 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
11924 (cond_umax@var{mode}): Document.
11925 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
11926 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
11927 (cond_umin_optab, cond_umax_optab): New optabs.
11928 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
11929 (COND_IOR, COND_XOR): New internal functions.
11930 * internal-fn.h (get_conditional_internal_fn): Declare.
11931 * internal-fn.c (cond_binary_direct): New macro.
11932 (expand_cond_binary_optab_fn): Likewise.
11933 (direct_cond_binary_optab_supported_p): Likewise.
11934 (get_conditional_internal_fn): New function.
11935 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
11936 Cope with reduction statements that are vectorized as calls rather
11938 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
11939 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
11940 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
11941 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11942 (UNSPEC_COND_EOR): New unspecs.
11943 (optab): Add mappings for them.
11944 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
11945 (sve_int_op, sve_fp_op): New int attributes.
11947 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11948 Alan Hayward <alan.hayward@arm.com>
11949 David Sherwood <david.sherwood@arm.com>
11951 * optabs.def (while_ult_optab): New optab.
11952 * doc/md.texi (while_ult@var{m}@var{n}): Document.
11953 * internal-fn.def (WHILE_ULT): New internal function.
11954 * internal-fn.h (direct_internal_fn_supported_p): New override
11955 that takes two types as argument.
11956 * internal-fn.c (while_direct): New macro.
11957 (expand_while_optab_fn): New function.
11958 (convert_optab_supported_p): Likewise.
11959 (direct_while_optab_supported_p): New macro.
11960 * wide-int.h (wi::udiv_ceil): New function.
11961 * tree-vectorizer.h (rgroup_masks): New structure.
11962 (vec_loop_masks): New typedef.
11963 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
11964 and fully_masked_p.
11965 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
11966 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
11967 (vect_max_vf): New function.
11968 (slpeel_make_loop_iterate_ntimes): Delete.
11969 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
11970 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
11971 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
11972 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
11973 internal-fn.h, stor-layout.h and optabs-query.h.
11974 (vect_set_loop_mask): New function.
11975 (add_preheader_seq): Likewise.
11976 (add_header_seq): Likewise.
11977 (interleave_supported_p): Likewise.
11978 (vect_maybe_permute_loop_masks): Likewise.
11979 (vect_set_loop_masks_directly): Likewise.
11980 (vect_set_loop_condition_masked): Likewise.
11981 (vect_set_loop_condition_unmasked): New function, split out from
11982 slpeel_make_loop_iterate_ntimes.
11983 (slpeel_make_loop_iterate_ntimes): Rename to..
11984 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
11985 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
11986 (vect_do_peeling): Update call accordingly.
11987 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
11989 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11990 mask_compare_type, can_fully_mask_p and fully_masked_p.
11991 (release_vec_loop_masks): New function.
11992 (_loop_vec_info): Use it to free the loop masks.
11993 (can_produce_all_loop_masks_p): New function.
11994 (vect_get_max_nscalars_per_iter): Likewise.
11995 (vect_verify_full_masking): Likewise.
11996 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
11997 retries, and free the mask rgroups before retrying. Check loop-wide
11998 reasons for disallowing fully-masked loops. Make the final decision
11999 about whether use a fully-masked loop or not.
12000 (vect_estimate_min_profitable_iters): Do not assume that peeling
12001 for the number of iterations will be needed for fully-masked loops.
12002 (vectorizable_reduction): Disable fully-masked loops.
12003 (vectorizable_live_operation): Likewise.
12004 (vect_halve_mask_nunits): New function.
12005 (vect_double_mask_nunits): Likewise.
12006 (vect_record_loop_mask): Likewise.
12007 (vect_get_loop_mask): Likewise.
12008 (vect_transform_loop): Handle the case in which the final loop
12009 iteration might handle a partial vector. Call vect_set_loop_condition
12010 instead of slpeel_make_loop_iterate_ntimes.
12011 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
12012 (check_load_store_masking): New function.
12013 (prepare_load_store_mask): Likewise.
12014 (vectorizable_store): Handle fully-masked loops.
12015 (vectorizable_load): Likewise.
12016 (supportable_widening_operation): Use vect_halve_mask_nunits for
12018 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
12019 (vect_gen_while): New function.
12020 * config/aarch64/aarch64.md (umax<mode>3): New expander.
12021 (aarch64_uqdec<mode>): New insn.
12023 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12024 Alan Hayward <alan.hayward@arm.com>
12025 David Sherwood <david.sherwood@arm.com>
12027 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
12028 (reduc_xor_scal_optab): New optabs.
12029 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
12030 (reduc_xor_scal_@var{m}): Document.
12031 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
12032 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
12033 internal functions.
12034 * fold-const-call.c (fold_const_call): Handle them.
12035 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
12036 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
12037 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
12038 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
12039 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
12040 (UNSPEC_XORV): New unspecs.
12041 (optab): Add entries for them.
12042 (BITWISEV): New int iterator.
12043 (bit_reduc_op): New int attributes.
12045 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12046 Alan Hayward <alan.hayward@arm.com>
12047 David Sherwood <david.sherwood@arm.com>
12049 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
12050 * internal-fn.def (VEC_SHL_INSERT): New internal function.
12051 * optabs.def (vec_shl_insert_optab): New optab.
12052 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
12053 (duplicate_and_interleave): Likewise.
12054 * tree-vect-loop.c: Include internal-fn.h.
12055 (neutral_op_for_slp_reduction): New function, split out from
12056 get_initial_defs_for_reduction.
12057 (get_initial_def_for_reduction): Handle option 2 for variable-length
12058 vectors by loading the neutral value into a vector and then shifting
12059 the initial value into element 0.
12060 (get_initial_defs_for_reduction): Replace the code argument with
12061 the neutral value calculated by neutral_op_for_slp_reduction.
12062 Use gimple_build_vector for constant-length vectors.
12063 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
12064 but the first group_size elements have a neutral value.
12065 Use duplicate_and_interleave otherwise.
12066 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
12067 Update call to get_initial_defs_for_reduction. Handle SLP
12068 reductions for variable-length vectors by creating one vector
12069 result for each scalar result, with the elements associated
12070 with other scalar results stubbed out with the neutral value.
12071 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
12072 Require IFN_VEC_SHL_INSERT for double reductions on
12073 variable-length vectors, or SLP reductions that have
12074 a neutral value. Require can_duplicate_and_interleave_p
12075 support for variable-length unchained SLP reductions if there
12076 is no neutral value, such as for MIN/MAX reductions. Also require
12077 the number of vector elements to be a multiple of the number of
12078 SLP statements when doing variable-length unchained SLP reductions.
12079 Update call to vect_create_epilog_for_reduction.
12080 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
12081 and remove initial values.
12082 (duplicate_and_interleave): Make public.
12083 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
12084 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
12086 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12087 Alan Hayward <alan.hayward@arm.com>
12088 David Sherwood <david.sherwood@arm.com>
12090 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
12091 (can_duplicate_and_interleave_p): New function.
12092 (vect_get_and_check_slp_defs): Take the vector of statements
12093 rather than just the current one. Remove excess parentheses.
12094 Restriction rejectinon of vect_constant_def and vect_external_def
12095 for variable-length vectors to boolean types, or types for which
12096 can_duplicate_and_interleave_p is false.
12097 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
12098 (duplicate_and_interleave): New function.
12099 (vect_get_constant_vectors): Use gimple_build_vector for
12100 constant-length vectors and suitable variable-length constant
12101 vectors. Use duplicate_and_interleave for other variable-length
12102 vectors. Don't defer the update when inserting new statements.
12104 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12105 Alan Hayward <alan.hayward@arm.com>
12106 David Sherwood <david.sherwood@arm.com>
12108 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
12109 min_profitable_iters doesn't go negative.
12111 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12112 Alan Hayward <alan.hayward@arm.com>
12113 David Sherwood <david.sherwood@arm.com>
12115 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
12116 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
12117 * optabs.def (vec_mask_load_lanes_optab): New optab.
12118 (vec_mask_store_lanes_optab): Likewise.
12119 * internal-fn.def (MASK_LOAD_LANES): New internal function.
12120 (MASK_STORE_LANES): Likewise.
12121 * internal-fn.c (mask_load_lanes_direct): New macro.
12122 (mask_store_lanes_direct): Likewise.
12123 (expand_mask_load_optab_fn): Handle masked operations.
12124 (expand_mask_load_lanes_optab_fn): New macro.
12125 (expand_mask_store_optab_fn): Handle masked operations.
12126 (expand_mask_store_lanes_optab_fn): New macro.
12127 (direct_mask_load_lanes_optab_supported_p): Likewise.
12128 (direct_mask_store_lanes_optab_supported_p): Likewise.
12129 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
12131 (vect_load_lanes_supported): Likewise.
12132 * tree-vect-data-refs.c (strip_conversion): New function.
12133 (can_group_stmts_p): Likewise.
12134 (vect_analyze_data_ref_accesses): Use it instead of checking
12135 for a pair of assignments.
12136 (vect_store_lanes_supported): Take a masked_p parameter.
12137 (vect_load_lanes_supported): Likewise.
12138 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
12139 vect_store_lanes_supported and vect_load_lanes_supported.
12140 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
12141 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
12142 parameter. Don't allow gaps for masked accesses.
12143 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
12144 and vect_load_lanes_supported.
12145 (get_load_store_type): Take a masked_p parameter and update
12146 call to get_group_load_store_type.
12147 (vectorizable_store): Update call to get_load_store_type.
12148 Handle IFN_MASK_STORE_LANES.
12149 (vectorizable_load): Update call to get_load_store_type.
12150 Handle IFN_MASK_LOAD_LANES.
12152 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12153 Alan Hayward <alan.hayward@arm.com>
12154 David Sherwood <david.sherwood@arm.com>
12156 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
12158 * config/aarch64/aarch64-protos.h
12159 (aarch64_sve_struct_memory_operand_p): Declare.
12160 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
12161 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
12162 (VPRED, vpred): Handle SVE structure modes.
12163 * config/aarch64/constraints.md (Utx): New constraint.
12164 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
12165 (aarch64_sve_struct_nonimmediate_operand): New predicates.
12166 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
12167 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
12168 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
12169 structure modes. Split into pieces after RA.
12170 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
12171 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
12173 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
12174 SVE structure modes.
12175 (aarch64_classify_address): Likewise.
12176 (sizetochar): Move earlier in file.
12177 (aarch64_print_operand): Handle SVE register lists.
12178 (aarch64_array_mode): New function.
12179 (aarch64_sve_struct_memory_operand_p): Likewise.
12180 (TARGET_ARRAY_MODE): Redefine.
12182 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12183 Alan Hayward <alan.hayward@arm.com>
12184 David Sherwood <david.sherwood@arm.com>
12186 * target.def (array_mode): New target hook.
12187 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
12188 * doc/tm.texi: Regenerate.
12189 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
12190 * hooks.c (hook_optmode_mode_uhwi_none): New function.
12191 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
12192 targetm.array_mode.
12193 * stor-layout.c (mode_for_array): Likewise. Support polynomial
12196 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12197 Alan Hayward <alan.hayward@arm.com>
12198 David Sherwood <david.sherwood@arm.com>
12200 * fold-const.c (fold_binary_loc): Check the argument types
12201 rather than the result type when testing for a vector operation.
12203 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12205 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
12206 * doc/tm.texi: Regenerate.
12208 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12209 Alan Hayward <alan.hayward@arm.com>
12210 David Sherwood <david.sherwood@arm.com>
12212 * doc/invoke.texi (-msve-vector-bits=): Document new option.
12213 (sve): Document new AArch64 extension.
12214 * doc/md.texi (w): Extend the description of the AArch64
12215 constraint to include SVE vectors.
12216 (Upl, Upa): Document new AArch64 predicate constraints.
12217 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
12219 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
12220 (msve-vector-bits=): New option.
12221 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
12222 SVE when these are disabled.
12223 (sve): New extension.
12224 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
12225 modes. Adjust their number of units based on aarch64_sve_vg.
12226 (MAX_BITSIZE_MODE_ANY_MODE): Define.
12227 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
12228 aarch64_addr_query_type.
12229 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
12230 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
12231 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
12232 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
12233 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
12234 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
12235 (aarch64_simd_imm_zero_p): Delete.
12236 (aarch64_check_zero_based_sve_index_immediate): Declare.
12237 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
12238 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
12239 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
12240 (aarch64_sve_float_mul_immediate_p): Likewise.
12241 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
12242 rather than an rtx.
12243 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
12244 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
12245 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
12246 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
12247 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
12248 (aarch64_regmode_natural_size): Likewise.
12249 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
12250 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
12252 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
12253 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
12254 for VG and the SVE predicate registers.
12255 (V_ALIASES): Add a "z"-prefixed alias.
12256 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
12257 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
12258 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
12259 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
12260 (REG_CLASS_NAMES): Add entries for them.
12261 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
12262 and the predicate registers.
12263 (aarch64_sve_vg): Declare.
12264 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
12265 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
12266 (REGMODE_NATURAL_SIZE): Define.
12267 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
12269 * config/aarch64/aarch64.c: Include cfgrtl.h.
12270 (simd_immediate_info): Add a constructor for series vectors,
12271 and an associated step field.
12272 (aarch64_sve_vg): New variable.
12273 (aarch64_dbx_register_number): Handle VG and the predicate registers.
12274 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
12275 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
12276 (VEC_ANY_DATA, VEC_STRUCT): New constants.
12277 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
12278 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
12279 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
12280 (aarch64_get_mask_mode): New functions.
12281 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
12282 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
12283 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
12284 predicate modes and predicate registers. Explicitly restrict
12285 GPRs to modes of 16 bytes or smaller. Only allow FP registers
12286 to store a vector mode if it is recognized by
12287 aarch64_classify_vector_mode.
12288 (aarch64_regmode_natural_size): New function.
12289 (aarch64_hard_regno_caller_save_mode): Return the original mode
12291 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
12292 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
12293 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
12294 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
12296 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
12297 does not overlap dest if the function is frame-related. Handle
12299 (aarch64_split_add_offset): New function.
12300 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
12301 them aarch64_add_offset.
12302 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
12303 and update call to aarch64_sub_sp.
12304 (aarch64_add_cfa_expression): New function.
12305 (aarch64_expand_prologue): Pass extra temporary registers to the
12306 functions above. Handle the case in which we need to emit new
12307 DW_CFA_expressions for registers that were originally saved
12308 relative to the stack pointer, but now have to be expressed
12309 relative to the frame pointer.
12310 (aarch64_output_mi_thunk): Pass extra temporary registers to the
12312 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
12313 IP0 and IP1 values for SVE frames.
12314 (aarch64_expand_vec_series): New function.
12315 (aarch64_expand_sve_widened_duplicate): Likewise.
12316 (aarch64_expand_sve_const_vector): Likewise.
12317 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
12318 Handle SVE constants. Use emit_move_insn to move a force_const_mem
12319 into the register, rather than emitting a SET directly.
