1 2019-05-29 Jakub Jelinek <jakub@redhat.com>
4 * lto-streamer.h (LTO_minor_version): Bump to 2.
6 Backported from mainline
7 2019-05-16 Jakub Jelinek <jakub@redhat.com>
10 * tree-core.h (struct tree_decl_common): Document
11 decl_nonshareable_flag for PARM_DECLs.
12 * tree.h (DECL_HIDDEN_STRING_LENGTH): Define.
13 * calls.c (expand_call): Don't try tail call if caller
14 has any DECL_HIDDEN_STRING_LENGTH PARM_DECLs that are or might be
15 passed on the stack and callee needs to pass any arguments on the
17 * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Use
18 else if instead of series of mutually exclusive ifs. Handle
19 DECL_HIDDEN_STRING_LENGTH for PARM_DECLs.
20 * tree-streamer-out.c (pack_ts_decl_common_value_fields): Likewise.
22 2019-05-28 John David Anglin <danglin@gcc.gnu.org>
24 * config/pa/pa.c (hppa_profile_hook): Remove offset adjustment.
26 2019-05-27 Eric Botcazou <ebotcazou@adacore.com>
28 * dwarf2out.c (resolve_args_picking_1): Deal with DW_OP_GNU_addr_index
29 and DW_OP_GNU_const_index opcodes.
31 2019-05-24 John David Anglin <danglin@gcc.gnu.org>
34 * config/pa/pa.c (pa_can_change_mode_class): Accept mode changes from
35 DImode to SImode in floating-point registers on 64-bit target.
36 * config/pa/pa.md (umulsidi3): Change nonimmediate_operand to
37 register_operand in xmpyu patterns.
39 2019-05-24 Martin Liska <mliska@suse.cz>
41 Backport from mainline
42 2019-05-23 Martin Liska <mliska@suse.cz>
45 * gimplify.c (gimplify_target_expr): Skip TREE_STATIC target
46 expression similarly to gimplify_decl_expr.
48 2019-05-23 Uroš Bizjak <ubizjak@gmail.com>
50 Backported from mainline
51 2019-05-21 Uroš Bizjak <ubizjak@gmail.com>
53 * config/i386/cpuid.h (__cpuid): For 32bit targets, zero
54 %ebx and %ecx bafore calling cpuid with leaf 1 or
55 non-constant leaf argument.
57 2019-05-21 Uroš Bizjak <ubizjak@gmail.com>
60 * config/i386/i386.md (anddi_1 to andsi_1_zext splitter):
61 Avoid calling gen_lowpart with CONST operand.
63 2019-05-21 Sebastian Huber <sebastian.huber@embedded-brains.de>
65 Backported from mainline
66 2019-05-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
68 * config/arm/t-rtems: Replace ARMv7-M multilibs with Cortex-M
71 2019-05-20 Jeff Law <law@redhat.com>
73 Backport from mainline
74 2019-05-17 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
76 * config/mips/mips.c (mips_dwarf_frame_reg_mode): Replace
77 TARGET_FLOAT64 with !TARGET_FLOAT32, thus handling both fp64
80 2019-05-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
82 Backport from mainline.
83 2019-05-07 Kelvin Nilsen <kelvin@gcc.gnu.org>
86 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
87 In handling of ALTIVEC_BUILTIN_VEC_INSERT, use modular arithmetic
88 to compute vector element selector for both constant and variable
91 2019-05-15 David Edelsohn <dje.gcc@gmail.com>
93 Backport from mainline
94 2019-04-11 David Edelsohn <dje.gcc@gmail.com>
95 * xcoffout.h (xcoff_private_rodata_section_name): Declare.
96 * xcoffout.c (xcoff_private_rodata_section_name): Define.
97 * config/rs6000/rs6000.c (rs6000_xcoff_asm_init_sections): Create
98 read_only_private_data_section using coff_private_rodata_section_name.
99 (rs6000_xcoff_file_start): Generate coff_private_rodata_section_name.
101 2018-12-04 David Edelsohn <dje.gcc@gmail.com>
102 2018-12-13 David Edelsohn <dje.gcc@gmail.com>
104 * config/rs6000/rs6000.c (rs6000_function_arg): Don't pass aggregates
105 in FPRs on AIX. Ensure type is non-NULL.
106 (rs6000_arg_partial_bytes): Same.
108 2019-05-14 Martin Liska <mliska@suse.cz>
110 Backport from mainline
111 2019-05-13 Martin Liska <mliska@suse.cz>
113 PR gcov-profile/90380
114 * gcov.c (handle_cycle): Do not support zero cycle count,
115 it should not be possible.
116 (path_contains_zero_cycle_arc): New function.
117 (circuit): Ignore zero cycle arc counts.
119 2019-05-14 Martin Liska <mliska@suse.cz>
121 Backport from mainline
122 2019-05-13 Martin Liska <mliska@suse.cz>
124 PR gcov-profile/90380
125 * gcov.c (enum loop_type): Remove the enum and
127 (handle_cycle): Assert that we should not reach
129 (circuit): Use loop_found instead of a tri-state loop_type.
130 (get_cycles_count): Do not handle NEGATIVE_LOOP as it can't
133 2019-05-11 Andreas Tobler <andreast@gcc.gnu.org>
135 Backport from mainline.
136 2019-04-25 Andreas Tobler <andreast@gcc.gnu.org>
138 * config/i386/freebsd64.h: Add bits for 32-bit multilib support.
139 * config/i386/t-freebsd64: New file.
140 * config.gcc: Add the t-freebsd64 for multilib support.
142 2019-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
144 Backport from mainline.
145 2019-05-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
148 * config/rs6000/rs6000.c (rs6000_expand_vector_extract): Add
149 handling of V1TImode.
151 2019-05-06 Richard Biener <rguenther@suse.de>
154 Backport from mainline
155 2019-03-11 Jakub Jelinek <jakub@redhat.com>
159 * vr-values.c (vr_values::update_value_range): If
160 old_vr->varying_p (), don't update it, make new_vr also VARYING
163 2019-05-03 Richard Biener <rguenther@suse.de>
165 Backport from mainline
166 2019-03-14 Richard Biener <rguenther@suse.de>
169 * fold-const.c (operand_equal_p): For INDIRECT_REF check
170 that the access types are similar.
172 2019-03-07 Richard Biener <rguenther@suse.de>
174 PR tree-optimization/89595
175 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Take
176 stmt iterator as reference, take boolean output parameter to
177 indicate whether the stmt was removed and thus the iterator
179 (dom_opt_dom_walker::before_dom_children): Re-iterate over
180 stmts created by folding.
182 2019-03-14 Richard Biener <rguenther@suse.de>
184 PR tree-optimization/89710
185 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Use
188 2019-04-15 Richard Biener <rguenther@suse.de>
190 PR tree-optimization/90071
191 * tree-ssa-reassoc.c (init_range_entry): Do not pick up
192 abnormal operands from def stmts.
194 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
197 * expr.c (expand_assignment): Assert that bitpos is positive.
198 (store_field): Likewise
199 (expand_expr_real_1): Make sure that bitpos is positive.
200 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
203 2019-03-14 Richard Biener <rguenther@suse.de>
206 * config/i386/i386.c (make_resolver_func): Properly set
207 DECL_CONTEXT on the RESULT_DECL.
208 * config/rs6000/rs6000.c (make_resolver_func): Likewise.
210 2019-04-30 Jakub Jelinek <jakub@redhat.com>
212 Backported from mainline
213 2019-04-24 Jakub Jelinek <jakub@redhat.com>
216 * rtl.c (classify_insn): Return JUMP_INSN for asm goto.
217 * emit-rtl.c (try_split): Copy over REG_LABEL_TARGET.
220 * config/i386/i386.c (ix86_expand_sse_fp_minmax): Force if_true into
221 a register if both if_true and if_false are MEMs.
223 PR tree-optimization/90208
224 * tree-cfg.c (remove_bb): Move forced labels from removed bbs
225 after labels of new_bb, not before them.
227 2019-04-16 Jakub Jelinek <jakub@redhat.com>
229 PR rtl-optimization/90082
230 * dce.c (can_delete_call): New function.
231 (deletable_insn_p, mark_insn): Use it.
233 PR tree-optimization/90090
234 * tree-ssa-math-opts.c (is_division_by): Ignore divisions that can
236 (is_division_by_square): Likewise. Formatting fix.
238 2019-04-12 Jakub Jelinek <jakub@redhat.com>
240 PR rtl-optimization/89965
241 * dce.c: Include rtl-iter.h.
242 (struct check_argument_load_data): New type.
243 (check_argument_load): New function.
244 (find_call_stack_args): Check for loads from stack slots still tracked
245 in sp_bytes and punt if any is found.
248 * varasm.c (assemble_start_function): Don't use tree_fits_uhwi_p
249 and gcc_unreachable if it fails, just call tree_to_uhwi which
250 verifies that too. Test TREE_CHAIN instead of list_length > 1.
251 Start warning message with a lower-case letter. Formatting fixes.
253 PR rtl-optimization/90026
254 * cfgcleanup.c (try_optimize_cfg): When removing empty bb with no
255 successors, look for BARRIERs inside of the whole BB_FOOTER chain
256 rather than just at the start of it. If e->src BB_FOOTER is not NULL
257 in cfglayout mode, use emit_barrier_after_bb.
259 2019-04-11 Jakub Jelinek <jakub@redhat.com>
261 PR rtl-optimization/89965
262 * dce.c (sp_based_mem_offset): New function.
263 (find_call_stack_args): Use sp_based_mem_offset.
265 2019-04-10 Jakub Jelinek <jakub@redhat.com>
268 * gimple-ssa-sprintf.c (target_to_host): Fix handling of targstr
269 with strlen in between hostsz-3 and hostsz-1 inclusive when no
270 translation is needed, and when translation is needed, only append
271 ... if the string length is hostsz or more bytes long. Avoid using
274 2019-04-09 Jakub Jelinek <jakub@redhat.com>
276 PR tree-optimization/89998
277 * gimple-ssa-sprintf.c (try_substitute_return_value): Use lhs type
278 instead of integer_type_node if possible, don't add ranges if return
279 type is not compatible with int.
280 * gimple-fold.c (gimple_fold_builtin_sprintf,
281 gimple_fold_builtin_snprintf): Use lhs type instead of hardcoded
284 2019-03-29 Jakub Jelinek <jakub@redhat.com>
287 * gimplify.c (gimplify_compound_literal_expr): Don't optimize a
288 non-addressable complit into its initializer if it is volatile.
290 2019-03-28 Jakub Jelinek <jakub@redhat.com>
293 * tree-inline.h (struct copy_body_data): Add
294 dont_remap_vla_if_no_change flag.
295 * tree-inline.c (remap_type_3, remap_type_2): New functions.
296 (remap_type): Don't remap vla types if id->dont_remap_vla_if_no_change
297 and remap_type_2 returns false.
298 * omp-low.c (new_omp_context): Set ctx->cb.dont_remap_vla_if_no_change.
300 2019-03-20 Jakub Jelinek <jakub@redhat.com>
303 * lra-constraints.c (process_alt_operands) <reg>: For BLKmode, don't
304 update this_alternative nor this_alternative_set.
306 2019-03-19 Jakub Jelinek <jakub@redhat.com>
308 PR rtl-optimization/89768
309 * loop-unroll.c (unroll_loop_constant_iterations): Use gen_int_mode
311 (unroll_loop_runtime_iterations): Likewise.
314 * gimplify.c (gimplify_asm_expr): For output argument with
315 TREE_ADDRESSABLE type, clear allows_reg if it allows memory, otherwise
319 * config/i386/i386.c (ix86_expand_floorceildf_32): In ceil
320 compensation use x2 += 1 instead of x2 -= -1 and when honoring
321 signed zeros, do another copysign after the compensation.
323 2019-03-15 Jakub Jelinek <jakub@redhat.com>
326 * dwarf2out.c (add_const_value_attribute): Return false for MINUS,
327 SIGN_EXTEND and ZERO_EXTEND.
329 2019-03-14 Jakub Jelinek <jakub@redhat.com>
332 * multiple_target.c (create_dispatcher_calls): Change
333 references_to_redirect from vector of ipa_ref * to vector of ipa_ref.
334 In the node->iterate_referring loop, push *ref rather than ref, call
335 ref->remove_reference () and always pass 0 to iterate_referring.
337 PR rtl-optimization/89679
338 * expmed.c (expand_mult_const): Don't add a REG_EQUAL note if it
339 would contain a paradoxical SUBREG.
341 PR tree-optimization/89703
342 * tree-ssa-strlen.c (valid_builtin_call): Punt if stmt call types
343 aren't compatible also with builtin_decl_explicit. Check pure
344 or non-pure status of BUILT_IN_STR{{,N}CMP,N{LEN,{CAT,CPY}{,_CHK}}}
345 and BUILT_IN_STPNCPY{,_CHK}.
347 2019-03-13 Jakub Jelinek <jakub@redhat.com>
350 * dwarf2out.c (size_of_die): For dw_val_class_view_list always use
352 (value_format): For dw_val_class_view_list never use DW_FORM_loclistx.
355 * omp-simd-clone.c (ipa_simd_modify_stmt_ops): Handle PHI args.
356 (ipa_simd_modify_function_body): Handle PHIs.
358 2019-03-12 Jakub Jelinek <jakub@redhat.com>
361 * builtins.c (expand_builtin_int_roundingfn,
362 expand_builtin_int_roundingfn_2): Return NULL_RTX instead of
363 gcc_unreachable if validate_arglist fails.
365 2019-03-09 Jakub Jelinek <jakub@redhat.com>
368 * attribs.c (handle_dll_attribute): Don't clear TREE_STATIC for
369 dllimport on VAR_DECLs with RECORD_TYPE or UNION_TYPE DECL_CONTEXT.
371 2019-03-05 Jakub Jelinek <jakub@redhat.com>
374 * config/rs6000/t-linux (MULTIARCH_DIRNAME): Set to non-empty only
378 * builtins.c (maybe_emit_free_warning): Punt if free doesn't have
379 exactly one argument.
381 2019-02-28 Jakub Jelinek <jakub@redhat.com>
384 * convert.c (convert_to_real_1, convert_to_integer_1): Punt for
385 builtins if they don't have a single scalar floating point argument.
388 2019-02-20 Jakub Jelinek <jakub@redhat.com>
391 * expr.c (expand_assignment): If result is a MEM, use change_address
392 instead of simplify_gen_subreg.
394 2019-02-20 Jakub Jelinek <jakub@redhat.com>
395 David Malcolm <dmalcolm@redhat.com>
398 * fold-const.c (decode_field_reference): Return NULL_TREE if
399 lang_hooks.types.type_for_size returns NULL. Check it before
400 overwriting *exp_. Use return NULL_TREE instead of return 0.
402 2019-02-20 Jakub Jelinek <jakub@redhat.com>
406 * toplev.c (do_compile): Double the emin/emax exponents to workaround
409 2019-02-19 Richard Biener <rguenther@suse.de>
412 * toplev.c (do_compile): Initialize mpfr's exponent range
413 based on available float modes.
415 2019-04-30 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
417 Backport from mainline.
418 2019-04-30 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
420 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
421 Define __ARM_FEATURE_ATOMICS.
423 2019-04-30 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
425 Backport from mainline
426 2018-12-11 Richard Earnshaw <Richard.Earnshaw@arm.com>
429 * config/aarch64/iterators.md (sizem1): Add sizes for
431 (Vbtype): Add SFmode mapping.
432 * config/aarch64/aarch64.md (copysigndf3, copysignsf3): Delete.
433 (copysign<GPF:mode>3): New expand pattern.
434 (copysign<GPF:mode>3_insn): New insn pattern.
436 2019-04-29 Richard Sandiford <richard.sandiford@arm.com>
438 Backport from mainline:
439 2019-01-25 Richard Sandiford <richard.sandiford@arm.com>
442 * varasm.c (output_constructor_bitfield): Use wi::extract_uhwi
443 instead of accessing TREE_INT_CST_ELT directly.
445 2019-04-29 Richard Sandiford <richard.sandiford@arm.com>
447 Backport from mainline:
448 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
451 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
452 aarch64_evpc_tbl guard.
454 2019-04-29 Richard Sandiford <richard.sandiford@arm.com>
456 Backport from mainline:
457 2019-03-08 Richard Sandiford <richard.sandiford@arm.com>
460 * dwarf2cfi.c (dwarf2out_frame_debug_expr): Use CONST_POLY_INT
461 instead of POLY_INT_CST.
463 2019-04-26 Roman Zhuykov <zhroma@ispras.ru>
465 Backport from mainline
466 2019-04-23 Roman Zhuykov <zhroma@ispras.ru>
468 PR rtl-optimization/87979
469 * modulo-sched.c (sms_schedule): Start ii value "mii" should
472 PR rtl-optimization/84032
473 * modulo-sched.c (ps_insn_find_column): Change condition so that
474 branch will always be the last insn in a row inside partial
477 2019-04-19 Pat Haugen <pthaugen@us.ibm.com>
479 Backport from mainline:
480 2019-04-16 Pat Haugen <pthaugen@us.ibm.com>
483 * config/rs6000/power9.md: Add store forwarding bypass.
485 2019-04-17 Martin Jambor <mjambor@suse.cz>
487 Backport from mainline
488 2019-03-10 Martin Jambor <mjambor@suse.cz>
490 PR tree-optimization/85762
491 PR tree-optimization/87008
492 PR tree-optimization/85459
493 * tree-sra.c (contains_vce_or_bfcref_p): New parameter, set the bool
494 it points to if there is a type changing MEM_REF. Adjust all callers.
495 (build_accesses_from_assign): Disable total scalarization if
496 contains_vce_or_bfcref_p returns true through the new parameter, for
499 2019-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
501 Backport from mainline
502 2019-03-15 Kelvin Nilsen <kelvin@gcc.gnu.org>
505 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
506 When handling vec_extract, use modular arithmetic to allow
507 constant selectors greater than vector length.
508 * config/rs6000/rs6000.c (rs6000_expand_vector_extract): Allow
509 V1TImode vectors to have constant selector values greater than 0.
510 Use modular arithmetic to compute vector index.
511 (rs6000_split_vec_extract_var): Use modular arithmetic to compute
512 index for in-memory vectors. Correct code generation for
513 in-register vectors. Use inner mode of vector rather than mode of
514 destination for move instruction.
515 (altivec_expand_vec_ext_builtin): Use modular arithmetic to
518 2019-04-12 Kelvin Nilsen <kelvin@gcc.gnu.org>
521 * config/rs6000/vsx.md (*vsx_extract_<mode>_<VS_scalar>mode_var):
522 Use QI inner mode with V16QI vector mode.
524 2019-04-15 Martin Jambor <mjambor@suse.cz>
526 Backport from mainline
527 2019-04-15 Martin Jambor <mjambor@suse.cz>
530 * cgraph.c (clone_of_p): Loop over clone chain for each step in
533 2019-04-15 Martin Liska <mliska@suse.cz>
535 Backport from mainline
536 2019-01-18 Martin Liska <mliska@suse.cz>
537 Richard Biener <rguenther@suse.de>
540 * cgraph.h (create_version_clone_with_body): Add new argument
542 * cgraphclones.c (cgraph_node::create_version_clone): Add
543 DECL_ATTRIBUTES to a newly created decl. And call
544 valid_attribute_p so that proper cl_target_optimization_node
545 is set for the newly created declaration.
546 * multiple_target.c (create_target_clone): Set DECL_ATTRIBUTES
548 (expand_target_clones): Do not call valid_attribute_p, it must
550 * tree-inline.c (copy_decl_for_dup_finish): Reset mode for
553 2019-04-11 Richard Biener <rguenther@suse.de>
555 PR tree-optimization/90018
556 * tree-vect-data-refs.c (vect_preserves_scalar_order_p):
557 Test both SLP and interleaving variants.
559 2019-04-10 Matthew Malcomson <matthew.malcomson@arm.com>
562 * config/arm/arm.c (neon_valid_immediate): Disallow VOIDmode parameter.
563 * config/arm/constraints.md (Dm, DN, Dn): Split previous Dn constraint
565 * config/arm/neon.md (*neon_mov<mode>): Account for TImode and DImode
566 differences directly.
567 (*smax<mode>3_neon, vashl<mode>3, vashr<mode>3_imm): Use Dm constraint.
569 2019-04-07 Uroš Bizjak <ubizjak@gmail.com>
572 * config/i386/i386.md (anddi_1 to andsi_1_zext splitter):
573 Avoid calling gen_lowpart with SYMBOL_REF and LABEL_REF operand.
575 2019-04-04 Martin Sebor <msebor@redhat.com>
578 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Bail
579 out if the number of arguments is less than expected.
581 2019-04-03 Richard Biener <rguenther@suse.de>
584 * lto-wrapper.c (run_gcc): Avoid implicit rules making
585 the all target phony.
587 2019-04-02 Uroš Bizjak <ubizjak@gmail.com>
591 * config/i386/i386.c (dimode_scalar_to_vector_candidate_p):
592 Return false for variable DImode shifts.
593 (dimode_scalar_chain::compute_convert_gain): Do not handle
594 register count operand in variable DImode shifts.
595 (dimode_scalar_chain::make_vector_copies): Remove support to copy
596 count argument of a variable shift instruction to a vector register.
597 (dimode_scalar_chain::convert_reg): Remove support to convert
598 count argument of a variable shift instruction.
600 2019-03-29 Kugan Vivekanandarajah <kuganv@linaro.org>
602 Backport from mainline
603 2019-03-29 Kugan Vivekanandarajah <kuganv@linaro.org>
604 Eric Botcazou <ebotcazou@adacore.com>
606 PR rtl-optimization/89862
607 * rtl.h (word_register_operation_p): Exclude CONST_INT from operations
608 that operates on the full registers for WORD_REGISTER_OPERATIONS
611 2019-03-28 Uroš Bizjak <ubizjak@gmail.com>
614 * config/i386/i386.c (dimode_scalar_chain::make_vector_copies):
615 Also process XEXP (src, 0) of a shift insn.
617 2019-03-28 Martin Liska <mliska@suse.cz>
619 Backport from mainline
620 2018-11-05 Martin Liska <mliska@suse.cz>
623 * doc/invoke.texi: Remove options that are
624 not disabled with -Os.
626 2019-03-26 Uroš Bizjak <ubizjak@gmail.com>
629 * config/i386/i386.c (dimode_scalar_chain::convert_reg):
630 Also process XEXP (src, 0) of a shift insn.
632 2019-03-25 Andreas Krebbel <krebbel@linux.ibm.com>
634 Backport from mainline
635 2019-03-20 Andreas Krebbel <krebbel@linux.ibm.com>
638 * config/s390/s390.c (global_not_special_regno_p): Move to make it
640 (s390_optimize_register_info): Use global_not_special_regno_p to
641 check for global regs.
643 2019-03-22 Bill Schmidt <wschmidt@linux.ibm.com>
645 Backport from mainline
646 2019-03-22 Bill Schmidt <wschmidt@linux.ibm.com>
648 * config/rs6000/mmintrin.h (_mm_sub_pi32): Fix typo.
650 2019-03-21 Bill Schmidt <wschmidt@linux.ibm.com>
652 Backport from mainline
653 2019-03-08 Bill Schmidt <wschmidt@linux.ibm.com>
655 * config/rs6000/rs6000-p8swap.c (rs6000_analyze_swaps): Rebuild
656 ud- and du-chains between phases.
658 2019-03-19 Richard Biener <rguenther@suse.de>
661 * opts.c (finish_options): Disable -gsplit-dwarf when doing LTO.
663 2019-03-19 Eric Botcazou <ebotcazou@adacore.com>
665 PR rtl-optimization/89753
666 * loop-unroll.c (decide_unroll_constant_iterations): Make guard for
667 explicit unrolling factor even more robust.
669 2019-03-18 Martin Sebor <msebor@redhat.com>
672 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range):
673 Handle anti-ranges the same as no range at all.
675 2019-03-18 Martin Jambor <mjambor@suse.cz>
677 Backport from mainline
678 2019-03-18 Martin Jambor <mjambor@suse.cz>
680 PR tree-optimization/89546
681 * tree-sra.c (propagate_subaccesses_across_link): Requeue new_acc if
682 any propagation to its children took place.
684 2019-03-17 H.J. Lu <hongjiu.lu@intel.com>
686 Backport from mainline
687 2019-03-14 H.J. Lu <hongjiu.lu@intel.com>
690 * config/i386/i386.c (ix86_print_operand): Handle 'M' to add
691 addr32 prefix to VSIB address for X32.
692 * config/i386/sse.md (*avx512pf_gatherpf<mode>sf_mask): Prepend
694 (*avx512pf_gatherpf<mode>df_mask): Likewise.
695 (*avx512pf_scatterpf<mode>sf_mask): Likewise.
696 (*avx512pf_scatterpf<mode>df_mask): Likewise.
697 (*avx2_gathersi<mode>): Prepend "%M3" to opcode.
698 (*avx2_gathersi<mode>_2): Prepend "%M2" to opcode.
699 (*avx2_gatherdi<mode>): Prepend "%M3" to opcode.
700 (*avx2_gatherdi<mode>_2): Prepend "%M2" to opcode.
701 (*avx2_gatherdi<mode>_3): Prepend "%M3" to opcode.
702 (*avx2_gatherdi<mode>_4): Prepend "%M2" to opcode.`
703 (*avx512f_gathersi<mode>): Prepend "%M4" to opcode.
704 (*avx512f_gathersi<mode>_2): Prepend "%M3" to opcode.
705 (*avx512f_gatherdi<mode>): Prepend "%M4" to opcode.
706 (*avx512f_gatherdi<mode>_2): Prepend "%M3" to opcode.
707 (*avx512f_scattersi<mode>): Prepend "%M0" to opcode.
708 (*avx512f_scatterdi<mode>): Likewise.
710 2019-03-15 Richard Biener <rguenther@suse.de>
712 Backport from mainline
713 2018-11-30 Jakub Jelinek <jakub@redhat.com>
716 * params.def (PARAM_LOGICAL_OP_NON_SHORT_CIRCUIT): New param.
717 * tree-ssa-ifcombine.c (ifcombine_ifandif): If
718 --param logical-op-non-short-circuit is present, override
719 LOGICAL_OP_NON_SHORT_CIRCUIT value from the param.
720 * fold-const.c (fold_range_test, fold_truth_andor): Likewise.
722 2019-03-15 Martin Liska <mliska@suse.cz>
725 * doc/invoke.texi: Remove -fdump-class-hierarchy option.
727 2019-03-14 Martin Jambor <mjambor@suse.cz>
729 Backport from mainline
730 2019-03-07 Martin Jambor <mjambor@suse.cz>
733 * ipa-cp.c (perform_estimation_of_a_value): Account zero time benefit
734 for extern inline functions.
736 2019-03-14 Richard Biener <rguenther@suse.de>
739 * tree-scalar-evolution.c (get_loop_exit_condition): Use
741 * tree-ssa-loop-ivcanon.c (canonicalize_loop_induction_variables):
742 Use gimple_location_safe.
744 2019-03-13 Vladimir Makarov <vmakarov@redhat.com>
747 * lra-constraints.c (inherit_in_ebb): Update
748 potential_reload_hard_regs along with live_hard_regs.
750 2019-03-14 Richard Biener <rguenther@suse.de>
752 Backport from mainline
753 2019-03-13 Richard Biener <rguenther@suse.de>
756 * tree-scalar-evolution.c (simplify_peeled_chrec): Do not
757 throw FP expressions at tree-affine.
759 2019-03-01 Richard Biener <rguenther@suse.de>
762 * tree-cfgcleanup.h (cleanup_tree_cfg): Add SSA update flags
763 argument, defaulted to zero.
764 * passes.c (execute_function_todo): Pass down SSA update flags
766 * tree-cfgcleanup.c: Include tree-into-ssa.h and tree-cfgcleanup.h.
767 (cleanup_tree_cfg_1): After cleanup_control_flow_pre update SSA
769 (cleanup_tree_cfg_noloop): Pass down SSA update flags.
770 (cleanup_tree_cfg): Get and pass down SSA update flags.
772 2019-02-18 Richard Biener <rguenther@suse.de>
774 PR tree-optimization/89296
775 * tree-ssa-loop-ch.c (ch_base::copy_headers): Restrict setting
776 of no-warning flag to cases that might emit the bogus warning.
778 2019-03-13 Andreas Krebbel <krebbel@linux.ibm.com>
780 Backport from mainline
781 2019-03-11 Andreas Krebbel <krebbel@linux.ibm.com>
783 * config/s390/s390.c (s390_expand_builtin): Do the copy_to_reg not
784 only on the else branch.
786 2019-03-13 Andreas Krebbel <krebbel@linux.ibm.com>
788 Backport from mainline
789 2019-02-07 Andreas Krebbel <krebbel@linux.ibm.com>
791 * config/s390/s390-builtin-types.def: Add new types.
792 * config/s390/s390-builtins.def: (s390_vec_xl, s390_vec_xld2)
793 (s390_vec_xlw4): Make the memory operand into a const pointer.
794 (s390_vec_xld2, s390_vec_xlw4): Add a variant for single precision
796 * config/s390/s390-c.c (s390_expand_overloaded_builtin): Generate
797 a new vector type with the alignment of the scalar memory operand.
799 2019-03-12 Richard Biener <rguenther@suse.de>
801 PR tree-optimization/89664
802 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Properly
803 free the occurance tree after the early out.
805 2019-03-12 Andre Vieira <andre.simoesdiasvieira@arm.com>
807 Backport from mainline
808 2019-03-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
810 * config/arm/arm.h (TARGET_FP16_TO_DOUBLE): Add TARGET_VFP_DOUBLE
813 2019-03-11 Eric Botcazou <ebotcazou@adacore.com>
815 PR rtl-optimization/89588
816 * loop-unroll.c (decide_unroll_constant_iterations): Make guard for
817 explicit unrolling factor more robust.
819 2019-03-11 Martin Liska <mliska@suse.cz>
821 Backport from mainline
822 2019-03-08 Martin Liska <mliska@suse.cz>
825 * config/i386/i386.c (ix86_option_override_internal): Disable
826 jump tables when retpolines are used.
828 2019-03-11 Martin Liska <mliska@suse.cz>
830 Backport from mainline
831 2019-02-11 David Malcolm <dmalcolm@redhat.com>
834 * input.c (selftest::test_line_offset_overflow): New selftest.
835 (selftest::input_c_tests): Call it.
837 2019-03-08 Martin Jambor <mjambor@suse.cz>
839 Backport from mainline
840 2019-03-07 Martin Jambor <mjambor@suse.cz>
843 * cgraph.h (cgraph_node): New inline method former_thunk_p.
844 * cgraph.c (cgraph_node::dump): Dump a note if node is a former thunk.
845 (clone_of_p): Treat expanded thunks like thunks, be optimistic if they
846 have multiple callees. At the end check if declarations match as
847 opposed to cgraph_nodes.
849 2019-03-06 Tamar Christina <tamar.christina@arm.com>
852 2019-02-25 Tamar Christina <tamar.christina@arm.com>
855 * common/config/aarch64/aarch64-common.c
856 (struct aarch64_option_extension): Add is_synthetic.
857 (all_extensions): Use it.
858 (TARGET_OPTION_INIT_STRUCT): Define hook.
859 (struct gcc_targetm_common): Moved to end.
860 (all_extensions_by_on): New.
861 (opt_ext_cmp, typedef opt_ext): New.
862 (aarch64_option_init_struct): New.
863 (aarch64_contains_opt): New.
864 (aarch64_get_extension_string_for_isa_flags): Output smallest set.
865 * config/aarch64/aarch64-option-extensions.def
866 (AARCH64_OPT_EXTENSION): Explicitly include AES and SHA2 in crypto.
867 (fp, simd, crc, lse, fp16, rcpc, rdma, dotprod, aes, sha2, sha3,
869 Set is_synthetic to false.
870 (crypto): Set is_synthetic to true.
871 * config/aarch64/driver-aarch64.c (AARCH64_OPT_EXTENSION): Add
874 2019-03-06 Tamar Christina <tamar.christina@arm.com>
877 2019-02-28 Tamar Christina <tamar.christina@arm.com>
880 * config/aarch64/aarch64-option-extensions.def: Document it.
881 * config/aarch64/driver-aarch64.c (host_detect_local_cpu): Skip feature
884 2019-03-06 Xiong Hu Luo <luoxhu@linux.ibm.com>
886 Backport of r268834 from mainline to gcc-8-branch.
887 2019-02-13 Xiong Hu Luo <luoxhu@linux.vnet.ibm.com>
889 * config/rs6000/altivec.h (vec_sbox_be, vec_cipher_be,
890 vec_cipherlast_be, vec_ncipher_be, vec_ncipherlast_be): New #defines.
891 * config/rs6000/crypto.md (CR_vqdi): New define_mode_iterator.
892 (crypto_vsbox_<mode>, crypto_<CR_insn>_<mode>): New define_insns.