12320 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
12321 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
12322 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
12323 (offset_9bit_signed_scaled_p): New functions.
12324 (aarch64_replicate_bitmask_imm): New function.
12325 (aarch64_bitmask_imm): Use it.
12326 (aarch64_cannot_force_const_mem): Reject expressions involving
12327 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
12328 (aarch64_classify_index): Handle SVE indices, by requiring
12329 a plain register index with a scale that matches the element size.
12330 (aarch64_classify_address): Handle SVE addresses. Assert that
12331 the mode of the address is VOIDmode or an integer mode.
12332 Update call to aarch64_classify_symbol.
12333 (aarch64_classify_symbolic_expression): Update call to
12334 aarch64_classify_symbol.
12335 (aarch64_const_vec_all_in_range_p): New function.
12336 (aarch64_print_vector_float_operand): Likewise.
12337 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
12338 "vN" for FP registers with SVE modes. Handle (const ...) vectors
12339 and the FP immediates 1.0 and 0.5.
12340 (aarch64_print_address_internal): Handle SVE addresses.
12341 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
12342 (aarch64_regno_regclass): Handle predicate registers.
12343 (aarch64_secondary_reload): Handle big-endian reloads of SVE
12345 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
12346 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
12347 (aarch64_convert_sve_vector_bits): New function.
12348 (aarch64_override_options): Use it to handle -msve-vector-bits=.
12349 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
12350 rather than an rtx.
12351 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
12352 Handle SVE vector and predicate modes. Accept VL-based constants
12353 that need only one temporary register, and VL offsets that require
12354 no temporary registers.
12355 (aarch64_conditional_register_usage): Mark the predicate registers
12356 as fixed if SVE isn't available.
12357 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
12358 Return true for SVE vector and predicate modes.
12359 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
12360 rather than an unsigned int. Handle SVE modes.
12361 (aarch64_preferred_simd_mode): Update call accordingly. Handle
12363 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
12365 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
12366 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
12367 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
12368 (aarch64_sve_float_mul_immediate_p): New functions.
12369 (aarch64_sve_valid_immediate): New function.
12370 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
12371 Explicitly reject structure modes. Check for INDEX constants.
12372 Handle PTRUE and PFALSE constants.
12373 (aarch64_check_zero_based_sve_index_immediate): New function.
12374 (aarch64_simd_imm_zero_p): Delete.
12375 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
12376 vector modes. Accept constants in the range of CNT[BHWD].
12377 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
12378 ask for an Advanced SIMD mode.
12379 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
12380 (aarch64_simd_vector_alignment): Handle SVE predicates.
12381 (aarch64_vectorize_preferred_vector_alignment): New function.
12382 (aarch64_simd_vector_alignment_reachable): Use it instead of
12384 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
12385 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
12387 (MAX_VECT_LEN): Delete.
12388 (expand_vec_perm_d): Add a vec_flags field.
12389 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
12390 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
12391 (aarch64_evpc_ext): Don't apply a big-endian lane correction
12393 (aarch64_evpc_rev): Rename to...
12394 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
12395 (aarch64_evpc_rev_global): New function.
12396 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
12397 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
12399 (aarch64_evpc_sve_tbl): New function.
12400 (aarch64_expand_vec_perm_const_1): Update after rename of
12401 aarch64_evpc_rev. Handle SVE permutes too, trying
12402 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
12403 than aarch64_evpc_tbl.
12404 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
12405 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
12406 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
12407 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
12408 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
12409 (aarch64_expand_sve_vcond): New functions.
12410 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
12411 of aarch64_vector_mode_p.
12412 (aarch64_dwarf_poly_indeterminate_value): New function.
12413 (aarch64_compute_pressure_classes): Likewise.
12414 (aarch64_can_change_mode_class): Likewise.
12415 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
12416 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
12417 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
12418 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
12419 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
12420 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
12421 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
12422 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
12424 (Dn, Dl, Dr): Accept const as well as const_vector.
12425 (Dz): Likewise. Compare against CONST0_RTX.
12426 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
12427 of "vector" where appropriate.
12428 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
12429 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
12430 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
12431 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
12432 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
12433 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
12434 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
12435 (v_int_equiv): Extend to SVE modes.
12436 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
12438 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
12439 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
12440 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
12441 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
12442 (SVE_COND_FP_CMP): New int iterators.
12443 (perm_hilo): Handle the new unpack unspecs.
12444 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
12446 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
12447 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
12448 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
12449 (aarch64_equality_operator, aarch64_constant_vector_operand)
12450 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
12451 (aarch64_sve_nonimmediate_operand): Likewise.
12452 (aarch64_sve_general_operand): Likewise.
12453 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
12454 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
12455 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
12456 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
12457 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
12458 (aarch64_sve_float_arith_immediate): Likewise.
12459 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
12460 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
12461 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
12462 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
12463 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
12464 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
12465 (aarch64_sve_float_arith_operand): Likewise.
12466 (aarch64_sve_float_arith_with_sub_operand): Likewise.
12467 (aarch64_sve_float_mul_operand): Likewise.
12468 (aarch64_sve_vec_perm_operand): Likewise.
12469 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
12470 (aarch64_mov_operand): Accept const_poly_int and const_vector.
12471 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
12472 as well as const_vector.
12473 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
12474 in file. Use CONST0_RTX and CONSTM1_RTX.
12475 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
12476 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
12477 Use aarch64_simd_imm_zero.
12478 * config/aarch64/aarch64-sve.md: New file.
12479 * config/aarch64/aarch64.md: Include it.
12480 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
12481 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
12482 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
12483 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
12484 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
12485 (sve): New attribute.
12486 (enabled): Disable instructions with the sve attribute unless
12488 (movqi, movhi): Pass CONST_POLY_INT operaneds through
12489 aarch64_expand_mov_immediate.
12490 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
12491 CNT[BHSD] immediates.
12492 (movti): Split CONST_POLY_INT moves into two halves.
12493 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
12494 Split additions that need a temporary here if the destination
12495 is the stack pointer.
12496 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
12497 (*add<mode>3_poly_1): New instruction.
12498 (set_clobber_cc): New expander.
12500 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12502 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
12503 parameter and use it instead of GET_MODE_SIZE (innermode). Use
12504 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
12505 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
12506 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
12507 Change innermode from fixed_mode_size to machine_mode.
12508 (simplify_subreg): Update call accordingly. Handle a constant-sized
12509 subreg of a variable-length CONST_VECTOR.
12511 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12512 Alan Hayward <alan.hayward@arm.com>
12513 David Sherwood <david.sherwood@arm.com>
12515 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
12516 (add_offset_to_base): New function, split out from...
12517 (create_mem_ref): ...here. When handling a scale other than 1,
12518 check first whether the address is valid without the offset.
12519 Add it into the base if so, leaving the index and scale as-is.
12521 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12524 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
12525 fold_for_warn before checking if arg2 is INTEGER_CST.
12527 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
12529 * config/rs6000/predicates.md (load_multiple_operation): Delete.
12530 (store_multiple_operation): Delete.
12531 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
12532 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
12533 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
12534 guarded by TARGET_STRING.
12535 (rs6000_output_load_multiple): Delete.
12536 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
12537 OPTION_MASK_STRING / TARGET_STRING handling.
12538 (print_operand) <'N', 'O'>: Add comment that these are unused now.
12539 (const rs6000_opt_masks) <"string">: Change mask to 0.
12540 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
12541 (MASK_STRING): Delete.
12542 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
12544 (load_multiple): Delete.
12551 (store_multiple): Delete.
12558 (movmemsi_8reg): Delete.
12559 (corresponding unnamed define_insn): Delete.
12560 (movmemsi_6reg): Delete.
12561 (corresponding unnamed define_insn): Delete.
12562 (movmemsi_4reg): Delete.
12563 (corresponding unnamed define_insn): Delete.
12564 (movmemsi_2reg): Delete.
12565 (corresponding unnamed define_insn): Delete.
12566 (movmemsi_1reg): Delete.
12567 (corresponding unnamed define_insn): Delete.
12568 * config/rs6000/rs6000.opt (mno-string): New.
12569 (mstring): Replace by deprecation warning stub.
12570 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
12572 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12574 * regrename.c (regrename_do_replace): If replacing the same
12575 reg multiple times, try to reuse last created gen_raw_REG.
12578 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
12579 main to workaround a bug in GDB.
12581 2018-01-12 Tom de Vries <tom@codesourcery.com>
12584 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
12586 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
12588 PR rtl-optimization/80481
12589 * ira-color.c (get_cap_member): New function.
12590 (allocnos_conflict_by_live_ranges_p): Use it.
12591 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
12592 (setup_slot_coalesced_allocno_live_ranges): Ditto.
12594 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
12597 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
12598 (*saddl_se_1): Ditto.
12599 (*ssubsi_1): Ditto.
12600 (*ssubl_se_1): Ditto.
12602 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12604 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
12605 rather than wi::to_widest for DR_INITs.
12606 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
12607 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
12608 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
12610 (vect_analyze_group_access_1): Note that here.
12612 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12614 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
12615 polynomial type sizes.
12617 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12619 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
12620 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
12621 (gimple_add_tmp_var): Likewise.
12623 2018-01-12 Martin Liska <mliska@suse.cz>
12625 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
12626 (gimple_alloc_sizes): Likewise.
12627 (dump_gimple_statistics): Use PRIu64 in printf format.
12628 * gimple.h: Change uint64_t to int.
12630 2018-01-12 Martin Liska <mliska@suse.cz>
12632 * tree-core.h: Use uint64_t instead of int.
12633 * tree.c (tree_node_counts): Likewise.
12634 (tree_node_sizes): Likewise.
12635 (dump_tree_statistics): Use PRIu64 in printf format.
12637 2018-01-12 Martin Liska <mliska@suse.cz>
12639 * Makefile.in: As qsort_chk is implemented in vec.c, add
12640 vec.o to linkage of gencfn-macros.
12641 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
12642 passing the info to record_node_allocation_statistics.
12643 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
12645 * ggc-common.c (struct ggc_usage): Add operator== and use
12646 it in operator< and compare function.
12647 * mem-stats.h (struct mem_usage): Likewise.
12648 * vec.c (struct vec_usage): Remove operator< and compare
12649 function. Can be simply inherited.
12651 2018-01-12 Martin Jambor <mjambor@suse.cz>
12654 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
12655 * tree-ssa-math-opts.c: Include domwalk.h.
12656 (convert_mult_to_fma_1): New function.
12657 (fma_transformation_info): New type.
12658 (fma_deferring_state): Likewise.
12659 (cancel_fma_deferring): New function.
12660 (result_of_phi): Likewise.
12661 (last_fma_candidate_feeds_initial_phi): Likewise.
12662 (convert_mult_to_fma): Added deferring logic, split actual
12663 transformation to convert_mult_to_fma_1.
12664 (math_opts_dom_walker): New type.
12665 (math_opts_dom_walker::after_dom_children): New method, body moved
12666 here from pass_optimize_widening_mul::execute, added deferring logic
12668 (pass_optimize_widening_mul::execute): Moved most of code to
12669 math_opts_dom_walker::after_dom_children.
12670 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
12671 * config/i386/i386.c (ix86_option_override_internal): Added
12672 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
12674 2018-01-12 Richard Biener <rguenther@suse.de>
12677 * dwarf2out.c (gen_variable_die): Do not reset old_die for
12678 inline instance vars.
12680 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
12683 * config/rx/rx.c (rx_is_restricted_memory_address):
12684 Handle SUBREG case.
12686 2018-01-12 Richard Biener <rguenther@suse.de>
12688 PR tree-optimization/80846
12689 * target.def (split_reduction): New target hook.
12690 * targhooks.c (default_split_reduction): New function.
12691 * targhooks.h (default_split_reduction): Declare.
12692 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
12693 target requests first reduce vectors by combining low and high
12695 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
12696 (get_vectype_for_scalar_type_and_size): Export.
12697 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
12698 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
12699 * doc/tm.texi: Regenerate.
12700 * config/i386/i386.c (ix86_split_reduction): Implement
12701 TARGET_VECTORIZE_SPLIT_REDUCTION.
12703 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12706 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
12707 in PIC mode except for TARGET_VXWORKS_RTP.
12708 * config/sparc/sparc.c: Include cfgrtl.h.
12709 (TARGET_INIT_PIC_REG): Define.
12710 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
12711 (sparc_pic_register_p): New predicate.
12712 (sparc_legitimate_address_p): Use it.
12713 (sparc_legitimize_pic_address): Likewise.
12714 (sparc_delegitimize_address): Likewise.
12715 (sparc_mode_dependent_address_p): Likewise.
12716 (gen_load_pcrel_sym): Remove 4th parameter.
12717 (load_got_register): Adjust call to above. Remove obsolete stuff.
12718 (sparc_expand_prologue): Do not call load_got_register here.
12719 (sparc_flat_expand_prologue): Likewise.
12720 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
12721 (sparc_use_pseudo_pic_reg): New function.
12722 (sparc_init_pic_reg): Likewise.
12723 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12724 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
12726 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
12728 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
12729 Add item for branch_cost.
12731 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12733 PR rtl-optimization/83565
12734 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
12735 not extend the result to a larger mode for rotate operations.
12736 (num_sign_bit_copies1): Likewise.
12738 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12741 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
12743 Use values-Xc.o for -pedantic.
12744 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
12746 2018-01-12 Martin Liska <mliska@suse.cz>
12749 * ipa-devirt.c (final_warning_record::grow_type_warnings):
12751 (possible_polymorphic_call_targets): Use it.
12752 (ipa_devirt): Likewise.
12754 2018-01-12 Martin Liska <mliska@suse.cz>
12756 * profile-count.h (enum profile_quality): Use 0 as invalid
12757 enum value of profile_quality.
12759 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
12761 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
12762 -mext-string options.
12764 2018-01-12 Richard Biener <rguenther@suse.de>
12766 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
12767 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
12768 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
12770 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
12772 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
12774 * configure.ac (--with-long-double-format): Add support for the
12775 configuration option to change the default long double format on
12777 * config.gcc (powerpc*-linux*-*): Likewise.
12778 * configure: Regenerate.
12779 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
12780 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
12781 used without modification.
12783 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12785 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
12786 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
12787 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
12788 MISC_BUILTIN_SPEC_BARRIER.
12789 (rs6000_init_builtins): Likewise.
12790 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
12792 (speculation_barrier): New define_insn.
12793 * doc/extend.texi: Document __builtin_speculation_barrier.
12795 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12798 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
12799 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
12800 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
12802 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
12803 integral modes instead of "ss" and "sd".
12804 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
12805 vectors with 32-bit and 64-bit elements.
12806 (vecdupssescalarmodesuffix): New mode attribute.
12807 (vec_dup<mode>): Use it.
12809 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
12812 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
12813 frame if argument is passed on stack.
12815 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12818 * ree.c (combine_reaching_defs): Optimize also
12819 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
12820 reg2=any_extend(exp); reg1=reg2;, formatting fix.
12822 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12824 PR middle-end/83189
12825 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
12827 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12829 PR middle-end/83718
12830 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
12831 after they are computed.
12833 2018-01-11 Bin Cheng <bin.cheng@arm.com>
12835 PR tree-optimization/83695
12836 * gimple-loop-linterchange.cc
12837 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
12838 reset cached scev information after interchange.
12839 (pass_linterchange::execute): Remove call to scev_reset_htab.
12841 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12843 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
12844 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
12845 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
12846 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
12847 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
12848 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
12849 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
12850 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
12851 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
12852 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
12853 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
12854 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
12855 (V_lane_reg): Likewise.
12856 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
12858 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
12859 (vfmal_lane_low<mode>_intrinsic,
12860 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
12861 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
12862 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
12863 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
12864 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
12865 vfmsl_lane_high<mode>_intrinsic): New define_insns.
12867 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12869 * config/arm/arm-cpus.in (fp16fml): New feature.
12870 (ALL_SIMD): Add fp16fml.
12871 (armv8.2-a): Add fp16fml as an option.
12872 (armv8.3-a): Likewise.
12873 (armv8.4-a): Add fp16fml as part of fp16.
12874 * config/arm/arm.h (TARGET_FP16FML): Define.
12875 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
12877 * config/arm/arm-modes.def (V2HF): Define.
12878 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
12879 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
12880 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
12881 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
12882 vfmsl_low, vfmsl_high): New set of builtins.
12883 * config/arm/iterators.md (PLUSMINUS): New code iterator.
12884 (vfml_op): New code attribute.
12885 (VFMLHALVES): New int iterator.
12886 (VFML, VFMLSEL): New mode attributes.
12887 (V_reg): Define mapping for V2HF.
12888 (V_hi, V_lo): New mode attributes.
12889 (VF_constraint): Likewise.
12890 (vfml_half, vfml_half_selector): New int attributes.
12891 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
12893 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
12894 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
12896 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
12897 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
12898 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
12899 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
12901 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
12902 Document new effective target and option set.
12904 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12906 * config/arm/arm-cpus.in (armv8_4): New feature.
12907 (ARMv8_4a): New fgroup.
12908 (armv8.4-a): New arch.
12909 * config/arm/arm-tables.opt: Regenerate.
12910 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
12911 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
12912 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
12913 Add matching rules for -march=armv8.4-a and extensions.
12914 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
12916 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
12919 * config/rx/rx.md (BW): New mode attribute.
12920 (sync_lock_test_and_setsi): Add mode suffix to insn output.
12922 2018-01-11 Richard Biener <rguenther@suse.de>
12924 PR tree-optimization/83435
12925 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
12926 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
12927 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
12929 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12930 Alan Hayward <alan.hayward@arm.com>
12931 David Sherwood <david.sherwood@arm.com>
12933 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
12935 (aarch64_classify_address): Initialize it. Track polynomial offsets.
12936 (aarch64_print_address_internal): Use it to check for a zero offset.
12938 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12939 Alan Hayward <alan.hayward@arm.com>
12940 David Sherwood <david.sherwood@arm.com>
12942 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
12943 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
12944 Return a poly_int64 rather than a HOST_WIDE_INT.
12945 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
12946 rather than a HOST_WIDE_INT.
12947 * config/aarch64/aarch64.h (aarch64_frame): Protect with
12948 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
12949 hard_fp_offset, frame_size, initial_adjust, callee_offset and
12950 final_offset from HOST_WIDE_INT to poly_int64.
12951 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
12952 to_constant when getting the number of units in an Advanced SIMD
12954 (aarch64_builtin_vectorized_function): Check for a constant number
12956 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
12958 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
12959 attribute instead of GET_MODE_NUNITS.
12960 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
12961 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
12962 GET_MODE_SIZE for fixed-size registers.
12963 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
12964 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
12965 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
12966 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
12967 (aarch64_print_operand, aarch64_print_address_internal)
12968 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
12969 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
12970 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
12971 Handle polynomial GET_MODE_SIZE.
12972 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
12973 wider than SImode without modification.
12974 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
12975 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
12976 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
12977 passing and returning SVE modes.
12978 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
12979 rather than GEN_INT.
12980 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
12981 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
12982 (aarch64_allocate_and_probe_stack_space): Likewise.
12983 (aarch64_layout_frame): Cope with polynomial offsets.
12984 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
12985 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
12986 polynomial offsets.
12987 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
12988 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
12989 poly_int64 rather than a HOST_WIDE_INT.
12990 (aarch64_get_separate_components, aarch64_process_components)
12991 (aarch64_expand_prologue, aarch64_expand_epilogue)
12992 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
12993 (aarch64_anchor_offset): New function, split out from...
12994 (aarch64_legitimize_address): ...here.
12995 (aarch64_builtin_vectorization_cost): Handle polynomial
12996 TYPE_VECTOR_SUBPARTS.
12997 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
12999 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
13000 number of elements from the PARALLEL rather than the mode.
13001 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
13002 rather than GET_MODE_BITSIZE.
13003 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
13004 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
13005 (aarch64_expand_vec_perm_const_1): Handle polynomial
13006 d->perm.length () and d->perm elements.
13007 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
13008 Apply to_constant to d->perm elements.
13009 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
13010 polynomial CONST_VECTOR_NUNITS.
13011 (aarch64_move_pointer): Take amount as a poly_int64 rather
13013 (aarch64_progress_pointer): Avoid temporary variable.
13014 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
13015 the mode attribute instead of GET_MODE.
13017 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13018 Alan Hayward <alan.hayward@arm.com>
13019 David Sherwood <david.sherwood@arm.com>
13021 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
13022 x exists before using it.
13023 (aarch64_add_constant_internal): Rename to...
13024 (aarch64_add_offset_1): ...this. Replace regnum with separate
13025 src and dest rtxes. Handle the case in which they're different,
13026 including when the offset is zero. Replace scratchreg with an rtx.
13027 Use 2 additions if there is no spare register into which we can
13028 move a 16-bit constant.
13029 (aarch64_add_constant): Delete.
13030 (aarch64_add_offset): Replace reg with separate src and dest
13031 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
13032 Use aarch64_add_offset_1.
13033 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
13034 an rtx rather than an int. Take the delta as a poly_int64
13035 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
13036 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
13037 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
13038 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
13039 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
13040 and aarch64_add_sp.
13041 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
13042 aarch64_add_constant.
13044 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13046 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
13047 Use scalar_float_mode.
13049 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13051 * config/aarch64/aarch64-simd.md
13052 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
13053 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
13054 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
13055 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
13056 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
13057 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
13058 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
13059 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
13060 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
13061 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
13063 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13066 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
13067 targ_options->x_arm_arch_string is non NULL.
13069 2018-01-11 Tamar Christina <tamar.christina@arm.com>
13071 * config/aarch64/aarch64.h
13072 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
13074 2018-01-11 Sudakshina Das <sudi.das@arm.com>
13077 * expmed.c (emit_store_flag_force): Swap if const op0
13078 and change VOIDmode to mode of op0.
13080 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13082 PR rtl-optimization/83761
13083 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
13084 than bytes to mode_for_size.
13086 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13088 PR middle-end/83189
13089 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
13090 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
13093 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13095 PR middle-end/83575
13096 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
13097 when in layout mode.
13098 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
13099 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
13102 2018-01-10 Michael Collison <michael.collison@arm.com>
13104 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
13105 * config/aarch64/aarch64-option-extension.def: Add
13106 AARCH64_OPT_EXTENSION of 'fp16fml'.
13107 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13108 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
13109 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
13110 * config/aarch64/constraints.md (Ui7): New constraint.
13111 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
13112 (VFMLA_SEL_W): Ditto.
13115 (VFMLA16_LOW): New int iterator.
13116 (VFMLA16_HIGH): Ditto.
13117 (UNSPEC_FMLAL): New unspec.
13118 (UNSPEC_FMLSL): Ditto.
13119 (UNSPEC_FMLAL2): Ditto.
13120 (UNSPEC_FMLSL2): Ditto.
13121 (f16mac): New code attribute.
13122 * config/aarch64/aarch64-simd-builtins.def
13123 (aarch64_fmlal_lowv2sf): Ditto.
13124 (aarch64_fmlsl_lowv2sf): Ditto.
13125 (aarch64_fmlalq_lowv4sf): Ditto.
13126 (aarch64_fmlslq_lowv4sf): Ditto.
13127 (aarch64_fmlal_highv2sf): Ditto.
13128 (aarch64_fmlsl_highv2sf): Ditto.
13129 (aarch64_fmlalq_highv4sf): Ditto.
13130 (aarch64_fmlslq_highv4sf): Ditto.
13131 (aarch64_fmlal_lane_lowv2sf): Ditto.
13132 (aarch64_fmlsl_lane_lowv2sf): Ditto.
13133 (aarch64_fmlal_laneq_lowv2sf): Ditto.
13134 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
13135 (aarch64_fmlalq_lane_lowv4sf): Ditto.
13136 (aarch64_fmlsl_lane_lowv4sf): Ditto.
13137 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
13138 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
13139 (aarch64_fmlal_lane_highv2sf): Ditto.
13140 (aarch64_fmlsl_lane_highv2sf): Ditto.
13141 (aarch64_fmlal_laneq_highv2sf): Ditto.
13142 (aarch64_fmlsl_laneq_highv2sf): Ditto.
13143 (aarch64_fmlalq_lane_highv4sf): Ditto.
13144 (aarch64_fmlsl_lane_highv4sf): Ditto.
13145 (aarch64_fmlalq_laneq_highv4sf): Ditto.
13146 (aarch64_fmlsl_laneq_highv4sf): Ditto.
13147 * config/aarch64/aarch64-simd.md:
13148 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
13149 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
13150 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
13151 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
13152 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
13153 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
13154 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
13155 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
13156 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
13157 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
13158 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
13159 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
13160 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
13161 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
13162 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
13163 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
13164 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
13165 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
13166 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
13167 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
13168 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
13169 (vfmlsl_low_u32): Ditto.
13170 (vfmlalq_low_u32): Ditto.
13171 (vfmlslq_low_u32): Ditto.
13172 (vfmlal_high_u32): Ditto.
13173 (vfmlsl_high_u32): Ditto.
13174 (vfmlalq_high_u32): Ditto.
13175 (vfmlslq_high_u32): Ditto.
13176 (vfmlal_lane_low_u32): Ditto.
13177 (vfmlsl_lane_low_u32): Ditto.
13178 (vfmlal_laneq_low_u32): Ditto.
13179 (vfmlsl_laneq_low_u32): Ditto.
13180 (vfmlalq_lane_low_u32): Ditto.
13181 (vfmlslq_lane_low_u32): Ditto.
13182 (vfmlalq_laneq_low_u32): Ditto.
13183 (vfmlslq_laneq_low_u32): Ditto.
13184 (vfmlal_lane_high_u32): Ditto.
13185 (vfmlsl_lane_high_u32): Ditto.
13186 (vfmlal_laneq_high_u32): Ditto.
13187 (vfmlsl_laneq_high_u32): Ditto.
13188 (vfmlalq_lane_high_u32): Ditto.
13189 (vfmlslq_lane_high_u32): Ditto.
13190 (vfmlalq_laneq_high_u32): Ditto.
13191 (vfmlslq_laneq_high_u32): Ditto.
13192 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
13193 (AARCH64_FL_FOR_ARCH8_4): New.
13194 (AARCH64_ISA_F16FML): New ISA flag.
13195 (TARGET_F16FML): New feature flag for fp16fml.
13196 (doc/invoke.texi): Document new fp16fml option.
13198 2018-01-10 Michael Collison <michael.collison@arm.com>
13200 * config/aarch64/aarch64-builtins.c:
13201 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
13202 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13203 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
13204 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
13205 (AARCH64_ISA_SHA3): New ISA flag.
13206 (TARGET_SHA3): New feature flag for sha3.
13207 * config/aarch64/iterators.md (sha512_op): New int attribute.
13208 (CRYPTO_SHA512): New int iterator.
13209 (UNSPEC_SHA512H): New unspec.
13210 (UNSPEC_SHA512H2): Ditto.
13211 (UNSPEC_SHA512SU0): Ditto.
13212 (UNSPEC_SHA512SU1): Ditto.
13213 * config/aarch64/aarch64-simd-builtins.def
13214 (aarch64_crypto_sha512hqv2di): New builtin.
13215 (aarch64_crypto_sha512h2qv2di): Ditto.
13216 (aarch64_crypto_sha512su0qv2di): Ditto.
13217 (aarch64_crypto_sha512su1qv2di): Ditto.
13218 (aarch64_eor3qv8hi): Ditto.
13219 (aarch64_rax1qv2di): Ditto.
13220 (aarch64_xarqv2di): Ditto.
13221 (aarch64_bcaxqv8hi): Ditto.
13222 * config/aarch64/aarch64-simd.md:
13223 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
13224 (aarch64_crypto_sha512su0qv2di): Ditto.
13225 (aarch64_crypto_sha512su1qv2di): Ditto.
13226 (aarch64_eor3qv8hi): Ditto.
13227 (aarch64_rax1qv2di): Ditto.
13228 (aarch64_xarqv2di): Ditto.
13229 (aarch64_bcaxqv8hi): Ditto.
13230 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
13231 (vsha512h2q_u64): Ditto.
13232 (vsha512su0q_u64): Ditto.
13233 (vsha512su1q_u64): Ditto.
13234 (veor3q_u16): Ditto.
13235 (vrax1q_u64): Ditto.
13236 (vxarq_u64): Ditto.
13237 (vbcaxq_u16): Ditto.
13238 * config/arm/types.md (crypto_sha512): New type attribute.
13239 (crypto_sha3): Ditto.
13240 (doc/invoke.texi): Document new sha3 option.
13242 2018-01-10 Michael Collison <michael.collison@arm.com>
13244 * config/aarch64/aarch64-builtins.c:
13245 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
13246 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13247 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
13248 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
13249 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
13250 (AARCH64_ISA_SM4): New ISA flag.
13251 (TARGET_SM4): New feature flag for sm4.
13252 * config/aarch64/aarch64-simd-builtins.def
13253 (aarch64_sm3ss1qv4si): Ditto.
13254 (aarch64_sm3tt1aq4si): Ditto.
13255 (aarch64_sm3tt1bq4si): Ditto.
13256 (aarch64_sm3tt2aq4si): Ditto.
13257 (aarch64_sm3tt2bq4si): Ditto.
13258 (aarch64_sm3partw1qv4si): Ditto.
13259 (aarch64_sm3partw2qv4si): Ditto.
13260 (aarch64_sm4eqv4si): Ditto.
13261 (aarch64_sm4ekeyqv4si): Ditto.
13262 * config/aarch64/aarch64-simd.md:
13263 (aarch64_sm3ss1qv4si): Ditto.
13264 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
13265 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
13266 (aarch64_sm4eqv4si): Ditto.
13267 (aarch64_sm4ekeyqv4si): Ditto.
13268 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
13269 (sm3part_op): Ditto.
13270 (CRYPTO_SM3TT): Ditto.