893 * config/rs6000/rs6000-builtin.def (VSBOX_BE): New BU_CRYPTO_1.
894 (VCIPHER_BE, VCIPHERLAST_BE, VNCIPHER_BE, VNCIPHERLAST_BE):
896 * config/rs6000/rs6000.c (builtin_function_type)
897 <CRYPTO_BUILTIN_VSBOX_BE, CRYPTO_BUILTIN_VCIPHER_BE,
898 CRYPTO_BUILTIN_VCIPHERLAST_BE, CRYPTO_BUILTIN_VNCIPHER_BE,
899 CRYPTO_BUILTIN_VNCIPHERLAST_BE>: New switch options.
900 * doc/extend.texi (vec_sbox_be, vec_cipher_be, vec_cipherlast_be,
901 vec_ncipher_be, vec_ncipherlast_be): New builtin functions.
903 2019-03-05 Richard Biener <rguenther@suse.de>
905 Backport from mainline
906 2019-02-26 Richard Biener <rguenther@suse.de>
908 PR tree-optimization/89505
909 * tree-ssa-structalias.c (compute_dependence_clique): Make sure
910 to handle restrict pointed-to vars with multiple subvars
913 2019-03-01 Tamar Christina <tamar.christina@arm.com>
916 * config/aarch64/aarch64-option-extensions.def (fp, simd, crypto,
917 fp16): Collapse line.
919 2019-03-01 Richard Biener <rguenther@suse.de>
921 Backport from mainline
922 2019-03-01 Jakub Jelinek <jakub@redhat.com>
925 * dwarf2out.c (output_comdat_type_unit): Add ATTRIBUTE_UNUSED to
926 early_lto_debug argument.
928 2019-02-27 Richard Biener <rguenther@suse.de>
931 * dwarf2out.c (use_debug_types): Disable when in_lto_p.
933 2019-02-27 Richard Biener <rguenther@suse.de>
936 * dwarf2out.c (output_comdat_type_unit): Add early_lto_debug
937 parameter, prefix section name with .gnu.debuglto_ if true.
938 (dwarf2out_finish): Pass false to output_comdat_type_unit.
939 (dwarf2out_early_finish): Pass true to output_comdat_type_unit.
941 2019-02-27 Richard Biener <rguenther@suse.de>
944 * dwarf2out.c (size_of_die): Key on AT_ref (a)->comdat_type_p
945 rather than on use_debug_types, doing what output_die does.
946 (value_format): Likewise.
948 2019-02-28 Eric Botcazou <ebotcazou@adacore.com>
950 PR tree-optimization/89536
951 * tree-ssa-dom.c (edge_info::derive_equivalences) <BIT_NOT_EXPR>: Test
952 only whether bit #0 of the value is 0 instead of the entire value.
954 2019-02-28 Li Jia He <helijia@linux.ibm.com>
957 2019-02-20 Li Jia He <helijia@linux.ibm.com>
960 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin)
961 <case ALTIVEC_BUILTIN_VSPLTISB, ALTIVEC_BUILTIN_VSPLTISH,
962 ALTIVEC_BUILTIN_VSPLTISW>: Don't convert the operand before
965 2019-02-27 Uroš Bizjak <ubizjak@gmail.com>
968 * config/i386/i386.c (ix86_atomic_assign_expand_fenv): Check
969 TARGET_SSE in addition to TARGET_SSE_MATH.
971 (ix86_excess_precision): Ditto.
972 (ix86_float_exceptions_rounding_supported_p): Ditto.
973 (use_rsqrt_p): Ditto.
974 * config/i386/sse.md (rsqrt<mode>2): Ditto.
976 2019-02-27 Jakub Jelinek <jakub@redhat.com>
978 Backported from mainline
979 2019-02-18 Jakub Jelinek <jakub@redhat.com>
982 * config/s390/s390.c (s390_indirect_branch_attrvalue,
983 s390_indirect_branch_settings): Define unconditionally.
984 (s390_set_current_function): Likewise, but guard the whole body except
985 the s390_indirect_branch_settings call with
986 #if S390_USE_TARGET_ATTRIBUTE.
987 (TARGET_SET_CURRENT_FUNCTION): Redefine unconditionally.
989 2019-02-26 Eric Botcazou <ebotcazou@adacore.com>
991 Backport from mainline
992 2019-02-19 Eric Botcazou <ebotcazou@adacore.com>
994 * rtlanal.c (get_initial_register_offset): Fall back to the estimate
995 as long as the epilogue isn't completed.
997 2019-02-26 Eric Botcazou <ebotcazou@adacore.com>
999 * tree-ssa-dom.c (edge_info::derive_equivalences) <BIT_IOR_EXPR>: Fix
1000 and move around comment.
1001 <BIT_AND_EXPR>: Likewise.
1002 <BIT_NOT_EXPR>: Add specific handling for boolean types.
1004 2019-02-24 Alan Modra <amodra@gmail.com>
1007 * config/rs6000/rs6000.md (<bd>_<mode> split): Check for an int
1008 output reg on add insn.
1009 (<bd>tf_<mode> split): Likewise. Match predicates with insn.
1011 2019-02-22 Jakub Jelinek <jakub@redhat.com>
1013 * BASE-VER: Set to 8.3.1.
1015 2019-02-22 Release Manager
1017 * GCC 8.3.0 released.
1019 2019-02-15 Richard Biener <rguenther@suse.de>
1020 Jakub Jelinek <jakub@redhat.com>
1022 PR tree-optimization/89278
1023 * tree-loop-distribution.c: Include tree-eh.h.
1024 (generate_memset_builtin, generate_memcpy_builtin): Call
1025 rewrite_to_non_trapping_overflow on builtin->size before passing it
1026 to force_gimple_operand_gsi.
1028 2019-02-15 Jakub Jelinek <jakub@redhat.com>
1031 * optc-save-gen.awk: Handle optimize_fast like optimize_size or
1033 * opth-gen.awk: Likewise.
1035 2019-02-14 David Malcolm <dmalcolm@redhat.com>
1037 Backport of r267957 from trunk.
1038 2019-01-15 David Malcolm <dmalcolm@redhat.com>
1041 * tree.c (build_function_type): Assert that arg_types is not
1044 2019-02-14 Jakub Jelinek <jakub@redhat.com>
1046 PR rtl-optimization/89354
1047 * combine.c (make_extraction): Punt if extraction_mode is narrower
1050 2019-02-14 David Malcolm <dmalcolm@redhat.com>
1052 Backport of r263339 from trunk.
1053 2018-08-06 David Malcolm <dmalcolm@redhat.com>
1055 * function-tests.c (selftest::test_expansion_to_rtl): Call
1056 free_after_compilation.
1058 2019-02-14 David Malcolm <dmalcolm@redhat.com>
1060 Backport of r263295 from trunk.
1061 2018-08-03 David Malcolm <dmalcolm@redhat.com>
1063 * doc/gcov.texi (-x): Remove duplicate "to".
1064 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
1065 (-Wif-not-aligned): Remove duplicate "is".
1066 (-flto): Remove duplicate "the".
1067 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
1068 duplicate "v5.00.b".
1069 (MSP430 Options): Remove duplicate "and" from the description
1070 of "-mgprel-sec=regexp".
1071 (x86 Options): Remove duplicate copies of "vmldLog102" and
1072 vmlsLog104 from description of "-mveclibabi=type".
1074 2019-02-14 David Malcolm <dmalcolm@redhat.com>
1076 Backport of r263275 from trunk.
1077 2018-08-02 David Malcolm <dmalcolm@redhat.com>
1079 * diagnostic-show-locus.c (diagnostic_show_locus): Use
1080 pp_take_prefix when saving the existing prefix.
1081 * diagnostic.c (diagnostic_append_note): Likewise.
1082 * langhooks.c (lhd_print_error_function): Likewise.
1083 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
1084 param's type. Free the existing prefix.
1085 (pp_take_prefix): New function.
1086 (pretty_printer::pretty_printer): Drop the prefix parameter.
1087 Rename the length parameter to match the comment.
1088 (pretty_printer::~pretty_printer): Free the prefix.
1089 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
1091 (struct pretty_printer): Drop the "const" from "prefix" field's
1092 type and clarify memory management.
1093 (pp_set_prefix): Drop the "const" from the 2nd param.
1094 (pp_take_prefix): New decl.
1096 2019-02-14 Segher Boessenkool <segher@kernel.crashing.org>
1099 2018-08-31 Segher Boessenkool <segher@kernel.crashing.org>
1103 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
1105 2019-02-14 Segher Boessenkool <segher@kernel.crashing.org>
1108 2019-01-18 Segher Boessenkool <segher@kernel.crashing.org>
1111 * config/rs6000/rs6000.md (*movsi_from_df): Allow only register
1114 2019-02-14 Segher Boessenkool <segher@kernel.crashing.org>
1117 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
1119 PR rtl-optimization/85805
1120 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
1121 value for hard registers if that was written in the same mode.
1123 2019-02-14 Martin Liska <mliska@suse.cz>
1125 Backport from mainline
1126 2019-02-14 Martin Liska <mliska@suse.cz>
1128 PR rtl-optimization/89242
1129 * dce.c (delete_unmarked_insns): Call free_dominance_info we
1130 process a transformation.
1132 2019-02-14 Martin Liska <mliska@suse.cz>
1134 Backport from mainline
1135 2019-02-13 Martin Liska <mliska@suse.cz>
1138 * cfgrtl.c (remove_barriers_from_footer): New function.
1139 (try_redirect_by_replacing_jump): Use it.
1140 (cfg_layout_redirect_edge_and_branch): Likewise.
1142 2019-02-14 Martin Liska <mliska@suse.cz>
1144 Backport from mainline
1145 2019-02-11 Martin Liska <mliska@suse.cz>
1148 * ipa-cp.c (build_toporder_info): Remove usage of a param.
1149 * ipa-inline.c (inline_small_functions): Likewise.
1150 * ipa-pure-const.c (propagate_pure_const): Likewise.
1151 (propagate_nothrow): Likewise.
1152 * ipa-reference.c (propagate): Likewise.
1153 * ipa-utils.c (struct searchc_env): Remove unused field.
1154 (searchc): Always search across AVAIL_INTERPOSABLE.
1155 (ipa_reduced_postorder): Always allow AVAIL_INTERPOSABLE as
1156 the only called IPA pure const can properly not propagate
1157 across interposable boundary.
1158 * ipa-utils.h (ipa_reduced_postorder): Remove param.
1160 2019-02-14 Jakub Jelinek <jakub@redhat.com>
1162 PR tree-optimization/89314
1163 * fold-const.c (fold_binary_loc): Cast strlen argument to
1164 const char * before dereferencing it. Formatting fixes.
1166 Backported from mainline
1167 2019-02-13 Jakub Jelinek <jakub@redhat.com>
1170 * tree-ssa-structalias.c (set_uids_in_ptset): Or in vi->is_heap_var
1171 into pt->vars_contains_escaped_heap instead of setting
1172 pt->vars_contains_escaped_heap to it.
1175 * optabs.c (prepare_cmp_insn): Use UINTVAL (size) instead of
1176 INTVAL (size), compare it to GET_MODE_MASK instead of
1177 1 << GET_MODE_BITSIZE.
1180 * config/i386/predicates.md (x86_64_immediate_operand): Allow
1181 TLS UNSPECs offsetted by signed 32-bit CONST_INT even with
1184 2019-02-09 Jakub Jelinek <jakub@redhat.com>
1187 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
1188 If !node->definition and TYPE_ARG_TYPES is non-NULL, use
1189 TYPE_ARG_TYPES instead of DECL_ARGUMENTS.
1191 2019-02-13 Bill Schmidt <wschmidt@linux.ibm.com>
1193 Backport from mainline
1194 2019-02-11 Bill Schmidt <wschmidt@linux.ibm.com>
1196 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Shift-right
1197 and shift-left vector built-ins need to include a TRUNC_MOD_EXPR
1198 for correct semantics.
1200 2019-02-13 Richard Biener <rguenther@suse.de>
1202 Backport from mainline
1203 2019-01-08 Richard Biener <rguenther@suse.de>
1205 PR tree-optimization/86554
1206 * tree-ssa-sccvn.c (visit_nary_op): When value-numbering to
1207 expressions with different overflow behavior make sure there's an
1208 available expression on the path.
1210 2019-02-13 Martin Liska <mliska@suse.cz>
1213 * lto-streamer.h (LTO_minor_version): Bump version due
1216 2019-02-13 Richard Biener <rguenther@suse.de>
1218 Backport from mainline
1219 2019-02-12 Richard Biener <rguenther@suse.de>
1221 PR tree-optimization/89253
1222 * tree-ssa-loop-split.c (tree_ssa_split_loops): Check we can
1225 2019-02-08 Richard Biener <rguenther@suse.de>
1228 * tree-data-ref.c (initialize_matrix_A): Fail if constant
1230 (analyze_subscript_affine_affine): Handle failure from
1231 initialize_matrix_A.
1233 2019-01-28 Richard Biener <rguenther@suse.de>
1235 PR tree-optimization/88739
1236 * tree-ssa-sccvn.c (vn_reference_lookup_3): Avoid generating
1237 BIT_FIELD_REFs of non-mode-precision integral operands.
1239 2019-02-12 Jan Hubicka <hubicka@ucw.cz>
1241 Backport from mainline:
1242 2019-02-09 Jan Hubicka <hubicka@ucw.cz>
1244 * ipa-visibility.c (localize_node): Also do not localize
1245 LDPR_PREVAILING_DEF_IRONLY_EXP.
1247 2019-02-11 Stefan Agner <stefan@agner.ch>
1249 Backport from mainline.
1250 2019-01-10 Stefan Agner <stefan@agner.ch>
1253 * config/arm/arm.c (arm_option_override_internal): Force
1254 opts->x_inline_asm_unified to true only if TARGET_THUMB2_P.
1256 2019-02-09 Jan Hubicka <hubicka@ucw.cz>
1258 Backport from mainline:
1259 2019-01-05 Jan Hubicka <hubicka@ucw.cz>
1263 2017-05-22 Jan Hubicka <hubicka@ucw.cz>
1265 * ipa-inline.c (edge_badness): Use inlined_time instead of
1266 inline_summaries->get.
1268 2019-02-09 Aaron Sawdey <acsawdey@linux.ibm.com>
1270 Backported from mainline
1271 2019-02-05 Aaron Sawdey <acsawdey@linux.ibm.com>
1274 * config/rs6000/rs6000.md (<bd>tf_<mode>): Generate a local label
1275 for the long branch case.
1277 2019-02-05 Aaron Sawdey <acsawdey@linux.ibm.com>
1280 * config/rs6000/rs6000-string.c (do_ifelse, expand_cmp_vec_sequence,
1281 expand_compare_loop, expand_block_compare_gpr,
1282 expand_strncmp_align_check, expand_strncmp_gpr_sequence): Insert
1283 REG_BR_PROB notes in inline expansion of memcmp/strncmp. Add
1284 #include "profile-count.h" and "predict.h" for types and functions
1285 needed to work with REG_BR_PROB notes.
1287 2019-02-09 Aaron Sawdey <acsawdey@linux.ibm.com>
1289 * config/rs6000/rs6000-string.c (expand_compare_loop,
1290 expand_block_compare): Insert REG_BR_PROB notes in inline expansion of
1293 2019-02-09 Jakub Jelinek <jakub@redhat.com>
1295 Backported from mainline
1296 2019-01-16 David Malcolm <dmalcolm@redhat.com>
1299 * combine.c (delete_noop_moves): Convert to "bool" return,
1300 returning true if any edges are eliminated.
1301 (combine_instructions): Also return true if delete_noop_moves
1304 2019-02-08 Jakub Jelinek <jakub@redhat.com>
1306 PR rtl-optimization/89234
1307 * except.c (copy_reg_eh_region_note_forward): Return if note_or_insn
1308 is a NOTE, CODE_LABEL etc. - rtx_insn * other than INSN_P.
1309 (copy_reg_eh_region_note_backward): Likewise.
1311 2019-02-09 Alan Modra <amodra@gmail.com>
1314 * config/rs6000/rs6000.c (save_reg_p): Match logic in
1315 rs6000_emit_prologue emitting pic_offset_table setup.
1317 2019-02-08 Richard Biener <rguenther@suse.de>
1319 Backport from mainline
1320 2018-11-13 Richard Biener <rguenther@suse.de>
1322 PR tree-optimization/86991
1323 * tree-vect-loop.c (vect_is_slp_reduction): Delay reduction
1324 group building until we have successfully detected the SLP
1326 (vect_is_simple_reduction): Remove fixup code here.
1328 2019-02-08 Richard Biener <rguenther@suse.de>
1330 Backport from mainline
1331 2019-02-03 Richard Biener <rguenther@suse.de>
1334 * dwarf2out.c (copy_ancestor_tree): Register non-stubs as
1337 2019-01-29 Richard Biener <rguenther@suse.de>
1340 * dwarf2out.c (collect_skeleton_dies): New helper.
1341 (copy_decls_for_unworthy_types): Call it.
1342 (build_abbrev_table): Do not try to replace
1343 DW_AT_signature refs with local refs.
1345 2019-01-17 Richard Biener <rguenther@suse.de>
1348 * dwarf2out.c (want_pubnames): Never generate pubnames sections
1349 and friends for the LTO part of debug info.
1351 2019-01-24 Richard Biener <rguenther@suse.de>
1354 * tree-streamer-out.c (write_ts_decl_common_tree_pointers):
1355 When in "legacy" debug mode make sure to reset self-origins.
1357 2019-01-15 Richard Biener <rguenther@suse.de>
1360 * dwarf2out.c (gen_member_die): Do not generate inheritance
1363 2018-11-22 Richard Biener <rguenther@suse.de>
1367 * lto-streamer-out.c (lto_is_streamable): Allow CALL_EXPRs
1368 which can appear in size expressions.
1369 * tree-streamer-in.c (unpack_ts_base_value_fields): Stream
1370 CALL_EXPR_BY_DESCRIPTOR.
1371 (streamer_read_tree_bitfields): Stream CALL_EXPR_IFN.
1372 * tree-streamer-out.c (pack_ts_base_value_fields): Stream
1373 CALL_EXPR_BY_DESCRIPTOR.
1374 (streamer_write_tree_bitfields): Stream CALL_EXPR_IFN.
1376 2019-02-08 Christophe Lyon <christophe.lyon@linaro.org>
1378 Backport from mainline
1379 2019-01-07 Christophe Lyon <christophe.lyon@linaro.org>
1382 * doc/install.texi (with-multilib-list): Document for aarch64.
1384 2019-02-08 Richard Biener <rguenther@suse.de>
1386 Backport from mainline
1387 2018-12-10 Richard Biener <rguenther@suse.de>
1389 PR tree-optimization/88427
1390 * vr-values.c (vr_values::extract_range_from_phi_node):
1391 Handle symbolic ranges conservatively when trying to drop
1394 2018-11-28 Richard Biener <rguenther@suse.de>
1396 PR tree-optimization/88217
1397 * vr-values.c (vr_values::extract_range_from_phi_node): Make
1398 sure to handle results > +INF and < -INF correctly when
1399 trying to drop down to +INF - 1 or -INF + 1.
1401 2018-11-23 Richard Biener <rguenther@suse.de>
1403 PR tree-optimization/88149
1404 * tree-vect-slp.c (vect_slp_analyze_node_operations): Detect
1405 the case where there are two different def types for the
1406 same operand at different operand position in the same stmt.
1408 2019-02-07 Jakub Jelinek <jakub@redhat.com>
1410 Backported from mainline
1411 2019-02-05 Jakub Jelinek <jakub@redhat.com>
1414 * dce.c (delete_unmarked_insns): Don't remove no-op moves if they
1415 can throw, non-call exceptions are enabled and we can't delete
1416 dead exceptions or alter cfg. Set must_clean if
1417 delete_insn_and_edges returns true, don't set it blindly for calls.
1419 PR rtl-optimization/89195
1420 * combine.c (make_extraction): For MEMs, don't extract bytes outside
1421 of the original MEM.
1424 * optabs.c (prepare_cmp_insn): Pass x and y to
1425 emit_block_comp_via_libcall rather than XEXP (x, 0) and XEXP (y, 0).
1427 2019-02-02 Jakub Jelinek <jakub@redhat.com>
1430 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
1431 Punt with warning on aggregate return or argument types. Ignore
1432 type/mode checking for uniform arguments.
1434 2019-02-01 Jakub Jelinek <jakub@redhat.com>
1436 PR tree-optimization/88107
1437 * tree-cfg.c (find_outermost_region_in_block): Add ALL argument,
1438 instead of assertion that eh_region_outermost is non-NULL, if it
1439 is NULL, set *ALL to true and return NULL.
1440 (move_sese_region_to_fn): Adjust caller, if all is set, call
1441 duplicate_eh_regions with NULL region.
1443 2019-01-31 Jakub Jelinek <jakub@redhat.com>
1446 * ipa-inline.c (sanitize_attrs_match_for_inline_p): Allow inlining
1447 always_inline callees into no_sanitize_address callers.
1449 2019-01-30 Jakub Jelinek <jakub@redhat.com>
1452 * config/i386/i386.c (ix86_warn_parameter_passing_abi): Don't warn
1453 for arguments to functions that are TU-local and shouldn't be
1454 referenced by assembly.
1456 2019-01-29 Jakub Jelinek <jakub@redhat.com>
1460 * omp-simd-clone.c (simd_clone_clauses_extract)
1461 <case OMP_CLAUSE_ALIGNED>: Ignore clauses with NULL
1462 OMP_CLAUSE_ALIGNED_ALIGNMENT.
1464 2019-01-28 Jakub Jelinek <jakub@redhat.com>
1467 * gimplify.c (gimplify_omp_for): When adding OMP_CLAUSE_*_GIMPLE_SEQ
1468 for lastprivate/linear IV, push gimplify context around gimplify_assign
1469 and, if it needed any temporaries, pop it into a gimple bind around the
1473 * doc/invoke.texi (-mclwb, -mprfchw, -mrdpid, -mrdseed, -msgx,
1474 -madx, -mhle, -mavx5124fmaps, -mavx512vnni, -mavx5124vnniw): Document
1476 (bmi2): Add missing @opindex.
1477 * doc/extend.texi (x86 target attribute): Move fma4, lwp, ssse3
1478 options alphabetically. Add missing 3dnow, 3dnowa, adx, avx, avx2,
1479 avx5124fmaps, avx5124vnniw, avx512bitalg, avx512bw, avx512cd,
1480 avx512dq, avx512er, avx512f, avx512ifma, avx512pf, avx512vbmi,
1481 avx512vbmi2, avx512vl, avx512vnni, avx512vpopcntdq, bmi, bmi2,
1482 clflushopt, clwb, clzero, crc32, cx16, f16c, fma, fsgsbase,
1483 fxsr, gfni, hle, lzcnt, movbe, movdir64b, movdiri, mwaitx, pconfig,
1484 pku, prefetchwt1, prfchw, rdpid, rdrnd, rdseed, rtm, sahf,
1485 sgx, sha, shstk, tbm, vaes, vpclmulqdq, wbnoinvd, xsave,
1486 xsavec, xsaveopt and xsaves options.
1488 2019-01-27 Jakub Jelinek <jakub@redhat.com>
1491 * config/i386/sse.md
1492 (<mask_codefor>avx512dq_shuf_<shuffletype>64x2_1<mask_name>,
1493 avx512f_shuf_<shuffletype>64x2_1<mask_name>): Ensure the
1494 first constants in pairs are multiples of 2. Formatting fixes.
1495 (avx512vl_shuf_<shuffletype>32x4_1<mask_name>,
1496 avx512vl_shuf_<shuffletype>32x4_1<mask_name>): Ensure the
1497 first constants in each quadruple are multiples of 4. Formatting fixes.
1499 2019-01-24 Jakub Jelinek <jakub@redhat.com>
1501 PR tree-optimization/88964
1502 * gimple-loop-interchange.cc (loop_cand::analyze_induction_var): Also
1503 punt if HONOR_SNANS (chrec).
1505 2019-01-23 Jakub Jelinek <jakub@redhat.com>
1507 PR tree-optimization/88964
1508 * gimple-loop-interchange.cc (loop_cand::analyze_induction_var): Use
1509 build_zero_cst instead of build_int_cst. Return false for loop
1510 invariants which honor signed zeros.
1512 2019-01-22 Jakub Jelinek <jakub@redhat.com>
1515 * config/rs6000/rs6000.c: Include tree-vrp.h and tree-ssanames.h.
1516 (rs6000_gimple_fold_builtin): If MEM_REF address doesn't satisfy
1517 is_gimple_mem_ref_addr predicate, force it into a SSA_NAME first.
1520 * gimplify.c (gimplify_omp_atomic): Handle bitfield atomics with
1521 non-integral DECL_BIT_FIELD_REPRESENTATIVEs.
1524 * optabs.c (add_equal_note): Add op0_mode argument, use it instead of
1526 (expand_binop_directly, expand_doubleword_clz,
1527 expand_doubleword_popcount, expand_ctz, expand_ffs,
1528 expand_unop_direct, maybe_emit_unop_insn): Adjust callers.
1530 PR rtl-optimization/49429
1532 PR rtl-optimization/86334
1534 * expr.c (emit_block_move_hints): Move marking of MEM_EXPRs
1535 addressable from here...
1536 (emit_block_op_via_libcall): ... to here.
1538 2019-01-17 Jakub Jelinek <jakub@redhat.com>
1541 * config/aarch64/arm_neon.h: Fix #pragma GCC target syntax - replace
1542 (("..."))) with ("..."). Use arch=armv8.2-a+sha3 instead of
1543 arch=armv8.2-a+crypto for vsha512hq_u64 etc. intrinsics.
1545 PR rtl-optimization/88870
1546 * dce.c (deletable_insn_p): Never delete const/pure calls that can
1547 throw if we can't alter the cfg or delete dead exceptions.
1548 (mark_insn): Don't call find_call_stack_args for such calls.
1550 2019-01-11 Jakub Jelinek <jakub@redhat.com>
1553 * passes.c (finish_optimization_passes): Call print_combine_total_stats
1554 inside of pass_combine_1 dump rather than pass_profile_1.
1556 2019-01-10 Jakub Jelinek <jakub@redhat.com>
1559 * attribs.c (handle_dll_attribute): Clear TREE_STATIC after setting
1562 2019-01-07 Jakub Jelinek <jakub@redhat.com>
1565 * dwarf2out.c (const_ok_for_output_1): Remove redundant call to
1566 const_not_ok_for_debug_p target hook.
1567 (mem_loc_descriptor) <case UNSPEC>: Only call const_ok_for_output_1
1568 on UNSPEC and subexpressions thereof if all subexpressions of the
1569 UNSPEC are CONSTANT_P.
1571 2019-01-05 Jakub Jelinek <jakub@redhat.com>
1574 * dwarf2out.c (const_ok_for_output_1): Reject MINUS that contains
1575 SYMBOL_REF, CODE_LABEL or UNSPEC in subexpressions of second argument.
1576 Reject PLUS that contains SYMBOL_REF, CODE_LABEL or UNSPEC in
1577 subexpressions of both operands.
1578 (mem_loc_descriptor): Handle UNSPEC if target hook acks it and all the
1579 subrtxes are CONSTANT_P.
1581 2019-02-07 Richard Biener <rguenther@suse.de>
1583 Backport from mainline
1584 2018-11-20 Richard Biener <rguenther@suse.de>
1586 PR tree-optimization/88105
1587 * tree-ssa-dom.c (pass_dominator::execute): Do not walk
1590 2018-11-28 Richard Biener <rguenther@suse.de>
1592 PR tree-optimization/88223
1593 * tree-ssa-sccvn.c (vn_reference_lookup_3): When skipping
1594 over a stored-same value may-alias store make sure to consider
1595 partial overlaps which are valid when TBAA reasonings do not
1596 apply and byte-granular overlaps are possible at all.
1598 2019-02-07 Richard Biener <rguenther@suse.de>
1600 Backport from mainline
1601 2018-12-11 Richard Biener <rguenther@suse.de>
1605 * tree-complex.c (update_complex_assignment): Properly transfer
1606 or clean EH info around gimple_assign_set_rhs_with_ops.
1608 2018-11-15 Richard Biener <rguenther@suse.de>
1610 PR tree-optimization/88030
1611 * tree-complex.c (need_eh_cleanup): New global.
1612 (update_complex_assignment): Mark blocks that need EH update.
1613 (expand_complex_comparison): Likewise.
1614 (tree_lower_complex): Allocate and deallocate need_eh_cleanup,
1615 perform EH cleanup and schedule CFG cleanup if that did anything.
1617 2018-11-08 Richard Biener <rguenther@suse.de>
1619 PR tree-optimization/87929
1620 * tree-complex.c (expand_complex_comparison): Clean EH.
1622 2019-02-06 Eric Botcazou <ebotcazou@adacore.com>
1624 * config/i386/i386.c (ix86_expand_prologue): Emit a memory blockage
1625 after restoring registers saved to allocate the frame on Windows.
1627 2019-02-06 Richard Biener <rguenther@suse.de>
1629 Backport from mainline
1630 2019-01-31 Richard Biener <rguenther@suse.de>
1632 PR tree-optimization/89135
1633 * tree-ssa-phiprop.c (pass_phiprop::execute): Skip blocks
1634 with abnormal preds.
1636 2019-01-18 Richard Biener <rguenther@suse.de>
1638 PR tree-optimization/88903
1639 * tree-vect-stmts.c (vectorizable_shift): Verify we see all
1640 scalar stmts a SLP shift amount is composed of when detecting
1643 2019-02-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
1645 Backport from mainline.
1646 2019-01-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
1647 * config/rs6000/rs6000-c.c (altivec-resolve_overloaded_builtin):
1648 Change handling of ALTIVEC_BUILTIN_VEC_EXTRACT. Coerce result to
1649 type of vector element when vec_extract is implemented by direct
1652 2019-02-05 Andreas Krebbel <krebbel@linux.ibm.com>
1654 Backport from mainline
1655 2019-02-05 Andreas Krebbel <krebbel@linux.ibm.com>
1658 * config/s390/s390.md: Remove load and test FP splitter.
1660 2019-02-04 Bill Schmidt <wshmidt@linux.ibm.com>
1663 Backport from mainline
1665 2019-01-30 Bill Schmidt <wschmidt@linux.ibm.com>
1668 * config/rs6000/vsx.md (*vsx_reduc_<VEC_reduc_name>_v4sf_scalar):
1669 Disable for little-endian.
1671 2019-01-22 Jakub Jelinek <jakub@redhat.com>
1674 * config/rs6000/vsx.md (*vsx_reduc_<VEC_reduc_name>_v2df_scalar):
1675 Disable for little endian.
1677 2019-02-04 Eric Botcazou <ebotcazou@adacore.com>
1679 * config/sparc/sparc.h: Remove superfluous blank lines.
1680 * config/sparc/sparc.c (global_offset_table_rtx): Rename into...
1681 (got_register_rtx): ...this.
1682 (sparc_got): Adjust to above renaming.
1683 (sparc_tls_got): Likewise.
1684 (sparc_output_mi_thunk): Likewise.
1685 (sparc_init_pic_reg): Likewise.
1686 (save_local_or_in_reg_p): Fix test on the GOT register.
1687 (USE_HIDDEN_LINKONCE): Move around.
1688 (get_pc_thunk_name): Likewise.
1689 (gen_load_pcrel_sym): Likewise.
1690 (load_got_register): Likewise.
1692 2019-02-01 Bin Cheng <bin.cheng@linux.alibaba.com>
1694 Backport from mainline
1695 2018-10-15 Bin Cheng <bin.cheng@linux.alibaba.com>
1697 PR tree-optimization/87022
1698 * tree-loop-distribution.c (pg_add_dependence_edges): Check all
1699 bits in dist vector rather than the first one.
1701 2019-02-01 Bin Cheng <bin.cheng@linux.alibaba.com>
1703 Backport from mainline
1704 2019-02-01 Bin Cheng <bin.cheng@linux.alibaba.com>
1706 PR tree-optimization/88932
1707 * tree-predcom.c (try_combine_chains): Get loop bbs in dom order.
1709 2018-01-31 Bill Schmidt <wschmidt@linux.ibm.com>
1711 Backport from mainline
1712 2018-01-31 Bill Schmidt <wschmidt@linux.ibm.com>
1714 PR tree-optimization/89008
1715 * gimple-ssa-strength-reduction.c (slsr_process_mul): Don't
1716 process anything of the form X * 0.
1718 2019-01-31 Richard Biener <rguenther@suse.de>
1720 Backport from mainline
1721 2019-01-31 Richard Biener <rguenther@suse.de>
1723 PR rtl-optimization/89115
1724 * lra.c (lra_rtx_hash): Properly hash CONST_INT values.