13271 (CRYPTO_SM3PART): Ditto.
13272 (UNSPEC_SM3SS1): New unspec.
13273 (UNSPEC_SM3TT1A): Ditto.
13274 (UNSPEC_SM3TT1B): Ditto.
13275 (UNSPEC_SM3TT2A): Ditto.
13276 (UNSPEC_SM3TT2B): Ditto.
13277 (UNSPEC_SM3PARTW1): Ditto.
13278 (UNSPEC_SM3PARTW2): Ditto.
13279 (UNSPEC_SM4E): Ditto.
13280 (UNSPEC_SM4EKEY): Ditto.
13281 * config/aarch64/constraints.md (Ui2): New constraint.
13282 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
13283 * config/arm/types.md (crypto_sm3): New type attribute.
13284 (crypto_sm4): Ditto.
13285 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
13286 (vsm3tt1aq_u32): Ditto.
13287 (vsm3tt1bq_u32): Ditto.
13288 (vsm3tt2aq_u32): Ditto.
13289 (vsm3tt2bq_u32): Ditto.
13290 (vsm3partw1q_u32): Ditto.
13291 (vsm3partw2q_u32): Ditto.
13292 (vsm4eq_u32): Ditto.
13293 (vsm4ekeyq_u32): Ditto.
13294 (doc/invoke.texi): Document new sm4 option.
13296 2018-01-10 Michael Collison <michael.collison@arm.com>
13298 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
13299 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
13300 (AARCH64_FL_FOR_ARCH8_4): New.
13301 (AARCH64_FL_V8_4): New flag.
13302 (doc/invoke.texi): Document new armv8.4-a option.
13304 2018-01-10 Michael Collison <michael.collison@arm.com>
13306 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13307 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
13308 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
13309 * config/aarch64/aarch64-option-extension.def: Add
13310 AARCH64_OPT_EXTENSION of 'sha2'.
13311 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
13312 (crypto): Disable sha2 and aes if crypto disabled.
13313 (crypto): Enable aes and sha2 if enabled.
13314 (simd): Disable sha2 and aes if simd disabled.
13315 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
13317 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
13318 (TARGET_SHA2): New feature flag for sha2.
13319 (TARGET_AES): New feature flag for aes.
13320 * config/aarch64/aarch64-simd.md:
13321 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
13322 conditional on TARGET_AES.
13323 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
13324 (aarch64_crypto_sha1hsi): Make pattern conditional
13326 (aarch64_crypto_sha1hv4si): Ditto.
13327 (aarch64_be_crypto_sha1hv4si): Ditto.
13328 (aarch64_crypto_sha1su1v4si): Ditto.
13329 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
13330 (aarch64_crypto_sha1su0v4si): Ditto.
13331 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
13332 (aarch64_crypto_sha256su0v4si): Ditto.
13333 (aarch64_crypto_sha256su1v4si): Ditto.
13334 (doc/invoke.texi): Document new aes and sha2 options.
13336 2018-01-10 Martin Sebor <msebor@redhat.com>
13338 PR tree-optimization/83781
13339 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
13342 2018-01-11 Martin Sebor <msebor@gmail.com>
13343 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13345 PR tree-optimization/83501
13346 PR tree-optimization/81703
13348 * tree-ssa-strlen.c (get_string_cst): Rename...
13349 (get_string_len): ...to this. Handle global constants.
13350 (handle_char_store): Adjust.
13352 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
13353 Jim Wilson <jimw@sifive.com>
13355 * config/riscv/riscv-protos.h (riscv_output_return): New.
13356 * config/riscv/riscv.c (struct machine_function): New naked_p field.
13357 (riscv_attribute_table, riscv_output_return),
13358 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
13359 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
13360 (riscv_compute_frame_info): Only compute frame->mask if not a naked
13362 (riscv_expand_prologue): Add early return for naked function.
13363 (riscv_expand_epilogue): Likewise.
13364 (riscv_function_ok_for_sibcall): Return false for naked function.
13365 (riscv_set_current_function): New.
13366 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
13367 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
13368 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
13369 * doc/extend.texi (RISC-V Function Attributes): New.
13371 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
13373 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
13374 check for 128-bit long double before checking TCmode.
13375 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
13376 128-bit long doubles before checking TFmode or TCmode.
13377 (FLOAT128_IBM_P): Likewise.
13379 2018-01-10 Martin Sebor <msebor@redhat.com>
13381 PR tree-optimization/83671
13382 * builtins.c (c_strlen): Unconditionally return zero for the empty
13384 Use -Warray-bounds for warnings.
13385 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
13386 for non-constant array indices with COMPONENT_REF, arrays of
13387 arrays, and pointers to arrays.
13388 (gimple_fold_builtin_strlen): Determine and set length range for
13389 non-constant character arrays.
13391 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
13393 PR middle-end/81897
13394 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
13397 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
13399 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
13401 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
13404 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
13405 VECTOR_MEM_ALTIVEC_OR_VSX_P.
13406 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
13407 indexed_or_indirect_operand predicate.
13408 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
13409 (*vsx_le_perm_load_v8hi): Likewise.
13410 (*vsx_le_perm_load_v16qi): Likewise.
13411 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
13412 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
13413 (*vsx_le_perm_store_v8hi): Likewise.
13414 (*vsx_le_perm_store_v16qi): Likewise.
13415 (eight unnamed splitters): Likewise.
13417 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
13419 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
13420 * config/rs6000/emmintrin.h: Likewise.
13421 * config/rs6000/mmintrin.h: Likewise.
13422 * config/rs6000/xmmintrin.h: Likewise.
13424 2018-01-10 David Malcolm <dmalcolm@redhat.com>
13427 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
13429 * tree.c (tree_nop_conversion): Return true for location wrapper
13431 (maybe_wrap_with_location): New function.
13432 (selftest::check_strip_nops): New function.
13433 (selftest::test_location_wrappers): New function.
13434 (selftest::tree_c_tests): Call it.
13435 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
13436 (maybe_wrap_with_location): New decl.
13437 (EXPR_LOCATION_WRAPPER_P): New macro.
13438 (location_wrapper_p): New inline function.
13439 (tree_strip_any_location_wrapper): New inline function.
13441 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
13444 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
13445 stack_realign_offset for the largest alignment of stack slot
13447 (ix86_find_max_used_stack_alignment): New function.
13448 (ix86_finalize_stack_frame_flags): Use it. Set
13449 max_used_stack_alignment if we don't realign stack.
13450 * config/i386/i386.h (machine_function): Add
13451 max_used_stack_alignment.
13453 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
13455 * config/arm/arm.opt (-mbranch-cost): New option.
13456 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
13459 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
13462 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
13463 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
13465 2018-01-10 Richard Biener <rguenther@suse.de>
13468 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
13469 early out so it also covers the case where we have a non-NULL
13472 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13474 PR tree-optimization/83753
13475 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
13476 for non-strided grouped accesses if the number of elements is 1.
13478 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13481 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
13482 * i386.h (TARGET_USE_GATHER): Define.
13483 * x86-tune.def (X86_TUNE_USE_GATHER): New.
13485 2018-01-10 Martin Liska <mliska@suse.cz>
13488 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
13489 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
13491 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
13492 CLEANUP_NO_PARTITIONING is not set.
13494 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13496 * doc/rtl.texi: Remove documentation of (const ...) wrappers
13497 for vectors, as a partial revert of r254296.
13498 * rtl.h (const_vec_p): Delete.
13499 (const_vec_duplicate_p): Don't test for vector CONSTs.
13500 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
13501 * expmed.c (make_tree): Likewise.
13504 * common.md (E, F): Use CONSTANT_P instead of checking for
13506 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
13507 checking for CONST_VECTOR.
13509 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13511 PR middle-end/83575
13512 * predict.c (force_edge_cold): Handle in more sane way edges
13513 with no prediction.
13515 2018-01-09 Carl Love <cel@us.ibm.com>
13517 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
13519 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
13520 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
13521 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
13522 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
13523 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
13524 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
13525 * config/rs6000/rs6000-protos.h: Add extern defition for
13526 rs6000_generate_float2_double_code.
13527 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
13529 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
13530 (float2_v2df): Add define_expand.
13532 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
13535 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
13536 op_mode in the force_to_mode call.
13538 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13540 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
13541 instead of checking each element individually.
13542 (aarch64_evpc_uzp): Likewise.
13543 (aarch64_evpc_zip): Likewise.
13544 (aarch64_evpc_ext): Likewise.
13545 (aarch64_evpc_rev): Likewise.
13546 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
13547 instead of checking each element individually. Return true without
13549 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
13550 whether all selected elements come from the same input, instead of
13551 checking each element individually. Remove calls to gen_rtx_REG,
13552 start_sequence and end_sequence and instead assert that no rtl is
13555 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13557 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
13558 order of HIGH and CONST checks.
13560 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13562 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
13563 if the destination isn't an SSA_NAME.
13565 2018-01-09 Richard Biener <rguenther@suse.de>
13567 PR tree-optimization/83668
13568 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
13570 (canonicalize_loop_form): ... here, renamed from ...
13571 (canonicalize_loop_closed_ssa_form): ... this and amended to
13572 swap successor edges for loop exit blocks to make us use
13573 the RPO order we need for initial schedule generation.
13575 2018-01-09 Joseph Myers <joseph@codesourcery.com>
13577 PR tree-optimization/64811
13578 * match.pd: When optimizing comparisons with Inf, avoid
13579 introducing or losing exceptions from comparisons with NaN.
13581 2018-01-09 Martin Liska <mliska@suse.cz>
13584 * asan.c (shadow_mem_size): Add gcc_assert.
13586 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
13588 Don't save registers in main().
13591 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
13592 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
13593 * config/avr/avr.c (avr_set_current_function): Don't error if
13594 naked, OS_task or OS_main are specified at the same time.
13595 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
13597 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
13599 * common/config/avr/avr-common.c (avr_option_optimization_table):
13600 Switch on -mmain-is-OS_task for optimizing compilations.
13602 2018-01-09 Richard Biener <rguenther@suse.de>
13604 PR tree-optimization/83572
13605 * graphite.c: Include cfganal.h.
13606 (graphite_transform_loops): Connect infinite loops to exit
13607 and remove fake edges at the end.
13609 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13611 * ipa-inline.c (edge_badness): Revert accidental checkin.
13613 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13616 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
13617 symbols; not inline clones.
13619 2018-01-09 Jakub Jelinek <jakub@redhat.com>
13622 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
13623 hard registers. Formatting fixes.
13625 PR preprocessor/83722
13626 * gcc.c (try_generate_repro): Pass
13627 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
13628 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
13631 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
13632 Kito Cheng <kito.cheng@gmail.com>
13634 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
13635 (riscv_leaf_function_p): Delete.
13636 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
13638 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13640 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
13642 (do_ifelse): New function.
13643 (do_isel): New function.
13644 (do_sub3): New function.
13645 (do_add3): New function.
13646 (do_load_mask_compare): New function.
13647 (do_overlap_load_compare): New function.
13648 (expand_compare_loop): New function.
13649 (expand_block_compare): Call expand_compare_loop() when appropriate.
13650 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
13651 option description.
13652 (-mblock-compare-inline-loop-limit): New option.
13654 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13657 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
13658 Reverse order of second and third operands in first alternative.
13659 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
13660 of first and second elements in UNSPEC_VPERMR vector.
13661 (altivec_expand_vec_perm_le): Likewise.
13663 2018-01-08 Jeff Law <law@redhat.com>
13665 PR rtl-optimizatin/81308
13666 * tree-switch-conversion.c (cfg_altered): New file scoped static.
13667 (process_switch): If group_case_labels makes a change, then set
13669 (pass_convert_switch::execute): If a switch is converted, then
13670 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
13672 PR rtl-optimization/81308
13673 * recog.c (split_all_insns): Conditionally cleanup the CFG after
13676 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
13678 PR target/83663 - Revert r255946
13679 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
13680 generation for cases where splatting a value is not useful.
13681 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
13682 across a vec_duplicate and a paradoxical subreg forming a vector
13683 mode to a vec_concat.
13685 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13687 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
13688 -march=armv8.3-a variants.
13689 * config/arm/t-multilib: Likewise.
13690 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
13692 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13694 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
13696 (cceq_ior_compare_complement): Give it a name so I can use it, and
13697 change boolean_or_operator predicate to boolean_operator so it can
13698 be used to generate a crand.
13699 (eqne): New code iterator.
13700 (bd/bd_neg): New code_attrs.
13701 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
13702 a single define_insn.
13703 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
13704 decrement (bdnzt/bdnzf/bdzt/bdzf).
13705 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
13706 with the new names of the branch decrement patterns, and added the
13707 names of the branch decrement conditional patterns.
13709 2018-01-08 Richard Biener <rguenther@suse.de>
13711 PR tree-optimization/83563
13712 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
13715 2018-01-08 Richard Biener <rguenther@suse.de>
13717 PR middle-end/83713
13718 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
13720 2018-01-08 Richard Biener <rguenther@suse.de>
13722 PR tree-optimization/83685
13723 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
13724 references to abnormals.
13726 2018-01-08 Richard Biener <rguenther@suse.de>
13729 * dwarf2out.c (output_indirect_strings): Handle empty
13730 skeleton_debug_str_hash.
13731 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
13733 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13735 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
13736 (emit_store_direct): Likewise.
13737 (arc_trampoline_adjust_address): Likewise.
13738 (arc_asm_trampoline_template): New function.
13739 (arc_initialize_trampoline): Use asm_trampoline_template.
13740 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
13741 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
13742 * config/arc/arc.md (flush_icache): Delete pattern.
13744 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13746 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
13747 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
13750 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13753 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
13754 by not USED_FOR_TARGET.
13755 (make_pass_resolve_sw_modes): Likewise.
13757 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13759 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
13762 2018-01-08 Richard Biener <rguenther@suse.de>
13764 PR middle-end/83580
13765 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
13767 2018-01-08 Richard Biener <rguenther@suse.de>
13769 PR middle-end/83517
13770 * match.pd ((t * 2) / 2) -> t): Add missing :c.
13772 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
13774 PR middle-end/81897
13775 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
13776 basic blocks with a small number of successors.
13777 (convert_control_dep_chain_into_preds): Improve handling of
13779 (dump_predicates): Split apart into...
13780 (dump_pred_chain): ...here...
13781 (dump_pred_info): ...and here.
13782 (can_one_predicate_be_invalidated_p): Add debugging printfs.
13783 (can_chain_union_be_invalidated_p): Improve check for invalidation
13785 (uninit_uses_cannot_happen): Avoid unnecessary if
13786 convert_control_dep_chain_into_preds yielded nothing.
13788 2018-01-06 Martin Sebor <msebor@redhat.com>
13790 PR tree-optimization/83640
13791 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
13792 subtracting negative offset from size.
13793 (builtin_access::overlap): Adjust offset bounds of the access to fall
13794 within the size of the object if possible.