1726 2019-01-30 Richard Biener <rguenther@suse.de>
1728 PR rtl-optimization/89115
1729 * opts.c (default_options_optimization): Reduce
1730 PARAM_MAX_DSE_ACTIVE_LOCAL_STORES by a factor of 10 at -O1.
1731 Make PARAM_LOOP_INVARIANT_MAX_BBS_IN_LOOP reduction relative
1734 2019-01-30 Kewen Lin <linkw@gcc.gnu.org>
1736 Backport from mainline.
1737 2019-01-17 Kewen Lin <linkw@gcc.gnu.org>
1739 * doc/extend.texi: Add four new prototypes for vec_ld and seven new
1740 prototypes for vec_st.
1741 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add entries
1742 for scalar address type variants of altivec_vec_ld/altivec_vec_st,
1743 mainly on signed/unsigned long long and double.
1745 2019-01-29 Vineet Gupta <vgupta@synopsys.com>
1747 Backport from mainline.
1748 * gcc/config.gcc: Force .init_array for ARC.
1750 2019-01-29 Vineet Gupta <vgupta@synopsys.com>
1752 Backport from mainline.
1753 * config/arc/atomic.md: Add operand to DMB instruction.
1755 2019-01-27 Uroš Bizjak <ubizjak@gmail.com>
1758 * rtl.h (prepare_copy_insn): New prototype.
1759 * gcse.c (prepare_copy_insn): New function, split out from
1760 process_insert_insn.
1761 (process_insert_insn): Use prepare_copy_insn.
1762 * store-motion.c (replace_store_insn): Use prepare_copy_insn
1763 instead of gen_move_insn.
1765 2019-01-25 Richard Biener <rguenther@suse.de>
1767 PR tree-optimization/86865
1768 * graphite-scop-detection.c (scop_detection::can_represent_loop):
1769 Reject non-do-while loops.
1771 2019-01-24 Uroš Bizjak <ubizjak@gmail.com>
1774 * config/i386/sse.md (sse2_cvtpi2pd): Add SSE alternatives.
1775 Disparage MMX alternative.
1776 (sse2_cvtpd2pi): Ditto.
1777 (sse2_cvttpd2pi): Ditto.
1779 2019-01-24 Richard Earnshaw <rearnsha@arm.com>
1782 * profile-count.h (profile_count): On ARM systems using GCC 6/7/8
1783 force the alignment of m_val.
1785 2019-01-24 Richard Biener <rguenther@suse.de>
1787 Backport from mainline
1788 2019-01-23 Richard Biener <rguenther@suse.de>
1790 PR tree-optimization/89008
1791 * tree-ssa-reassoc.c (eliminate_using_constants): For * 0 do
1792 not leave another stray operand.
1794 2019-01-22 Uroš Bizjak <ubizjak@gmail.com>
1797 * config/i386/i386.c (ix86_expand_builtin) [case IX86_BUILTIN_BEXTRI32,
1798 case IX86_BUILTIN_BEXTRI64]: Sanitize operands.
1800 2019-01-18 Richard Earnshaw <rearnsha@arm.com>
1803 * config/arm/arm-cpus.in (mp): New feature.
1805 (fgroup ARMv7ve): Add mp and sec features.
1806 (arch armv7-a): Add options to allow mp and sec extensions.
1807 (cpu generic-armv7-a): Add options to allow mp and sec extensions.
1808 (cpu cortex-a5, cpu cortex-7, cpu cortex-a9): Add mp and sec
1809 extenstions to the base architecture.
1810 (cpu cortex-a8): Add sec extension to the base architecture.
1811 (cpu marvell-pj4): Add mp and sec extensions to the base architecture.
1812 * config/arm/t-aprofile (MULTILIB_MATCHES): Map all armv7-a arch
1813 variants down to the base v7-a varaint.
1814 * config/arm/t-multilib (v7_a_arch_variants): New variable.
1815 * doc/invoke.texi (ARM Options): Add +mp and +sec to the list
1816 of permitted extensions for -march=armv7-a and for
1817 -mcpu=generic-armv7-a.
1819 2019-01-17 Uroš Bizjak <ubizjak@gmail.com>
1821 * config/alpha/alpha.c (alpha_gimplify_va_arg):
1822 Handle split indirect COMPLEX_TYPE arguments.
1824 2019-01-17 Eric Botcazou <ebotcazou@adacore.com>
1826 * gimple-ssa-isolate-paths.c (stmt_uses_name_in_undefined_way): Replace
1827 flag_non_call_exceptions with cfun->can_throw_non_call_exceptions.
1828 (stmt_uses_0_or_null_in_undefined_way): Likewise.
1829 * tree-ssa-alias.c (same_addr_size_stores_p): Likewise.
1831 2019-01-16 Aaron Sawdey <acsawdey@linux.ibm.com>
1833 Backported from mainline
1834 2019-01-03 Aaron Sawdey <acsawdey@linux.ibm.com>
1835 * config/rs6000/rs6000-string.c (expand_block_move): Don't use
1836 unaligned vsx and avoid lxvd2x/stxvd2x.
1837 (gen_lvx_v4si_move): New function.
1839 2019-01-16 Martin Jambor <mjambor@suse.cz>
1841 Backported from mainline
1842 2018-12-10 Martin Jambor <mjambor@suse.cz>
1845 * ipa-prop.c (determine_locally_known_aggregate_parts): Make sure
1846 we check pointers against pointers.
1848 2019-01-10 Jan Hubicka <hubicka@ucw.cz>
1850 Backported from mainline
1851 2018-12-27 Jan Hubicka <hubicka@ucw.cz>
1853 * ipa-devirt.c (polymorphic_call_target_d): Add n_odr_types.
1854 (polymorphic_call_target_hasher::hash): Hash it.
1855 (polymorphic_call_target_hasher::equal): Compare it.
1856 (possible_polymorphic_call_targets): Set it.
1857 * tree.c (free_lang_data): Rebuild type inheritance graph even on
1860 2019-01-03 Jan Hubicka <hubicka@ucw.cz>
1862 Backported from mainline
1863 2019-01-02 Richard Biener <rguenther@suse.de>
1866 * ipa-icf.h (sem_item_optimizer::sort_congruence_split): Declare.
1867 * ipa-icf.c (sem_item_optimizer::sort_congruence_split): New
1869 (sem_item_optimizer::do_congruence_step_f): Sort the congruence
1870 set after UIDs before splitting them.
1872 2019-01-03 Jan Hubicka <hubicka@ucw.cz>
1874 Backported from mainline
1875 2019-01-03 Jan Hubicka <hubicka@ucw.cz>
1877 PR tree-optimization/85574
1878 * tree-ssa-uncprop.c (struct equiv_hash_elt): Remove unused
1880 (struct ssa_equip_hash_traits): Declare.
1881 (val_ssa_equiv): Use custom hash traits using operand_equal_p.
1883 2019-01-09 Eric Botcazou <ebotcazou@adacore.com>
1885 * config/sparc/sparc.md (*tablejump_sp32): Merge into...
1886 (*tablejump_sp64): Likewise.
1887 (*tablejump<P:mode>): ...this.
1888 (*call_address_sp32): Merge into...
1889 (*call_address_sp64): Likewise.
1890 (*call_address<P:mode>): ...this.
1891 (*call_symbolic_sp32): Merge into...
1892 (*call_symbolic_sp64): Likewise.
1893 (*call_symbolic<P:mode>): ...this.
1894 (call_value): Remove constraint and add predicate.
1895 (*call_value_address_sp32): Merge into...
1896 (*call_value_address_sp64): Likewise.
1897 (*call_value_address<P:mode>): ...this.
1898 (*call_value_symbolic_sp32): Merge into...
1899 (*call_value_symbolic_sp64): Likewise.
1900 (*call_value_symbolic<P:mode>): ...this.
1901 (*sibcall_symbolic_sp32): Merge into...
1902 (*sibcall_symbolic_sp64): Likewise.
1903 (*sibcall_symbolic<P:mode>): ...this.
1904 (sibcall_value): Remove constraint and add predicate.
1905 (*sibcall_value_symbolic_sp32): Merge into...
1906 (*sibcall_value_symbolic_sp64): Likewise.
1907 (*sibcall_value_symbolic<P:mode>): ...this.
1908 (window_save): Minor tweak.
1909 (*branch_sp32): Merge into...
1910 (*branch_sp64): Likewise.
1911 (*branch<P:mode>): ...this.
1913 2019-01-09 Eric Botcazou <ebotcazou@adacore.com>
1914 James Clarke <jrtc27@jrtc27.com>
1917 * config/sparc/sparc.c (sparc_legitimize_tls_address): Only use Pmode
1918 consistently in TLS address generation and adjust code to the renaming
1919 of patterns. Mark calls to __tls_get_addr as const.
1920 * config/sparc/sparc.md (tgd_hi22): Turn into...
1921 (tgd_hi22<P:mode>): ...this and use Pmode throughout.
1922 (tgd_lo10): Turn into...
1923 (tgd_lo10<P:mode>): ...this and use Pmode throughout.
1924 (tgd_add32): Merge into...
1925 (tgd_add64): Likewise.
1926 (tgd_add<P:mode>): ...this and use Pmode throughout.
1927 (tldm_hi22): Turn into...
1928 (tldm_hi22<P:mode>): ...this and use Pmode throughout.
1929 (tldm_lo10): Turn into...
1930 (tldm_lo10<P:mode>): ...this and use Pmode throughout.
1931 (tldm_add32): Merge into...
1932 (tldm_add64): Likewise.
1933 (tldm_add<P:mode>): ...this and use Pmode throughout.
1934 (tldm_call32): Merge into...
1935 (tldm_call64): Likewise.
1936 (tldm_call<P:mode>): ...this and use Pmode throughout.
1937 (tldo_hix22): Turn into...
1938 (tldo_hix22<P:mode>): ...this and use Pmode throughout.
1939 (tldo_lox10): Turn into...
1940 (tldo_lox10<P:mode>): ...this and use Pmode throughout.
1941 (tldo_add32): Merge into...
1942 (tldo_add64): Likewise.
1943 (tldo_add<P:mode>): ...this and use Pmode throughout.
1944 (tie_hi22): Turn into...
1945 (tie_hi22<P:mode>): ...this and use Pmode throughout.
1946 (tie_lo10): Turn into...
1947 (tie_lo10<P:mode>): ...this and use Pmode throughout.
1948 (tie_ld64): Use DImode throughout.
1949 (tie_add32): Merge into...
1950 (tie_add64): Likewise.
1951 (tie_add<P:mode>): ...this and use Pmode throughout.
1952 (tle_hix22_sp32): Merge into...
1953 (tle_hix22_sp64): Likewise.
1954 (tle_hix22<P:mode>): ...this and use Pmode throughout.
1955 (tle_lox22_sp32): Merge into...
1956 (tle_lox22_sp64): Likewise.
1957 (tle_lox22<P:mode>): ...this and use Pmode throughout.
1958 (*tldo_ldub_sp32): Merge into...
1959 (*tldo_ldub_sp64): Likewise.
1960 (*tldo_ldub<P:mode>): ...this and use Pmode throughout.
1961 (*tldo_ldub1_sp32): Merge into...
1962 (*tldo_ldub1_sp64): Likewise.
1963 (*tldo_ldub1<P:mode>): ...this and use Pmode throughout.
1964 (*tldo_ldub2_sp32): Merge into...
1965 (*tldo_ldub2_sp64): Likewise.
1966 (*tldo_ldub2<P:mode>): ...this and use Pmode throughout.
1967 (*tldo_ldsb1_sp32): Merge into...
1968 (*tldo_ldsb1_sp64): Likewise.
1969 (*tldo_ldsb1<P:mode>): ...this and use Pmode throughout.
1970 (*tldo_ldsb2_sp32): Merge into...
1971 (*tldo_ldsb2_sp64): Likewise.
1972 (*tldo_ldsb2<P:mode>): ...this and use Pmode throughout.
1973 (*tldo_ldub3_sp64): Use DImode throughout.
1974 (*tldo_ldsb3_sp64): Likewise.
1975 (*tldo_lduh_sp32): Merge into...
1976 (*tldo_lduh_sp64): Likewise.
1977 (*tldo_lduh<P:mode>): ...this and use Pmode throughout.
1978 (*tldo_lduh1_sp32): Merge into...
1979 (*tldo_lduh1_sp64): Likewise.
1980 (*tldo_lduh1<P:mode>): ...this and use Pmode throughout.
1981 (*tldo_ldsh1_sp32): Merge into...
1982 (*tldo_ldsh1_sp64): Likewise.
1983 (*tldo_ldsh1<P:mode>): ...this and use Pmode throughout.
1984 (*tldo_lduh2_sp64): Use DImode throughout.
1985 (*tldo_ldsh2_sp64): Likewise.
1986 (*tldo_lduw_sp32): Merge into...
1987 (*tldo_lduw_sp64): Likewise.
1988 (*tldo_lduw<P:mode>): ...this and use Pmode throughout.
1989 (*tldo_lduw1_sp64): Use DImode throughout.
1990 (*tldo_ldsw1_sp64): Likewise.
1991 (*tldo_ldx_sp64): Likewise.
1992 (*tldo_stb_sp32): Merge into...
1993 (*tldo_stb_sp64): Likewise.
1994 (*tldo_stb<P:mode>): ...this and use Pmode throughout.
1995 (*tldo_sth_sp32): Merge into...
1996 (*tldo_sth_sp64): Likewise.
1997 (*tldo_sth<P:mode>): ...this and use Pmode throughout.
1998 (*tldo_stw_sp32): Merge into...
1999 (*tldo_stw_sp64): Likewise.
2000 (*tldo_stw<P:mode>): ...this and use Pmode throughout.
2001 (*tldo_stx_sp64): Use DImode throughout.
2003 2019-01-09 Eric Botcazou <ebotcazou@adacore.com>
2005 * doc/invoke.texi (-Os): Add reference to -finline-functions.
2006 (-finline-functions): Remove reference to -O2, add references
2007 to -fprofile-use and -fauto-profile.
2009 2019-01-08 Jakub Jelinek <jakub@redhat.com>
2011 Backported from mainline
2012 2019-01-05 Jakub Jelinek <jakub@redhat.com>
2016 * expr.c (expand_assignment): For calls returning VLA structures
2017 if to_rtx is not a MEM, force it into a stack temporary.
2019 2019-01-04 Jakub Jelinek <jakub@redhat.com>
2022 * config/i386/i386.c (ix86_expand_divmod_libfunc): Use mode instead
2023 of GET_MODE (opN) as modes of the libcall arguments.
2025 2019-01-03 Jakub Jelinek <jakub@redhat.com>
2028 * dwarf2out.c (modified_type_die): If type is equal to sizetype,
2029 change it to qualified_type.
2031 2018-12-21 Jakub Jelinek <jakub@redhat.com>
2035 * omp-expand.c (extract_omp_for_update_vars): Regimplify the condition
2037 (expand_omp_for_generic): Don't clobber t temporary for ordered loops.
2040 * config/i386/sse.md (*avx512pf_gatherpf<mode>sf_mask,
2041 *avx512pf_gatherpf<mode>df_mask, *avx512pf_scatterpf<mode>sf_mask,
2042 *avx512pf_scatterpf<mode>df_mask): Use %X5 instead of %5 for
2044 (gatherq_mode): Remove mode iterator.
2045 (*avx512f_gathersi<mode>, *avx512f_gathersi<mode>_2): Use X instead
2047 (*avx512f_gatherdi<mode>): Use X instead of <gatherq_mode>.
2048 (*avx512f_gatherdi<mode>_2, *avx512f_scattersi<mode>,
2049 *avx512f_scatterdi<mode>): Use %X5 for -masm=intel.
2051 PR rtl-optimization/88563
2052 * expr.c (expand_expr_real_2) <case WIDEN_MULT_EXPR>: Swap innermode
2053 and mode arguments to convert_modes. Likewise swap mode and word_mode
2054 arguments. Handle both arguments with VOIDmode before convert_modes
2055 of one of them. Formatting fixes.
2057 2018-12-19 Jakub Jelinek <jakub@redhat.com>
2060 * config/i386/vpclmulqdqintrin.h (_mm256_clmulepi64_epi128): Enable
2061 for -mavx -mvpclmulqdq rather than just for -mavx512vl -mvpclmulqdq.
2063 2018-12-14 Jakub Jelinek <jakub@redhat.com>
2067 * expr.h (categorize_ctor_elements): Add p_unique_nz_elts argument.
2068 * expr.c (categorize_ctor_elements_1): Likewise. Compute it like
2069 p_nz_elts, except don't multiply it by mult. Adjust recursive call.
2070 Fix up COMPLEX_CST handling.
2071 (categorize_ctor_elements): Add p_unique_nz_elts argument, initialize
2072 it and pass it through to categorize_ctor_elements_1.
2073 (mostly_zeros_p, all_zeros_p): Adjust categorize_ctor_elements callers.
2074 * gimplify.c (gimplify_init_constructor): Likewise. Don't force
2075 ctor into readonly data section if num_unique_nonzero_elements is
2076 smaller or equal to 1/8 of num_nonzero_elements and size is >= 64
2079 2018-12-13 Jakub Jelinek <jakub@redhat.com>
2081 PR rtl-optimization/88470
2082 * cfgcleanup.c (outgoing_edges_match): If the function is
2083 shrink-wrapped and bb1 ends with a JUMP_INSN with a single fake
2084 edge to EXIT, return false.
2086 PR rtl-optimization/88416
2087 * valtrack.c (cleanup_auto_inc_dec): Handle pre/post-inc/dec/modify
2088 even if !AUTO_INC_DEC.
2090 2018-12-07 Jakub Jelinek <jakub@redhat.com>
2093 * final.c (rest_of_handle_final): Don't call collect_fn_hard_reg_usage
2094 for functions with naked attribute.
2096 2018-12-06 Jakub Jelinek <jakub@redhat.com>
2099 * config/aarch64/aarch64.c (aarch64_print_address_internal): Don't
2100 call output_operand_lossage on VOIDmode CONST_INTs. After
2101 output_operand_lossage do return false.
2103 2018-12-05 Jakub Jelinek <jakub@redhat.com>
2105 PR tree-optimization/87360
2106 * gimple-loop-jam.c (tree_loop_unroll_and_jam): On failure to analyze
2107 data dependencies, don't return false, just continue.
2109 2018-11-29 Jakub Jelinek <jakub@redhat.com>
2112 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): For
2113 vec_add and vec_sub builtins, perform PLUS_EXPR or MINUS_EXPR
2114 in unsigned_type_for instead of vector integral type where overflow
2117 2018-11-26 Jakub Jelinek <jakub@redhat.com>
2120 * dwarf2out.c (secname_for_decl): For functions with
2121 DECL_SECTION_NAME if in_cold_section_p, try to return
2122 current_function_section's name if it is a named section.
2124 2018-11-22 Jakub Jelinek <jakub@redhat.com>
2128 * config/i386/i386.c (ix86_option_override_internal): Default
2129 ix86_stack_protector_guard to SSP_TLS only if TARGET_THREAD_SSP_OFFSET
2131 * config/i386/i386.md (stack_protect_set, stack_protect_set_<mode>,
2132 stack_protect_test, stack_protect_test_<mode>): Use empty condition
2133 instead of TARGET_SSP_TLS_GUARD.
2135 2018-11-20 Jakub Jelinek <jakub@redhat.com>
2137 PR tree-optimization/87895
2138 * omp-simd-clone.c (ipa_simd_modify_function_body): When removing
2139 or replacing GIMPLE_RETURN, set EDGE_FALLTHRU on the edge to EXIT.
2140 (simd_clone_adjust): Don't set EDGE_FALLTHRU here. In a loop that
2141 redirects edges to EXIT to edges to incr_bb, iterate while EXIT
2142 has any preds and always use EDGE_PRED (, 0).
2144 2018-11-19 Jakub Jelinek <jakub@redhat.com>
2146 PR tree-optimization/88071
2147 * tree-vect-loop.c (vectorize_fold_left_reduction): Pass true instead
2148 of false as last argument to gsi_remove.
2149 * tree-vect-stmts.c (vect_finish_replace_stmt): Pass true instead of
2150 false as last argument to gsi_replace.
2153 * omp-expand.c: Don't include debug.h.
2154 (adjust_context_and_scope): Add REGION argument. Find DECL_CONTEXT
2155 from innermost outer parallel, task, or target that has a
2156 child_fn set, or, if there is no such outer region, use
2157 current_function_decl. Do the DECL_CONTEXT adjustment regardless of
2158 whether a suitable BLOCK is found or not.
2159 (expand_parallel_call): Don't call adjust_context_and_scope here.
2160 (grid_expand_target_grid_body): Revert 2017-01-25 changes.
2161 (expand_omp_taskreg, expand_omp_target): Likewise. Call
2162 adjust_context_and_scope.
2163 * dwarf2out.c (dwarf2out_early_global_decl): For
2164 decl_function_context recurse instead of calling dwarf2out_decl.
2166 2018-11-16 Jakub Jelinek <jakub@redhat.com>
2168 PR rtl-optimization/87475
2169 * cfgrtl.c (patch_jump_insn): Allow redirection failure for
2170 CROSSING_JUMP_P insns.
2171 (cfg_layout_redirect_edge_and_branch): Don't ICE if ret is NULL.
2173 2018-11-15 Jakub Jelinek <jakub@redhat.com>
2175 PR rtl-optimization/88018
2176 * cfgrtl.c (fixup_abnormal_edges): Guard moving insns to fallthru edge
2177 on the presence of fallthru edge, rather than if it is a USE or not.
2179 2018-11-13 Jakub Jelinek <jakub@redhat.com>
2181 PR tree-optimization/87898
2182 * omp-simd-clone.c (ipa_simd_modify_function_body): Remove debug stmts
2183 where the first argument was changed into a non-decl.
2185 2019-01-03 Jan Hubicka <hubicka@ucw.cz>
2187 Backport from mainline
2188 2019-01-02 Jan Hubicka <hubicka@ucw.cz>
2191 * varpool.c (varpool_node::ctor_useable_for_folding_p): Also return
2192 false at WPA time when body was removed.
2194 2019-01-03 Jan Hubicka <hubicka@ucw.cz>
2196 Backport from mainline
2197 2018-08-29 Jan Hubicka <jh@suse.cz>
2201 * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
2202 * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
2204 2019-01-04 Aaron Sawdey <acsawdey@linux.ibm.com>
2206 Backport from mainline
2207 2018-11-28 Aaron Sawdey <acsawdey@linux.ibm.com>
2209 * config/rs6000/rs6000-string.c (expand_block_clear): Change how
2210 we determine if unaligned vsx is ok.
2212 2019-01-03 Jan Hubicka <hubicka@ucw.cz>
2214 Backport from mainline
2215 2019-01-03 Jan Hubicka <hubicka@ucw.cz>
2217 * ipa-utils.c (scale_ipa_profile_for_fn): Break out from ...
2218 (ipa_merge_profiles): ... here; do not ICE on thunks and aliases.
2220 2019-01-03 Martin Liska <mliska@suse.cz>
2222 Backport from mainline
2223 2019-01-01 Jan Hubicka <hubicka@ucw.cz>
2225 * coverage.c (get_coverage_counts): Use current_function_decl.
2226 * profile.c (read_thunk_profile): New function.
2227 (branch_prob): Add THUNK parameter.
2228 * tree-profile.c (tree_profiling): Handle thunks.
2229 * value-prof.c (init_node_map): Handle thunks.
2230 * value-prof.h (branch_prob): Upate prototype.
2231 (read_thunk_profile): Declare.
2233 2019-01-03 Iain Sandoe <iain@sandoe.co.uk>
2236 2018-12-23 Iain Sandoe <iain@sandoe.co.uk>
2238 backport from mainline.
2239 2018-12-12 Segher Boessenkool <segher@kernel.crashing.org>
2240 Iain Sandoe <iain@sandoe.co.uk>
2243 * config/rs6000/rs6000.c (save_reg_p): Do not save the picbase reg
2244 unless it has been used.
2245 (first_reg_to_save): Remove dead code.
2247 2019-01-02 Segher Boessenkool <segher@kernel.crashing.org>
2250 2018-12-06 Segher Boessenkool <segher@kernel.crashing.org>
2253 * doc/extend.texi (Basic Asm): Update grammar.
2254 (Extended Asm): Update grammar.
2257 2018-12-06 Segher Boessenkool <segher@kernel.crashing.org>
2259 * doc/extend.texi (Using Assembly Language with C): Document asm inline.
2260 (Size of an asm): Fix typo. Document asm inline.
2261 * gimple-pretty-print.c (dump_gimple_asm): Handle asm inline.
2262 * gimple.h (enum gf_mask): Add GF_ASM_INLINE.
2263 (gimple_asm_set_volatile): Fix typo.
2264 (gimple_asm_inline_p): New.
2265 (gimple_asm_set_inline): New.
2266 * gimplify.c (gimplify_asm_expr): Propagate the asm inline flag from
2268 * ipa-icf-gimple.c (func_checker::compare_gimple_asm): Compare the
2269 gimple_asm_inline_p flag, too.
2270 * tree-core.h (tree_base): Document that protected_flag is ASM_INLINE_P
2272 * tree-inline.c (estimate_num_insns): If gimple_asm_inline_p return
2273 a minimum size for an asm.
2274 * tree.h (ASM_INLINE_P): New.
2276 2018-12-28 Thomas Schwinge <thomas@codesourcery.com>
2277 Julian Brown <julian@codesourcery.com>
2279 * gimplify.c (gimplify_scan_omp_clauses): Fix known_eq typo/bug.
2281 2018-12-27 Martin Liska <mliska@suse.cz>
2283 Backport from mainline
2284 2018-12-15 Jan Hubicka <hubicka@ucw.cz>
2287 * ipa-polymorphic-call.c
2288 (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Handle
2289 arguments of thunks correctly.
2290 (ipa_polymorphic_call_context::get_dynamic_context): Be ready for
2291 NULL instance pinter.
2292 * lto-cgraph.c (lto_output_node): Always stream thunk info.
2294 2018-12-23 Iain Sandoe <iain@sandoe.co.uk>
2296 backport from mainline.
2297 2018-12-12 Segher Boessenkool <segher@kernel.crashing.org>
2298 Iain Sandoe <iain@sandoe.co.uk>
2301 * config/rs6000/rs6000.c (save_reg_p): Do not save the picbase reg
2302 unless it has been used.
2303 (first_reg_to_save): Remove dead code.
2305 2018-12-23 Iain Sandoe <iain@sandoe.co.uk>
2307 Backport from mainline
2308 2018-12-06 Iain Sandoe <iain@sandoe.co.uk>
2311 * config/darwin.h (TARGET_WEAK_NOT_IN_ARCHIVE_TOC) Remove, use the
2313 * config/rs6000/darwin7.h (TARGET_WEAK_NOT_IN_ARCHIVE_TOC): New.
2315 2018-12-23 Iain Sandoe <iain@sandoe.co.uk>
2317 Backport from mainline
2318 2018-12-06 Iain Sandoe <iain@sandoe.co.uk>
2321 * config/i386/darwin.h (STACK_BOUNDARY): Remove macro.
2322 * config/i386/i386.c (ix86_compute_frame_layout): Ensure at least 128b
2323 stack alignment in non-leaf functions.
2325 2018-12-21 Uros Bizjak <ubizjak@gmail.com>
2327 Backport from mainline
2328 2018-12-10 Uros Bizjak <ubizjak@gmail.com>
2331 * config/i386/i386.c (ix86_expand_sse_cmp): For vector modes,
2332 check operand 1 with vector_operand predicate.
2333 (ix86_expand_sse_movcc): For vector modes, check op_true with
2334 vector_operand, not nonimmediate_operand.
2336 2018-12-19 Bill Schmidt <wschmidt@linux.ibm.com>
2338 Backport from mainline
2339 2018-12-18 Bill Schmidt <wschmidt@linux.ibm.com>
2341 * doc/extend.texi (PowerPC Altivec/VSX Built-in Functions):
2342 Describe when a typedef name can be used as the type specifier for
2343 a vector type, and when it cannot.
2345 2018-12-19 Segher Boessenkool <segher@kernel.crashing.org>
2348 2018-12-19 Segher Boessenkool <segher@kernel.crashing.org>
2351 * config/rs6000/vsx.md (*vsx_extract_<P:mode>_<VSX_D:mode>_load):
2352 Require TARGET_POWERPC64.
2354 2018-12-17 Senthil Kumar Selvaraj <senthilkumar.selvaraj@microchip.com>
2357 2018-12-17 Senthil Kumar Selvaraj <senthilkumar.selvaraj@microchip.com>
2359 PR rtl-optimization/88253
2360 * combine.c (combine_simplify_rtx): Test for side-effects before
2361 substituting by zero.
2363 2018-12-15 Segher Boessenkool <segher@kernel.crashing.org>
2366 2018-12-14 Segher Boessenkool <segher@kernel.crashing.org>
2368 PR rtl-optimization/88001
2369 * function.c (match_asm_constraints_1): Don't invalidly share RTL.
2371 2018-12-15 Jan Hubicka <jh@suse.cz>
2373 * ipa.c (cgraph_build_static_cdtor_1): Add OPTIMIZATION and TARGET
2375 (cgraph_build_static_cdtor): Update.
2376 (build_cdtor): Use OPTIMIZATION and TARGET of the first real cdtor
2379 2018-12-15 Jan Hubicka <jh@suse.cz>
2381 Backport from mainline
2382 2018-11-02 Richard Biener <rguenther@suse.de>
2384 * dwarf2out.c (build_abbrev_table): Guard lookup_external_ref call
2387 2018-12-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2389 Backport from mainline
2390 2018-12-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2392 * doc/invoke.texi (-msve-vector-bits): Clarify -msve-vector-bits=128
2395 2018-12-14 Andre Vieira <andre.simoesdiasvieira@arm.com>
2397 Backport from mainline
2399 * config/arm/arm-cpus.in (armv7-r): Add FP16conv configurations.
2400 (cortex-r7, cortex-r8): Update fpu and add new configuration.
2401 * doc/invoke.texi (armv7-r): Add two new vfp options.
2402 (nofp.dp): Add cortex-r7 and cortex-r8 to the list of targets that
2403 support this option.
2405 2018-12-13 H.J. Lu <hongjiu.lu@intel.com>
2407 Backport from mainline
2408 2018-11-05 Xuepeng Guo <xuepeng.guo@intel.com>
2411 * config/i386/emmintrin.h (__v16qs): New to cope with option
2413 (_mm_cmpeq_epi8): Replace __v16qi with __v16qs.
2414 (_mm_cmplt_epi8): Likewise.
2415 (_mm_cmpgt_epi8): Likewise.
2417 2018-12-13 Andreas Krebbel <krebbel@linux.ibm.com>
2419 Backport from mainline
2420 2018-12-13 Andreas Krebbel <krebbel@linux.ibm.com>
2422 * config/s390/s390-builtins.def (s390_vec_double_s64): Map to
2423 s390_vec_double_s64 instead of s390_vcdgb.
2424 (s390_vec_double_u64): Map to s390_vec_double_u64 instead of
2427 2018-12-13 Andreas Krebbel <krebbel@linux.ibm.com>
2429 Backport from mainline
2430 2018-12-13 Andreas Krebbel <krebbel@linux.ibm.com>
2432 * config/s390/vx-builtins.md ("vec_ctd_s64", "vec_ctd_u64")
2433 ("vec_ctsl", "vec_ctul"): Replace 0 with VEC_NOINEXACT.
2434 ("vec_double_s64", "vec_double_u64"): Replace 4 with VEC_INEXACT.
2436 2018-12-13 Eric Botcazou <ebotcazou@adacore.com>
2438 Backport from mainline
2439 2018-12-11 Eric Botcazou <ebotcazou@adacore.com>
2441 * config/rs6000/vxworks.h (RS6000_STARTING_FRAME_OFFSET): Define,
2442 accounting for STACK_BOUNDARY 128.
2443 (STACK_DYNAMIC_OFFSET): Likewise.
2445 2018-12-12 Peter Bergner <bergner@linux.ibm.com>
2447 Backport from mainline
2448 2018-12-07 Peter Bergner <bergner@linux.ibm.com>
2451 * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow
2452 -mabi=ieeelongdouble and -mabi=ibmlongdouble without -mlong-double-128.
2453 Do not error for -mabi=ibmlongdouble and no ISA 2.06 support.
2454 * doc/invoke.texi: Document -mabi=ibmlongdouble and -mabi=ieeelongdouble
2455 require -mlong-double-128.
2457 2018-12-05 Richard Biener <rguenther@suse.de>
2459 Backport from mainline
2460 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
2462 PR tree-optimization/87288
2463 * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
2464 into account when determining PEELING_FOR_NITERS.
2466 2018-12-05 Iain Sandoe <iain@sandoe.co.uk>
2468 Backport from mainline
2469 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
2471 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
2472 emit a diagnostic that it is not supported and reset the option.
2473 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
2474 supported and consume the option. (ASM_FINAL_SPEC): New.
2476 2018-12-04 Iain Sandoe <iain@sandoe.co.uk>
2478 Backport from mainline
2479 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
2482 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
2483 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
2484 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
2486 2018-11-29 Peter Bergner <bergner@linux.ibm.com>
2488 Backport from mainline
2489 2018-11-29 Peter Bergner <bergner@linux.ibm.com>
2492 * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow
2493 -mabi=ieeelongdouble without both -mpopcntd and -mvsx.
2495 2018-11-28 Aaron Sawdey <acsawdey@linux.ibm.com>
2497 Backport from mainline
2498 2018-10-25 Aaron Sawdey <acsawdey@linux.ibm.com>
2500 * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
2501 a shorter sequence with fewer branches.
2502 (emit_final_str_compare_gpr): Ditto.
2504 Backport from mainline to allow the above code to go in:
2505 2018-06-14 Aaron Sawdey <acsawdey@linux.ibm.com>
2507 * config/rs6000/rs6000-string.c (do_and3, do_and3_mask,
2508 do_cmpb3, do_rotl3): New functions.
2510 2018-11-28 Richard Biener <rguenther@suse.de>
2512 PR tree-optimization/79351
2513 * tree-ssa-sccvn.c (vn_reference_lookup_3): For assignments from
2514 empty CONSTRUCTORs ensure the store is at a constant position.
2516 2018-11-28 Richard Biener <rguenther@suse.de>
2518 PR tree-optimization/88229
2519 * tree-ssa.c (non_rewritable_mem_ref_base): Check DECL_SIZE_UNIT
2520 is an INTEGER_CST before accessing it so.
2522 2018-11-26 Andreas Krebbel <krebbel@linux.ibm.com>
2524 Backport from mainline
2525 2018-11-20 Andreas Krebbel <krebbel@linux.ibm.com>
2527 * config/s390/s390.md ("clztidi2"): Swap the RTX's written to the
2528 DImode parts of the target operand.
2530 2018-11-26 Andreas Krebbel <krebbel@linux.ibm.com>
2532 Backport from mainline
2533 2018-11-26 Andreas Krebbel <krebbel@linux.ibm.com>
2535 * doc/invoke.texi: Document z14/arch12 -march option.
2537 2018-11-23 Tamar Christina <tamar.christina@arm.com>
2539 Backported from mainline.
2540 2018-08-16 Tamar Christina <tamar.christina@arm.com>
2543 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
2544 (mov<mov>): ..this and enable unconditionally.
2546 2018-11-22 Uros Bizjak <ubizjak@gmail.com>
2548 Backport from mainline
2549 2018-11-16 Uros Bizjak <ubizjak@gmail.com>
2552 * config/i386/sse.md (UNSPEC_MOVDI_TO_SSE): New UNSPEC.
2553 (movdi_to_sse): Rewrite using UNSPEC_MOVDI_TO_SSE unspec.
2555 2018-11-20 Eric Botcazou <ebotcazou@adacore.com>
2557 PR rtl-optimization/85925
2558 * rtl.h (word_register_operation_p): New predicate.
2559 * combine.c (record_dead_and_set_regs_1): Only apply specific handling
2560 for WORD_REGISTER_OPERATIONS targets to word_register_operation_p RTX.
2561 * rtlanal.c (nonzero_bits1): Likewise. Adjust couple of comments.
2562 (num_sign_bit_copies1): Likewise.
2564 2018-11-19 Martin Liska <mliska@suse.cz>
2566 PR gcov-profile/88045
2567 * coverage.c (coverage_begin_function): One can't relly on the
2568 fact that function ends on the same line it starts or later.
2570 2018-11-18 Uros Bizjak <ubizjak@gmail.com>
2572 Backport from mainline
2573 2018-11-11 Uros Bizjak <ubizjak@gmail.com>
2576 * config/i386/i386.h (STACK_BOUNDARY): Use TARGET_64BIT_MS_ABI
2577 instead of (TARGET_64BIT && ix86_abi == MS_ABI).
2578 * config/i386/darwin.h (STACK_BOUNDARY): Ditto.
2579 * config/i386/cygming.h (STACK_BOUNDARY): Remove.
2581 2018-11-16 Eric Botcazou <ebotcazou@adacore.com>
2583 * tree-switch-conversion.c (emit_case_decision_tree): Pass location
2584 of switch statement to emit_case_nodes.
2585 (do_jump_if_equal): Add LOC parameter and set it on the newly built
2586 GIMPLE comparison statement.
2587 (emit_cmp_and_jump_insns): Likewise.
2588 (emit_case_nodes): Add LOC parameter and pass it to do_jump_if_equal
2589 as well as recursive calls.
2591 2018-11-15 Nathan Sidwell <nathan@acm.org>
2595 * dwarf2out.c (dwarf2out_finish): Apply resolve_addr to comdat
2598 2018-11-14 Paul A. Clarke <pc@us.ibm.com>
2600 Backport from mainline
2601 * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
2603 2018-11-14 Claudiu Zissulescu <claziss@synopsys.com>
2605 Backport from mainline
2607 * config/arc/arc.c (arc_return_address_register): Fix
2610 2018-11-14 Claudiu Zissulescu <claziss@synopsys.com>
2612 Backport from mainline
2613 * config/arc/arc.h (FUNCTION_PROFILER): Redefine to empty.
2614 * config/arc/elf.h (PROFILE_HOOK): Define.
2615 * config/arc/linux.h (PROFILE_HOOK): Likewise.
2617 2018-11-14 Claudiu Zissulescu <claziss@synopsys.com>
2619 Backport from mainline
2620 * config/arc/arc.c (hwloop_optimize): Bailout when detecting a
2621 jump table data in the text section.
2623 2018-11-14 Claudiu Zissulescu <claziss@synopsys.com>
2625 Backport from mainline
2626 * config/arc/arc.c (arc_eh_return_address_location): Repurpose it
2627 to fit the eh_return pattern.
2628 * config/arc/arc.md (eh_return): Define.
2629 (VUNSPEC_ARC_EH_RETURN): Likewise.
2630 * config/arc/arc-protos.h (arc_eh_return_address_location): Match
2632 * config/arc/arc.h (EH_RETURN_HANDLER_RTX): Remove it.
2634 2018-11-14 Claudiu Zissulescu <claziss@synopsys.com>
2636 Backport from mainline
2637 2018-09-17 Alexey Brodkin <abrodkin@synopsys.com>
2639 * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
2641 2018-11-14 Claudiu Zissulescu <claziss@synopsys.com>
2643 Backport from mainline
2644 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
2646 2018-11-11 Uros Bizjak <ubizjak@gmail.com>
2648 Backport from mainline
2649 2018-11-04 Uros Bizjak <ubizjak@gmail.com>
2652 * cfgexpand.c (pass_expand::execute): Move the call to
2653 finish_eh_generation in front of the call to expand_stack_alignment.
2655 2018-11-07 Max Filippov <jcmvbkbc@gmail.com>
2657 Backport from mainline
2658 2018-11-05 Max Filippov <jcmvbkbc@gmail.com>
2660 * config/xtensa/uclinux.h (XTENSA_ALWAYS_PIC): Change to 0.
2662 2018-11-05 Jakub Jelinek <jakub@redhat.com>
2664 PR tree-optimization/87859
2665 * gimple-ssa-store-merging.c (struct merged_store_group): Add
2666 first_nonmergeable_order member.
2667 (merged_store_group::merged_store_group): Initialize them.
2668 (imm_store_chain_info::coalesce_immediate_stores): Don't merge
2669 stores with order >= first_nonmergeable_order.
2670 Set merged_store->first_nonmergeable_order if we've skipped any
2671 stores. Attempt to merge overlapping INTEGER_CST stores that
2672 we would otherwise skip.
2675 * match.pd (X + Y < X): Don't optimize if TYPE_OVERFLOW_SANITIZED.
2677 Backported from mainline
2678 2018-10-20 Jakub Jelinek <jakub@redhat.com>
2681 * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
2683 2018-10-19 Jakub Jelinek <jakub@redhat.com>
2687 * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
2688 depend closely nested inside of loop with ordered clause with
2691 2018-10-29 Richard Biener <rguenther@suse.de>
2693 Backport from mainline
2694 2018-09-26 Richard Biener <rguenther@suse.de>
2698 * tree-inline.c (expand_call_inline): When the location
2699 of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
2700 or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
2701 the inserted BLOCK to make inlined_function_outer_scope_p
2703 * dwarf2out.c (add_call_src_coords_attributes): Do not add
2704 coords for reserved locations.
2706 2018-10-28 Iain Sandoe <iain@sandoe.co.uk>
2708 Backport from mainline
2709 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2714 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
2715 (dwarf2out_switch_text_section): Generate a local label for the second
2716 function sub-section and apply it as the second FDE start label.
2717 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
2718 second sub-section start.
2720 2018-10-28 Iain Sandoe <iain@sandoe.co.uk>
2722 Backport from mainline
2723 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
2726 (darwin_function_switched_text_sections): Delete.
2727 * gcc/config/darwin.h
2728 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
2730 2018-10-28 Iain Sandoe <iain@sandoe.co.uk>
2732 backport from mainline.
2733 2018-10-28 Iain Sandoe <iain@sandoe.co.uk>
2736 * config/rs6000/darwin.h (STACK_BOUNDARY): New.
2737 (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
2738 (STACK_DYNAMIC_OFFSET): Likewise.
2740 2018-10-26 Bill Schmidt <wschmidt@linux.ibm.com>
2742 Backport from mainline
2743 2018-10-19 Bill Schmidt <wschmidt@linux.ibm.com>
2745 PR tree-optimization/87473
2746 * gimple-ssa-strength-reduction.c (record_phi_increments_1): For
2747 phi arguments identical to the base expression of the phi
2748 candidate, record a phi-adjust increment of zero minus the index
2749 expression of the hidden basis.
2750 (phi_incr_cost_1): For phi arguments identical to the base
2751 expression of the phi candidate, the difference to compare against
2752 the increment is zero minus the index expression of the hidden
2753 basis, and there is no potential savings from replacing the (phi)
2755 (ncd_with_phi): For phi arguments identical to the base expression
2756 of the phi candidate, the difference to compare against the
2757 increment is zero minus the index expression of the hidden basis.
2758 (all_phi_incrs_profitable_1): For phi arguments identical to the
2759 base expression of the phi candidate, the increment to be checked
2760 for profitability is zero minus the index expression of the hidden
2763 2018-10-25 Richard Biener <rguenther@suse.de>
2765 PR tree-optimization/87665
2766 PR tree-optimization/87745
2767 * tree-vectorizer.h (get_earlier_stmt): Remove.
2768 (get_later_stmt): Pick up UID from the original non-pattern stmt.
2770 2018-10-24 Richard Biener <rguenther@suse.de>
2772 PR tree-optimization/87665
2773 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
2776 2018-10-23 Richard Biener <rguenther@suse.de>
2778 PR tree-optimization/87700
2779 * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
2781 2018-10-19 Richard Biener <rguenther@suse.de>
2784 Backport from mainline
2785 2018-07-12 Richard Biener <rguenther@suse.de>
2787 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
2788 (vn_lookup_simplify_result): Remove recursion limit applied
2790 (vn_nary_build_or_lookup_1): Adjust.
2791 (try_to_simplify): Likewise.
2792 * gimple-match-head.c (gimple_resimplify1): Instead apply one
2794 (gimple_resimplify2): Likewise.
2795 (gimple_resimplify3): Likewise.
2796 (gimple_resimplify4): Likewise.
2798 2018-10-19 Andreas Krebbel <krebbel@linux.ibm.com>
2800 Backport from mainline
2801 2018-10-15 Andreas Krebbel <krebbel@linux.ibm.com>
2803 * config/s390/s390.c (s390_expand_vec_init): Force vector element
2804 into reg if it isn't a general operand.
2806 2018-10-18 Richard Biener <rguenther@suse.de>
2810 2018-02-07 Richard Biener <rguenther@suse.de>
2812 PR tree-optimization/84204
2813 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2816 2018-10-17 Eric Botcazou <ebotcazou@adacore.com>
2819 * fold-const.c (fold_truth_andor_1): If the right side is not constant,
2820 bail out if both sides do not have the same storage order.
2822 2018-10-17 Richard Biener <rguenther@suse.de>
2824 Backport from mainline
2825 2018-10-08 Richard Sandiford <richard.sandiford@arm.com>
2828 * gimple-ssa-backprop.c (backprop::intersect_uses): Use
2829 FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
2831 2018-10-16 Richard Biener <rguenther@suse.de>
2833 Backport from mainline
2834 2018-10-08 Richard Biener <rguenther@suse.de>
2836 PR tree-optimization/63155
2837 * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
2838 (ssa_propagation_engine::ssa_propagate): Remove redundant
2839 bitmap bit clearing.
2841 2018-10-05 Richard Biener <rguenther@suse.de>
2843 PR tree-optimization/63155
2844 * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
2845 vertical space in dumpfiles.
2846 * tree-ssa-propagate.h
2847 (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
2848 * tree-ssa-propagate.c (cfg_blocks_back): New global.
2849 (ssa_edge_worklist_back): Likewise.
2850 (curr_order): Likewise.
2851 (cfg_blocks_get): Remove abstraction.
2852 (cfg_blocks_add): Likewise.
2853 (cfg_blocks_empty_p): Likewise.
2854 (add_ssa_edge): Add to current or next worklist based on
2856 (add_control_edge): Likewise.
2857 (ssa_propagation_engine::process_ssa_edge_worklist): Fold
2859 (ssa_propagation_engine::ssa_propagate): ... here. Unify
2860 iteration from CFG and SSA edge worklist so we process
2861 everything in RPO order, prioritizing forward progress
2863 (ssa_prop_init): Allocate new worklists, do not dump
2865 (ssa_prop_fini): Free new worklists.
2867 2018-09-24 Richard Biener <rguenther@suse.de>
2869 PR tree-optimization/63155
2870 * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
2871 the worklist when the edge of the respective argument isn't
2874 2018-10-16 Wilco Dijkstra <wdijkstr@arm.com>
2876 Backported from mainline
2878 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
2879 Use HOST_WIDE_INT_1U for shift.
2881 2018-10-16 Richard Biener <rguenther@suse.de>
2883 Backport from mainline
2884 2018-09-18 Richard Biener <rguenther@suse.de>
2887 * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
2888 (compute_samebase_partition_bases): Likewise.
2889 (coalesce_ssa_name): Always use compute_optimized_partition_bases.
2890 (gimple_can_coalesce_p): Simplify.
2892 2018-10-16 Richard Biener <rguenther@suse.de>
2894 Backport from mainline
2895 2018-10-15 Richard Biener <rguenther@suse.de>
2898 * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
2899 (visit_loadstore): When a used restrict tag escaped verify that
2900 the points-to solution of "other" pointers do not include
2902 (compute_dependence_clique): If a used restrict tag escaped
2903 communicated that down to visit_loadstore.
2905 2018-10-01 Richard Biener <rguenther@suse.de>
2907 PR tree-optimization/87465
2908 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
2909 causing branch miscounts.
2911 2018-10-11 Jakub Jelinek <jakub@redhat.com>
2913 Backported from mainline
2914 2018-10-10 Jakub Jelinek <jakub@redhat.com>
2917 * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
2918 to special_args set.
2920 2018-09-26 Jakub Jelinek <jakub@redhat.com>
2923 * config/i386/i386.c: Include debug.h and dwarf2out.h.
2924 (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
2927 2018-09-12 Jakub Jelinek <jakub@redhat.com>
2928 Andreas Krebbel <krebbel@linux.ibm.com>
2930 PR tree-optimization/86844
2931 * gimple-ssa-store-merging.c
2932 (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
2933 there are any overlapping stores in between them, make sure they are
2934 also coalesced or we give up completely.
2936 2018-10-09 Will Schmidt <will_schmidt@vnet.ibm.com>
2938 Backport from trunk.
2939 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
2942 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
2943 around folding of vec_sl to handle out of range shift values.
2945 2018-10-09 H.J. Lu <hongjiu.lu@intel.com>
2947 Backport from mainline
2948 2018-09-29 H.J. Lu <hongjiu.lu@intel.com>
2951 * config/i386/i386.c (construct_container): Use TImode for
2952 BLKmode values in 2 integer registers.
2954 2018-10-08 H.J. Lu <hongjiu.lu@intel.com>
2956 Backport from mainline
2957 2018-10-08 H.J. Lu <hongjiu.lu@intel.com>
2960 * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
2961 Defined with __builtin_ia32_vfmaddsubpd512_mask.
2963 2018-10-05 H.J. Lu <hongjiu.lu@intel.com>
2965 Backport from mainline
2966 2018-10-05 H.J. Lu <hongjiu.lu@intel.com>
2969 * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
2970 assembler for -mavx.
2971 * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
2973 2018-10-03 Uros Bizjak <ubizjak@gmail.com>
2975 Backport from mainline
2976 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
2978 * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
2979 (GET_SSE_REGNO): Rename from SSE_REGNO. Update all uses for rename.
2981 2018-10-03 Jonathan Wakely <jwakely@redhat.com>
2984 * doc/invoke.texi (Link Options): Fix formatting and grammar.
2986 2018-10-03 Martin Liska <mliska@suse.cz>
2988 Backport from mainline
2989 2018-10-03 Martin Liska <mliska@suse.cz>
2991 PR gcov-profile/86109
2992 * coverage.c (coverage_begin_function): Do not
2993 mark lambdas as artificial.
2994 * tree-core.h (struct GTY): Remove tm_clone_flag
2995 and introduce new lambda_function.
2996 * tree.h (DECL_LAMBDA_FUNCTION): New macro.
2998 2018-10-02 H.J. Lu <hongjiu.lu@intel.com>
3000 Backport from mainline
3001 2018-09-24 H.J. Lu <hongjiu.lu@intel.com>
3004 * config/i386/i386.c (rest_of_insert_endbranch): Set
3005 endbr_queued_at_entrance to true and don't insert ENDBR if
3006 x86_function_profiler will be called.
3007 (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
3009 * config/i386/i386.h (machine_function): Add
3010 endbr_queued_at_entrance.
3012 2018-10-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3014 Backport from mainline
3015 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3017 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
3018 if starting source register is not even.
3020 2018-09-29 Jakub Jelinek <jakub@redhat.com>
3023 * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
3024 __m512d type for __A argument rather than __m512.
3026 2018-09-27 Michael Meissner <meissner@linux.ibm.com>
3028 Backport from mainline
3029 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
3032 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
3033 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
3036 2018-09-25 Martin Liska <mliska@suse.cz>
3038 Backport from mainline
3039 2018-09-24 Martin Liska <mliska@suse.cz>
3042 * asan.c: Make asan_handled_variables extern.
3044 * cfgexpand.c (expand_stack_vars): Make sure
3045 a representative is unpoison if another
3046 variable in the partition is handled by
3047 use-after-scope sanitization.
3049 2018-09-24 Andrew Pinski <apinski@marvell.com>
3051 * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
3052 access prev before checking it for NULLness in the
3053 AARCH64_FUSE_CMP_BRANCH case.
3055 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
3057 * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
3058 if the call takes a static chain.
3060 2018-09-19 John David Anglin <danglin@gcc.gnu.org>
3062 * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
3063 (atomic_storehi): Likewise.
3064 (atomic_storesi): Likewise.
3065 (atomic_loaddi): Restore compare and swap exchange loop code.
3067 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
3070 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
3072 PR rtl-optimization/86882
3073 * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
3075 2018-09-18 Martin Liska <mliska@suse.cz>
3077 Backport from mainline
3078 2018-09-17 Martin Liska <mliska@suse.cz>
3080 PR gcov-profile/85871
3081 * gcov.c (output_intermediate_file): Fix out of bounds
3084 2018-09-16 Hans-Peter Nilsson <hp@bitrange.com>
3087 * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
3088 non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
3089 (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
3090 leaf_function_p, instead use has_hard_reg_initial_val.
3092 2018-09-14 John David Anglin <danglin@gcc.gnu.org>
3095 * dojump.c (do_compare_and_jump): Canonicalize function pointers
3096 when one operand is a function pointer. Use POINTER_TYPE_P and
3097 FUNC_OR_METHOD_TYPE_P.
3098 * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
3099 * fold-const.c (build_range_check): Likewise.
3100 * match.pd (simple_comparison): Likewise.
3102 2018-09-14 Segher Boessenkool <segher@kernel.crashing.org>
3105 2018-09-14 Segher Boessenkool <segher@kernel.crashing.org>
3108 * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
3111 2018-09-14 Carl Love <cel@us.ibm.com>
3114 2018-09-14 Carl Love <cel@us.ibm.com>
3116 * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
3117 * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
3119 2018-09-12 Segher Boessenkool <segher@kernel.crashing.org>
3122 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
3125 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
3128 2018-09-12 Segher Boessenkool <segher@kernel.crashing.org>
3131 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
3133 PR rtl-optimization/86771
3134 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
3135 of two SETs into those two SETs, one to be placed at i2, if that SETs
3136 destination is modified between i2 and i3.
3138 2018-09-12 Andreas Krebbel <krebbel@linux.ibm.com>
3140 Backport from mainline
3141 2018-09-12 Andreas Krebbel <krebbel@linux.ibm.com>
3143 * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
3145 ("trunc<BFP:mode><DFP_ALL:mode>2")
3146 ("trunc<DFP_ALL:mode><BFP:mode>2")
3147 ("extend<BFP:mode><DFP_ALL:mode>2")
3148 ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
3149 according to the target operand type.
3151 2018-09-12 Jakub Jelinek <jakub@redhat.com>
3154 * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
3155 BIT_AND_EXPR's second operand is a power of two. Formatting fix.
3157 2018-09-05 Nathan Sidwell <nathan@acm.org>
3160 * stor-layout.c (place_field): Scan forwards to check last
3161 bitfield when ms_bitfield_placement is in effect.
3163 2018-09-05 Richard Biener <rguenther@suse.de>
3166 * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
3169 2018-09-05 Jakub Jelinek <jakub@redhat.com>
3171 Backported from mainline
3172 2018-09-04 Jakub Jelinek <jakub@redhat.com>
3175 * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
3176 OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
3177 OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
3178 (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
3179 and OPTION_MASK_ISA_XSAVEC_UNSET.
3181 2018-08-31 Jakub Jelinek <jakub@redhat.com>
3184 * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
3185 gen_int_mode. Formatting fixes.
3187 2018-08-11 Jakub Jelinek <jakub@redhat.com>
3189 PR tree-optimization/86835
3190 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
3191 new_stmt after def_gsi, make sure to insert new_square_stmt after
3192 that stmt, not 2 stmts before it.
3194 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
3196 Backport from mainline
3197 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
3199 * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
3200 XOR operations in NAND case.
3202 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
3204 * doc/invoke.texi (Option Summary): Add -Waligned-new.
3206 2018-09-03 Tom de Vries <tdevries@suse.de>
3208 backport from trunk:
3209 2018-06-21 Tom de Vries <tdevries@suse.de>
3211 PR tree-optimization/85859
3212 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
3213 test with comment from bb_no_side_effects_p.
3215 2018-09-01 Michael Matz <matz@suse.de>
3217 Backport from mainline
3218 PR tree-optimization/87074
3219 * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
3220 PHIs for outer-loop uses.
3222 2018-08-31 Richard Biener <rguenther@suse.de>
3224 Backport from mainline
3225 2018-08-27 Richard Biener <rguenther@suse.de>
3227 PR tree-optimization/86927
3228 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
3229 use const cond reduction code.
3231 2018-08-23 Richard Biener <rguenther@suse.de>
3234 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
3237 2018-08-22 Richard Biener <rguenther@suse.de>
3239 PR tree-optimization/86945
3240 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
3242 2018-08-17 Richard Biener <rguenther@suse.de>
3245 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
3246 across a va-arg-pack using call adjust its return value accordingly.
3248 2018-08-02 Richard Biener <rguenther@suse.de>
3250 PR tree-optimization/86816
3251 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
3252 which checks for value availability before querying it.
3253 (gvn_uses_equal): Use it.
3254 (same_succ_hash): Likewise.
3255 (gimple_equal_p): Likewise.
3257 2018-07-17 Richard Biener <rguenther@suse.de>
3260 * dwarf2out.c (init_sections_and_labels): Always generate
3261 a debug_line_str_section for early LTO debug.
3262 (dwarf2out_finish): Reset debug_line_str_hash output early.
3263 Bump counter for extra dwarf5 .debug_loc labels to not conflict
3265 (dwarf2out_early_finish): Output debug_line_str.
3267 2018-08-29 Luis Machado <luis.machado@linaro.org>
3269 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static global.
3270 (qdf24xx_tunings): Set vector cost structure to qdf24xx_vector_cost.
3272 2018-08-29 Luis Machado <luis.machado@linaro.org>
3274 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
3275 <register_sextend>: Set to 3.
3277 2018-08-27 Martin Sebor <msebor@redhat.com>
3279 PR tree-optimization/86914
3280 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
3282 2018-08-28 Jakub Jelinek <jakub@redhat.com>
3285 * calls.c (maybe_warn_nonstring_arg): Punt early if
3286 warn_stringop_overflow is zero. Don't call get_range_strlen
3287 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
3288 Only use lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly
3289 increment lenrng[0].
3291 2018-08-27 Jakub Jelinek <jakub@redhat.com>
3293 PR rtl-optimization/87065
3294 * combine.c (simplify_if_then_else): Formatting fix.
3295 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
3297 (known_cond): Don't return const_true_rtx for vector modes. Use
3298 CONST0_RTX instead of const0_rtx. Formatting fixes.
3300 2018-08-25 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3302 Backport from mainline
3304 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
3305 with all enabled __intN types.
3307 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
3309 2018-08-21 H.J. Lu <hongjiu.lu@intel.com>
3311 Backport from mainline
3312 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
3315 * config/i386/i386.md (eh_return): Always update EH return
3316 address in word_mode.
3318 2018-08-17 John David Anglin <danglin@gcc.gnu.org>
3320 Backport from mainline
3321 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
3323 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
3324 Update comment for atomic instructions.
3325 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
3326 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
3328 (atomic_loaddi): Revise fence expansion to only emit fence prior to
3329 load for __ATOMIC_SEQ_CST model.
3330 (atomic_loaddi_1): Remove float register target.
3331 (atomic_storedi): Handle CONST_INT values.
3332 (atomic_storedi_1): Remove float register source. Add special case
3334 (memory_barrier): New expander and insn.
3336 2018-08-16 Tamar Christina <tamar.christina@arm.com>
3338 Backport from mainline
3339 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
3342 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
3345 2018-08-16 Martin Liska <mliska@suse.cz>
3347 Backport from mainline
3348 2018-08-02 Martin Liska <mliska@suse.cz>
3350 PR gcov-profile/86817
3351 * gcov.c (process_all_functions): New function.
3353 (process_file): Move functions processing to
3354 process_all_functions.
3356 2018-08-14 Siddhesh Poyarekar <siddhesh@sourceware.org>
3358 Backport from mainline
3359 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
3361 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
3363 (falkor_am_1_gtov_gtov): ... a new insn reservation.
3365 2018-08-13 Liu Hao <lh_mouse@126.com>
3367 Backport from mainline
3368 2018-08-13 Liu Hao <lh_mouse@126.com>
3370 * pretty-print.c (eat_esc_sequence): Swap the foreground and
3371 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
3372 and clear it thereafter, as it only works for DBCS.
3374 2018-08-13 Liu Hao <lh_mouse@126.com>
3376 Backport from mainline
3377 2018-08-13 Liu Hao <lh_mouse@126.com>
3379 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
3380 handle returned by _get_osf_handle().
3382 2018-08-14 Richard Sandiford <richard.sandiford@arm.com>
3384 Backport from mainline
3385 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
3387 PR tree-optimization/86871
3388 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
3389 instead of gimple_assign_lhs.
3391 2018-08-07 H.J. Lu <hongjiu.lu@intel.com>
3393 Backport from mainline
3394 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
3397 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
3398 cfun->machine->max_used_stack_alignment if needed.
3400 2018-08-02 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3402 Backport from mainline
3403 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3406 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
3407 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
3408 requested variable alignment.
3409 (expand_one_ssa_partition): Likewise.
3410 (expand_one_var): Likewise.
3412 2018-08-02 Andreas Schwab <schwab@linux-m68k.org>
3414 Backport from mainline
3415 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
3418 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
3419 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
3421 2018-08-01 Richard Biener <rguenther@suse.de>
3424 * graphite.h: Include isl/id.h and isl/space.h to allow build
3427 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
3430 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3431 mode for COFF targets.
3432 * defaults.h (TARGET_COFF): Define.
3433 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
3434 TARGET_COFF): Define.
3435 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
3436 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
3437 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
3439 2018-07-29 Bernd Edlinger <bernd.edlinger@hotmail.de>
3441 Backport from mainline
3442 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
3445 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
3447 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
3449 * config/pa/pa.c (pa_output_addr_vec): Align address table.
3450 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
3451 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
3453 2018-07-26 Uros Bizjak <ubizjak@gmail.com>
3455 Backport from mainline
3456 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
3459 * expmed.c (emit_store_flag): Do not emit setcc followed by a
3460 conditional move when trapping comparison was split to a
3461 non-trapping one (and vice versa).
3463 2018-07-26 Jakub Jelinek <jakub@redhat.com>
3465 Backported from mainline
3466 2018-07-24 Jakub Jelinek <jakub@redhat.com>
3469 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
3470 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
3471 and abs_d == d, do the power of two handling if profitable.
3473 2018-07-17 Jakub Jelinek <jakub@redhat.com>
3476 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
3477 to _looptemp_ clauses, other than the first two.
3480 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
3481 and cond temporaries don't have reference type if iterator has
3482 pointer type. For init use &for_pre_body instead of pre_p if
3483 for_pre_body is non-empty.
3485 2018-07-26 Jakub Jelinek <jakub@redhat.com>
3488 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
3489 declare target to variables if they have always,{to,from,tofrom} map
3492 * BASE-VER: Set to 8.2.1.
3494 2018-07-26 Release Manager
3496 * GCC 8.2.0 released.
3498 2018-07-18 Martin Sebor <msebor@redhat.com>
3501 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
3502 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
3503 Handle integer subtraction.
3504 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
3505 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
3506 * doc/invoke.texi (-Wstringop-truncation): Update.
3508 gcc/testsuite/ChangeLog:
3510 2018-07-16 Claudiu Zissulescu <claziss@synopsys.com>
3512 Backport from mainline
3513 2017-03-24 Claudiu Zissulescu <claziss@synopsys.com>
3515 * config/arc/arc-protos.h (arc_pad_return): Remove.
3516 * config/arc/arc.c (machine_function): Remove force_short_suffix
3518 (arc_print_operand): Adjust printing of '&'.
3519 (arc_verify_short): Remove conditional printing of short suffix.
3520 (arc_final_prescan_insn): Remove reference to size_reason.
3521 (pad_return): New function.
3522 (arc_reorg): Call pad_return.
3523 (arc_pad_return): Remove.
3524 (arc_init_machine_status): Remove reference to force_short_suffix.
3525 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
3526 (attr length): When attribute iscompact is true force to 2
3527 regardless; in the case of maybe check if we want to force the
3528 instruction to have 4 bytes length.
3529 (nopv): Change it to generate 4 byte long nop as well.
3530 (blockage): New pattern.
3531 (simple_return): Remove call to arc_pad_return.
3532 (p_return_i): Likewise.
3534 2018-07-19 Richard Biener <rguenther@suse.de>
3536 Backport from mainline
3537 2018-07-13 Richard Biener <rguenther@suse.de>
3540 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
3541 instead of get_context_die.
3543 2018-07-11 Richard Biener <rguenther@suse.de>
3546 * dwarf2out.c (init_sections_and_labels): Use
3547 output_asm_line_debug_info consistently.
3548 (dwarf2out_early_finish): Likewise.
3549 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
3552 2018-07-16 Richard Biener <rguenther@suse.de>
3554 Backport from mainline
3555 2018-07-13 Richard Biener <rguenther@suse.de>
3558 * match.pd (addr1 - addr2): Allow either of the operand to
3561 2018-06-15 Richard Biener <rguenther@suse.de>
3564 * tree-cfg.c (move_stmt_op): unshare invariant addresses
3565 before adjusting their block.
3567 2018-06-06 Richard Biener <rguenther@suse.de>
3569 PR tree-optimization/85935
3570 * graphite-scop-detection.c (find_params_in_bb): Analyze
3571 condition operands with respect to the correct loop. Assert
3572 the analysis doesn't fail.
3574 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
3576 PR tree-optimization/86514
3577 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
3578 conversion to a boolean type from a type with greater precision.
3580 2018-07-15 Bill Schmidt <wschmidt@linux.ibm.com>
3582 Backport from mainline
3583 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
3584 Steve Munroe <munroesj52@gmail.com>
3586 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
3587 (_mm_andnot_si128): Likewise.
3588 (_mm_or_si128): Likewise.
3589 (_mm_xor_si128): Likewise.