13796 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
13798 PR rtl-optimization/83699
13799 * expmed.c (extract_bit_field_1): Restrict the vector usage of
13800 extract_bit_field_as_subreg to cases in which the extracted
13801 value is also a vector.
13803 * lra-constraints.c (process_alt_operands): Test for the equivalence
13804 substitutions when detecting a possible reload cycle.
13806 2018-01-06 Jakub Jelinek <jakub@redhat.com>
13809 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
13810 by default if flag_selective_schedling{,2}. Formatting fixes.
13812 PR rtl-optimization/83682
13813 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
13814 if it has non-VECTOR_MODE element mode.
13815 (vec_duplicate_p): Likewise.
13817 PR middle-end/83694
13818 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
13819 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
13821 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13824 * config/i386/i386-builtin.def
13825 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
13826 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
13827 Require also OPTION_MASK_ISA_AVX512F in addition to
13828 OPTION_MASK_ISA_GFNI.
13829 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
13830 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
13831 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
13832 to OPTION_MASK_ISA_GFNI.
13833 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
13834 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
13835 OPTION_MASK_ISA_AVX512BW.
13836 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
13837 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
13838 addition to OPTION_MASK_ISA_GFNI.
13839 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
13840 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
13841 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
13842 to OPTION_MASK_ISA_GFNI.
13843 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
13844 a requirement for all ISAs rather than any of them with a few
13846 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
13848 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
13849 bitmasks to be enabled with 3 exceptions, instead of requiring any
13850 enabled ISA with lots of exceptions.
13851 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
13852 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
13853 Change avx512bw in isa attribute to avx512f.
13854 * config/i386/sgxintrin.h: Add license boilerplate.
13855 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
13856 to __AVX512F__ and __AVX512VL to __AVX512VL__.
13857 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
13858 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
13860 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
13861 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
13862 temporarily sse2 rather than sse if not enabled already.
13865 * config/i386/sse.md (VI248_VLBW): Rename to ...
13866 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
13867 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
13868 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
13869 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
13870 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
13871 mode iterator instead of VI248_VLBW.
13873 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
13875 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
13876 (record_modified): Skip clobbers; add debug output.
13877 (param_change_prob): Use sreal frequencies.
13879 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13881 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
13882 punt for user-aligned variables.
13884 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13886 * tree-chrec.c (chrec_contains_symbols): Return true for
13889 2018-01-05 Sudakshina Das <sudi.das@arm.com>
13892 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
13893 of (x|y) == x for BICS pattern.
13895 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13897 PR tree-optimization/83605
13898 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
13899 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
13902 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
13904 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
13905 * config/epiphany/rtems.h: New file.
13907 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13908 Uros Bizjak <ubizjak@gmail.com>
13911 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
13912 QIreg_operand instead of register_operand predicate.
13913 * config/i386/i386.c (ix86_rop_should_change_byte_p,
13914 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
13915 comments instead of -fmitigate[-_]rop.
13917 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13920 * cgraphunit.c (symbol_table::compile): Switch to text_section
13921 before calling assembly_start debug hook.
13922 * run-rtl-passes.c (run_rtl_passes): Likewise.
13925 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13927 * tree-vrp.c (extract_range_from_binary_expr_1): Check
13928 range_int_cst_p rather than !symbolic_range_p before calling
13929 extract_range_from_multiplicative_op_1.
13931 2018-01-04 Jeff Law <law@redhat.com>
13933 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
13934 redundant test in assertion.
13936 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13938 * doc/rtl.texi: Document machine_mode wrapper classes.
13940 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13942 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
13943 using tree_to_uhwi.
13945 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13947 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
13948 the VEC_PERM_EXPR fold to fail.
13950 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13953 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
13954 to switched_sections.
13956 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13959 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
13960 test for d.testing.
13962 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
13965 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
13966 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
13968 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13971 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
13972 is BLKmode and bitpos not zero or mode change is needed.
13974 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13977 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
13980 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
13983 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
13984 instead of MULT rtx. Update all corresponding splitters.
13985 (*saddl_se): Ditto.
13986 (*ssub<modesuffix>): Ditto.
13987 (*ssubl_se): Ditto.
13988 (*cmp_sadd_di): Update split patterns.
13989 (*cmp_sadd_si): Ditto.
13990 (*cmp_sadd_sidi): Ditto.
13991 (*cmp_ssub_di): Ditto.
13992 (*cmp_ssub_si): Ditto.
13993 (*cmp_ssub_sidi): Ditto.
13994 * config/alpha/predicates.md (const23_operand): New predicate.
13995 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
13996 Look for ASHIFT, not MULT inner operand.
13997 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13999 2018-01-04 Martin Liska <mliska@suse.cz>
14001 PR gcov-profile/83669
14002 * gcov.c (output_intermediate_file): Add version to intermediate
14004 * doc/gcov.texi: Document new field 'version' in intermediate
14005 file format. Fix location of '-k' option of gcov command.
14007 2018-01-04 Martin Liska <mliska@suse.cz>
14010 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
14012 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14014 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
14016 2018-01-03 Martin Sebor <msebor@redhat.com>
14018 PR tree-optimization/83655
14019 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
14020 checking calls with invalid arguments.
14022 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14024 * tree-vect-stmts.c (vect_get_store_rhs): New function.
14025 (vectorizable_mask_load_store): Delete.
14026 (vectorizable_call): Return false for masked loads and stores.
14027 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
14028 instead of gimple_assign_rhs1.
14029 (vectorizable_load): Handle IFN_MASK_LOAD.
14030 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
14032 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14034 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
14036 (vectorizable_mask_load_store): ...here.
14037 (vectorizable_load): ...and here.
14039 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14041 * tree-vect-stmts.c (vect_build_all_ones_mask)
14042 (vect_build_zero_merge_argument): New functions, split out from...
14043 (vectorizable_load): ...here.
14045 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14047 * tree-vect-stmts.c (vect_check_store_rhs): New function,
14049 (vectorizable_mask_load_store): ...here.
14050 (vectorizable_store): ...and here.
14052 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14054 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
14056 (vectorizable_mask_load_store): ...here.
14058 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14060 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
14061 (vect_model_store_cost): Take a vec_load_store_type instead of a
14063 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
14064 (vect_model_store_cost): Take a vec_load_store_type instead of a
14066 (vectorizable_mask_load_store): Update accordingly.
14067 (vectorizable_store): Likewise.
14068 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
14070 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14072 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
14073 IFN_MASK_LOAD calls here rather than...
14074 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
14076 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14077 Alan Hayward <alan.hayward@arm.com>
14078 David Sherwood <david.sherwood@arm.com>
14080 * expmed.c (extract_bit_field_1): For vector extracts,
14081 fall back to extract_bit_field_as_subreg if vec_extract
14084 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14085 Alan Hayward <alan.hayward@arm.com>
14086 David Sherwood <david.sherwood@arm.com>
14088 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
14089 they are variable or constant sized.
14090 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
14091 slots for constant-sized data.
14093 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14094 Alan Hayward <alan.hayward@arm.com>
14095 David Sherwood <david.sherwood@arm.com>
14097 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
14098 handling COND_EXPRs with boolean comparisons, try to find a better
14099 basis for the mask type than the boolean itself.
14101 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14103 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
14104 is calculated and how it can be overridden.
14105 * genmodes.c (max_bitsize_mode_any_mode): New variable.
14106 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
14108 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
14111 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14112 Alan Hayward <alan.hayward@arm.com>
14113 David Sherwood <david.sherwood@arm.com>
14115 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
14116 Remove the mode argument.
14117 (aarch64_simd_valid_immediate): Remove the mode and inverse
14119 * config/aarch64/iterators.md (bitsize): New iterator.
14120 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
14121 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
14122 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
14123 aarch64_simd_valid_immediate.
14124 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
14125 (aarch64_reg_or_bic_imm): Likewise.
14126 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
14127 with an insn_type enum and msl with a modifier_type enum.
14128 Replace element_width with a scalar_mode. Change the shift
14129 to unsigned int. Add constructors for scalar_float_mode and
14130 scalar_int_mode elements.
14131 (aarch64_vect_float_const_representable_p): Delete.
14132 (aarch64_can_const_movi_rtx_p)
14133 (aarch64_simd_scalar_immediate_valid_for_move)
14134 (aarch64_simd_make_constant): Update call to
14135 aarch64_simd_valid_immediate.
14136 (aarch64_advsimd_valid_immediate_hs): New function.
14137 (aarch64_advsimd_valid_immediate): Likewise.
14138 (aarch64_simd_valid_immediate): Remove mode and inverse
14139 arguments. Rewrite to use the above. Use const_vec_duplicate_p
14140 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
14141 and aarch64_float_const_representable_p on the result.
14142 (aarch64_output_simd_mov_immediate): Remove mode argument.
14143 Update call to aarch64_simd_valid_immediate and use of
14144 simd_immediate_info.
14145 (aarch64_output_scalar_simd_mov_immediate): Update call
14148 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14149 Alan Hayward <alan.hayward@arm.com>
14150 David Sherwood <david.sherwood@arm.com>
14152 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
14153 (mode_nunits): Likewise CONST_MODE_NUNITS.
14154 * machmode.def (ADJUST_NUNITS): Document.
14155 * genmodes.c (mode_data::need_nunits_adj): New field.
14156 (blank_mode): Update accordingly.
14157 (adj_nunits): New variable.
14158 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
14160 (emit_mode_size_inline): Set need_bytesize_adj for all modes
14161 listed in adj_nunits.
14162 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
14163 listed in adj_nunits. Don't emit case statements for such modes.
14164 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
14165 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
14166 nothing if adj_nunits is nonnull.
14167 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
14168 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
14169 (emit_mode_fbit): Update use of print_maybe_const_decl.
14170 (emit_move_size): Likewise. Treat the array as non-const
14172 (emit_mode_adjustments): Handle adj_nunits.
14174 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14176 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
14177 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
14178 (VECTOR_MODES): Use it.
14179 (make_vector_modes): Take the prefix as an argument.
14181 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14182 Alan Hayward <alan.hayward@arm.com>
14183 David Sherwood <david.sherwood@arm.com>
14185 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
14186 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
14187 for MODE_VECTOR_BOOL.
14188 * machmode.def (VECTOR_BOOL_MODE): Document.
14189 * genmodes.c (VECTOR_BOOL_MODE): New macro.
14190 (make_vector_bool_mode): New function.
14191 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
14193 * lto-streamer-in.c (lto_input_mode_table): Likewise.
14194 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
14196 * stor-layout.c (int_mode_for_mode): Likewise.
14197 * tree.c (build_vector_type_for_mode): Likewise.
14198 * varasm.c (output_constant_pool_2): Likewise.
14199 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
14200 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
14201 for MODE_VECTOR_BOOL.
14202 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
14203 of mode class checks.
14204 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
14205 instead of a list of mode class checks.
14206 (expand_vector_scalar_condition): Likewise.
14207 (type_for_widest_vector_mode): Handle BImode as an inner mode.
14209 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14210 Alan Hayward <alan.hayward@arm.com>
14211 David Sherwood <david.sherwood@arm.com>
14213 * machmode.h (mode_size): Change from unsigned short to
14215 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
14216 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
14217 or if measurement_type is not polynomial.
14218 (fixed_size_mode::includes_p): Check for constant-sized modes.
14219 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
14220 return a poly_uint16 rather than an unsigned short.
14221 (emit_mode_size): Change the type of mode_size from unsigned short
14222 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
14223 (emit_mode_adjustments): Cope with polynomial vector sizes.
14224 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14226 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14228 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
14229 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
14230 * caller-save.c (setup_save_areas): Likewise.
14231 (replace_reg_with_saved_mem): Likewise.
14232 * calls.c (emit_library_call_value_1): Likewise.
14233 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
14234 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
14235 (gen_lowpart_for_combine): Likewise.
14236 * convert.c (convert_to_integer_1): Likewise.
14237 * cse.c (equiv_constant, cse_insn): Likewise.
14238 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
14239 (cselib_subst_to_values): Likewise.
14240 * dce.c (word_dce_process_block): Likewise.
14241 * df-problems.c (df_word_lr_mark_ref): Likewise.
14242 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
14243 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
14244 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
14245 (rtl_for_decl_location): Likewise.
14246 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
14247 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
14248 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
14249 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
14250 (expand_expr_real_1): Likewise.
14251 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
14252 (pad_below): Likewise.
14253 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
14254 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
14255 * ira.c (get_subreg_tracking_sizes): Likewise.
14256 * ira-build.c (ira_create_allocno_objects): Likewise.
14257 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
14258 (ira_sort_regnos_for_alter_reg): Likewise.
14259 * ira-costs.c (record_operand_costs): Likewise.
14260 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
14261 (resolve_simple_move): Likewise.
14262 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
14263 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
14264 (lra_constraints): Likewise.
14265 (CONST_POOL_OK_P): Reject variable-sized modes.
14266 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
14267 (add_pseudo_to_slot, lra_spill): Likewise.
14268 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14269 * optabs-query.c (get_best_extraction_insn): Likewise.
14270 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14271 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
14272 (expand_mult_highpart, valid_multiword_target_p): Likewise.
14273 * recog.c (offsettable_address_addr_space_p): Likewise.
14274 * regcprop.c (maybe_mode_change): Likewise.
14275 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
14276 * regrename.c (build_def_use): Likewise.
14277 * regstat.c (dump_reg_info): Likewise.
14278 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
14279 (find_reloads, find_reloads_subreg_address): Likewise.
14280 * reload1.c (eliminate_regs_1): Likewise.
14281 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
14282 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
14283 (simplify_binary_operation_1, simplify_subreg): Likewise.
14284 * targhooks.c (default_function_arg_padding): Likewise.
14285 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
14286 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
14287 (verify_gimple_assign_ternary): Likewise.
14288 * tree-inline.c (estimate_move_cost): Likewise.
14289 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14290 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
14291 (get_address_cost_ainc): Likewise.
14292 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
14293 (vect_supportable_dr_alignment): Likewise.
14294 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14295 (vectorizable_reduction): Likewise.
14296 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
14297 (vectorizable_operation, vectorizable_load): Likewise.
14298 * tree.c (build_same_sized_truth_vector_type): Likewise.
14299 * valtrack.c (cleanup_auto_inc_dec): Likewise.
14300 * var-tracking.c (emit_note_insn_var_location): Likewise.
14301 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
14302 (ADDR_VEC_ALIGN): Likewise.
14304 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14305 Alan Hayward <alan.hayward@arm.com>
14306 David Sherwood <david.sherwood@arm.com>
14308 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
14310 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
14311 or if measurement_type is polynomial.
14312 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
14313 * combine.c (make_extraction): Likewise.
14314 * dse.c (find_shift_sequence): Likewise.
14315 * dwarf2out.c (mem_loc_descriptor): Likewise.
14316 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
14317 (extract_bit_field, extract_low_bits): Likewise.
14318 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
14319 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
14320 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
14321 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
14322 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
14323 * reload.c (find_reloads): Likewise.
14324 * reload1.c (alter_reg): Likewise.