3591 2018-07-14 Martin Sebor <msebor@redhat.com>
3593 PR tree-optimization/86274
3594 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
3596 (format_floating): Correct handling of infinities and NaNs.
3598 2018-07-13 H.J. Lu <hongjiu.lu@intel.com>
3600 Backport from mainline
3601 2018-07-13 H.J. Lu <hongjiu.lu@intel.com>
3602 Sunil K Pandey <sunil.k.pandey@intel.com>
3605 * config/i386/i386.c (m_CORE_AVX512): New.
3606 (m_CORE_AVX2): Likewise.
3607 (m_CORE_ALL): Add m_CORE_AVX2.
3608 * config/i386/x86-tune.def: Replace m_HASWELL with m_CORE_AVX2.
3609 Replace m_SKYLAKE_AVX512 with m_CORE_AVX512 on avx256_optimal
3610 and remove the rest of m_SKYLAKE_AVX512.
3612 2018-07-13 Marek Polacek <polacek@redhat.com>
3615 2018-07-03 Marek Polacek <polacek@redhat.com>
3618 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
3620 2018-07-12 Richard Biener <rguenther@suse.de>
3623 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
3624 Remove -mieee-fp handling.
3626 2018-07-12 Jakub Jelinek <jakub@redhat.com>
3628 PR tree-optimization/86492
3629 * gimple-ssa-store-merging.c
3630 (imm_store_chain_info::coalesce_immediate_stores): Call
3631 check_no_overlap even for the merge_overlapping case.
3633 2018-07-11 Jakub Jelinek <jakub@redhat.com>
3635 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
3636 Use __mmask64 type instead of __mmask8 for __M argument.
3637 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
3638 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
3640 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
3641 __mmask16 for __M argument.
3642 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
3643 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
3644 to __mmask16 instead of __mmask8.
3645 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
3646 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
3647 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
3648 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
3649 instead of __mmask16 for __U argument.
3650 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
3651 __mmask16 instead of __mmask8 for __U argument.
3652 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
3654 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
3656 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
3658 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
3660 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
3662 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
3664 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
3666 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
3668 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
3670 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
3672 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
3673 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
3674 return type as well as __M argument type and all casts from __mmask8
3676 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
3677 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
3678 return type as well as __M argument type and all casts from __mmask8
3680 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
3681 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
3682 return type as well as __M argument type and all casts from __mmask8
3684 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
3685 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
3686 return type as well as __M argument type and all casts from __mmask8
3688 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
3689 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
3692 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
3694 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
3695 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
3698 2018-07-05 Martin Sebor <msebor@redhat.com>
3701 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
3702 than its domain to compute its the upper bound of a char array.
3704 2018-07-03 Carl Love <cel@us.ibm.com>
3707 2018-07-03 Carl Love <cel@us.ibm.com>
3709 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
3710 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
3711 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
3712 VSX_BUILTIN_DOUBLEL_V4SF.
3714 2018-07-04 Luis Machado <luis.machado@linaro.org>
3716 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
3718 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
3720 2018-07-02 Segher Boessenkool <segher@kernel.crashing.org>
3723 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
3726 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
3727 ieee128_float_type_node to long_double_type_node unless
3728 TARGET_LONG_DOUBLE_128 is set.
3730 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
3732 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
3733 -fno-omit-frame-pointer when not optimizing.
3735 2018-06-30 Alexandre Oliva <aoliva@redhat.com>
3738 * dwarf2out.c (loc_list_has_views): Adjust comments.
3739 (dw_loc_list): Split single cross-partition range with
3742 2018-06-29 Kelvin Nilsen <kelvin@gcc.gnu.org>
3744 Backport from mainline
3745 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3747 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
3748 behavior of vec_pack (vector double, vector double) to match
3749 behavior of vec_float2 (vector double, vector double).
3751 2018-06-29 Jakub Jelinek <jakub@redhat.com>
3753 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
3755 2018-06-28 Kelvin Nilsen <kelvin@gcc.gnu.org>
3757 Backport from mainline
3758 2017-05-23 Segher Boessenkool <segher@kernel.crashing.org>
3760 * doc/sourcebuild.texi (Endianness): New subsubsection.
3762 2018-06-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
3764 Backport from mainline
3765 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
3767 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
3768 behavior of vec_packsu (vector unsigned long long, vector unsigned
3769 long long) to match behavior of vec_packs with same signature.
3771 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
3774 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
3776 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
3777 unaligned vsx for 16B memset.
3779 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
3782 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
3785 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
3788 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
3791 2018-05-08 Segher Boessenkool <segher@kernel.crashing.org>
3793 PR rtl-optimization/85645
3794 * regrename.c (build_def_use): Also kill the chains that include the
3795 destination of a REG_CFA_REGISTER note.
3797 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
3800 2018-05-08 Segher Boessenkool <segher@kernel.crashing.org>
3802 PR rtl-optimization/85645
3803 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
3804 insn that has a REG_CFA_REGISTER note.
3806 2018-06-26 Robin Dapp <rdapp@linux.ibm.com>
3808 * config/s390/s390.h (enum processor_flags): Do not use
3809 default tune parameter when -march was specified.
3811 2018-06-26 Jakub Jelinek <jakub@redhat.com>
3814 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
3815 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
3818 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
3820 Back port from trunk
3821 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
3823 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
3824 added IEEE/IBM long double multilib support on PowerPC little
3825 endian Linux systems.
3826 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
3827 (MULTILIB_DEFAULTS): Likewise.
3828 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3830 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
3831 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
3832 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
3834 2018-06-25 Segher Boessenkool <segher@kernel.crashing.org>
3837 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
3839 * config/rs6000/rs6000.md (abs<mode>2): Handle IFmode.
3841 2018-05-02 Jan Hubicka <jh@suse.cz>
3843 Backport from mainline
3844 2018-04-30 Jan Hubicka <jh@suse.cz>
3846 * lto-wrapper.c (ltrans_priorities): New static var.
3847 (cmp_priority): New.
3848 (run_gcc): Read priorities and if doing parallel build order
3849 the Makefile by them.
3851 2018-06-25 Segher Boessenkool <segher@kernel.crashing.org>
3854 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
3857 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
3858 ieee128 argument takes up only one (vector) register, not two (floating
3861 2018-06-23 Richard Sandiford <richard.sandiford@linaro.org>
3863 PR tree-optimization/85989
3864 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
3866 (backprop::intersect_uses): Check it when deciding whether this
3867 is a backedge reference.
3868 (backprop::process_block): Add each phi to m_visited_phis
3869 after visiting it, then clear it at the end.
3871 2018-06-22 David Edelsohn <dje.gcc@gmail.com>
3873 Backport from mainline
3875 2018-06-19 Tony Reix <tony.reix@atos.com>
3876 Damien Bergamini <damien.bergamini@atos.com>
3877 David Edelsohn <dje.gcc@gmail.com>
3879 * collect2.c (static_obj): New variable.
3880 (static_libs): New variable.
3881 (is_in_list): Uncomment declaration.
3882 (main): Track AIX libraries linked statically.
3883 (is_in_list): Uncomment definition.
3884 (scan_prog_file): Don't add AIX shared libraries initializer
3885 to constructor list if linking statically.
3887 2018-06-22 Michael Meissner <meissner@linux.ibm.com>
3889 Back port from trunk
3890 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
3892 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
3893 conversion insn that shows up when pr85657-3.c is compiled using
3894 IEEE 128-bit long double.
3896 Back port from trunk
3897 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
3899 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
3900 multiply and divide external functions from being created more
3903 Back port from trunk
3904 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
3906 * config/rs6000/rs6000.md (neg<mode>2_internal): Use the correct
3907 mode to check whether the mode is IBM extended.
3909 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
3912 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
3913 floating point modes, so that IFmode is numerically greater than
3914 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
3915 to declare the ordering. This prevents IFmode from being
3916 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
3917 machine. Include rs6000-modes.h to share the fractional values
3918 between genmodes* and the rest of the compiler.
3922 * config/rs6000/rs6000-modes.h: New file.
3923 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
3924 meaning of rs6000_long_double_size so that 126..128 selects an
3925 appropriate 128-bit floating point type.
3926 (rs6000_option_override_internal): Likewise.
3927 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
3928 (TARGET_LONG_DOUBLE_128): Change the meaning of
3929 rs6000_long_double_size so that 126..128 selects an appropriate
3930 128-bit floating point type.
3931 (LONG_DOUBLE_TYPE_SIZE): Update comment.
3932 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
3933 source and destination to match the standard usage.
3934 (truncifkf2): Likewise.
3935 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
3936 ISA 2.07 to use an explicit clobber, instead of passing in a
3938 (copysign<mode>3_soft): Likewise.
3940 Back port from trunk
3941 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
3943 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
3946 Back port from trunk
3947 2018-06-04 Michael Meissner <meissner@linux.ibm.com>
3949 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
3950 track if we pass or return IEEE 128-bit floating point.
3951 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
3952 C++ mangling that is compatible with GCC 8.1.
3953 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
3954 (init_cumulative_args): Note if we pass or return IEEE 128-bit
3955 floating point types.
3956 (rs6000_function_arg_advance_1): Likewise.
3957 (rs6000_mangle_type): Optionally generate mangled names that match
3958 what GCC 8.1 generated for IEEE 128-bit floating point types.
3959 (rs6000_globalize_decl_name): If we have an external function that
3960 passes or returns IEEE 128-bit types, generate a weak reference
3961 from the mangled name used in GCC 8.1 to the current mangled
3963 (rs6000_init_builtins): Make __ibm128 use the long double type if
3964 long double is IBM extended double. Make __float128 use the long
3965 double type if long double is IEEE 128-bit.
3968 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
3969 macro for __ibm128 built-in functions.
3970 (PACK_IF): Add __ibm128 pack/unpack functions.
3971 (UNPACK_IF): Likewise.
3972 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
3973 enable long double built-in functions if long double is IEEE
3974 128-bit floating point.
3975 (rs6000_invalid_builtin): Update long double built-in function
3977 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
3978 functions, adjust the built-in function to use the long double
3979 built-in function if __ibm128 and long double are the same type.
3980 * doc/extend.texi (PowerPC builtins): Update documention for
3981 __builtin_{,un}pack_longdouble. Add documentation for
3982 __builtin_{,un}pack_ibm128.
3984 Back port from trunk
3985 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
3987 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
3988 the 128-bit floating point types. Fix function comment.
3990 2018-06-22 Michael Meissner <meissner@linux.ibm.com>
3992 Back port from trunk
3993 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
3996 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
3997 define __ibm128 as long double.
3998 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
3999 as a distinct type with IEEE 128-bit floating point is supported.
4000 (init_float128_ieee): Fix up conversions between IFmode and IEEE
4001 128-bit types to use the correct functions.
4002 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
4003 convert between 128-bit floating point types that have different
4004 modes but the same representation, instead of using gen_lowpart to
4006 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
4008 (IFKF_reg): New attributes to give the register constraints for
4010 (extend<mode>tf2_internal): New insns to mark an explicit
4011 conversion between 128-bit floating point types that have a
4012 different mode but share the same representation.
4014 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4017 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
4018 -x assembler-with-cpp.
4020 2018-06-21 Sebastian Huber <sebastian.huber@embedded-brains.de>
4022 Backported from mainline
4023 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
4025 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
4026 multilibs for *-*-rtems*.
4027 * config/riscv/t-rtems: New file.
4029 2018-06-20 Jakub Jelinek <jakub@redhat.com>
4031 Backported from mainline
4032 2018-06-16 Jakub Jelinek <jakub@redhat.com>
4034 PR rtl-optimization/86108
4035 * bb-reorder.c (create_forwarder_block): Renamed to ...
4036 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
4037 jump from new landing pad to the second part.
4038 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
4041 2018-06-20 Jakub Jelinek <jakub@redhat.com>
4044 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
4047 PR tree-optimization/86231
4048 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
4049 anti-range don't overwrite *vr0min before using it to compute *vr0max.
4051 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
4053 Backport from mainline
4054 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
4056 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
4058 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
4060 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
4062 * gimplify.c (gimplify_init_constructor): Really never clear for an
4063 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
4065 2018-06-18 Martin Sebor <msebor@redhat.com>
4068 * calls.c (alloc_max_size): Correct a logic error/typo.
4069 Treat excessive arguments as infinite. Warn for invalid arguments.
4070 * doc/invoke.texi (-Walloc-size-larger-than): Update.
4072 2018-06-16 Jakub Jelinek <jakub@redhat.com>
4075 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
4076 documented as preserved for backward compatibility only.
4077 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
4079 2018-06-15 Jakub Jelinek <jakub@redhat.com>
4082 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
4083 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
4084 Only call store_expr for halves if the mode is the same.
4087 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
4088 Fix up comment formatting.
4090 2018-06-14 Jakub Jelinek <jakub@redhat.com>
4093 * match.pd ((A +- CST1) +- CST2): Punt if last resort
4094 unsigned_type_for returns NULL.
4097 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
4098 subregs of multi-word pseudos unless the float mode has word size.
4100 2018-06-14 Richard Biener <rguenther@suse.de>
4103 * tree-vect-generic.c (build_word_mode_vector_type): Remove
4104 duplicate and harmful type_hash_canon.
4105 * tree.c (type_hash_canon): Assert we didn't find ourselves.
4107 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
4109 Backport from mainline
4110 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
4112 * config/rtems.h (STDINT_LONG32): Define.
4114 2018-05-05 Roland McGrath <mcgrathr@google.com>
4117 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
4118 any section for which we don't know a specific type it should have,
4119 regardless of name. Previously this was done only for the exact
4120 names ".init_array", ".fini_array", and ".preinit_array".
4121 (default_elf_asm_named_section): Add comment about
4122 relationship with default_section_type_flags and SECTION_NOTYPE.
4123 (get_section): Don't consider it a type conflict if one side has
4124 SECTION_NOTYPE and the other doesn't, as long as neither has the
4125 SECTION_BSS et al used in the default_section_type_flags logic.
4127 2018-06-13 Martin Sebor <msebor@redhat.com>
4129 PR tree-optimization/86114
4130 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
4132 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
4134 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
4137 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
4138 offsets for register save directives. Emit a second batch of save
4139 directives, if need be, when the function accesses prior frames.
4141 2018-06-11 Jason Merrill <jason@redhat.com>
4143 PR c++/80485 - inline function non-zero address.
4144 * symtab.c (nonzero_address): Check DECL_COMDAT.
4146 2018-06-11 Peter Bergner <bergner@vnet.ibm.com>
4148 Backport from mainline
4149 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
4152 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
4155 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
4158 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
4161 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
4162 on the correct operand.
4163 (*movdi_internal64): Ditto.
4165 2018-06-08 Carl Love <cel@us.ibm.com>
4167 Backport from mainline
4168 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
4169 vextract_fp_from_shortl): Add BE support.
4171 2018-06-08 Carl Love <cel@us.ibm.com>
4173 Backport from mainline
4174 * gcc/config/rs6000/vsx.md (first_match_index_<mode>):
4175 Calculate index using natureal element order.
4176 (first_match_or_eos_index_<mode>):
4177 Calculate index using natural element order.
4178 (first_match_index_<mode>):
4179 Calculate index using natural element order.
4180 (first_match_or_eos_index_<mode>):
4181 Calculate index using natural order.
4182 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
4183 for BE and LE modes.
4184 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
4185 P9V_BUILTIN_VCLZLSBB_V16QI.
4186 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
4189 2018-06-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4191 Backport from mainline
4192 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4195 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
4196 qualifier_void_pointer and qualifier_const_void_pointer.
4197 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
4198 (arm_init_builtins): Handle the above.
4199 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
4200 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
4201 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
4204 2018-06-07 Uros Bizjak <ubizjak@gmail.com>
4207 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
4208 Generate SImode target register for null target.
4209 <case IX86_BUILTIN_XGETBV>: Ditto.
4210 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
4211 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
4213 2018-06-07 Peter Bergner <bergner@vnet.ibm.com>
4215 Backport from mainline
4216 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
4219 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
4220 Don't handle -mcpu=power8 if -mpower9-vector is also used.
4222 2018-06-06 Richard Biener <rguenther@suse.de>
4224 Backport from mainline
4225 2018-06-04 Richard Biener <rguenther@suse.de>
4227 PR tree-optimization/86038
4228 * tracer.c (find_best_successor): Check probability for
4229 being initialized, bail out if not.
4231 2018-05-30 Richard Biener <rguenther@suse.de>
4233 PR tree-optimization/85964
4234 * tracer.c (better_p): Drop initialized count check, we only
4235 call the function with initialized counts now.
4236 (find_best_successor): Do find a best edge if one
4237 has uninitialized count.
4238 (find_best_predecessor): Likewise. Do BB frequency check only
4239 if count is initialized.
4241 2018-05-28 Richard Biener <rguenther@suse.de>
4243 PR tree-optimization/85934
4244 * tree-vect-generic.c (expand_vector_operations_1): Hoist
4245 vector boolean check before scalar optimization.
4247 2018-05-22 Richard Biener <rguenther@suse.de>
4249 PR tree-optimization/85863
4250 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
4251 comparisons when vectype is specified.
4252 (vectorizable_condition): Do not specify vectype for
4253 vect_is_simple_cond when SLP vectorizing.
4255 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
4257 Backport from mainline
4258 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
4260 * config/s390/s390-builtin-types.def: Add void function type.
4261 * config/s390/s390-builtins.def: Use the function type for the
4264 2018-06-04 Martin Sebor <msebor@redhat.com>
4267 * calls.c (maybe_warn_nonstring_arg): Use string length to set
4268 or ajust the presumed bound on an operation to avoid unnecessary
4271 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
4274 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
4275 of bits to ignore when comparing architectures.
4277 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4280 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
4282 2018-06-01 Bill Schmidt <wschmidt@linux.ibm.com>
4284 PR tree-optimization/85712
4285 Backport from mainline:
4286 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
4288 PR tree-optimization/85712
4289 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
4291 (alloc_cand_and_find_basis): Initialize first_interp field.
4292 (slsr_process_mul): Modify first_interp field.
4293 (slsr_process_add): Likewise.
4294 (slsr_process_cast): Modify first_interp field for each new
4296 (slsr_process_copy): Likewise.
4297 (dump_candidate): Dump first_interp field.
4298 (replace_mult_candidate): Process all interpretations, not just
4300 (replace_rhs_if_not_dup): Likewise.
4301 (replace_one_candidate): Likewise.
4303 Backport from mainline:
4304 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
4306 PR tree-optimization/85712
4307 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
4308 this candidate has already been replaced in-situ by a copy.
4310 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
4313 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
4314 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
4316 (sse4_1_round<mode>2): Use nonimmediate_operand
4317 for operand 1 predicate.
4319 2018-05-31 Jakub Jelinek <jakub@redhat.com>
4322 * bb-reorder.c (pass_partition_blocks::gate): Return false for
4323 functions with naked attribute.
4325 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
4328 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
4329 and movx for Haswell.
4331 2018-05-30 Jakub Jelinek <jakub@redhat.com>
4333 Backported from mainline
4334 2018-05-14 Jakub Jelinek <jakub@redhat.com>
4337 * config/i386/i386.md: Disallow non-commutative arithmetics in
4338 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
4339 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
4340 in the peephole2 before it.
4342 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4345 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
4346 after cmpelim optimization.
4348 2018-05-28 H.J. Lu <hongjiu.lu@intel.com>
4350 Backport from mainline
4351 2018-05-26 H.J. Lu <hongjiu.lu@intel.com>
4355 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
4357 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
4361 * varasm.c (assemble_alias): Check ifunc_resolver only on
4364 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
4367 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
4369 (cgraph_node::create_alias): Likewise.
4370 (cgraph_node::get_availability): Check ifunc_resolver instead
4371 of looking up ifunc attribute.
4372 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
4373 * varasm.c (do_assemble_alias): Likewise.
4374 (assemble_alias): Likewise.
4375 (default_binds_local_p_3): Likewise.
4376 * cgraph.h (cgraph_node): Add ifunc_resolver.
4377 (cgraph_node::only_called_directly_or_aliased_p): Return false
4379 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
4381 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
4382 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
4383 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
4384 instead of looking up ifunc attribute.
4386 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
4388 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
4390 2018-05-25 Richard Biener <rguenther@suse.de>
4393 * tree-dump.c (dequeue_and_dump): Remove access to removed
4394 operand 2 of a SWITCH_EXPR.
4396 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
4398 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
4399 Add {q} suffix to insn mnemonic.
4401 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
4404 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
4405 when memory input operand is handled.
4407 2018-05-23 Bin Cheng <bin.cheng@arm.com>
4409 Backport from mainline
4410 2018-05-17 Bin Cheng <bin.cheng@arm.com>
4411 Richard Biener <rguenther@suse.de>
4413 PR tree-optimization/85793
4414 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
4415 for VMAT_ELEMENTWISE.
4417 2018-05-23 Richard Biener <rguenther@suse.de>
4420 * tree-data-ref.c (create_runtime_alias_checks): Defer
4421 and ignore overflow warnings.
4423 2018-05-23 Martin Liska <mliska@suse.cz>
4425 Backport from mainline
4426 2018-05-23 Yury Gribov <tetra2005@gmail.com>
4428 PR tree-optimization/85822
4429 * tree-vrp.c (is_masked_range_test): Fix handling of negative
4432 2018-05-23 Martin Liska <mliska@suse.cz>
4434 Backport from mainline
4435 2018-05-11 Martin Liska <mliska@suse.cz>
4438 * doc/extend.texi: Document LLVM style format for no_sanitize
4441 2018-05-22 Martin Sebor <msebor@redhat.com>
4444 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
4446 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
4448 PR tree-optimization/85814
4449 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
4450 a null return from get_strinfo when unsharing the next
4451 strinfo in the chain.
4453 2018-05-21 Pat Haugen <pthaugen@us.ibm.com>
4455 Backport from mainline
4456 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
4457 Segher Boessenkool <segher@kernel.crashing.org>
4460 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest operand.
4462 2018-05-18 Toon Moene <toon@moene.org>
4464 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
4465 directly after that of -floop-interchange. Indicate that both
4466 options are enabled by default when specifying -O3.
4468 2018-05-11 Martin Jambor <mjambor@suse.cz>
4471 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
4474 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4477 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
4479 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4482 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
4484 (cortex-m0): Use armv6s-m isa.
4485 (cortex-m0plus): Likewise.
4486 (cortex-m1): Likewise.
4487 (cortex-m0.small-multiply): Likewise.
4488 (cortex-m0plus.small-multiply): Likewise.
4489 (cortex-m1.small-multiply): Likewise.
4491 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
4493 PR rtl-optimization/85638
4494 * bb-reorder.c: Include common/common-target.h.
4495 (create_forwarder_block): New function extracted from...
4496 (fix_up_crossing_landing_pad): ...here. Rename into...
4497 (dw2_fix_up_crossing_landing_pad): ...this.
4498 (sjlj_fix_up_crossing_landing_pad): New function.
4499 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
4500 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
4501 from both partitions and exit the loop after one iteration.
4503 2018-05-08 Richard Biener <rguenther@suse.de>
4505 Backport from mainline
4506 2018-05-04 Richard Biener <rguenther@suse.de>
4509 * fold-const.c (negate_expr_p): Restrict negation of operand
4510 zero of a division to when we know that can happen without
4512 (fold_negate_expr_1): Likewise.
4514 2018-05-03 Richard Biener <rguenther@suse.de>
4516 PR tree-optimization/85615
4517 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
4518 to loops not nested in BBs loop father to avoid creating multi-entry
4521 2018-05-02 Richard Biener <rguenther@suse.de>
4524 * gimplify.c (gimplify_save_expr): When in SSA form allow
4525 SAVE_EXPRs to compute to SSA vars.
4527 2018-05-02 Richard Biener <rguenther@suse.de>
4529 PR tree-optimization/85597
4530 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
4531 do not use split vect_get_vec_defs call but call vect_get_slp_defs
4534 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
4537 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
4538 (check_arch): Likewise.
4539 (check_fpu): Return the result rather than printing it.
4540 (end arch): Fix operator precedence.
4541 (end cpu): Likewise.
4542 (END): Print the result from check_fpu.
4544 2018-05-08 Richard Biener <rguenther@suse.de>
4546 Backport from mainline
4547 2018-05-04 Richard Biener <rguenther@suse.de>
4549 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
4550 * gimple-ssa-store-merging.c
4551 (imm_store_chain_info::output_merged_store): Remove redundant create,
4552 release split_store vector contents on failure.
4553 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
4554 scalar stmt vector on cache hit.
4556 2018-05-04 Richard Biener <rguenther@suse.de>
4559 * tree-complex.c (tree_lower_complex): Handle removed blocks.
4561 2018-05-07 Chung-Ju Wu <jasonwucj@gmail.com>
4563 Backport from mainline.
4564 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4566 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
4567 GET_MODE_MASK before any checking.
4568 (nds32_can_use_bset_p): Likewise.
4569 (nds32_can_use_btgl_p): Likewise.
4571 2018-05-07 Chung-Ju Wu <jasonwucj@gmail.com>
4573 Backport from mainline.
4574 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4576 * config/nds32/nds32-doubleword.md: New define_split pattern for
4577 illegal register number.
4579 2018-05-07 Chung-Ju Wu <jasonwucj@gmail.com>
4581 Backport from mainline.
4582 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4584 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
4586 2018-05-07 Chung-Ju Wu <jasonwucj@gmail.com>
4588 Backport from mainline.
4589 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4591 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
4593 2018-05-06 Jakub Jelinek <jakub@redhat.com>
4596 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
4597 the type is addressable. Don't force op into register if it has
4600 2018-05-04 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4602 Backport from trunk.
4603 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4606 * config/aarch64/constraints.md (Usg): Limit to 31.
4609 2018-05-02 Jakub Jelinek <jakub@redhat.com>
4611 * BASE-VER: Set to 8.1.1.
4613 2018-05-02 Release Manager
4615 * GCC 8.1.0 released.
4617 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
4619 PR tree-optimization/85586
4620 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
4621 exit early for statements in the same group if the accesses are
4624 2018-05-01 Jakub Jelinek <jakub@redhat.com>
4627 * doc/install.texi2html: Replace _002d with - and _002a with * in
4628 generated html files using sed.
4630 2018-04-27 Martin Jambor <mjambor@suse.cz>
4633 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
4634 the jump function allows for passing through aggregate values.
4636 2018-04-27 Jeff Law <law@redhat.com>
4638 * config.gcc: Mark tile* targets as deprecated/obsolete.
4640 2018-04-27 Jakub Jelinek <jakub@redhat.com>
4642 PR tree-optimization/85529
4643 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
4644 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
4645 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
4646 zero extension or masking of the MSB bit.
4647 (optimize_range_tests): Add FIRST_BB argument, pass it through
4648 to optimize_range_tests_var_bound.
4649 (maybe_optimize_range_tests, reassociate_bb): Adjust
4650 optimize_range_tests callers.
4652 2018-04-26 Richard Biener <rguenther@suse.de>
4653 Jakub Jelinek <jakub@redhat.com>
4655 * cgraph.h (symbol_table): Just declare debug method here.
4656 * symtab.c (symbol_table::debug): Define.
4658 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
4660 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
4662 2018-04-25 H.J. Lu <hongjiu.lu@intel.com>
4664 Backport from mainline
4665 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
4668 * config/i386/i386.c (ix86_expand_builtin): Change memory
4669 operand to XI, extend p0 to Pmode.
4670 * config/i386/i386.md: Change unspec volatile and operand
4671 1 mode to XI, change operand 0 mode to P.
4673 2018-04-25 Jakub Jelinek <jakub@redhat.com>
4675 * DEV-PHASE: Set to prerelease.
4677 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
4679 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
4680 in __abskf2 and __powikf2.
4682 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4685 * config/aarch64/constraints.md (Usg, Usj): New constraints.
4686 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
4687 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
4688 Use the above on operand 2. Reindent.
4689 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
4691 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
4694 * common/config/i386/i386-common.c (ix86_handle_option): Don't
4696 * config/i386/i386.opt (mcet): Removed.
4697 * doc/install.texi: Remove -mcet documentation.
4698 * doc/invoke.texi: Likewise.
4700 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
4703 * doc/install.texi: Remove -mcet from bootstrap-cet.
4705 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4708 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
4709 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
4713 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
4714 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
4715 containing a CONST_VECTOR.
4717 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
4719 * doc/install.texi: Update newlib dependency for nvptx.
4721 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4724 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
4725 instead of INTVAL when shifting x left.
4727 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
4729 PR tree-optimization/85478
4730 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
4731 vect_grouped_store_supported for single element vectors.
4733 2018-04-24 Richard Biener <rguenther@suse.de>
4736 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
4737 load cost increase to the case of non-constant step.
4739 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4742 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
4743 destination if any_malformed_asm.
4745 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
4748 * expr.c (store_field): In the bitfield case, if the value comes from
4749 a function call and is returned in registers by means of a PARALLEL,
4750 do not change the mode of the temporary unless BLKmode and VOIDmode.
4752 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
4754 PR rtl-optimization/85423
4755 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
4756 dependencies to debug insns when the previous insn is non-debug.
4758 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
4760 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
4761 enums into a single definition.
4762 (fls): Fix predicates and printing.
4765 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
4767 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
4768 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
4769 and short u6 immediate.
4770 (check_if_valid_sleep_operand): Remove.
4771 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
4773 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4775 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
4776 flag_always_save_lp condition.
4777 * config/nds32/nds32.opt (malways-save-lp): New option.
4779 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4781 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
4782 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
4783 * config/nds32/nds32.h
4784 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
4785 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
4787 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4789 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
4790 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
4792 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4793 Chung-Ju Wu <jasonwucj@gmail.com>
4795 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
4797 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
4798 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
4800 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4802 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
4804 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4806 * config/nds32/nds32-protos.h (nds32_data_alignment,
4807 nds32_local_alignment): Declare.
4808 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
4809 nds32_local_alignment): New functions.
4810 (TARGET_CONSTANT_ALIGNMENT): Define.
4811 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
4813 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4815 * config/nds32/nds32.c
4816 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
4817 (TARGET_MODES_TIEABLE_P): Likewise.
4819 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4821 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
4824 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
4825 Chung-Ju Wu <jasonwucj@gmail.com>
4827 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4828 for unaligned access.
4829 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4831 * config/nds32/nds32-intrinsic.md: Likewise.
4832 * config/nds32/nds32_intrinsic.h: Likewise.
4833 * config/nds32/nds32.h (nds32_builtins): Likewise.
4834 * config/nds32/nds32.opt (munaligned-access): New option.
4835 * config/nds32/nds32.c (nds32_asm_file_start): Display
4836 flag_unaligned_access status.
4838 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4840 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
4841 -mno-relax is present.
4842 * config/riscv/linux.h (LINK_SPEC): Ditto.
4844 2018-04-20 Martin Sebor <msebor@redhat.com>
4847 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
4849 (gimple_fold_builtin_stxcpy_chk): Same.
4850 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
4852 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
4855 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
4856 __powikf2 when long double is IEEE 128-bit.
4858 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4860 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
4861 step to make sure stack always aligned.
4863 2018-04-20 Carl Love <cel@us.ibm.com>
4866 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
4867 size check for arg0.
4869 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
4870 Tom de Vries <tom@codesourcery.com>
4873 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
4874 Emit insns for calls too.
4875 (nvptx_find_par): Always look for worker-level predecessor insn.
4876 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
4878 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
4879 (nvptx_process_pars): Propagate frames for calls.
4881 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
4884 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
4886 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
4887 (ix86_handle_option): Don't handle OPT_mibt.
4888 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
4890 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
4892 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
4893 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
4894 (ix86_target_macros): Define __CET__ with flag_cf_protection
4895 for -fcf-protection.
4896 * config/i386/i386.c (isa2_opts): Remove -mibt.
4897 * config/i386/i386.h (TARGET_IBT): Removed.
4898 (TARGET_IBT_P): Likewise.
4899 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
4900 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
4901 * config/i386/i386.opt (mcet): Update help message.
4904 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
4905 -mcet as an alias for -mshstk.
4907 2018-04-20 Richard Biener <rguenther@suse.de>
4910 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
4911 complexity by forcing a single use of the multiply operand.
4913 2018-04-20 Martin Jambor <mjambor@suse.cz>
4916 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
4917 recursion dependency to only apply to non-clones.
4919 2018-04-20 Martin Jambor <mjambor@suse.cz>
4922 * ipa-cp.c (create_specialized_node): Check that clones of
4923 self-recursive edges exist during IPA-CP.
4925 2018-04-19 Toon Moene <toon@moene.org>
4927 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
4930 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4932 PR tree-optimization/85467
4933 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
4934 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
4935 VECTOR_CST element to type.
4937 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4940 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
4941 * config/i386/i386.md (builtin_setjmp_setup): Removed.
4942 (builtin_longjmp): Likewise.
4943 (save_stack_nonlocal): New pattern.
4944 (restore_stack_nonlocal): Likewise.