14325 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
14326 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
14327 * tree-if-conv.c (predicate_mem_writes): Likewise.
14328 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
14329 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
14330 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
14331 * valtrack.c (dead_debug_insert_temp): Likewise.
14332 * varasm.c (mergeable_constant_section): Likewise.
14333 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
14335 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14336 Alan Hayward <alan.hayward@arm.com>
14337 David Sherwood <david.sherwood@arm.com>
14339 * expr.c (expand_assignment): Cope with polynomial mode sizes
14340 when assigning to a CONCAT.
14342 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14343 Alan Hayward <alan.hayward@arm.com>
14344 David Sherwood <david.sherwood@arm.com>
14346 * machmode.h (mode_precision): Change from unsigned short to
14348 (mode_to_precision): Return a poly_uint16 rather than an unsigned
14350 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
14351 or if measurement_type is not polynomial.
14352 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
14353 in which the mode is already known to be a scalar_int_mode.
14354 * genmodes.c (emit_mode_precision): Change the type of mode_precision
14355 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
14357 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14358 for GET_MODE_PRECISION.
14359 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14360 for GET_MODE_PRECISION.
14361 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
14363 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
14364 (expand_field_assignment, make_extraction): Likewise.
14365 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
14366 (get_last_value): Likewise.
14367 * convert.c (convert_to_integer_1): Likewise.
14368 * cse.c (cse_insn): Likewise.
14369 * expr.c (expand_expr_real_1): Likewise.
14370 * lra-constraints.c (simplify_operand_subreg): Likewise.
14371 * optabs-query.c (can_atomic_load_p): Likewise.
14372 * optabs.c (expand_atomic_load): Likewise.
14373 (expand_atomic_store): Likewise.
14374 * ree.c (combine_reaching_defs): Likewise.
14375 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
14376 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
14377 * tree.h (type_has_mode_precision_p): Likewise.
14378 * ubsan.c (instrument_si_overflow): Likewise.
14380 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14381 Alan Hayward <alan.hayward@arm.com>
14382 David Sherwood <david.sherwood@arm.com>
14384 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
14385 polynomial numbers of units.
14386 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
14387 (valid_vector_subparts_p): New function.
14388 (build_vector_type): Remove temporary shim and take the number
14389 of units as a poly_uint64 rather than an int.
14390 (build_opaque_vector_type): Take the number of units as a
14391 poly_uint64 rather than an int.
14392 * tree.c (build_vector_from_ctor): Handle polynomial
14393 TYPE_VECTOR_SUBPARTS.
14394 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
14395 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
14396 (build_vector_from_val): If the number of units is variable,
14397 use build_vec_duplicate_cst for constant operands and
14398 VEC_DUPLICATE_EXPR otherwise.
14399 (make_vector_type): Remove temporary is_constant ().
14400 (build_vector_type, build_opaque_vector_type): Take the number of
14401 units as a poly_uint64 rather than an int.
14402 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
14404 * cfgexpand.c (expand_debug_expr): Likewise.
14405 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
14406 (store_constructor, expand_expr_real_1): Likewise.
14407 (const_scalar_mask_from_tree): Likewise.
14408 * fold-const-call.c (fold_const_reduction): Likewise.
14409 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
14410 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
14411 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
14412 (fold_relational_const): Likewise.
14413 (native_interpret_vector): Likewise. Change the size from an
14414 int to an unsigned int.
14415 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
14416 TYPE_VECTOR_SUBPARTS.
14417 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
14418 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
14419 duplicating a non-constant operand into a variable-length vector.
14420 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
14421 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
14422 * ipa-icf.c (sem_variable::equals): Likewise.
14423 * match.pd: Likewise.
14424 * omp-simd-clone.c (simd_clone_subparts): Likewise.
14425 * print-tree.c (print_node): Likewise.
14426 * stor-layout.c (layout_type): Likewise.
14427 * targhooks.c (default_builtin_vectorization_cost): Likewise.
14428 * tree-cfg.c (verify_gimple_comparison): Likewise.
14429 (verify_gimple_assign_binary): Likewise.
14430 (verify_gimple_assign_ternary): Likewise.
14431 (verify_gimple_assign_single): Likewise.
14432 * tree-pretty-print.c (dump_generic_node): Likewise.
14433 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14434 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
14435 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
14436 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
14437 (vect_shift_permute_load_chain): Likewise.
14438 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
14439 (expand_vector_condition, optimize_vector_constructor): Likewise.
14440 (lower_vec_perm, get_compute_type): Likewise.
14441 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14442 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
14443 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
14444 (vect_recog_mask_conversion_pattern): Likewise.
14445 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
14446 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
14447 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14448 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
14449 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
14450 (vectorizable_shift, vectorizable_operation, vectorizable_store)
14451 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
14452 (supportable_widening_operation): Likewise.
14453 (supportable_narrowing_operation): Likewise.
14454 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
14456 * varasm.c (output_constant): Likewise.
14458 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14459 Alan Hayward <alan.hayward@arm.com>
14460 David Sherwood <david.sherwood@arm.com>
14462 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
14463 so that both the length == 3 and length != 3 cases set up their
14464 own permute vectors. Add comments explaining why we know the
14465 number of elements is constant.
14466 (vect_permute_load_chain): Likewise.
14468 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14469 Alan Hayward <alan.hayward@arm.com>
14470 David Sherwood <david.sherwood@arm.com>
14472 * machmode.h (mode_nunits): Change from unsigned char to
14474 (ONLY_FIXED_SIZE_MODES): New macro.
14475 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
14476 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
14477 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
14479 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
14480 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
14481 or if measurement_type is not polynomial.
14482 * genmodes.c (ZERO_COEFFS): New macro.
14483 (emit_mode_nunits_inline): Make mode_nunits_inline return a
14485 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
14486 Use ZERO_COEFFS when emitting initializers.
14487 * data-streamer.h (bp_pack_poly_value): New function.
14488 (bp_unpack_poly_value): Likewise.
14489 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14490 for GET_MODE_NUNITS.
14491 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14492 for GET_MODE_NUNITS.
14493 * tree.c (make_vector_type): Remove temporary shim and make
14494 the real function take the number of units as a poly_uint64
14495 rather than an int.
14496 (build_vector_type_for_mode): Handle polynomial nunits.
14497 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
14498 * emit-rtl.c (const_vec_series_p_1): Likewise.
14499 (gen_rtx_CONST_VECTOR): Likewise.
14500 * fold-const.c (test_vec_duplicate_folding): Likewise.
14501 * genrecog.c (validate_pattern): Likewise.
14502 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
14503 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14504 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
14505 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
14506 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
14507 * rtlanal.c (subreg_get_info): Likewise.
14508 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14509 (vect_grouped_load_supported): Likewise.
14510 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
14511 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
14512 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
14513 (simplify_const_unary_operation, simplify_binary_operation_1)
14514 (simplify_const_binary_operation, simplify_ternary_operation)
14515 (test_vector_ops_duplicate, test_vector_ops): Likewise.
14516 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
14517 instead of CONST_VECTOR_NUNITS.
14518 * varasm.c (output_constant_pool_2): Likewise.
14519 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
14520 explicit-encoded elements in the XVEC for variable-length vectors.
14522 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14524 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
14526 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14527 Alan Hayward <alan.hayward@arm.com>
14528 David Sherwood <david.sherwood@arm.com>
14530 * coretypes.h (fixed_size_mode): Declare.
14531 (fixed_size_mode_pod): New typedef.
14532 * builtins.h (target_builtins::x_apply_args_mode)
14533 (target_builtins::x_apply_result_mode): Change type to
14534 fixed_size_mode_pod.
14535 * builtins.c (apply_args_size, apply_result_size, result_vector)
14536 (expand_builtin_apply_args_1, expand_builtin_apply)
14537 (expand_builtin_return): Update accordingly.
14539 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14541 * cse.c (hash_rtx_cb): Hash only the encoded elements.
14542 * cselib.c (cselib_hash_rtx): Likewise.
14543 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
14544 CONST_VECTOR encoding.
14546 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14547 Jeff Law <law@redhat.com>
14550 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
14551 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
14552 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
14553 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
14556 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
14557 explicitly probe *sp in a noreturn function if there were any callee
14558 register saves or frame pointer is needed.
14560 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14563 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
14564 BLKmode for ternary, binary or unary expressions.
14567 * var-tracking.c (delete_vta_debug_insn): New inline function.
14568 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
14569 insns from get_insns () to NULL instead of each bb separately.
14570 Use delete_vta_debug_insn. No longer static.
14571 (vt_debug_insns_local, variable_tracking_main_1): Adjust
14572 delete_vta_debug_insns callers.
14573 * rtl.h (delete_vta_debug_insns): Declare.
14574 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
14575 instead of variable_tracking_main.
14577 2018-01-03 Martin Sebor <msebor@redhat.com>
14579 PR tree-optimization/83603
14580 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
14581 arguments past the endof the argument list in functions declared
14582 without a prototype.
14583 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
14584 Avoid checking when arguments are null.
14586 2018-01-03 Martin Sebor <msebor@redhat.com>
14589 * doc/extend.texi (attribute const): Fix a typo.
14590 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
14591 issuing -Wsuggest-attribute for void functions.
14593 2018-01-03 Martin Sebor <msebor@redhat.com>
14595 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
14596 offset_int::from instead of wide_int::to_shwi.
14597 (maybe_diag_overlap): Remove assertion.
14598 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
14599 * gimple-ssa-sprintf.c (format_directive): Same.
14600 (parse_directive): Same.
14601 (sprintf_dom_walker::compute_format_length): Same.
14602 (try_substitute_return_value): Same.
14604 2018-01-03 Jeff Law <law@redhat.com>
14606 PR middle-end/83654
14607 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
14608 non-constant residual for zero at runtime and avoid probing in
14609 that case. Reorganize code for trailing problem to mirror handling
14612 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14614 PR tree-optimization/83501
14615 * tree-ssa-strlen.c (get_string_cst): New.
14616 (handle_char_store): Call get_string_cst.
14618 2018-01-03 Martin Liska <mliska@suse.cz>
14620 PR tree-optimization/83593
14621 * tree-ssa-strlen.c: Include tree-cfg.h.
14622 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
14623 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
14624 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
14626 (strlen_dom_walker::before_dom_children): Call
14627 gimple_purge_dead_eh_edges. Dump tranformation with details
14629 (strlen_dom_walker::before_dom_children): Update call by adding
14630 new argument cleanup_eh.
14631 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
14633 2018-01-03 Martin Liska <mliska@suse.cz>
14636 * cif-code.def (VARIADIC_THUNK): New enum value.
14637 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
14640 2018-01-03 Jan Beulich <jbeulich@suse.com>
14642 * sse.md (mov<mode>_internal): Tighten condition for when to use
14643 vmovdqu<ssescalarsize> for TI and OI modes.
14645 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14647 Update copyright years.
14649 2018-01-03 Martin Liska <mliska@suse.cz>
14652 * ipa-visibility.c (function_and_variable_visibility): Skip
14653 functions with noipa attribure.
14655 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14657 * gcc.c (process_command): Update copyright notice dates.
14658 * gcov-dump.c (print_version): Ditto.
14659 * gcov.c (print_version): Ditto.
14660 * gcov-tool.c (print_version): Ditto.
14661 * gengtype.c (create_file): Ditto.
14662 * doc/cpp.texi: Bump @copying's copyright year.
14663 * doc/cppinternals.texi: Ditto.
14664 * doc/gcc.texi: Ditto.
14665 * doc/gccint.texi: Ditto.
14666 * doc/gcov.texi: Ditto.
14667 * doc/install.texi: Ditto.
14668 * doc/invoke.texi: Ditto.
14670 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14672 * vector-builder.h (vector_builder::m_full_nelts): Change from
14673 unsigned int to poly_uint64.
14674 (vector_builder::full_nelts): Update prototype accordingly.
14675 (vector_builder::new_vector): Likewise.
14676 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
14677 (vector_builder::operator ==): Likewise.
14678 (vector_builder::finalize): Likewise.
14679 * int-vector-builder.h (int_vector_builder::int_vector_builder):
14680 Take the number of elements as a poly_uint64 rather than an
14682 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
14683 from unsigned int to poly_uint64.
14684 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
14685 (vec_perm_indices::new_vector): Likewise.
14686 (vec_perm_indices::length): Likewise.
14687 (vec_perm_indices::nelts_per_input): Likewise.
14688 (vec_perm_indices::input_nelts): Likewise.
14689 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
14690 number of elements per input as a poly_uint64 rather than an
14691 unsigned int. Use the original encoding for variable-length
14692 vectors, rather than clamping each individual element.
14693 For the second and subsequent elements in each pattern,
14694 clamp the step and base before clamping their sum.
14695 (vec_perm_indices::series_p): Handle polynomial element counts.
14696 (vec_perm_indices::all_in_range_p): Likewise.
14697 (vec_perm_indices_to_tree): Likewise.
14698 (vec_perm_indices_to_rtx): Likewise.
14699 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
14700 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
14701 (tree_vector_builder::new_binary_operation): Handle polynomial
14702 element counts. Return false if we need to know the number
14703 of elements at compile time.
14704 * fold-const.c (fold_vec_perm): Punt if the number of elements
14705 isn't known at compile time.
14707 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14709 * vec-perm-indices.h (vec_perm_builder): Change element type
14710 from HOST_WIDE_INT to poly_int64.
14711 (vec_perm_indices::element_type): Update accordingly.
14712 (vec_perm_indices::clamp): Handle polynomial element_types.
14713 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14714 (vec_perm_indices::all_in_range_p): Likewise.
14715 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
14717 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
14718 polynomial vec_perm_indices element types.
14719 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
14720 * fold-const.c (fold_vec_perm): Likewise.
14721 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
14722 * tree-vect-generic.c (lower_vec_perm): Likewise.
14723 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14724 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
14725 element type to HOST_WIDE_INT.
14727 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14728 Alan Hayward <alan.hayward@arm.com>
14729 David Sherwood <david.sherwood@arm.com>
14731 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
14732 rather than an int. Use plus_constant.
14733 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
14734 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
14736 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14737 Alan Hayward <alan.hayward@arm.com>
14738 David Sherwood <david.sherwood@arm.com>
14740 * calls.c (emit_call_1, expand_call): Change struct_value_size from
14741 a HOST_WIDE_INT to a poly_int64.
14743 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14744 Alan Hayward <alan.hayward@arm.com>
14745 David Sherwood <david.sherwood@arm.com>
14747 * calls.c (load_register_parameters): Cope with polynomial
14748 mode sizes. Require a constant size for BLKmode parameters
14749 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
14750 forces a parameter to be padded at the lsb end in order to
14751 fill a complete number of words, require the parameter size
14752 to be ordered wrt UNITS_PER_WORD.
14754 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14755 Alan Hayward <alan.hayward@arm.com>
14756 David Sherwood <david.sherwood@arm.com>
14758 * reload1.c (spill_stack_slot_width): Change element type
14759 from unsigned int to poly_uint64_pod.
14760 (alter_reg): Treat mode sizes as polynomial.
14762 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14763 Alan Hayward <alan.hayward@arm.com>
14764 David Sherwood <david.sherwood@arm.com>
14766 * reload.c (complex_word_subreg_p): New function.
14767 (reload_inner_reg_of_subreg, push_reload): Use it.
14769 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14770 Alan Hayward <alan.hayward@arm.com>
14771 David Sherwood <david.sherwood@arm.com>
14773 * lra-constraints.c (process_alt_operands): Reject matched
14774 operands whose sizes aren't ordered.
14775 (match_reload): Refer to this check here.
14777 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14778 Alan Hayward <alan.hayward@arm.com>
14779 David Sherwood <david.sherwood@arm.com>
14781 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
14782 that the mode size is in the set {1, 2, 4, 8, 16}.
14784 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14785 Alan Hayward <alan.hayward@arm.com>
14786 David Sherwood <david.sherwood@arm.com>
14788 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
14789 Use plus_constant instead of gen_rtx_PLUS.
14791 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14792 Alan Hayward <alan.hayward@arm.com>
14793 David Sherwood <david.sherwood@arm.com>
14795 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
14796 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
14797 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
14798 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
14799 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
14800 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
14801 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
14802 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
14803 * config/i386/i386.c (ix86_push_rounding): ...this new function.
14804 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
14806 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
14807 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
14808 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
14809 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
14810 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
14811 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
14812 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
14813 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
14814 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
14815 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
14817 * expr.c (emit_move_resolve_push): Treat the input and result
14818 of PUSH_ROUNDING as a poly_int64.
14819 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
14820 (emit_push_insn): Likewise.
14821 * lra-eliminations.c (mark_not_eliminable): Likewise.
14822 * recog.c (push_operand): Likewise.
14823 * reload1.c (elimination_effects): Likewise.
14824 * rtlanal.c (nonzero_bits1): Likewise.
14825 * calls.c (store_one_arg): Likewise. Require the padding to be
14826 known at compile time.
14828 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14829 Alan Hayward <alan.hayward@arm.com>
14830 David Sherwood <david.sherwood@arm.com>
14832 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
14833 Use plus_constant instead of gen_rtx_PLUS.
14835 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14836 Alan Hayward <alan.hayward@arm.com>
14837 David Sherwood <david.sherwood@arm.com>
14839 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
14840 rather than an int.
14842 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14843 Alan Hayward <alan.hayward@arm.com>
14844 David Sherwood <david.sherwood@arm.com>
14846 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
14847 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
14848 via stack temporaries. Treat the mode size as polynomial too.
14850 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14851 Alan Hayward <alan.hayward@arm.com>
14852 David Sherwood <david.sherwood@arm.com>
14854 * expr.c (expand_expr_real_2): When handling conversions involving
14855 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
14856 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
14857 as a poly_uint64 too.
14859 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14860 Alan Hayward <alan.hayward@arm.com>
14861 David Sherwood <david.sherwood@arm.com>
14863 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
14865 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14866 Alan Hayward <alan.hayward@arm.com>
14867 David Sherwood <david.sherwood@arm.com>
14869 * combine.c (can_change_dest_mode): Handle polynomial
14870 REGMODE_NATURAL_SIZE.
14871 * expmed.c (store_bit_field_1): Likewise.
14872 * expr.c (store_constructor): Likewise.
14873 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
14874 and polynomial REGMODE_NATURAL_SIZE.
14875 (gen_lowpart_common): Likewise.
14876 * reginfo.c (record_subregs_of_mode): Likewise.
14877 * rtlanal.c (read_modify_subreg_p): Likewise.
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 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
14884 numbers of elements.
14886 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14887 Alan Hayward <alan.hayward@arm.com>
14888 David Sherwood <david.sherwood@arm.com>
14890 * match.pd: Cope with polynomial numbers of vector elements.
14892 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14893 Alan Hayward <alan.hayward@arm.com>
14894 David Sherwood <david.sherwood@arm.com>
14896 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
14897 in a POINTER_PLUS_EXPR.
14899 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14900 Alan Hayward <alan.hayward@arm.com>
14901 David Sherwood <david.sherwood@arm.com>
14903 * omp-simd-clone.c (simd_clone_subparts): New function.
14904 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
14905 (ipa_simd_modify_function_body): Likewise.
14907 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14908 Alan Hayward <alan.hayward@arm.com>
14909 David Sherwood <david.sherwood@arm.com>
14911 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
14912 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
14913 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
14914 (expand_vector_condition, vector_element): Likewise.
14915 (subparts_gt): New function.
14916 (get_compute_type): Use subparts_gt.
14917 (count_type_subparts): Delete.
14918 (expand_vector_operations_1): Use subparts_gt instead of
14919 count_type_subparts.
14921 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14922 Alan Hayward <alan.hayward@arm.com>
14923 David Sherwood <david.sherwood@arm.com>
14925 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
14926 (vect_compile_time_alias): ...this new function. Do the calculation
14927 on poly_ints rather than trees.
14928 (vect_prune_runtime_alias_test_list): Update call accordingly.
14930 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14931 Alan Hayward <alan.hayward@arm.com>
14932 David Sherwood <david.sherwood@arm.com>
14934 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
14936 (vect_schedule_slp_instance): Likewise.
14938 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14939 Alan Hayward <alan.hayward@arm.com>
14940 David Sherwood <david.sherwood@arm.com>
14942 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
14943 constant and extern definitions for variable-length vectors.
14944 (vect_get_constant_vectors): Note that the number of units
14945 is known to be constant.
14947 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14948 Alan Hayward <alan.hayward@arm.com>
14949 David Sherwood <david.sherwood@arm.com>
14951 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
14952 of units as polynomial. Choose between WIDE and NARROW based
14955 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14956 Alan Hayward <alan.hayward@arm.com>
14957 David Sherwood <david.sherwood@arm.com>
14959 * tree-vect-stmts.c (simd_clone_subparts): New function.
14960 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
14962 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14963 Alan Hayward <alan.hayward@arm.com>
14964 David Sherwood <david.sherwood@arm.com>
14966 * tree-vect-stmts.c (vectorizable_call): Treat the number of
14967 vectors as polynomial. Use build_index_vector for
14968 IFN_GOMP_SIMD_LANE.
14970 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14971 Alan Hayward <alan.hayward@arm.com>
14972 David Sherwood <david.sherwood@arm.com>
14974 * tree-vect-stmts.c (get_load_store_type): Treat the number of
14975 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
14976 for variable-length vectors.
14977 (vectorizable_mask_load_store): Treat the number of units as
14978 polynomial, asserting that it is constant if the condition has
14979 already been enforced.
14980 (vectorizable_store, vectorizable_load): Likewise.
14982 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14983 Alan Hayward <alan.hayward@arm.com>
14984 David Sherwood <david.sherwood@arm.com>
14986 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
14987 of units as polynomial. Punt if we can't tell at compile time
14988 which vector contains the final result.
14990 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14991 Alan Hayward <alan.hayward@arm.com>
14992 David Sherwood <david.sherwood@arm.com>
14994 * tree-vect-loop.c (vectorizable_induction): Treat the number
14995 of units as polynomial. Punt on SLP inductions. Use an integer
14996 VEC_SERIES_EXPR for variable-length integer reductions. Use a
14997 cast of such a series for variable-length floating-point
15000 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15001 Alan Hayward <alan.hayward@arm.com>
15002 David Sherwood <david.sherwood@arm.com>
15004 * tree.h (build_index_vector): Declare.
15005 * tree.c (build_index_vector): New function.
15006 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
15007 of units as polynomial, forcibly converting it to a constant if
15008 vectorizable_reduction has already enforced the condition.
15009 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
15010 to create a {1,2,3,...} vector.
15011 (vectorizable_reduction): Treat the number of units as polynomial.
15012 Choose vectype_in based on the largest scalar element size rather
15013 than the smallest number of units. Enforce the restrictions
15016 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15017 Alan Hayward <alan.hayward@arm.com>
15018 David Sherwood <david.sherwood@arm.com>
15020 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
15021 number of units as polynomial.
15023 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15024 Alan Hayward <alan.hayward@arm.com>
15025 David Sherwood <david.sherwood@arm.com>
15027 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
15028 * target.def (autovectorize_vector_sizes): Return the vector sizes
15029 by pointer, using vector_sizes rather than a bitmask.
15030 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
15031 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
15032 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
15034 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
15035 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
15036 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
15037 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
15038 * omp-general.c (omp_max_vf): Likewise.
15039 * omp-low.c (omp_clause_aligned_alignment): Likewise.
15040 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
15041 * tree-vect-loop.c (vect_analyze_loop): Likewise.
15042 * tree-vect-slp.c (vect_slp_bb): Likewise.
15043 * doc/tm.texi: Regenerate.
15044 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
15046 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
15047 the vector size as a poly_uint64 rather than an unsigned int.
15048 (current_vector_size): Change from an unsigned int to a poly_uint64.
15049 (get_vectype_for_scalar_type): Update accordingly.
15050 * tree.h (build_truth_vector_type): Take the size and number of
15051 units as a poly_uint64 rather than an unsigned int.
15052 (build_vector_type): Add a temporary overload that takes
15053 the number of units as a poly_uint64 rather than an unsigned int.
15054 * tree.c (make_vector_type): Likewise.
15055 (build_truth_vector_type): Take the number of units as a poly_uint64
15056 rather than an unsigned int.
15058 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15059 Alan Hayward <alan.hayward@arm.com>
15060 David Sherwood <david.sherwood@arm.com>
15062 * target.def (get_mask_mode): Take the number of units and length
15063 as poly_uint64s rather than unsigned ints.
15064 * targhooks.h (default_get_mask_mode): Update accordingly.
15065 * targhooks.c (default_get_mask_mode): Likewise.
15066 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
15067 * doc/tm.texi: Regenerate.
15069 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15070 Alan Hayward <alan.hayward@arm.com>
15071 David Sherwood <david.sherwood@arm.com>
15073 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
15074 * omp-general.c (omp_max_vf): Likewise.
15075 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
15076 (expand_omp_simd): Handle polynomial safelen.
15077 * omp-low.c (omplow_simd_context): Add a default constructor.
15078 (omplow_simd_context::max_vf): Change from int to poly_uint64.
15079 (lower_rec_simd_input_clauses): Update accordingly.
15080 (lower_rec_input_clauses): Likewise.
15082 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15083 Alan Hayward <alan.hayward@arm.com>
15084 David Sherwood <david.sherwood@arm.com>
15086 * tree-vectorizer.h (vect_nunits_for_cost): New function.
15087 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
15088 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
15089 (vect_analyze_slp_cost): Likewise.
15090 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
15091 (vect_model_load_cost): Likewise.
15093 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15094 Alan Hayward <alan.hayward@arm.com>
15095 David Sherwood <david.sherwood@arm.com>
15097 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
15098 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
15099 from an unsigned int * to a poly_uint64_pod *.
15100 (calculate_unrolling_factor): New function.
15101 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
15103 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15104 Alan Hayward <alan.hayward@arm.com>
15105 David Sherwood <david.sherwood@arm.com>
15107 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
15108 from an unsigned int to a poly_uint64.
15109 (_loop_vec_info::slp_unrolling_factor): Likewise.
15110 (_loop_vec_info::vectorization_factor): Change from an int
15112 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
15113 (vect_get_num_vectors): New function.
15114 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
15115 (vect_get_num_copies): Use vect_get_num_vectors.
15116 (vect_analyze_data_ref_dependences): Change max_vf from an int *
15117 to an unsigned int *.
15118 (vect_analyze_data_refs): Change min_vf from an int * to a
15120 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
15121 than an unsigned HOST_WIDE_INT.
15122 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
15123 (vect_analyze_data_ref_dependence): Change max_vf from an int *
15124 to an unsigned int *.
15125 (vect_analyze_data_ref_dependences): Likewise.
15126 (vect_compute_data_ref_alignment): Handle polynomial vf.
15127 (vect_enhance_data_refs_alignment): Likewise.
15128 (vect_prune_runtime_alias_test_list): Likewise.
15129 (vect_shift_permute_load_chain): Likewise.
15130 (vect_supportable_dr_alignment): Likewise.
15131 (dependence_distance_ge_vf): Take the vectorization factor as a
15132 poly_uint64 rather than an unsigned HOST_WIDE_INT.
15133 (vect_analyze_data_refs): Change min_vf from an int * to a
15135 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
15136 vfm1 as a poly_uint64 rather than an int. Make the same change
15137 for the returned bound_scalar.
15138 (vect_gen_vector_loop_niters): Handle polynomial vf.
15139 (vect_do_peeling): Likewise. Update call to
15140 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
15141 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
15143 * tree-vect-loop.c (vect_determine_vectorization_factor)
15144 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
15145 (vect_get_known_peeling_cost): Likewise.
15146 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
15147 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
15148 (vect_transform_loop): Likewise. Use the lowest possible VF when
15149 updating the upper bounds of the loop.
15150 (vect_min_worthwhile_factor): Make static. Return an unsigned int
15151 rather than an int.
15152 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
15153 polynomial unroll factors.
15154 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
15155 (vect_make_slp_decision): Likewise.
15156 (vect_supported_load_permutation_p): Likewise, and polynomial
15158 (vect_analyze_slp_cost): Handle polynomial vf.
15159 (vect_slp_analyze_node_operations): Likewise.
15160 (vect_slp_analyze_bb_1): Likewise.
15161 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
15162 than an unsigned HOST_WIDE_INT.
15163 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
15164 (vectorizable_load): Handle polynomial vf.
15165 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
15167 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
15169 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15170 Alan Hayward <alan.hayward@arm.com>
15171 David Sherwood <david.sherwood@arm.com>
15173 * match.pd: Handle bit operations involving three constants
15174 and try to fold one pair.
15176 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15178 * tree-vect-loop-manip.c: Include gimple-fold.h.
15179 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
15180 niters_maybe_zero parameters. Handle other cases besides a step of 1.
15181 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
15182 Add a path that uses a step of VF instead of 1, but disable it
15184 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
15185 and niters_no_overflow parameters. Update calls to
15186 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
15187 Create a new SSA name if the latter choses to use a ste other
15188 than zero, and return it via niters_vector_mult_vf_var.
15189 * tree-vect-loop.c (vect_transform_loop): Update calls to
15190 vect_do_peeling, vect_gen_vector_loop_niters and
15191 slpeel_make_loop_iterate_ntimes.
15192 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
15193 (vect_gen_vector_loop_niters): Update declarations after above changes.
15195 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
15197 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
15198 128-bit round to integer instructions.
15199 (ceil<mode>2): Likewise.
15200 (btrunc<mode>2): Likewise.
15201 (round<mode>2): Likewise.
15203 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15205 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
15206 unaligned VSX load/store on P8/P9.
15207 (expand_block_clear): Allow the use of unaligned VSX
15208 load/store on P8/P9.
15210 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15212 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
15214 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
15215 swap associated with both a load and a store.