4946 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4949 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4950 Replace ASM_OUTPUT_LABEL with fprintf.
4952 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4955 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4956 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
4957 * config/i386/i386-c.c (ix86_target_macros_internal): Also
4958 define __IBT__ and __SHSTK__ for -fcf-protection.
4959 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
4961 (ix86_trampoline_init): Likewise.
4962 (x86_output_mi_thunk): Likewise.
4963 (ix86_notrack_prefixed_insn_p): Likewise.
4964 (ix86_option_override_internal): Don't disallow -fcf-protection.
4965 * config/i386/i386.md (rdssp<mode>): Also enable for
4967 (incssp<mode>): Likewise.
4968 (nop_endbr): Likewise.
4969 * config/i386/i386.opt (mcet): Change help message to built-in
4973 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
4974 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
4975 enable CET built-in functions.
4977 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
4979 * common/config/i386/i386-common.c
4980 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
4981 OPTION_MASK_ISA_MOVDIRI_UNSET,
4982 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
4983 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
4984 * config.gcc (movdirintrin.h): New header.
4985 * config/i386/cpuid.h (bit_MOVDIRI,
4986 bit_MOVDIR64B): New bits.
4987 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
4989 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
4990 (VOID, PVOID, PCVOID)): New function types.
4991 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
4992 __builtin_ia32_directstoreu_u64,
4993 __builtin_ia32_movdir64b): New builtins.
4994 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
4995 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
4997 (ix86_valid_target_attribute_inner_p): Ditto.
4998 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
4999 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
5000 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
5001 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
5002 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
5003 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
5004 (movdiri<mode>, movdir64b_<mode>): New.
5005 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
5006 * config/i386/immintrin.h: Include movdirintrin.h.
5007 * config/i386/movdirintrin.h: New file.
5008 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
5010 2018-04-19 Richard Biener <rguenther@suse.de>
5013 * cfg.c (clear_bb_flags): When loop state says we have
5014 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
5016 2018-04-19 Richard Biener <rguenther@suse.de>
5018 PR tree-optimization/84737
5019 * tree-vect-data-refs.c (vect_copy_ref_info): New function
5020 copying restrict info.
5021 (vect_setup_realignment): Use it.
5022 * tree-vectorizer.h (vect_copy_ref_info): Declare.
5023 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
5024 the first DR to all generated stores.
5025 (vectorizable_load): Likewise for loads.
5027 2018-04-19 Jakub Jelinek <jakub@redhat.com>
5029 PR tree-optimization/85446
5030 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
5031 the integral and pointer types to have the same precision.
5033 * doc/install.texi: Document --disable-cet being the default and
5036 2018-04-18 Martin Liska <mliska@suse.cz>
5038 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
5041 2018-04-18 Martin Liska <mliska@suse.cz>
5044 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5047 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5048 arguments if they are comparable.
5050 2018-04-18 Martin Liska <mliska@suse.cz>
5053 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5056 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5059 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
5062 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
5063 ENDBR after calling __morestack.
5065 2018-04-18 David Malcolm <dmalcolm@redhat.com>
5068 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
5069 by using gcc_base_ver to generate a gcc_driver_version, and use
5070 it when generating GCC_DRIVER_NAME.
5071 * configure: Regenerate.
5073 2018-04-18 Jakub Jelinek <jakub@redhat.com>
5076 * config.gcc: Obsolete powerpc*-*-*spe*.
5078 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5081 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
5082 (stabstr_D): Change type of unum from unsigned int to
5083 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
5086 2018-04-17 Jim Wilson <jimw@sifive.com>
5089 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
5090 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
5091 Set arg_pointer_offset after using pretend_args_size.
5093 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5095 PR rtl-optimization/85431
5096 * dse.c (record_store): Ignore zero width stores.
5099 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
5100 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
5101 __builtin_stack_restore rather than after it.
5102 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
5103 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
5104 argument instead of virtual_dynamic_stack_rtx.
5106 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
5108 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
5110 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
5111 Add note to error message to explain internal mapping of overloaded
5112 built-in function name to non-overloaded built-in function name.
5113 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
5116 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
5119 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
5120 where the inputs overlap with the output.
5122 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5124 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
5125 (=v, v) alternative and explicit "memory" attribute.
5126 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
5127 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5129 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
5130 "sselog1" type instead of "sselog".
5131 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
5132 "sselog". Remove explicit "memory" attribute.
5133 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
5134 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5136 (vec_extract_hi_v32hi): Merge all alternatives into one, use
5137 "sselog1" type instead of "sselog". Remove explicit "memory"
5139 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
5140 use "sselog1" type instead of "sselog". Remove explicit "memory"
5142 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
5143 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5145 (vec_extract_hi_v64qi): Merge all alternatives into one, use
5146 "sselog1" type instead of "sselog". Remove explicit "memory"
5148 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
5149 use "sselog1" type instead of "sselog". Remove explicit "memory"
5153 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
5156 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
5159 2018-04-17 Martin Jambor <mjambor@suse.cz>
5162 * ipa-cp.c (create_specialized_node): Call
5163 expand_all_artificial_thunks if necessary.
5165 2018-04-17 Martin Liska <mliska@suse.cz>
5168 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
5169 in message, remote space in between '_G' and '('.
5171 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5174 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
5175 avx512f_vmcmp<mode>3<round_saeonly_name>,
5176 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
5177 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
5178 avx512f_rndscale<mode><round_saeonly_name>,
5179 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
5180 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
5181 Use %<iptr>2 instead of %2 for -masm=intel.
5182 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
5183 avx512f_vcvttss2usi<round_saeonly_name>,
5184 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
5186 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
5187 avx512f_vcvttsd2usi<round_saeonly_name>,
5188 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
5189 Use %q1 instead of %1 for -masm=intel.
5190 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
5191 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
5192 of %3 for -masm=intel.
5193 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
5195 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
5197 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
5199 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
5200 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
5202 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
5204 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
5205 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
5206 %g1 and one with %0 and %1.
5207 (avx512er_vmrcp28<mode><round_saeonly_name>,
5208 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
5210 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
5211 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
5212 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
5213 of %0 and %{%4%} for -masm=intel.
5214 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
5215 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
5216 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
5217 order of %0 and %{%5%}%{z%} for -masm=intel.
5219 2018-04-17 Jan Hubicka <jh@suse.cz>
5222 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
5224 2018-04-17 Martin Liska <mliska@suse.cz>
5227 * multiple_target.c (create_dispatcher_calls): Set apostrophes
5228 for target_clone error message. Make default implementation
5229 clone to be a local declaration.
5230 (separate_attrs): Add new argument and check for an empty
5232 (expand_target_clones): Handle it.
5233 (ipa_target_clone): Make redirection just for target_clones
5236 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
5237 Tom de Vries <tom@codesourcery.com>
5240 * omp-expand.c (expand_oacc_for): Add dummy false branch for
5241 tiled basic blocks without omp continue statements.
5243 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
5246 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
5247 vec_extract expression as having side effects to make sure it gets
5250 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
5253 * config/i386/i386.c (get_builtin_code_for_version): Check
5256 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
5259 * gcc/config.gcc: Support "skylake".
5260 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
5262 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
5263 (processor_target_table): Add "skylake".
5264 (ix86_option_override_internal): Add "skylake".
5265 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
5266 PROCESSOR_CANNONLAKE.
5267 (get_builtin_code_for_version): Fix priority for
5268 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
5269 PROCESSOR_SKYLAKE-AVX512.
5270 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
5271 (processor_type): Add PROCESSOR_SKYLAKE.
5273 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
5274 Jason Merrill <jason@redhat.com>
5277 * convert.c (convert_to_integer_1): Use direct recursion for
5278 enumeral types and types with a precision less than the number
5279 of bits in their mode.
5281 2018-04-16 Julia Koval <julia.koval@intel.com>
5284 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
5285 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
5287 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
5290 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
5291 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
5292 and -mno-direct-move.
5294 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
5297 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
5298 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
5299 Ensure negative shifts result in {0}.
5301 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
5303 PR rtl-optimization/79916
5304 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
5305 regs (if any) to define how to gnerate SD moves when LRA is in
5308 2018-04-13 Jakub Jelinek <jakub@redhat.com>
5310 PR rtl-optimization/85393
5311 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
5312 * except.c (expand_dw2_landing_pad_for_region): Make static.
5313 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
5314 a label and unconditional jump to old_bb, rather than
5315 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
5318 PR rtl-optimization/85376
5319 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
5320 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
5321 instead of a specific value.
5323 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
5324 Bin Cheng <bin.cheng@arm.com>
5326 PR tree-optimization/82965
5327 PR tree-optimization/83991
5328 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
5329 by_profile_only parameter.
5330 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
5331 information if the loop was predicted to iterate too many times.
5332 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
5334 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
5337 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
5340 2018-04-13 Martin Liska <mliska@suse.cz>
5341 Jakub Jelinek <jakub@redhat.com>
5344 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
5345 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
5346 * builtins.c (expand_builtin_memory_copy_args): Use
5347 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
5348 handle dest_addr == pc_rtx.
5350 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
5353 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
5354 asked to not generate direct moves.
5355 (fix_trunc<mode>si2_stfiwx): Similar.
5356 (fix_trunc<mode>si2_internal): Similar.
5358 2018-04-12 Jakub Jelinek <jakub@redhat.com>
5361 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
5362 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
5363 lookup if dest in some wider mode is known to be const0_rtx and
5364 if so, record permanent equivalence for it to be ZERO_EXTEND of
5365 the narrower mode destination.
5367 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
5369 * lto-streamer-out.c (output_function): Revert 259346.
5370 * omp-expand.c (expand_oacc_for): Likewise.
5372 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
5374 PR rtl-optimization/85354
5375 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
5376 * sel-sched.c (sel_global_init): ... here.
5378 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
5381 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
5382 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
5383 mode for PE-COFF targets.
5384 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
5385 (i386_pe_asm_lto_end): Likewise.
5386 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
5387 (TARGET_ASM_LTO_END): Likewise.
5388 * config/i386/winnt.c (saved_debug_info_level): New static variable.
5389 (i386_pe_asm_lto_start): New function.
5390 (i386_pe_asm_lto_end): Likewise.
5392 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
5393 Richard Biener <rguenther@suse.de>
5396 * lto-streamer-out.c (output_function): Fix CFG loop state before
5398 * omp-expand.c (expand_oacc_for): Handle calls to internal
5399 functions like regular functions.
5401 2018-04-12 Richard Biener <rguenther@suse.de>
5404 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
5405 for the early LTO debug to properly generate references to it
5406 during DIE emission. Do not re-use that for the skeleton for
5408 (dwarf2out_early_finish): Likewise.
5410 2018-04-12 Jakub Jelinek <jakub@redhat.com>
5413 * config/i386/sse.md
5414 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
5415 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
5416 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
5417 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
5418 and output is a reg, avoid creating invalid lowpart subreg, but
5419 instead split into a 512-bit move. Don't split if not AVX512VL,
5420 input is xmm16+ reg and output is a mem.
5421 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
5422 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
5423 xmm16+ reg and output is a mem.
5425 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5427 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
5428 also for flag_dwarf2_cfi_asm.
5430 2018-04-12 Jakub Jelinek <jakub@redhat.com>
5432 PR rtl-optimization/85342
5433 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
5434 a bool scalar var inside of the loop instead. Don't try to update
5435 recog_data.operand after failed apply_change_group.
5437 2018-04-12 Tom de Vries <tom@codesourcery.com>
5440 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
5441 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
5442 array with flexible array member as array without given dimension.
5443 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
5444 argument for undefined param to true.
5446 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
5449 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
5450 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
5451 from PowerPC section.
5452 * config/rs6000/sysv4.opt (mcall-): Improve help text.
5453 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
5454 help text that is too long.
5455 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
5456 help text that is too long.
5457 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
5458 help text that is too long.
5460 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
5462 * config/alpha/alpha.md (stack_probe_internal): Rename
5463 from "probe_stack". Update all callers.
5465 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
5467 PR rtl-optimization/84566
5468 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
5469 sched_macro_fuse_insns.
5471 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
5474 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
5475 (compute_block_dependences): ... from here.
5477 2018-04-11 Jakub Jelinek <jakub@redhat.com>
5479 PR tree-optimization/85331
5480 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
5481 from int to HOST_WIDE_INT.
5483 2018-04-11 Martin Jambor <mjambor@suse.cz>
5486 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
5487 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
5488 not the same as the source val.
5489 (cgraph_edge_brings_value_p): New parameter.
5490 (gather_edges_for_value): Pass destination value to
5491 cgraph_edge_brings_value_p.
5492 (perhaps_add_new_callers): Likewise.
5493 (get_info_about_necessary_edges): Likewise and exclude values brought
5494 only by self-recursive edges.
5495 (create_specialized_node): Redirect only clones of self-calling edges.
5496 (+self_recursive_pass_through_p): New function.
5497 (find_more_scalar_values_for_callers_subset): Use it.
5498 (find_aggregate_values_for_callers_subset): Likewise.
5499 (known_aggs_to_agg_replacement_list): Removed.
5500 (decide_whether_version_node): Re-calculate known constants for all
5501 remaining context clones.
5503 2018-04-11 Richard Biener <rguenther@suse.de>
5506 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
5507 from early DWARF output.
5508 (dwarf2out_early_finish): Output line info unconditionally into
5509 early DWARF and add reference to it.
5511 2018-04-11 Jakub Jelinek <jakub@redhat.com>
5514 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
5515 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
5516 other than V2DFmode using iptr mode attribute.
5517 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
5519 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
5521 PR rtl-optimization/84659
5522 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
5524 2018-04-11 Jakub Jelinek <jakub@redhat.com>
5527 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
5529 (output_loc_list): Pass address of a dummy size variable even in the
5530 locview handling loop.
5531 (index_location_lists): Add comment on why skip_loc_list_entry can't
5534 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
5537 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
5540 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
5543 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
5544 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
5545 and -mstring-compare-inline-limit.
5547 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
5550 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
5551 for stack clash protection in a register whenever we need it to be in
5554 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
5556 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
5557 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
5559 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
5562 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
5564 (mlong-double-): Ditto.
5565 * config/rs6000/sysv4.opt (msdata=): Ditto.
5566 (mtls-size=): Ditto.
5568 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
5570 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
5571 erroneous entries for
5572 "vector int vec_ldl (int, long int *)", and
5573 "vector unsigned int vec_ldl (int, unsigned long int *)".
5574 Add comments and entries for
5575 "vector bool char vec_ldl (int, bool char *)",
5576 "vector bool short vec_ldl (int, bool short *)",
5577 "vector bool int vec_ldl (int, bool int *)",
5578 "vector bool long long vec_ldl (int, bool long long *)",
5579 "vector pixel vec_ldl (int, pixel *)",
5580 "vector long long vec_ldl (int, long long *)",
5581 "vector unsigned long long vec_ldl (int, unsigned long long *)".
5582 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
5583 type tree bool_long_long_type_node and correct definition of
5584 bool_V2DI_type_node to make reference to this new type tree.
5585 (rs6000_mangle_type): Replace erroneous reference to
5586 bool_long_type_node with bool_long_long_type_node.
5587 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
5588 comments to emphasize sign distinctions for char and int types and
5589 replace RS6000_BTI_bool_long constant with
5590 RS6000_BTI_bool_long_long constant. Also add comment to restrict
5591 use of RS6000_BTI_pixel.
5592 (bool_long_type_node): Remove this macro definition.
5593 (bool_long_long_type_node): New macro definition
5595 2018-04-10 Jakub Jelinek <jakub@redhat.com>
5597 PR rtl-optimization/85300
5598 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
5599 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
5600 simplify_unary_operation fails.
5602 2018-04-10 Martin Liska <mliska@suse.cz>
5604 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
5605 cgraph_edge and ipa_ref.
5607 2018-04-10 Jakub Jelinek <jakub@redhat.com>
5611 * config/i386/sse.md
5612 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
5613 computation of the VEC_MERGE selector from mask.
5614 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
5615 Fix decoding of the VEC_MERGE selector into mask.
5617 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
5619 PR tree-optimization/85286
5620 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
5622 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
5624 * final.c (final_1): Set insn_last_address as well as
5625 insn_current_address.
5627 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5630 * explow.c (emit_stack_probe): Call validize_mem on memory location
5631 before passing it to gen_probe_stack. Create address operand and
5632 legitimize it for the probe_stack_address case.
5634 2018-04-09 Jan Hubicka <jh@suse.cz>
5637 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
5638 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
5639 * tree.c (free_lang_data_in_type): Fix handling of binfos;
5641 (free_lang_data): Rebuild type inheritance graph.
5643 2018-04-09 Martin Sebor <msebor@redhat.com>
5645 * invoke.texi (-finline-small-functions): Mention other optimization
5647 (-findirect-inlining, -fpartial-inlining): Same.
5648 (-finline-functions-called-once): Same.
5649 (-freorder-blocks-and-partition): Same.
5651 2018-04-09 Jan Hubicka <jh@suse.cz>
5654 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
5655 jumps; choose last target that matches the criteria (i.e.
5656 no partition changes for non-crossing jumps).
5657 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
5658 support for redirecting crossing jumps to non-crossing.
5660 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
5662 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
5663 also for naked functions.
5665 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
5667 * config/arc/arc.md (add_shift): New pattern.
5668 (add_shift2): Likewise.
5669 (sub_shift): Likewise.
5670 (sub_shift_cmp0_noout): Likewise.
5671 (compare_si_ashiftsi): Likewise.
5672 (xbfu_cmp0_noout): New combine pattern.
5673 (xbfu_cmp0"): Likewise.
5674 (movsi_set_cc_insn): Place the predicable variant first.
5675 (commutative_binary_cmp0_noout): Remove clobber.
5676 (commutative_binary_cmp0): New pattern.
5677 (noncommutative_binary_cmp0): Likewise.
5678 (noncommutative_binary_cmp0_noout): Likewise.
5679 (noncommutative_binary_comparison_result_used): Removed.
5680 (rsub_cmp0): New pattern.
5681 (rsub_cmp0_noout): Likewise.
5682 (extzvsi): Changed, keep only meaningful variants.
5683 (SQH, SEZ): New iterators.
5684 (SQH_postfix): New mode attribute.
5685 (SEZ_prefix): New code attribute.
5686 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
5687 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
5688 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
5690 (noncommutative_operator): Check the availability of barrel
5693 2018-04-09 Richard Biener <rguenther@suse.de>
5695 PR tree-optimization/85284
5696 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
5697 Only use the niter constraining form of simple_iv when the exit
5700 2018-04-09 Tom de Vries <tom@codesourcery.com>
5703 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
5704 (define_expand "*memory_barrier"): New define_expand.
5705 (define_insn "memory_barrier"): New insn.
5707 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5709 PR rtl-optimization/80463
5710 PR rtl-optimization/83972
5711 PR rtl-optimization/83480
5713 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
5714 correct producer for the insn.
5715 (tidy_control_flow): Fixup seqnos in case of debug insns.
5717 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5719 PR rtl-optimization/83913
5721 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
5722 different sched-times when merging exprs.
5724 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5726 PR rtl-optimization/83962
5728 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
5729 tidy_fallthru_edge and tidy_control_flow.
5731 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5733 PR rtl-optimization/83530
5735 * sel-sched.c (force_next_insn): New global variable.
5736 (remove_insn_for_debug): When force_next_insn is true, also leave only
5737 next insn in the ready list.
5738 (sel_sched_region): When the region wasn't scheduled, make another pass
5739 over it with force_next_insn set to 1.
5741 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
5743 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
5745 * config/nds32/constants.md (unspec_volatile_element): Add enum values
5746 for interrupt control.
5747 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5748 functions for interrupt control.
5749 * config/nds32/nds32-intrinsic.md: Likewise.
5750 * config/nds32/nds32_intrinsic.h: Likewise.
5751 * config/nds32/nds32.h (nds32_builtins): Likewise.
5753 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
5755 * config/nds32/nds32.c (nds32_init_machine_status,
5756 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
5757 strict_aligned_p field.
5758 (nds32_expand_to_rtl_hook): New function.
5759 (TARGET_EXPAND_TO_RTL_HOOK): Define.
5760 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
5762 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5763 Chung-Ju Wu <jasonwucj@gmail.com>
5765 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
5766 * config/nds32/nds32-n7.md: New file.
5767 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
5768 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
5770 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
5771 * config/nds32/nds32.md (pipeline_model): Add n7.
5772 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
5773 * config/nds32/pipelines.md: Include n7 settings.
5775 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5776 Chung-Ju Wu <jasonwucj@gmail.com>
5778 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
5779 * config/nds32/nds32-e8.md: New file.
5780 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
5781 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
5783 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
5784 * config/nds32/nds32.md (pipeline_model): Add e8.
5785 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
5786 * config/nds32/pipelines.md: Include e8 settings.
5788 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5789 Chung-Ju Wu <jasonwucj@gmail.com>
5791 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
5792 * config/nds32/nds32-n8.md: New file.
5793 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
5794 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
5796 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
5797 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
5798 * config/nds32/nds32.md (pipeline_model): Add n8.
5799 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
5800 * config/nds32/pipelines.md: Include n8 settings.
5802 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5803 Chung-Ju Wu <jasonwucj@gmail.com>
5805 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
5806 * config/nds32/nds32-n9-2r1w.md: New file.
5807 * config/nds32/nds32-n9-3r2w.md: New file.
5808 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
5809 nds32_register_ports): New or modify for cpu n9.
5810 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
5812 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
5813 * config/nds32/nds32-utils.c: New file.
5814 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
5815 TARGET_MUL_SLOW): Define.
5816 * config/nds32/nds32.md (pipeline_model): New attribute.
5817 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
5818 New options that support cpu n9.
5819 * config/nds32/pipelines.md: Include n9 settings.
5820 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
5822 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
5824 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
5825 information if necessary.
5826 (output_cond_branch_compare_zero): Likewise.
5827 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
5828 (nds32_target_alignment): Refine for alignment.
5829 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
5830 (FUNCTION_BOUNDARY): Modify.
5831 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
5833 * config/nds32/nds32.opt (malways-align, malign-functions): New.
5835 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
5837 * config/nds32/constants.md (unspec_volatile_element): Add values for
5838 TLB operation and data prefetch.
5839 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5840 functions for TLB operation and data prefetch.
5841 * config/nds32/nds32-intrinsic.md: Likewise.
5842 * config/nds32/nds32_intrinsic.h: Likewise.
5843 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
5844 (nds32_print_operand): Likewise.
5845 * config/nds32/nds32.h (nds32_builtins): Likewise.
5847 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
5848 Andrew Pinski <pinsika@gcc.gnu.org>
5851 * match.pd: Use constant_boolean_node of correct type instead of
5852 boolean_true_node or boolean_false_node for simplifying
5853 pointer comparisons to zero.
5855 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5857 PR tree-optimization/80021
5858 * tree.c (verify_type_variant): Make error call in verify_variant_match
5859 translatable and remove final full stop.
5861 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5863 * config/nds32/constants.md (unspec_volatile_element): Add
5864 UNSPEC_VOLATILE_EH_RETURN.
5865 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
5866 nds32_output_stack_pop): Support dwarf exception handling process.
5867 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
5868 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
5869 exception handling process.
5870 (nds32_compute_stack_frame): Likewise.
5871 (nds32_return_addr_rtx): Likewise.
5872 (nds32_initial_elimination_offset): Likewise.
5873 (nds32_expand_prologue): Likewise.
5874 (nds32_expand_epilogue): Likewise.
5875 (nds32_dynamic_chain_address): New function.
5876 * config/nds32/nds32.h (machine_function): Add fields for dwarf
5878 (DYNAMIC_CHAIN_ADDRESS): Define.
5879 (EH_RETURN_DATA_REGNO): Define.
5880 (EH_RETURN_STACKADJ_RTX): Define.
5881 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
5882 patterns for dwarf exception handling.
5884 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5886 * config/nds32/nds32.h: Clean up obsolete macros.
5888 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5890 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5891 Add enum values for particular instructions.
5892 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5893 particular intrinsic functions.
5894 * config/nds32/nds32-intrinsic.md: Likewise.
5895 * config/nds32/nds32_intrinsic.h: Likewise.
5896 * config/nds32/nds32.h (nds32_builtins): Likewise.
5897 * config/nds32/nds32.md (type): Add pbsad and pbsada.
5898 (btst, ave): New patterns for particular instructions.
5900 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5902 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5903 Add enum values for atomic load/store and memory sync.
5904 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
5906 * config/nds32/nds32-intrinsic.md: Likewise.
5907 * config/nds32/nds32_intrinsic.h: Likewise.
5908 * config/nds32/nds32.h (nds32_builtins): Likewise.
5910 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5912 PR tree-optimization/85257
5913 * fold-const.c (native_encode_vector): If not all elts could fit
5914 and off is -1, return 0 rather than offset.
5915 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
5916 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
5917 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
5918 adjust buffer in native_interpret_expr call.
5920 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5922 * config/nds32/constants.md (unspec_volatile_element): Add cache
5923 control enum values.
5924 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
5925 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
5926 * config/nds32/nds32.c (nds32_cctl_names): New.
5927 (nds32_print_operand): Handle cache control register names.
5928 * config/nds32/nds32.h (nds32_builtins): New enum values.
5929 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
5931 * config/nds32/nds32.md (type): Add mmu.
5932 * config/nds32/pipelines.md (simple_insn): Add mmu.
5934 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5936 * config/nds32/nds32.md (type): Remove call.
5937 * config/nds32/pipelines.md (simple_insn): Likewise.
5939 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5941 * config/nds32/constants.md (unspec_volatile_element): Add
5942 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
5943 UNSPEC_VOLATILE_FMFCFG.
5944 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
5945 description for fmfcfg and fmfcsr.
5946 (bdesc_1arg): Add fmtcsr.
5947 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
5948 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
5949 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
5950 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
5951 unspec_fmfcfg): New patterns.
5952 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
5953 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
5954 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
5955 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
5956 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
5957 __nds32__fmfcfg): Define.
5959 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5961 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
5962 intrinsic register names.
5963 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
5964 intrinsic register enum values and macros.
5966 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5968 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
5969 for load/store addressing form.
5970 (nds32_print_operand_address): Likewise.
5972 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
5975 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
5976 based on LABEL_REF. Remove useless assertion.
5977 (pic_address_needs_scratch): Fix formatting.
5978 (sparc_legitimize_pic_address): Minor tweaks.
5979 (sparc_delegitimize_address): Adjust assertion accordingly.
5980 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
5981 into symbolic_operand.
5982 (movsi_high_pic_label_ref): Likewise.
5983 (movsi_lo_sum_pic_label_ref): Likewise.
5984 (movdi_pic_label_ref): Likewise.
5985 (movdi_high_pic_label_ref): Likewise.
5986 (movdi_lo_sum_pic_label_ref): Likewise.
5988 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
5990 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
5991 custom LIB_SPEC setup.
5993 2018-04-06 Ruslan Bukin <br@bsdpad.com>
5994 Kito Cheng <kito.cheng@gmail.com>
5996 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
5997 * config/riscv/freebsd.h: New.
5999 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6001 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
6002 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
6005 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6006 Kito Cheng <kito.cheng@gmail.com>
6008 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
6009 nds32_output_call, nds32_symbol_binds_local_p): New functions.
6010 * config/nds32/nds32-protos.h (nds32_output_call,
6011 nds32_output_return): Declare.
6012 * config/nds32/nds32.md: Refine all the call and return patterns.
6014 2018-04-06 Jakub Jelinek <jakub@redhat.com>
6017 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
6018 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
6020 PR rtl-optimization/84872
6021 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
6022 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
6025 2018-04-06 Tamar Christina <tamar.christina@arm.com>
6027 * expr.c (copy_blkmode_to_reg): Revert 254862.
6028 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
6030 2018-04-06 Richard Biener <rguenther@suse.de>
6033 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
6034 after seeing a component reference with an adjacent field. Treat
6035 refs to arrays at struct end of external decls similar to
6036 refs to unconstrained commons.
6038 2018-04-06 Jakub Jelinek <jakub@redhat.com>
6041 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
6042 look through SAVE_EXPRs with non-side-effects argument. Adjust
6044 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
6047 2018-04-06 Richard Biener <rguenther@suse.de>
6050 * alias.c (find_base_term): New wrapper around find_base_term
6051 unwinding CSELIB_VAL_PTR changes.
6052 (find_base_term): Do not restore CSELIB_VAL_PTR during the
6055 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6057 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
6059 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
6060 constant definitions.
6061 ("nop"): lr 0,0 -> nopr r0
6062 ("nop_lr0", "nop_lr1"): New insn definitions.
6064 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6066 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
6067 NDS32_V3PUSH_AVAILABLE_P macro.
6069 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
6070 Chung-Ju Wu <jasonwucj@gmail.com>
6072 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
6073 (nds32*-*-*): Add float and fpu_config into supported_defaults.
6074 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
6075 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
6076 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
6077 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
6078 * config/nds32/constraints.md: New constraints and checking for hard
6079 float configuration.
6080 * config/nds32/iterators.md: New mode iterator and attribute for hard
6081 float configuration.
6082 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
6084 * config/nds32/nds32-fpu.md: New file.
6085 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
6086 deal with hard float code generation.
6087 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
6089 (abi_type, float_reg_number): New enum type.
6090 * config/nds32/nds32-predicates.c: New predicates for hard float.
6091 * config/nds32/nds32-protos.h: Declare functions for hard float.
6092 * config/nds32/nds32.c: Implementation for hard float configuration.
6093 * config/nds32/nds32.h: Definitions for hard float configuration.
6094 * config/nds32/nds32.md: Include hard float machine description and
6095 modify patterns for hard float configuration.
6096 * config/nds32/nds32.opt: New options for hard float configuration.
6097 * config/nds32/predicates.md: New predicates for hard float
6100 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6102 * common/config/nds32/nds32-common.c
6103 (nds32_option_optimization_table): Enable -mreleax-hint by default.
6105 2018-04-05 Jakub Jelinek <jakub@redhat.com>
6108 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
6109 CONSTRUCTOR_ELT (ctor, ...)->value.
6111 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
6114 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
6116 2018-04-05 Tom de Vries <tom@codesourcery.com>
6119 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
6122 2018-04-05 Shiva Chen <shiva0217@gmail.com>
6123 Kito Cheng <kito.cheng@gmail.com>
6125 * config/nds32/constraints.md (U33): Fine-tune checking condition.
6126 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
6127 * config/nds32/nds32.h (nds32_16bit_address_type): Add
6128 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
6130 2018-04-05 Shiva Chen <shiva0217@gmail.com>
6131 Kito Cheng <kito.cheng@gmail.com>
6133 * config/nds32/constraints.md (Ufe): New memory constraint.
6134 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
6135 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
6136 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
6138 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
6139 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
6141 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6143 * config/nds32/nds32.md: Use optimize_size in the condition for
6144 alu-shift instructions.
6146 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6148 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
6150 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6152 * config/nds32/nds32.md (negsi2): Refine pattern.
6154 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
6155 Chung-Ju Wu <jasonwucj@gmail.com>
6157 * config/nds32/iterators.md (shift_rotate): New code iterator.
6158 (shift): New code attribute.
6159 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
6160 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
6161 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
6162 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
6163 bit-wise operations.
6164 (andsi3, *andsi3): Ditto.
6165 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
6166 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
6167 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
6168 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
6169 nds32_ior_operand, nds32_xor_operand): New predicates.
6171 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6173 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
6174 (addsi3, subsi3): ... this.
6176 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6178 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
6180 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6182 * config/nds32/nds32.md: Adjust indention.
6184 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
6186 * config/nds32/nds32.md (feature): New attribute.
6188 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6190 * config/nds32/nds32.md (subtype): New attribute.
6192 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
6195 * config/arm/arm-builtins.c (arm_expand_builtin): Change
6196 expansion to perform a bitwise AND of the argument followed by a
6197 boolean negation of the result.
6199 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
6201 PR rtl-optimization/84878
6202 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
6203 the basic block. Assert the use reference is not artificial and that
6204 it has an associated insn.
6206 2018-04-04 Michael Matz <matz@suse.de>
6208 * builtins.c (compute_objsize): Pass correct operand
6209 to array_at_struct_end_p.
6211 2018-04-04 Richard Biener <rguenther@suse.de>
6214 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
6215 from contexts for DINFO_LEVEL_TERSE and below.
6217 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6219 * config/nds32/nds32-doubleword.md (move_<mode>): Require
6220 resiter_operand condition.
6221 * config/nds32/nds32.md (*move<mode>): Ditto.
6223 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6224 Monk Chiang <sh.chiang04@gmail.com>
6226 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
6228 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6230 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
6232 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6233 Kito Cheng <kito.cheng@gmail.com>
6235 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
6236 nds32_cond_code_str, output_cond_branch,
6237 output_cond_branch_compare_zero, nds32_expand_cbranch,
6238 nds32_expand_cstore, nds32_expand_movcc,
6239 nds32_output_cbranchsi4_equality_zero,
6240 nds32_output_cbranchsi4_equality_reg,
6241 nds32_output_cbranchsi4_equality_reg_or_const_int,
6242 nds32_output_cbranchsi4_greater_less_zero: New functions.