15217 2018-01-02 Andrew Waterman <andrew@sifive.com>
15219 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
15220 * config/riscv/riscv.md (clear_cache): Use it.
15222 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
15224 * web.c: Remove out-of-date comment.
15226 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15228 * expr.c (fixup_args_size_notes): Check that any existing
15229 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
15230 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
15231 (emit_single_push_insn): ...here.
15233 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15235 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
15236 (const_vector_encoded_nelts): New function.
15237 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
15238 (const_vector_int_elt, const_vector_elt): Declare.
15239 * emit-rtl.c (const_vector_int_elt_1): New function.
15240 (const_vector_elt): Likewise.
15241 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
15242 of CONST_VECTOR_ELT.
15244 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15246 * expr.c: Include rtx-vector-builder.h.
15247 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
15248 directly on the tree encoding.
15249 (const_vector_from_tree): Likewise.
15250 * optabs.c: Include rtx-vector-builder.h.
15251 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
15252 sequence of "u" values.
15253 * vec-perm-indices.c: Include rtx-vector-builder.h.
15254 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
15255 directly on the vec_perm_indices encoding.
15257 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15259 * doc/rtl.texi (const_vector): Describe new encoding scheme.
15260 * Makefile.in (OBJS): Add rtx-vector-builder.o.
15261 * rtx-vector-builder.h: New file.
15262 * rtx-vector-builder.c: Likewise.
15263 * rtl.h (rtx_def::u2): Add a const_vector field.
15264 (CONST_VECTOR_NPATTERNS): New macro.
15265 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
15266 (CONST_VECTOR_DUPLICATE_P): Likewise.
15267 (CONST_VECTOR_STEPPED_P): Likewise.
15268 (CONST_VECTOR_ENCODED_ELT): Likewise.
15269 (const_vec_duplicate_p): Check for a duplicated vector encoding.
15270 (unwrap_const_vec_duplicate): Likewise.
15271 (const_vec_series_p): Check for a non-duplicated vector encoding.
15272 Say that the function only returns true for integer vectors.
15273 * emit-rtl.c: Include rtx-vector-builder.h.
15274 (gen_const_vec_duplicate_1): Delete.
15275 (gen_const_vector): Call gen_const_vec_duplicate instead of
15276 gen_const_vec_duplicate_1.
15277 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
15278 (gen_const_vec_duplicate): Use rtx_vector_builder.
15279 (gen_const_vec_series): Likewise.
15280 (gen_rtx_CONST_VECTOR): Likewise.
15281 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
15282 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
15283 Build a new vector rather than modifying a CONST_VECTOR in-place.
15284 (handle_special_swappables): Update call accordingly.
15285 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
15286 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
15287 Build a new vector rather than modifying a CONST_VECTOR in-place.
15288 (handle_special_swappables): Update call accordingly.
15290 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15292 * simplify-rtx.c (simplify_const_binary_operation): Use
15293 CONST_VECTOR_ELT instead of XVECEXP.
15295 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15297 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
15298 the selector elements to be different from the data elements
15299 if the selector is a VECTOR_CST.
15300 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
15301 ssizetype for the selector.
15303 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15305 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
15306 before testing each element individually.
15307 * tree-vect-generic.c (lower_vec_perm): Likewise.
15309 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15311 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
15312 * selftest-run-tests.c (selftest::run_tests): Call it.
15313 * vector-builder.h (vector_builder::operator ==): New function.
15314 (vector_builder::operator !=): Likewise.
15315 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
15316 (vec_perm_indices::all_from_input_p): New function.
15317 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
15318 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
15319 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
15320 instead of reading the VECTOR_CST directly. Detect whether both
15321 vector inputs are the same before constructing the vec_perm_indices,
15322 and update the number of inputs argument accordingly. Use the
15323 utility functions added above. Only construct sel2 if we need to.
15325 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15327 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
15328 the broadcast of the low byte.
15329 (expand_mult_highpart): Use an explicit encoding for the permutes.
15330 * optabs-query.c (can_mult_highpart_p): Likewise.
15331 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
15332 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15333 (vectorizable_bswap): Likewise.
15334 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
15335 explicit encoding for the power-of-2 permutes.
15336 (vect_permute_store_chain): Likewise.
15337 (vect_grouped_load_supported): Likewise.
15338 (vect_permute_load_chain): Likewise.
15340 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15342 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
15343 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
15344 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
15345 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
15346 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
15347 (vect_gen_perm_mask_any): Likewise.
15349 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15351 * int-vector-builder.h: New file.
15352 * vec-perm-indices.h: Include int-vector-builder.h.
15353 (vec_perm_indices): Redefine as an int_vector_builder.
15354 (auto_vec_perm_indices): Delete.
15355 (vec_perm_builder): Redefine as a stand-alone class.
15356 (vec_perm_indices::vec_perm_indices): New function.
15357 (vec_perm_indices::clamp): Likewise.
15358 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
15359 (vec_perm_indices::new_vector): New function.
15360 (vec_perm_indices::new_expanded_vector): Update for new
15361 vec_perm_indices class.
15362 (vec_perm_indices::rotate_inputs): New function.
15363 (vec_perm_indices::all_in_range_p): Operate directly on the
15364 encoded form, without computing elided elements.
15365 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
15366 encoding. Update for new vec_perm_indices class.
15367 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
15368 the given vec_perm_builder.
15369 (expand_vec_perm_var): Update vec_perm_builder constructor.
15370 (expand_mult_highpart): Use vec_perm_builder instead of
15371 auto_vec_perm_indices.
15372 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
15373 vec_perm_indices instead of auto_vec_perm_indices. Use a single
15374 or double series encoding as appropriate.
15375 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
15376 vec_perm_indices instead of auto_vec_perm_indices.
15377 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15378 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15379 (vect_permute_store_chain): Likewise.
15380 (vect_grouped_load_supported): Likewise.
15381 (vect_permute_load_chain): Likewise.
15382 (vect_shift_permute_load_chain): Likewise.
15383 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15384 (vect_transform_slp_perm_load): Likewise.
15385 (vect_schedule_slp_instance): Likewise.
15386 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15387 (vectorizable_mask_load_store): Likewise.
15388 (vectorizable_bswap): Likewise.
15389 (vectorizable_store): Likewise.
15390 (vectorizable_load): Likewise.
15391 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
15392 vec_perm_indices instead of auto_vec_perm_indices. Use
15393 tree_to_vec_perm_builder to read the vector from a tree.
15394 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
15395 vec_perm_builder instead of a vec_perm_indices.
15396 (have_whole_vector_shift): Use vec_perm_builder and
15397 vec_perm_indices instead of auto_vec_perm_indices. Leave the
15398 truncation to calc_vec_perm_mask_for_shift.
15399 (vect_create_epilog_for_reduction): Likewise.
15400 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
15401 from auto_vec_perm_indices to vec_perm_indices.
15402 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15403 instead of changing individual elements.
15404 (aarch64_vectorize_vec_perm_const): Use new_vector to install
15405 the vector in d.perm.
15406 * config/arm/arm.c (expand_vec_perm_d::perm): Change
15407 from auto_vec_perm_indices to vec_perm_indices.
15408 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15409 instead of changing individual elements.
15410 (arm_vectorize_vec_perm_const): Use new_vector to install
15411 the vector in d.perm.
15412 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
15413 Update vec_perm_builder constructor.
15414 (rs6000_expand_interleave): Likewise.
15415 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
15416 (rs6000_expand_interleave): Likewise.
15418 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15420 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
15421 to qimode could truncate the indices.
15422 * optabs.c (expand_vec_perm_var): Likewise.
15424 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15426 * Makefile.in (OBJS): Add vec-perm-indices.o.
15427 * vec-perm-indices.h: New file.
15428 * vec-perm-indices.c: Likewise.
15429 * target.h (vec_perm_indices): Replace with a forward class
15431 (auto_vec_perm_indices): Move to vec-perm-indices.h.
15432 * optabs.h: Include vec-perm-indices.h.
15433 (expand_vec_perm): Delete.
15434 (selector_fits_mode_p, expand_vec_perm_var): Declare.
15435 (expand_vec_perm_const): Declare.
15436 * target.def (vec_perm_const_ok): Replace with...
15437 (vec_perm_const): ...this new hook.
15438 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
15439 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
15440 * doc/tm.texi: Regenerate.
15441 * optabs.def (vec_perm_const): Delete.
15442 * doc/md.texi (vec_perm_const): Likewise.
15443 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
15444 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
15445 expand_vec_perm for constant permutation vectors. Assert that
15446 the mode of variable permutation vectors is the integer equivalent
15447 of the mode that is being permuted.
15448 * optabs-query.h (selector_fits_mode_p): Declare.
15449 * optabs-query.c: Include vec-perm-indices.h.
15450 (selector_fits_mode_p): New function.
15451 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
15452 is defined, instead of checking whether the vec_perm_const_optab
15453 exists. Use targetm.vectorize.vec_perm_const instead of
15454 targetm.vectorize.vec_perm_const_ok. Check whether the indices
15455 fit in the vector mode before using a variable permute.
15456 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
15457 vec_perm_indices instead of an rtx.
15458 (expand_vec_perm): Replace with...
15459 (expand_vec_perm_const): ...this new function. Take the selector
15460 as a vec_perm_indices rather than an rtx. Also take the mode of
15461 the selector. Update call to shift_amt_for_vec_perm_mask.
15462 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
15463 Use vec_perm_indices::new_expanded_vector to expand the original
15464 selector into bytes. Check whether the indices fit in the vector
15465 mode before using a variable permute.
15466 (expand_vec_perm_var): Make global.
15467 (expand_mult_highpart): Use expand_vec_perm_const.
15468 * fold-const.c: Includes vec-perm-indices.h.
15469 * tree-ssa-forwprop.c: Likewise.
15470 * tree-vect-data-refs.c: Likewise.
15471 * tree-vect-generic.c: Likewise.
15472 * tree-vect-loop.c: Likewise.
15473 * tree-vect-slp.c: Likewise.
15474 * tree-vect-stmts.c: Likewise.
15475 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
15477 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
15478 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
15479 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
15480 (aarch64_vectorize_vec_perm_const): ...this new function.
15481 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15482 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15483 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
15484 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
15485 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15486 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15487 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
15489 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
15490 check for NEON modes.
15491 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
15492 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
15493 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
15494 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
15496 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
15497 the old VEC_PERM_CONST conditions.
15498 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
15499 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
15500 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
15501 (ia64_vectorize_vec_perm_const_ok): Merge into...
15502 (ia64_vectorize_vec_perm_const): ...this new function.
15503 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
15504 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
15505 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
15506 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
15507 * config/mips/mips.c (mips_expand_vec_perm_const)
15508 (mips_vectorize_vec_perm_const_ok): Merge into...
15509 (mips_vectorize_vec_perm_const): ...this new function.
15510 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
15511 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
15512 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
15513 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
15514 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
15515 (rs6000_expand_vec_perm_const): Delete.
15516 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
15518 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15519 (altivec_expand_vec_perm_const_le): Take each operand individually.
15520 Operate on constant selectors rather than rtxes.
15521 (altivec_expand_vec_perm_const): Likewise. Update call to
15522 altivec_expand_vec_perm_const_le.
15523 (rs6000_expand_vec_perm_const): Delete.
15524 (rs6000_vectorize_vec_perm_const_ok): Delete.
15525 (rs6000_vectorize_vec_perm_const): New function.
15526 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15527 an element count and rtx array.
15528 (rs6000_expand_extract_even): Update call accordingly.
15529 (rs6000_expand_interleave): Likewise.
15530 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
15531 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
15532 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
15533 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
15534 (rs6000_expand_vec_perm_const): Delete.
15535 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15536 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15537 (altivec_expand_vec_perm_const_le): Take each operand individually.
15538 Operate on constant selectors rather than rtxes.
15539 (altivec_expand_vec_perm_const): Likewise. Update call to
15540 altivec_expand_vec_perm_const_le.
15541 (rs6000_expand_vec_perm_const): Delete.
15542 (rs6000_vectorize_vec_perm_const_ok): Delete.
15543 (rs6000_vectorize_vec_perm_const): New function. Remove stray
15544 reference to the SPE evmerge intructions.
15545 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15546 an element count and rtx array.
15547 (rs6000_expand_extract_even): Update call accordingly.
15548 (rs6000_expand_interleave): Likewise.
15549 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
15550 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
15552 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15554 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15556 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
15557 vector mode and that that mode matches the mode of the data
15559 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
15560 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
15561 directly using expand_vec_perm_1 when forcing selectors into
15563 (expand_vec_perm_var): New function, split out from expand_vec_perm.
15565 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15567 * optabs-query.h (can_vec_perm_p): Delete.
15568 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
15569 * optabs-query.c (can_vec_perm_p): Split into...
15570 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
15571 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
15572 particular selector is valid.
15573 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15574 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15575 (vect_grouped_load_supported): Likewise.
15576 (vect_shift_permute_load_chain): Likewise.
15577 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15578 (vect_transform_slp_perm_load): Likewise.
15579 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15580 (vectorizable_bswap): Likewise.
15581 (vect_gen_perm_mask_checked): Likewise.
15582 * fold-const.c (fold_ternary_loc): Likewise. Don't take
15583 implementations of variable permutation vectors into account
15584 when deciding which selector to use.
15585 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
15586 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
15587 with a false third argument.
15588 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
15589 to test whether the constant selector is valid and can_vec_perm_var_p
15590 to test whether a variable selector is valid.
15592 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15594 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
15595 * optabs-query.c (can_vec_perm_p): Likewise.
15596 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
15597 instead of vec_perm_indices.
15598 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
15599 (vect_gen_perm_mask_checked): Likewise,
15600 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
15601 (vect_gen_perm_mask_checked): Likewise,
15603 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15605 * optabs-query.h (qimode_for_vec_perm): Declare.
15606 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
15607 (qimode_for_vec_perm): ...this new function.
15608 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
15610 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15612 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
15613 does not have a conditional at the top.
15615 2018-01-02 Richard Biener <rguenther@suse.de>
15617 * ipa-inline.c (big_speedup_p): Fix expression.
15619 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15622 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
15625 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15629 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
15630 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
15631 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
15632 cond_taken_branch_cost 3->4.
15634 2018-01-01 Jakub Jelinek <jakub@redhat.com>
15636 PR tree-optimization/83581
15637 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
15638 TODO_cleanup_cfg if any changes have been made.
15640 PR middle-end/83608
15641 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
15642 convert_modes if target mode has the right side, but different mode
15645 PR middle-end/83609
15646 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
15647 last argument when extracting from CONCAT. If either from_real or
15648 from_imag is NULL, use expansion through memory. If result is not
15649 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
15650 the parts directly to inner mode, if even that fails, use expansion
15653 PR middle-end/83623
15654 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
15655 check for bswap in mode rather than HImode and use that in expand_unop
15658 Copyright (C) 2018 Free Software Foundation, Inc.
15660 Copying and distribution of this file, with or without modification,
15661 are permitted in any medium without royalty provided the copyright
15662 notice and this notice are preserved.