6243 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
6244 nds32_expand_cstore, nds32_expand_movcc,
6245 nds32_output_cbranchsi4_equality_zero,
6246 nds32_output_cbranchsi4_equality_reg,
6247 nds32_output_cbranchsi4_equality_reg_or_const_int,
6248 nds32_output_cbranchsi4_greater_less_zero): Declare.
6249 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
6250 nds32_rimm11s_operand): New predicates.
6251 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
6252 * config/nds32/nds32.md: Rewrite all the branch and conditional move
6255 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6257 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
6258 * config/nds32/nds32.md: Ditto.
6259 * config/nds32/pipelines.md: Ditto.
6261 2018-04-04 Richard Biener <rguenther@suse.de>
6263 PR tree-optimization/85168
6264 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
6265 propagating abnormals.
6267 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6269 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
6271 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6272 Kito Cheng <kito.cheng@gmail.com>
6274 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
6275 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
6276 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
6277 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
6278 * config/nds32/nds32.md (sibcall_internal): New.
6279 (sibcall_register): Remove.
6280 (sibcall_immediate): Remove.
6281 (sibcall_value_internal): New.
6282 (sibcall_value_register): Remove.
6283 (sibcall_value_immediate): Remove.
6284 * config/nds32/predicates.md (nds32_general_register_operand): New.
6285 (nds32_call_address_operand): New.
6287 2018-04-03 Jakub Jelinek <jakub@redhat.com>
6289 PR rtl-optimization/85167
6290 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
6291 bb_defs if *split_p, instead preinitialize it to NULL.
6293 PR tree-optimization/85156
6294 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
6295 evaluating the argument multiple times.
6297 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
6299 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
6301 (_mm_cvtpd_ps): Likewise.
6302 (_mm_cvttpd_epi32): Likewise.
6303 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
6304 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
6305 vector, pixel, and bool following altivec.h include.
6307 2018-04-03 Martin Sebor <msebor@redhat.com>
6309 * doc/extend.texi (Common Function Attributes): Clarify.
6310 (const attribute): Likewise.
6311 (pure attribute): Likewise.
6313 2018-04-03 Jakub Jelinek <jakub@redhat.com>
6316 * config/i386/i386.c (ix86_expand_vector_set): Use
6317 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
6319 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
6321 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
6322 instructions when changing rounding bits to preserve precision bits
6323 in the x87 control word.
6325 2018-04-03 Martin Liska <mliska@suse.cz>
6327 PR tree-optimization/82491
6328 * rtl.h (strip_offset_and_add): Replace += suboffset with
6329 poly_uint64 () + suboffset.
6331 2018-03-29 Martin Liska <mliska@suse.cz>
6332 Martin Jambor <mjambor@suse.cz>
6335 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
6336 param_type is not an integral or pointer type.
6338 2018-04-03 Richard Biener <rguenther@suse.de>
6340 * sese.h (recompute_all_dominators): Remove.
6342 2018-04-02 Martin Sebor <msebor@redhat.com>
6344 * doc/invoke.texi (-Wrestrict): Fix typos.
6346 2018-04-02 Jim Wilson <jimw@sifive.com>
6348 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
6349 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
6350 (<optab>di3, <optab>si3_extend): Likewise.
6351 (<optab>si3_mask, <optab>si3_mask_1): New.
6352 (<optab>di3_mask, <optab>di3_mask_1): New.
6353 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
6354 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
6355 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
6357 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
6359 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
6362 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
6364 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
6365 (nds32_canonicalize_comparison): New function.
6367 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
6368 Kito Cheng <kito.cheng@gmail.com>
6369 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6371 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
6372 * config/nds32/constants.md (unspec_volatile_element): Add
6373 UNSPEC_VOLATILE_RELAX_GROUP.
6374 * config/nds32/nds32-relax-opt.c: New file.
6375 * config/nds32/nds32-predicates.c
6376 (nds32_symbol_load_store_p): New function.
6377 * config/nds32/nds32-protos.h
6378 (nds32_symbol_load_store_p): Declare function.
6379 (make_pass_nds32_relax_opt): Declare new rtl pass function.
6380 * config/nds32/nds32.c
6381 (nds32_register_pass): New function to register pass.
6382 (nds32_register_passes): New function to register passes.
6383 * config/nds32/nds32.md (relax_group): New pattern.
6384 * config/nds32/nds32.opt (mrelax-hint): New option.
6385 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
6387 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
6389 * config/nds32/t-nds32: Modify files dependency.
6391 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
6393 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
6394 (PROFILE_HOOK): Define its implementation.
6396 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
6398 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
6399 type and 32-bit size.
6401 2018-04-01 Jakub Jelinek <jakub@redhat.com>
6404 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
6405 (V_128_256): New mode iterator.
6406 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
6407 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
6408 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
6410 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
6411 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
6413 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
6416 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
6417 NaN inputs correctly.
6419 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
6422 * config/rs6000/vsx.md (??r): New mode attribute.
6423 (*vsx_mov<mode>_64bit): Use it.
6424 (*vsx_mov<mode>_32bit): Likewise.
6426 2018-03-30 Martin Sebor <msebor@redhat.com>
6428 PR tree-optimization/84818
6429 * builtins.c (check_access): Use warning_n.
6431 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6434 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
6436 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
6439 2018-03-30 Julia Koval <julia.koval@intel.com>
6442 * x86-tune.def (movx, partial_reg_dependency): Enable for
6445 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
6448 * lra-constraints.c (process_alt_operands): Move setting
6449 this_alternative_matches below.
6451 2018-03-29 Martin Liska <mliska@suse.cz>
6454 * doc/invoke.texi: Document how LTO works with debug info.
6455 Describe auto-load support of binutils. Mention 'x86-64'
6456 as valid option value of -march option.
6458 2018-03-29 Jakub Jelinek <jakub@redhat.com>
6460 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
6463 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
6464 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
6465 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
6468 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
6471 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
6472 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
6473 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
6474 for RS6000_BTM_POWERPC64.
6475 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
6476 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
6477 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
6482 2018-03-28 Carl Love <cel@us.ibm.com>
6485 2017-09-27 Carl Love <cel@us.ibm.com>
6487 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
6488 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
6489 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
6492 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6494 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
6495 instead of __vector bool.
6496 (_mm_max_pu8): Likewise.
6497 (_mm_min_pi16): Likewise.
6499 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
6502 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
6503 (DIVWEUO): Likewise.
6505 (DIVDEUO): Likewise.
6506 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
6507 DIVWEUO and DIVDEUO.
6508 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
6509 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
6510 (div_extend): Likewise.
6511 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
6513 (__builtin_divweuo): Likewise.
6514 (__builtin_divdeo): Likewise.
6515 (__builtin_divdeuo): Likewise.
6517 2018-03-28 Jakub Jelinek <jakub@redhat.com>
6520 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
6521 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
6523 PR tree-optimization/82004
6524 * gimple-match-head.c (optimize_pow_to_exp): New function.
6525 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
6526 Don't fold to exp if optimize_pow_to_exp is false.
6528 2018-03-28 Martin Liska <mliska@suse.cz>
6531 * calls.c (initialize_argument_information): Fix trailing space.
6532 * common.opt: Fix typo and provide better explanation for
6533 -fsanitize-coverage option.
6534 * config/i386/i386.opt: Fix typo.
6536 2018-03-28 Jakub Jelinek <jakub@redhat.com>
6537 Martin Liska <mliska@suse.cz>
6540 * gimplify.c (asan_poison_variable): Don't do the check for
6541 gimplify_omp_ctxp here.
6542 (gimplify_decl_expr): Do it here.
6543 (gimplify_target_expr): Likewise.
6545 2018-03-28 Martin Liska <mliska@suse.cz>
6548 * config/i386/i386.c (ix86_function_arg_advance): Do not call
6549 chkp_type_bounds_count if MPX is not enabled.
6551 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
6553 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
6555 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
6558 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
6559 function to create the function decl for complex long double
6560 multiply and divide for -mabi=ieeelongdouble.
6561 (init_float128_ieee): Call it.
6563 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
6566 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
6567 -fcf-protection=branch -mibt.
6568 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
6570 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6573 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
6575 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
6578 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
6579 extern array declarations.
6581 2018-03-27 Richard Biener <rguenther@suse.de>
6584 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
6585 explicit single_use checks.
6587 2018-03-27 Richard Biener <rguenther@suse.de>
6589 PR tree-optimization/85082
6590 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
6593 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6595 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
6596 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
6597 Turn on fasynchronous-unwind-tables and funwind-tables.
6599 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
6602 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
6603 (*bmi_blsr_<mode>_ccz): Ditto.
6605 2018-03-26 Tom de Vries <tom@codesourcery.com>
6607 PR tree-optimization/85063
6608 * omp-general.c (offloading_function_p): New function. Factor out
6610 * omp-offload.c (pass_omp_target_link::gate): ... here.
6611 * omp-general.h (offloading_function_p): Declare.
6612 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
6613 with attribute omp declare target for offloading functions.
6615 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
6617 PR tree-optimization/84005
6618 * tree-data-ref.h (get_base_for_alignment): Declare.
6619 * tree-data-ref.c (get_base_for_alignment_1): New function.
6620 (get_base_for_alignment): Likewise.
6621 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
6622 get_base_for_alignment to find a suitable base object, instead
6623 of always using drb->base_address.
6625 2018-03-23 Jakub Jelinek <jakub@redhat.com>
6628 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
6629 known size by default.
6631 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
6634 * lra-constraints.c (process_alt_operands): Don't match BLKmode
6635 and non BLKmode operands.
6637 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6640 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
6641 Clean up attributes.
6643 2018-03-23 Richard Biener <rguenther@suse.de>
6646 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
6647 we are going to emit early debug for LTO.
6649 2018-03-23 Jakub Jelinek <jakub@redhat.com>
6652 * function.c (match_asm_constraints_1): Don't optimize if input
6653 doesn't satisfy general_operand predicate for output's mode.
6656 * alias.c (write_dependence_p): Don't require for x_canonicalized
6657 non-VOIDmode if x has VOIDmode.
6660 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
6661 just don't try to optimize it rather than assert it never happens.
6663 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
6665 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
6666 macro expansions for definition of ST_INTERNAL_<mode> and
6667 LD_INTERNAL_<mode> builtins.
6668 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
6670 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
6672 (altivec_expand_st_builtin): Likewise.
6673 (altivec_expand_builtin): Remove calls to deleted functions.
6674 (rs6000_address_for_altivec): Delete this function.
6675 * config/rs6000/vector.md: Remove expands for
6676 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
6678 2018-03-22 Sudakshina Das <sudi.das@arm.com>
6681 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
6682 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
6683 re-computing once computed.
6684 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
6685 (arm_init_machine_status): Initialize
6686 machine->static_chain_stack_bytes.
6688 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
6691 * doc/extend.texi: Add four new prototypes for vec_ld.
6692 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
6693 definitions for more logical presentation.
6694 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
6695 entries for V1TI variants of __builtin_altivec_ld builtin.
6696 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
6697 handling of V1TI variant of LVX icode pattern.
6698 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
6699 (rs6000_gimple_fold_builtin): Likewise.
6700 (altivec_init_builtins): Add code to define
6701 __builtin_altivec_lvx_v1ti function.
6703 2018-03-22 Jakub Jelinek <jakub@redhat.com>
6706 * function.c (match_asm_constraints_1): Don't do the optimization
6707 if input isn't a REG, SUBREG, MEM or constant.
6709 2018-03-22 Tom de Vries <tom@codesourcery.com>
6711 PR tree-optimization/84956
6712 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
6713 bb_has_abnormal_pred.
6715 2018-03-22 Jakub Jelinek <jakub@redhat.com>
6718 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
6719 DECL_INITIAL (decl) to decl at the end.
6720 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
6723 2018-03-21 Joseph Myers <joseph@codesourcery.com>
6725 * doc/extend.texi (__builtin_tgmath): Document when complex
6726 integer types are treated as _Complex _Float64.
6728 2018-03-21 Tom de Vries <tom@codesourcery.com>
6730 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
6732 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6734 PR tree-optimization/84960
6735 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
6736 if it is ENTRY block, move them into single succ of ENTRY in that case.
6738 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
6740 PR tree-optimization/84811
6741 * poly-int.h (poly_span_traits): Remove the T3 parameter and
6742 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
6743 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
6744 (known_subrange_p): Update accordingly. Cast each value involved
6745 in the size comparison, rather than casting the result of the
6748 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6750 PR tree-optimization/84982
6751 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
6752 by flipping the least significant bit rather than all bits from
6753 bitpos to bitpos + bitsize - 1.
6755 2018-03-21 Nathan Sidwell <nathan@acm.org>
6757 * doc/extend.texi (Deprecated Features): Remove mention of
6758 long-deleted deprecations.
6760 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6763 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
6764 * configure: Regenerate.
6766 2018-03-21 Tom de Vries <tom@codesourcery.com>
6768 PR tree-optimization/83126
6769 * tree-parloops.c (num_phis): New function.
6770 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
6772 2018-03-21 Nathan Sidwell <nathan@acm.org>
6774 * doc/extend.texi (Deprecated Features): Update deprecated flags,
6775 mention anon-struct/union members and trailing attributes.
6777 2018-03-21 Bin Cheng <bin.cheng@arm.com>
6779 PR tree-optimization/84969
6780 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
6781 builtin memset partitions if they set different rhs values.
6783 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6785 PR rtl-optimization/84989
6786 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
6787 VEC_DUPLICATE with scalar result mode.
6789 2018-03-21 Martin Liska <mliska@suse.cz>
6792 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
6793 not intended return statement.
6795 2018-03-21 Martin Liska <mliska@suse.cz>
6798 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
6799 (chkp_find_bound_slots_1): Limit number of iterations.
6801 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
6804 * Minor grammar fixes for x86 options.
6806 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6809 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
6810 holding REG_CFA_RESTORE notes, instead turn them into a USE.
6812 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
6815 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
6816 (altivec_lvx_<mode>_1op): Likewise.
6817 (altivec_stvx_<mode>_2op): Likewise.
6818 (altivec_stvx_<mode>_1op): Likewise.
6819 (altivec_lvx_<VM2:mode>): New define_expand.
6820 (altivec_stvx_<VM2:mode>): Likewise.
6821 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
6822 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6823 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
6824 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6825 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
6826 (rs6000_gen_lvx): Likewise.
6827 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
6828 (altivec_expand_stv_builtin): Likewise.
6829 (altivec_expand_builtin): Likewise.
6830 * config/rs6000/vector.md: Likewise.
6832 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6835 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
6838 2018-03-20 Richard Biener <rguenther@suse.de>
6841 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
6842 sign-conversions as zero, fall back to standard scalar_stmt
6845 2018-03-20 Martin Liska <mliska@suse.cz>
6848 * predict.c (rebuild_frequencies): Handle case when we have
6849 PROFILE_ABSENT, but flag_guess_branch_prob is false.
6851 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6854 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
6855 flag_section_anchors.
6856 * varasm.c (use_blocks_for_decl_p): Remove hack for
6857 dw2_force_const_mem.
6860 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
6862 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
6863 be created, use lowpart_subreg of operands[0] rather than operands[0]
6865 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
6866 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
6867 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
6868 and n constraint instead of aarch64_shift_imm_di and Usd.
6869 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
6870 (*aarch64_<optab>_reg_minus<mode>3): ... this.
6872 2018-03-20 Sudakshina Das <sudi.das@arm.com>
6875 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
6876 to favor GPR over NEON registers.
6877 (<shift>di3_neon): Likewise.
6879 2018-03-20 Tom de Vries <tom@codesourcery.com>
6882 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
6883 (nvptx_process_pars): Emit bar.sync asap and alap.
6885 2018-03-20 Tom de Vries <tom@codesourcery.com>
6888 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
6889 seen_label if seen_label is already set.
6891 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6894 * config/i386/i386.c (fold_builtin_cpu): For features above 31
6895 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
6896 Use 1U instead of 1. Formatting fixes.
6899 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
6900 instead of TREE_TYPE (s1) for the return value.
6902 2018-03-19 Jakub Jelinek <jakub@redhat.com>
6904 PR tree-optimization/84946
6905 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
6906 bitsize + bitsize in poly_uint64 rather than poly_int64.
6909 * dwarf2asm.c: Include fold-const.c.
6910 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
6911 of decl rather than decl itself.
6913 PR rtl-optimization/84643
6914 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
6916 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
6919 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
6920 calling assemble_variable.
6922 2018-03-19 Sudakshina Das <sudi.das@arm.com>
6925 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
6926 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
6928 2018-03-19 Jim Wilson <jimw@sifive.com>
6931 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
6932 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
6933 (riscv_first_stack_step): Likewise.
6934 (riscv_option_override): Use STACK_BOUNDARY instead of
6936 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
6938 (BIGGEST_ALIGNMENT): Set to 128.
6939 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
6940 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
6943 2018-03-19 Richard Biener <rguenther@suse.de>
6945 PR tree-optimization/84933
6946 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
6947 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
6949 2018-03-19 Richard Biener <rguenther@suse.de>
6951 PR tree-optimization/84859
6952 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
6953 (cond_if_else_store_replacement): Perform sinking operation on
6954 single-store BBs regardless of MAX_STORES_TO_SINK setting.
6955 Generalize what a BB with a single eligible store is.
6957 2018-03-19 Richard Biener <rguenther@suse.de>
6959 PR tree-optimization/84929
6960 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
6961 chrec_is_positive against non-chrec arg.
6963 2018-03-19 Tamar Christina <tamar.christina@arm.com>
6966 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
6968 2018-03-18 Martin Liska <mliska@suse.cz>
6970 PR rtl-optimization/84635
6971 * regrename.c (build_def_use): Use matches_mode only when
6974 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
6976 PR tree-optimization/84913
6977 * tree-vect-loop.c (vectorizable_reduction): Don't try to
6978 vectorize chains of COND_EXPRs.
6980 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6982 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
6984 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6986 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
6988 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6990 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
6992 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6993 Kito Cheng <kito.cheng@gmail.com>
6995 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
6996 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
6997 (nds32_adjust_reg_alloc_order): New function.
6998 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
7000 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
7002 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
7003 nds32_print_operand, nds32_print_operand_address): Use
7004 HOST_WIDE_INT_PRINT_DEC instead.
7006 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
7008 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
7010 2018-03-17 Jakub Jelinek <jakub@redhat.com>
7013 * config/i386/i386.c (initial_ix86_tune_features,
7014 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
7016 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
7017 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
7018 rather than 1u << ix86_tune. Formatting fix.
7019 (ix86_option_override_internal): Change ix86_arch_mask from
7020 unsigned int to unsigned HOST_WIDE_INT, initialize to
7021 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
7022 (ix86_function_specific_restore): Likewise.
7024 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7027 * postreload.c (reload_combine_recognize_pattern): Perform
7028 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
7029 truncate_int_for_mode the result for the destination's mode.
7032 * hsa-gen.c (mem_type_for_type): Fix comment typo.
7033 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
7035 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
7038 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
7041 * lra-assigns.c (lra_split_hard_reg_for): Don't use
7042 regno_allocno_class_array and sorted_pseudos.
7043 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
7044 insns where regno is used.
7046 2018-03-16 Martin Liska <mliska@suse.cz>
7049 * multiple_target.c (create_dispatcher_calls): Redirect
7050 reference in the symbol table.
7052 2018-03-16 Martin Liska <mliska@suse.cz>
7055 * multiple_target.c (create_dispatcher_calls): Redirect also
7058 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7062 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
7063 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
7064 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
7066 2018-03-16 Julia Koval <julia.koval@intel.com>
7068 * doc/invoke.texi (Skylake Server): Add CLWB.
7069 Cannonlake): Remove CLWB.
7071 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7073 PR tree-optimization/84841
7074 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
7076 (FLOAT_ONE_CONST_TYPE): Define.
7077 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
7078 (sort_by_operand_rank): Put entries with higher constant_type last
7079 rather than first to match comments.
7081 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
7083 * config/nios2/nios2.md (movsi_internal): Fix thinko in
7086 2018-03-15 Jakub Jelinek <jakub@redhat.com>
7089 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
7090 check and use address of target always.
7092 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
7095 * config/i386/i386.c (indirect_thunk_needed): Update comments.
7096 (indirect_thunk_bnd_needed): Likewise.
7097 (indirect_thunks_used): Likewise.
7098 (indirect_thunks_bnd_used): Likewise.
7099 (indirect_return_needed): New.
7100 (indirect_return_bnd_needed): Likewise.
7101 (output_indirect_thunk_function): Add a bool argument for
7103 (output_indirect_thunk_function): Don't generate alias for
7104 function return thunk.
7105 (ix86_code_end): Call output_indirect_thunk_function to generate
7106 function return thunks.
7107 (ix86_output_function_return): Set indirect_return_bnd_needed
7108 and indirect_return_needed instead of indirect_thunk_bnd_needed
7109 and indirect_thunk_needed.
7111 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
7113 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
7114 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
7115 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
7117 2018-03-15 David Malcolm <dmalcolm@redhat.com>
7118 Paul Hua <paul.hua.gm@gmail.com>
7121 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
7123 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
7125 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
7126 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
7129 2018-03-15 Tamar Christina <tamar.christina@arm.com>
7132 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
7133 instead of GET_MODE_SIZE when comparing Units.
7135 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
7138 * varasm.c (hash_section): Return an unchangeble hash value
7139 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
7140 Return !aarch64_can_use_per_function_literal_pools_p ().
7142 2018-03-15 Jakub Jelinek <jakub@redhat.com>
7145 * optabs.c (emit_conditional_move): Pass address of cmode's copy
7146 rather than address of cmode as last argument to prepare_cmp_insn.
7148 2018-03-15 Julia Koval <julia.koval@intel.com>
7150 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
7151 F_AVX512VNNI, F_AVX512BITALG): New.
7153 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
7156 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
7157 insn for floating-point loads and stores.
7159 2018-03-14 Carl Love <cel@us.ibm.com>
7161 * config/rs6000/rs6000-c.c: Add macro definitions for
7162 ALTIVEC_BUILTIN_VEC_PERMXOR.
7163 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
7164 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
7165 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
7166 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
7168 * config/doc/extend.texi: Add prototypes for vec_permxor.
7170 2018-03-14 David Malcolm <dmalcolm@redhat.com>
7173 * diagnostic-show-locus.c (class layout_point): Convert m_line
7174 from int to linenum_type.
7175 (line_span::comparator): Use linenum "compare" function when
7176 comparing line numbers.
7177 (test_line_span): New function.
7178 (layout_range::contains_point): Convert param "row" from int to
7180 (layout_range::intersects_line_p): Likewise.
7181 (layout::will_show_line_p): Likewise.
7182 (layout::print_source_line): Likewise.
7183 (layout::should_print_annotation_line_p): Likewise.
7184 (layout::print_annotation_line): Likewise.
7185 (layout::print_leading_fixits): Likewise.
7186 (layout::annotation_line_showed_range_p): Likewise.
7187 (struct line_corrections): Likewise for field m_row.
7188 (line_corrections::line_corrections): Likewise for param "row".
7189 (layout::print_trailing_fixits): Likewise.
7190 (layout::get_state_at_point): Likewise.
7191 (layout::get_x_bound_for_row): Likewise.
7192 (layout::print_line): Likewise.
7193 (diagnostic_show_locus): Likewise for locals "last_line" and
7195 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
7196 * input.c (selftest::test_linenum_comparisons): New function.
7197 (selftest::input_c_tests): Call it.
7198 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
7199 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
7200 * selftest.h (ASSERT_GT): New macro.
7201 (ASSERT_GT_AT): New macro.
7202 (ASSERT_LT): New macro.
7203 (ASSERT_LT_AT): New macro.
7205 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
7207 PR rtl-optimization/84780
7208 * combine.c (distribute_links): Don't make a link based on pc_rtx.
7210 2018-03-14 Martin Liska <mliska@suse.cz>
7212 * tree.c (record_node_allocation_statistics): Use
7213 get_stats_node_kind.
7214 (get_stats_node_kind): New function extracted from
7215 record_node_allocation_statistics.
7216 (free_node): Use get_stats_node_kind.
7218 2018-03-14 Richard Biener <rguenther@suse.de>
7220 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
7221 that the value-set of ANTIC_IN doesn't grow.
7224 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7226 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7227 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7229 2018-03-14 Julia Koval <julia.koval@intel.com>
7231 * config.gcc (icelake-client, icelake-server): New.
7233 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
7234 (initial_ix86_arch_features): Ditto.
7235 (PTA_SKYLAKE): Add SGX.
7236 (PTA_ICELAKE): Remove.
7237 (PTA_ICELAKE_CLIENT): New.
7238 (PTA_ICELAKE_SERVER): New.
7239 (ix86_option_override_internal): Split up icelake on icelake client and
7241 (get_builtin_code_for_version): Ditto.
7242 (fold_builtin_cpu): Ditto.
7243 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
7244 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
7245 * config/i386/i386.h (processor_type): Ditto.
7246 * doc/invoke.texi: Ditto.
7248 2018-03-14 Jakub Jelinek <jakub@redhat.com>
7251 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
7252 INTEGER_CST offset, add it together with bitpos / 8 and
7253 sign extend based on POINTER_SIZE.
7257 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
7260 * config/i386/constraints.md (Yc): New register constraint.
7261 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
7262 Use Yc constraint for alternative 2 of operand 0. Remove
7263 preferred_for_speed attribute.
7265 2018-03-14 Richard Biener <rguenther@suse.de>
7267 PR tree-optimization/84830
7268 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
7269 with the old one to avoid oscillations.
7271 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
7274 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
7276 (assign_by_spills): Return a flag of reload assignment failure.
7277 Do not process the reload assignment failures. Do not spill other
7278 reload pseudos if they has the same reg class. Update n if
7280 (lra_assign): Add a return arg. Set up from the result of
7281 assign_by_spills call.
7282 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
7283 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
7284 usage_insns if it is not NULL.
7285 (spill_hard_reg_in_range): New function.
7286 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
7287 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
7288 function prototypes.
7289 (lra_assign): Change prototype.
7290 * lra.c (lra): Add code to deal with fails by splitting hard reg
7293 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
7295 * config/riscv/riscv.opt (mrelax): New option.
7296 * config/riscv/riscv.c (riscv_file_start): Emit ".option
7297 "norelax" when riscv_mrelax is disabled.
7298 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
7300 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7303 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
7304 reassociation for int modes.
7306 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
7308 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
7309 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
7311 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
7312 * config/aarch64/aarch64-sve.md
7313 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
7314 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
7315 (*extend<mode><Vwide>2): Rename to...
7316 (aarch64_sve_extend<mode><Vwide>2): ...this.
7317 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
7318 renaming the old pattern to...
7319 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
7321 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
7322 define_expand, renaming the old pattern to...
7323 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
7324 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
7325 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
7326 account when deciding which SVE instruction the optab should use.
7327 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
7329 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
7331 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
7332 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
7333 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
7334 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
7335 (tlsdesc_small_<mode>): Turn a define_expand and use
7336 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
7337 (tlsdesc_small_advsimd_<mode>): ...this.
7338 (tlsdesc_small_sve_<mode>): New pattern.
7340 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
7342 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
7343 (UNSPEC_UMUL_HIGHPART): New constants.
7344 (MUL_HIGHPART): New int iteraor.
7345 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
7346 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
7348 (*<su>mul<mode>3_highpart): New define_insn.
7350 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
7353 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
7356 2018-03-13 Martin Liska <mliska@suse.cz>
7359 * (sem_item_optimizer::sem_item_optimizer): Initialize new
7361 (sem_item_optimizer::~sem_item_optimizer): Release it.
7362 (sem_item_optimizer::merge_classes): Register variable aliases.
7363 (sem_item_optimizer::fixup_pt_set): New function.
7364 (sem_item_optimizer::fixup_points_to_sets): Likewise.
7365 * ipa-icf.h: Declare new variables and functions.
7367 2018-03-13 Jakub Jelinek <jakub@redhat.com>
7370 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
7371 integer_pow2p@2 and test integer_pow2p in condition.
7372 (A < 0 ? C : 0): Similarly for @1.
7375 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
7376 characters starting at p contain '\0' character, don't look beyond
7380 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
7381 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
7384 * reg-stack.c (change_stack): Change update_end var from int to
7385 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
7386 also call set_block_for_insn on the newly added insns and rescan.
7389 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
7390 on the last operand.
7393 * tree.c (stabilize_reference_1): Return save_expr (e) for
7394 STATEMENT_LIST even if it doesn't have side-effects.
7396 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
7398 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
7400 2018-03-12 Renlin Li <renlin.li@arm.com>
7402 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
7403 aarch64_output_scalar_simd_mov_immediate.
7405 2018-03-12 Martin Sebor <msebor@redhat.com>
7407 PR tree-optimization/83456
7408 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
7409 for perfectly overlapping calls to memcpy.
7410 (gimple_fold_builtin_memory_chk): Same.
7411 (gimple_fold_builtin_strcpy): Handle no-warning.
7412 (gimple_fold_builtin_stxcpy_chk): Same.
7413 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
7415 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
7417 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
7418 parameter. Use it for SFmode.
7419 (rs6000_function_arg_advance_1): Adjust.
7420 (rs6000_function_arg): Adjust.
7421 (rs6000_gimplify_va_arg): Pass false for that new parameter.
7423 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
7425 PR rtl-optimization/84169
7426 PR rtl-optimization/84780
7427 * combine.c (can_combine_p): Check for a 2-insn combination whether
7428 the destination register is used between the two insns, too.
7430 2018-03-12 Richard Biener <rguenther@suse.de>
7432 PR tree-optimization/84803
7433 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
7434 for refs DR analysis didn't process.
7436 2018-03-12 Richard Biener <rguenther@suse.de>
7438 PR tree-optimization/84777
7439 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
7440 force-vectorize loops ignore whether we are optimizing for size.
7442 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
7444 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
7445 (TARGET_MD_ASM_ADJUST): Define.
7447 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
7448 Kito Cheng <kito.cheng@gmail.com>
7449 Chung-Ju Wu <jasonwucj@gmail.com>
7451 * config/nds32/nds32.c (nds32_compute_stack_frame,
7452 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
7453 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
7454 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
7455 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
7456 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
7457 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
7458 * config/nds32/nds32.md (prologue, epilogue): Use macro
7459 NDS32_V3PUSH_AVAILABLE_P to do checking.
7461 2018-03-11 Jakub Jelinek <jakub@redhat.com>
7464 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
7465 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
7466 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
7467 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
7468 addition of most attributes on !orig_type_die or the attribute not
7469 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
7471 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
7472 Chung-Ju Wu <jasonwucj@gmail.com>
7474 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
7476 * config/nds32/nds32.opt (mvh): New option.
7478 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
7479 Chung-Ju Wu <jasonwucj@gmail.com>
7481 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
7483 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
7484 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
7487 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
7488 Chung-Ju Wu <jasonwucj@gmail.com>
7490 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
7492 * config/nds32/nds32-multiple.md (strlensi): New pattern.
7493 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
7495 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
7496 Kito Cheng <kito.cheng@gmail.com>
7497 Chung-Ju Wu <jasonwucj@gmail.com>
7499 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
7500 UNSPEC_FFMISM and UNSPEC_FLMISM.
7501 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
7502 for ffb, ffmism and flmism.
7503 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
7504 (unspec_ffmism): Ditto.
7505 (unspec_flmism): Ditto.
7506 (nds32_expand_builtin_impl): Check if string extension is available.
7507 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
7508 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
7510 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
7513 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
7516 * lra-assigns.c (assign_by_spills): Return a flag of reload
7517 assignment failure. Do not process the reload assignment
7518 failures. Do not spill other reload pseudos if they has the same
7520 (lra_assign): Add a return arg. Set up from the result of
7521 assign_by_spills call.
7522 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
7523 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
7524 usage_insns if it is not NULL.
7525 (spill_hard_reg_in_range): New function.
7526 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
7527 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
7528 function prototypes.
7529 (lra_assign): Change prototype.
7530 * lra.c (lra): Add code to deal with fails by splitting hard reg
7533 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
7536 * config/i386/i386.opt: Replace Enforcment with Enforcement.
7538 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
7541 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
7542 (dw_val_node): Add val_symbolic_view.
7543 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
7544 (symview_upper_bound): New.
7545 (new_line_info_table): Initialize symviews_since_reset.
7546 (dwarf2out_source_line): Count symviews_since_reset and set
7547 symview_upper_bound.
7548 (dw_val_equal_p): Handle symview.
7549 (add_AT_symview): New.
7550 (print_dw_val): Handle symview.
7551 (attr_checksum, attr_checksum_ordered): Likewise.
7552 (same_dw_val_p, size_of_die): Likewise.
7553 (value_format, output_die): Likewise.
7554 (add_high_low_attributes): Use add_AT_symview for entry_view.
7555 (dwarf2out_finish): Reset symview_upper_bound, clear
7558 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
7561 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
7562 Add strict argument and use it.
7563 (rs6000_split_multireg_move): Update for new strict argument.
7564 (mem_operand_gpr): Disallow all non-offsettable addresses.
7565 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
7567 2018-03-09 Jakub Jelinek <jakub@redhat.com>
7570 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
7571 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
7572 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
7575 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
7576 decl, use remap_type if we want to use the type.
7578 2018-03-09 Martin Sebor <msebor@redhat.com>
7580 PR tree-optimization/84526
7581 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
7583 (builtin_access::generic_overlap): Be prepared to handle non-array
7586 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
7588 PR rtl-optimization/84682
7589 * lra-constraints.c (process_address_1): Check is_address flag
7590 for address constraints.
7591 (process_alt_operands): Likewise.
7592 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
7593 preprocess_constraints.
7594 * recog.h (preprocess_constraints): Add oploc parameter.
7596 * recog.c (preprocess_constraints): Test address_operand for
7597 CT_ADDRESS constraints.
7599 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
7602 * lra-assigns.c (assign_by_spills): Return a flag of reload
7603 assignment failure. Do not process the reload assignment
7604 failures. Do not spill other reload pseudos if they has the same
7606 (lra_assign): Add a return arg. Set up from the result of
7607 assign_by_spills call.
7608 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
7609 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
7610 usage_insns if it is not NULL.
7611 (spill_hard_reg_in_range): New function.
7612 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
7613 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
7614 function prototypes.
7615 (lra_assign): Change prototype.
7616 * lra.c (lra): Add code to deal with fails by splitting hard reg
7619 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7622 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
7623 Accept complain bool parameter. Only emit errors if it is true.
7624 (arm_parse_cpu_option_name): Likewise.
7625 (arm_target_thumb_only): Adjust callers of the above.
7626 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
7627 prototype to take a default true bool parameter.
7628 (arm_parse_arch_option_name): Likewise.
7630 2018-03-09 David Malcolm <dmalcolm@redhat.com>
7631 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
7635 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
7636 * configure: Regenerate.
7637 * configure.ac ("linker --version-script option"): New.
7638 ("linker soname option"): New.
7640 2018-03-09 Richard Biener <rguenther@suse.de>
7642 PR tree-optimization/84775
7643 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
7644 immediate uses of predicate stmts and mark them modified.
7647 PR tree-optimization/84178
7648 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
7650 (version_loop_for_if_conversion): Delay update_ssa call.
7651 (tree_if_conversion): Delay update_ssa until after predicate
7654 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
7657 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
7658 when the function accesses prior frames.
7660 2018-03-08 Jakub Jelinek <jakub@redhat.com>
7663 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
7664 gen_llsym, otherwise call maybe_gen_llsym.
7667 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
7668 has ',' character inside of it.
7670 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7673 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
7674 as clobbering CC_REGNUM.
7676 2018-03-08 Richard Biener <rguenther@suse.de>
7679 * tree-scalar-evolution.c: Include tree-into-ssa.h.
7680 (follow_copies_to_constant): Do not follow SSA names registered
7683 2018-03-08 Richard Biener <rguenther@suse.de>
7685 PR tree-optimization/84178
7686 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
7688 (version_loop_for_if_conversion): Delay update_ssa call.
7689 (tree_if_conversion): Delay update_ssa until after predicate
7692 2018-03-08 David Malcolm <dmalcolm@redhat.com>
7694 PR tree-optimization/84178
7695 * tree-if-conv.c (release_bb_predicate): Remove the
7696 the assertion that the stmts have NULL use_ops.
7697 Discard the statements, asserting that they haven't
7698 yet been added to a BB.
7700 2018-03-08 Richard Biener <rguenther@suse.de>
7702 PR tree-optimization/84746
7703 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
7704 (phi_translate): Pass in destination ANTIC_OUT set.
7705 (phi_translate_1): Likewise. For a simplified result lookup
7706 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
7707 (phi_translate_set): Adjust.
7708 (do_pre_regular_insertion): Likewise.
7709 (do_pre_partial_partial_insertion): Likewise.
7711 2018-03-08 Martin Liska <mliska@suse.cz>
7713 PR gcov-profile/84735
7714 * doc/gcov.texi: Document usage of profile files.
7715 * gcov-io.h: Document changes in the format.
7717 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
7721 * dwarf2out.c (struct dw_line_info_table): Update comments for
7723 (FORCE_RESET_NEXT_VIEW): New.
7724 (FORCE_RESETTING_VIEW_P): New.
7725 (RESETTING_VIEW_P): Check for -1 too.
7726 (ZERO_VIEW_P): Likewise.
7727 (new_line_info_table): Force-reset next view.
7728 (dwarf2out_begin_function): Likewise.
7729 (dwarf2out_source_line): Simplify zero_view_p initialization.
7730 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
7731 view directly. Omit view when omitting .loc at line 0.
7733 2018-03-08 Jakub Jelinek <jakub@redhat.com>
7735 PR tree-optimization/84740
7736 * tree-switch-conversion.c (process_switch): Call build_constructors
7737 only if info.phi_count is non-zero.
7739 PR tree-optimization/84739
7740 * tree-tailcall.c (find_tail_calls): Check call arguments against
7741 DECL_ARGUMENTS (current_function_decl) rather than
7742 DECL_ARGUMENTS (func) when checking for tail recursion.
7744 2018-03-07 Jakub Jelinek <jakub@redhat.com>
7746 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
7747 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
7748 Volker Reichelt's entry and add entries for people that perform
7749 GCC fuzzy testing and report numerous bugs.
7751 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
7754 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
7755 readonly data in sdata, if that is disabled.
7756 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
7757 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
7758 -mreadonly-in-sdata option.
7760 2018-03-07 Martin Sebor <msebor@redhat.com>
7762 PR tree-optimization/84468
7763 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
7764 basic block when looking for nul assignment.
7766 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
7769 * except.h (output_function_exception_table): Adjust prototype.
7770 * except.c (output_function_exception_table): Remove FNNAME parameter
7771 and add SECTION parameter. Ouput one part of the table at a time.
7772 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
7773 the first part of the exception table and emit unwind directives.
7774 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
7775 (i386_pe_seh_cold_init): Likewise.
7776 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
7777 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
7778 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
7779 (ix86_output_call_insn): Emit a nop in one more case for SEH.
7780 * config/i386/winnt.c: Include except.h.
7781 (struct seh_frame_state): Add reg_offset, after_prologue and
7782 in_cold_section fields.
7783 (i386_pe_seh_end_prologue): Set seh->after_prologue.
7784 (i386_pe_seh_cold_init): New function.
7785 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
7786 to seh->in_cold_section.
7787 (seh_emit_push): Record the offset of the push.
7788 (seh_emit_save): Record the offet of the save.
7789 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
7790 Test seh->after_prologue to disregard the epilogue.
7791 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
7792 (i386_pe_end_cold_function): New function.
7794 2018-03-07 Jakub Jelinek <jakub@redhat.com>
7797 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
7798 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
7801 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
7803 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
7804 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
7807 * multiple_target.c: Include tree-inline.h and intl.h.
7808 (expand_target_clones): Diagnose and fail if node->definition and
7809 !tree_versionable_function_p (node->decl).
7811 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
7813 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
7815 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
7816 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
7817 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
7819 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7822 * combine.c (try_combine): Use reg_or_subregno instead of handling
7823 just paradoxical SUBREGs and REGs.
7825 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
7827 * config/arc/arc.c (arc_finalize_pic): Remove function.
7828 (arc_must_save_register): We use single base PIC register, remove
7829 checks to save/restore the PIC register.
7830 (arc_expand_prologue): Likewise.
7831 * config/arc/arc-protos.h (arc_set_default_type_attributes):
7833 (arc_verify_short): Likewise.
7834 (arc_attr_type): Likewise.
7835 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
7836 (walk_stores): Likewise.
7837 (arc_address_cost): Make it static.
7838 (arc_verify_short): Likewise.
7839 (branch_dest): Likewise.
7840 (arc_attr_type): Likewise.
7841 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
7842 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
7843 (arc_final_prescan_insn): Remove inserting the nops due to
7844 hardware hazards. It is done in reorg step.
7845 (insn_length_variant_t): Remove.
7846 (insn_length_parameters_t): Likewise.
7847 (arc_insn_length_parameters): Likewise.
7848 (arc_get_insn_variants): Likewise.
7849 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
7851 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7854 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
7857 PR tree-optimization/84687
7858 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
7860 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
7862 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7864 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
7865 Rename to ppc_speculation_barrier.
7866 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
7867 __builtin_ppc_speculation_barrier.
7869 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7872 * combine.c (combine_simplify_rtx): Don't try to simplify if
7873 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
7876 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
7878 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
7879 to 32 bytes when compiling for POWER9.
7881 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7884 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
7885 regparm >= 3 with no arg reg available also for calls with
7886 flag_force_indirect_call. Pass decl to ix86_function_regparm.
7889 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
7891 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
7893 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
7896 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
7898 2018-03-05 Richard Biener <rguenther@suse.de>
7900 PR tree-optimization/84486
7901 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
7902 When inserting a __builtin_assume_aligned call set the LHS
7903 SSA name alignment info accordingly.
7905 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
7907 PR tree-optimization/84114
7908 * config/aarch64/aarch64.c (aarch64_reassociation_width)
7909 Avoid reassociation of FLOAT_MODE addition.
7911 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
7913 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
7914 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
7915 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
7916 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
7917 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
7918 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
7919 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
7921 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
7922 __builtin_ia32_wbinvd): New builtins.
7923 (SPECIAL_ARGS2): New.
7924 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
7925 (SPECIAL_ARGS2): New.
7926 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
7927 (ix86_valid_target_attribute_inner_p): Ditto.
7928 (ix86_init_mmx_sse_builtins): Add special_args2.
7929 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
7930 TARGET_WBNOINVD_P): New.
7931 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
7932 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
7933 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
7934 * config/i386/immintrin.h (_wbinvd): New intrinsic.
7935 * config/i386/pconfigintrin.h: New file.
7936 * config/i386/wbnoinvdintrin.h: Ditto.
7937 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
7938 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
7940 2018-03-05 Richard Biener <rguenther@suse.de>
7942 PR tree-optimization/84670
7943 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7945 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7946 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7947 (compute_antic_aux): Only assert the number of values in ANTIC_IN
7948 doesn't grow if all successors (recursively) were visited at least
7951 2018-03-05 Richard Biener <rguenther@suse.de>
7953 PR tree-optimization/84650
7954 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
7955 if executed in the loop pipeline.
7957 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
7959 * doc/configfiles.texi (Configuration Files): Move info about
7960 conditionalizing $target-protos.h to...
7961 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
7962 differs from $target-protos.h.
7964 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
7965 Chung-Ju Wu <jasonwucj@gmail.com>
7967 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
7968 * config/nds32/nds32-multiple.md (setmemsi): Define.
7969 * config/nds32/nds32-memory-manipulation.c
7970 (nds32_gen_dup_4_byte_to_word_value): New.
7971 (emit_setmem_word_loop): New.
7972 (emit_setmem_byte_loop): New.
7973 (nds32_expand_setmem_loop): New.
7974 (nds32_expand_setmem_loop_v3m): New.
7975 (nds32_expand_setmem_unroll): New.
7976 (nds32_expand_setmem): New.
7978 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7979 Chung-Ju Wu <jasonwucj@gmail.com>
7981 * config/nds32/nds32-memory-manipulation.c
7982 (nds32_emit_load_store): New.
7983 (nds32_emit_post_inc_load_store): New.
7984 (nds32_emit_mem_move): New.
7985 (nds32_emit_mem_move_block): New.
7986 (nds32_expand_movmemsi_loop_unknown_size): New.
7987 (nds32_expand_movmemsi_loop_known_size): New.
7988 (nds32_expand_movmemsi_loop): New.
7989 (nds32_expand_movmemsi_unroll): New.
7990 (nds32_expand_movmemqi): Rename ...
7991 (nds32_expand_movmemsi): ... to this.
7992 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
7993 (movmemsi): ... to this.
7994 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
7995 (nds32_expand_movmemsi): ... to this.
7997 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7998 Monk Chiang <sh.chiang04@gmail.com>
7999 Chung-Ju Wu <jasonwucj@gmail.com>
8001 * config/nds32/nds32-protos.h
8002 (nds32_expand_load_multiple): New arguments.
8003 (nds32_expand_store_multiple): Ditto.
8004 (nds32_valid_multiple_load_store): Rename ...
8005 (nds32_valid_multiple_load_store_p): ... to this.
8006 * config/nds32/nds32-memory-manipulation.c
8007 (nds32_expand_load_multiple): Refine implementation.
8008 (nds32_expand_store_multiple): Ditto.
8009 * config/nds32/nds32-multiple.md
8010 (load_multiple): Update nds32_expand_load_multiple interface.
8011 (store_multiple): Update nds32_expand_store_multiple interface.
8012 * config/nds32/nds32-predicates.c
8013 (nds32_valid_multiple_load_store): Rename ...
8014 (nds32_valid_multiple_load_store_p): ... to this and refine
8016 * config/nds32/predicates.md
8017 (nds32_load_multiple_and_update_address_operation): New predicate.
8018 (nds32_store_multiple_and_update_address_operation): New predicate.
8020 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8021 Chung-Ju Wu <jasonwucj@gmail.com>
8023 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
8024 (combo): New attribute.
8025 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
8027 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
8029 * config/nds32/nds32.opt: Change -mcmodel= default value.
8031 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
8032 Monk Chiang <sh.chiang04@gmail.com>
8033 Chung-Ju Wu <jasonwucj@gmail.com>
8035 * config/nds32/constants.md (unspec_element): New enum.
8036 * config/nds32/constraints.md (Umw): New constraint.
8037 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
8038 * config/nds32/nds32-intrinsic.md: Likewise.
8039 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
8040 (nds32_valid_smw_lwm_base_p): New.
8041 (nds32_output_smw_single_word): New.
8042 (nds32_output_lmw_single_word): New.
8043 (nds32_expand_unaligned_load): New.
8044 (nds32_expand_unaligned_store): New.
8045 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
8046 (nds32_output_smw_single_word): Declare.
8047 (nds32_output_lmw_single_word): Declare.
8048 (nds32_expand_unaligned_load): Declare.
8049 (nds32_expand_unaligned_store): Declare.
8050 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
8051 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
8052 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
8053 NDS32_BUILTIN_UASTORE_DW.
8054 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
8057 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
8058 Kito Cheng <kito.cheng@gmail.com>
8059 Chung-Ju Wu <jasonwucj@gmail.com>
8061 * config/nds32/nds32-intrinsic.c
8062 (nds32_expand_builtin_null_ftype_reg): Delete.
8063 (nds32_expand_builtin_reg_ftype_imm): Ditto.
8064 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
8065 (nds32_read_argument): New.
8066 (nds32_legitimize_target): Ditto.
8067 (nds32_legitimize_argument): Ditto.
8068 (nds32_check_constant_argument): Ditto.
8069 (nds32_expand_unop_builtin): Ditto.
8070 (nds32_expand_unopimm_builtin): Ditto.
8071 (nds32_expand_binop_builtin): Ditto.
8072 (nds32_builtin_decl_impl): Ditto.
8073 (builtin_description): Ditto.
8074 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
8075 (nds32_init_builtins_impl): Ditto.
8076 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
8077 (nds32_builtin_decl): New.
8078 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
8079 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
8081 2018-03-02 Jeff Law <law@redhat.com>
8083 * reorg.c (stop_search_p): Handle DEBUG_INSN.
8084 (redundant_insn, fill_simple_delay_slots): Likewise.
8085 (fill_slots_from_thread): Likewise.
8086 * resource.c (mark_referenced_resources): Likewise.
8087 (mark_set_resources, find_dead_or_set_registers): Likewise.
8089 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8091 * substring-locations.h (format_warning_va): Formatting fix for
8093 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
8095 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
8096 * substring-locations.c: Include intl.h.
8097 (format_warning_va): Turned into small wrapper around
8098 format_warning_n_va, renamed to ...
8099 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
8100 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
8102 (format_warning_at_substring_n): New function.
8103 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
8104 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
8105 format_warning_at_substring with just a shorter name instead of
8106 const function pointer.
8107 (fmtwarn_n): New function.
8108 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
8109 appropriate, get rid of all the fmtstr temporaries, move conditionals
8110 with G_() wrapped string literals directly into fmtwarn arguments,
8111 cast dir.len to (int), formatting fixes.
8113 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
8115 * doc/invoke.texi: Remove "Cilk Plus" references.
8117 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8118 Richard Biener <rguenther@suse.de>
8121 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
8122 for error or warning attributes if CALL_FROM_THUNK_P is set.
8125 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8128 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
8129 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
8132 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
8133 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
8135 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
8136 instead of -1U in last predictors element's probability member.
8138 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
8141 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8142 arguments if they are comparable.
8144 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
8146 PR tree-optimization/84634
8147 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
8148 masks and masked_loop_p with a single loop_masks, making sure it's
8149 null for bb vectorization.
8151 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
8153 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
8154 (vect_analyze_data_ref_access): Use loop->safe_len rather than
8155 loop->force_vectorize to check whether there is no alias.
8157 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8160 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
8162 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
8164 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
8165 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
8166 instead of a loop around prev_real_insn.
8167 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
8171 * config/i386/i386.c (ix86_print_operand): Use conditional
8172 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
8175 2018-03-02 Richard Biener <rguenther@suse.de>
8177 PR tree-optimization/84427
8178 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
8179 (bitmap_set_subtract_values): Rewrite to handle multiple
8182 (prune_clobbered_mems): Likewise.
8183 (phi_translate): Take edge instead of pred/phiblock.
8184 (phi_translate_1): Likewise.
8185 (phi_translate_set): Likewise. Insert all translated
8186 exprs for a value into the set, keeping possibly multiple
8187 expressions per value.
8188 (compute_antic_aux): Adjust for phi_translate changes.
8189 When intersecting union the expressions and prune those
8190 not in the final value set, keeping possibly multiple
8191 expressions per value. Do not use value-insertion
8192 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
8193 all expressions. Add verification that the value-sets
8194 only shrink during iteration.
8195 (compute_partial_antic_aux): Adjust for the phi_translate changes.
8196 (do_pre_regular_insertion): Likewise.
8197 (do_pre_partial_partial_insertion): Likewise.
8199 2018-03-02 Richard Biener <rguenther@suse.de>
8202 * config/darwin.c (saved_debug_info_level): New static global.
8203 (darwin_asm_lto_start): Disable debug info generation for LTO out.
8204 (darwin_asm_lto_end): Restore debug info generation settings.
8206 2018-03-01 Martin Liska <mliska@suse.cz>
8209 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
8212 2018-03-01 Richard Biener <rguenther@suse.de>
8215 * dwarf2out.c (gen_variable_die): Properly handle late VLA
8216 type annotation with LTO when debug was disabled at compile-time.
8218 2018-03-01 Matthew Fortune <mfortune@gmail.com>
8220 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
8222 (mips_final_postscan_insn): Likewise.
8224 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
8226 PR rtl-optimization/84528
8227 * alias.c (init_alias_target): Add commentary.
8228 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
8229 a unique base value if the frame pointer is not eliminated
8230 to the stack pointer.
8232 2018-03-01 Tom de Vries <tom@codesourcery.com>
8234 PR rtl-optimization/83327
8235 * lra-int.h (hard_regs_spilled_into): Declare.
8236 * lra.c (hard_regs_spilled_into): Define.
8237 (init_reg_info): Init hard_regs_spilled_into.
8238 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
8239 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
8240 (process_bb_lives): Handle hard_regs_spilled_into.
8241 (lra_create_live_ranges_1): Before doing liveness propagation, clear
8242 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
8244 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
8246 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
8247 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
8248 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
8249 * config/rs6000/aix72.h: New file.
8251 2018-02-28 Jakub Jelinek <jakub@redhat.com>
8253 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
8254 instead of warning_at with conditional singular and plural messages
8258 * stor-layout.c (update_alignment_for_field): For
8259 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
8260 && !DECL_PACKED (field), do the alignment update, just use
8261 only desired_align instead of MAX (type_align, desired_align)
8263 (place_field): Don't do known_align < desired_align handling
8264 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
8265 is non-NULL, instead do it after rli->prev_field handling and
8266 only if not within a bitfield word. For DECL_PACKED (field)
8267 use type_align of BITS_PER_UNIT.
8269 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
8271 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
8272 superfluous parentheses and trailing spaces.
8274 2018-02-28 Richard Biener <rguenther@suse.de>
8276 PR tree-optimization/84584
8277 * graphite-scop-detection.c (scop_detection::add_scop): Discard
8278 SCoPs with fake exit edge.
8280 2018-02-28 Martin Liska <mliska@suse.cz>
8283 * timevar.c (timer::print): Fix format to properly print 100%
8286 2018-02-28 Richard Biener <rguenther@suse.de>
8289 * genmatch.c (capture_info::walk_match): Do not mark
8290 captured expressions without operands as expr_p given
8291 they act more like predicates and should be subject to
8292 "lost tail" side-effect preserving.
8294 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
8296 PR rtl-optimization/81611
8297 * auto-inc-dec.c (attempt_change): Move dead note from
8298 mem_insn if it's the next use of regno
8299 (find_address): Take address use of reg holding
8300 non-incremented value. Add parm to limit search to the named
8302 (merge_in_block): Attempt to use a mem insn that is the next
8303 use of the original regno.
8305 2018-02-27 Martin Sebor <msebor@redhat.com>
8308 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
8309 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
8311 2018-02-27 Martin Sebor <msebor@redhat.com>
8313 PR translation/84207
8314 * diagnostic-core.h (warning_n, error_n, inform_n): Change
8315 n argument to unsigned HOST_WIDE_INT.
8316 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
8317 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
8318 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
8319 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
8321 2018-02-27 Richard Biener <rguenther@suse.de>
8323 PR tree-optimization/84512
8324 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
8325 Do not use the estimate returned from record_stmt_cost for
8326 the scalar iteration cost but sum properly using add_stmt_cost.
8328 2018-02-27 Richard Biener <rguenther@suse.de>
8330 PR tree-optimization/84466
8331 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8332 Adjust last change to less strictly validate use operands.
8334 2018-02-27 Martin Liska <mliska@suse.cz>
8336 PR gcov-profile/84548
8337 * gcov.c (process_file): Allow partial overlap and consider it
8338 also as group functions.
8339 (output_lines): Properly calculate range of lines for a group.
8341 2018-02-27 Martin Liska <mliska@suse.cz>
8343 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
8344 'ggc' suffixes. Change first column width.
8345 (timer::print): Fix formatting of the column.
8347 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
8349 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
8350 preserve inline entry blocks for the sake of debug inline
8351 entry point markers alone.
8352 (remove_unused_locals): Suggest in comments a better place to
8353 force the preservation of inline entry blocks that are
8354 otherwise unused, but do not preserve them.
8356 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
8358 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
8360 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
8363 * config/i386/constraints.md (Bs): Replace
8364 ix86_indirect_branch_register with
8365 TARGET_INDIRECT_BRANCH_REGISTER.
8367 * config/i386/i386.md (indirect_jump): Likewise.
8368 (tablejump): Likewise.
8369 (*sibcall_memory): Likewise.
8370 (*sibcall_value_memory): Likewise.
8371 Peepholes of indirect call and jump via memory: Likewise.
8372 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
8373 (*sibcall_value_GOT_32): Likewise.
8374 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8375 (GOT_memory_operand): Likewise.
8376 (call_insn_operand): Likewise.
8377 (sibcall_insn_operand): Likewise.
8378 (GOT32_symbol_operand): Likewise.
8379 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
8381 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
8383 PR rtl-optimization/83496
8384 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
8385 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
8386 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
8387 redundant insn, if any.
8388 (relax_delay_slots): Likewise.
8389 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
8391 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
8393 PR tree-optimization/83965
8394 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
8395 that grouped statements are part of a reduction chain. Return
8396 true if the statement is not marked as a reduction itself but
8398 (vect_recog_dot_prod_pattern): Don't check whether the statement
8399 is part of a group here.
8400 (vect_recog_sad_pattern): Likewise.
8401 (vect_recog_widen_sum_pattern): Likewise.
8403 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
8406 * final.c (rest_of_clean_state): Also look for calls inside sequences.
8408 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
8411 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
8413 (ix86_output_indirect_function_return): New prototype.
8414 (ix86_split_simple_return_pop_internal): Likewise.
8415 * config/i386/i386.c (indirect_return_via_cx): New.
8416 (indirect_return_via_cx_bnd): Likewise.
8417 (indirect_thunk_name): Handle return va CX_REG.
8418 (output_indirect_thunk_function): Create alias for
8419 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
8420 (ix86_output_indirect_jmp): Remove the bool argument.
8421 (ix86_output_indirect_function_return): New function.
8422 (ix86_split_simple_return_pop_internal): Likewise.
8423 * config/i386/i386.md (*indirect_jump): Don't pass false
8424 to ix86_output_indirect_jmp.
8425 (*tablejump_1): Likewise.
8426 (simple_return_pop_internal): Change it to define_insn_and_split.
8427 Call ix86_split_simple_return_pop_internal to split it for
8429 (simple_return_indirect_internal): Call
8430 ix86_output_indirect_function_return instead of
8431 ix86_output_indirect_jmp.
8433 2018-02-26 Jakub Jelinek <jakub@redhat.com>
8436 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
8437 memset and value initialization afterwards.
8439 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
8441 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
8443 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8446 * common/config/aarch64/aarch64-common.c
8447 (aarch_option_optimization_table[]): Switch
8448 off fomit-frame-pointer
8450 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
8451 Chung-Ju Wu <jasonwucj@gmail.com>
8453 * config/nds32/nds32-multiple.md (load_multiple): Disallow
8455 (store_multiple): Ditto.
8457 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
8459 * config.gcc: Add --with-cpu support for nds32 target.
8460 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
8461 * config/nds32/nds32.opt: Add -mcpu= option.
8463 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
8465 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
8466 isel=yes): Warn for these deprecated options.
8468 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
8470 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
8471 ISA_2_5_MASKS_EMBEDDED.
8473 2018-02-23 Jakub Jelinek <jakub@redhat.com>
8475 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
8476 p->max as pointers rather than using iterative_hash_expr.
8478 2018-02-23 Carl Love <cel@us.ibm.com>
8480 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
8481 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
8483 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
8484 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
8485 P8V_BUILTIN_VEC_VUNSIGNED2.
8487 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
8490 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
8491 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
8493 * lra-constraints.c (curr_insn_transform): Set up
8494 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
8496 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
8497 * lra-eliminations.c (spill_pseudos): Ditto.
8498 (process_insn_for_elimination): Ditto.
8499 * lra-lives.c (reg_early_clobber_p): Use the new macros.
8500 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
8501 LRA_NON_CLOBBERED_ALT.
8503 2018-02-22 Martin Sebor <msebor@redhat.com>
8505 PR tree-optimization/84480
8506 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
8507 to maybe_diag_stxncpy_trunc. Call it.
8508 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
8509 from gimple_fold_builtin_strcpy. Print inlining stack.
8510 (handle_builtin_stxncpy): Print inlining stack.
8511 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
8513 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
8516 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
8517 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
8518 and -fcheck-pointer-bounds are used together.
8519 (indirect_thunk_prefix): New enum.
8520 (indirect_thunk_need_prefix): New function.
8521 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
8522 "_nt" instead of "_bnd" for NOTRACK prefix.
8523 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
8524 (output_indirect_thunk_function): Likewise.
8526 (ix86_code_end): Update output_indirect_thunk_function calls.
8527 (ix86_output_indirect_branch_via_reg): Replace
8528 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
8529 (ix86_output_indirect_branch_via_push): Likewise.
8530 (ix86_output_function_return): Likewise.
8531 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
8532 incompatible with -fcf-protection=branch and
8533 -fcheck-pointer-bounds.
8535 2018-02-22 Steve Ellcey <sellcey@cavium.com>
8538 * config/aarch64/aarch64.c (aarch64_print_address_internal):
8539 Change gcc_assert call to output_operand_lossage.
8541 2018-02-22 Steve Ellcey <sellcey@cavium.com>
8543 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
8545 2018-02-22 DJ Delorie <dj@redhat.com>
8546 Sebastian Perta <sebastian.perta@renesas.com>
8547 Oleg Endo <olegendo@gcc.gnu.org>
8549 * config/rx/rx.c (rx_rtx_costs): New function.
8550 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
8552 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
8554 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
8556 2018-02-22 Martin Liska <mliska@suse.cz>
8559 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
8560 Add "native" as a possible value.
8562 2018-02-22 Martin Liska <mliska@suse.cz>
8565 * config/i386/i386.c (ix86_option_override_internal):
8566 Add "native" as a possible value for -march and -mtune.
8568 2018-02-22 Jakub Jelinek <jakub@redhat.com>
8571 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
8572 to all type variants.
8574 PR tree-optimization/84503
8575 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
8576 width as info->bitpos + info->bitsize - start.
8577 (merged_store_group::merge_overlapping): Simplify width computation.
8578 (check_no_overlap): New function.
8579 (imm_store_chain_info::try_coalesce_bswap): Compute expected
8580 start + width and last_order of the group, fail if check_no_overlap
8582 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
8583 to group if check_no_overlap fails.
8585 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
8587 * config/rs6000/altivec.md: Delete contraint arguments to
8588 define_expand, define_split, and define_peephole2, and in
8589 define_insn_and_split if always unused.
8590 * config/rs6000/darwin.md: Ditto.
8591 * config/rs6000/dfp.md: Ditto.
8592 * config/rs6000/rs6000.md: Ditto.
8593 * config/rs6000/sync.md: Ditto.
8594 * config/rs6000/vector.md: Ditto.
8595 * config/rs6000/vsx.md: Ditto.
8597 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
8599 * config/rs6000/altivec.md: Write output control strings as braced
8600 blocks instead of double-quoted strings.
8601 * config/rs6000/darwin.md: Ditto.
8602 * config/rs6000/rs6000.md: Ditto.
8603 * config/rs6000/vector.md: Ditto.
8604 * config/rs6000/vsx.md: Ditto.
8606 2018-02-21 Jason Merrill <jason@redhat.com>
8608 PR c++/84314 - ICE with templates and fastcall attribute.
8609 * attribs.c (build_type_attribute_qual_variant): Remove assert.
8611 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
8613 * ipa-cp.c (determine_versionability): Fix comment typos.
8615 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
8618 * ipa-cp.c (determine_versionability): Do not version functions caling
8621 2018-02-21 Martin Liska <mliska@suse.cz>
8624 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
8625 Add "native" as a possible value.
8626 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
8627 the macro when native cpu detection is available.
8629 2018-02-21 Martin Liska <mliska@suse.cz>
8632 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
8633 Add "native" as a possible value.
8634 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
8635 when native cpu detection is available.
8637 2018-02-21 Jakub Jelinek <jakub@redhat.com>
8638 Martin Sebor <msebor@redhat.com>
8640 PR tree-optimization/84478
8641 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
8643 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
8644 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
8645 support which is conservatively correct, for 2 only stay conservative
8646 for maxlen. Formatting and comment capitalization fixes. Add STRICT
8647 argument to the 2 argument get_range_strlen, adjust 6 arg
8648 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
8650 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
8651 (gimple_fold_builtin_strlen): Pass true as last argument to
8654 2018-02-20 Martin Sebor <msebor@redhat.com>
8657 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
8658 (builtin_memref::set_base_and_offset): Same. Handle inner references.
8659 (builtin_memref::builtin_memref): Factor out parts into
8660 set_base_and_offset and call it.
8662 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
8665 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
8666 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
8667 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
8668 search at the associated MODE_INT.
8670 2018-02-20 Jeff Law <law@redhat.com>
8673 PR tree-optimization/81592
8675 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
8676 for range data rather than using global data.
8677 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
8678 range data rather than using global data.
8679 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
8680 pass it to children as needed.
8681 (struct directive::fmtresult): Similarly.
8682 (struct directive::set_width): Similarly.
8683 (struct directive::set_precision): Similarly.
8684 (format_integer, format_directive, parse_directive): Similarly.
8685 (format_none): Accept unnamed vr_values parameter.
8686 (format_percent, format_floating, format_character): Similarly.
8687 (format_string, format_plain): Similarly.
8688 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
8689 the EVRP range analyzer for range data rather than using global data.
8690 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
8691 gimple-ssa-evrp-analyze.h
8692 (class sprintf_dom_walker): Add after_dom_children member function.
8693 Add evrp_range_analyzer member.
8694 (sprintf_dom_walker::before_dom_children): Call into the EVRP
8695 range analyzer as needed.
8696 (sprintf_dom_walker::after_dom_children): New member function.
8697 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
8699 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
8700 (evrp_range_analyzer::pop_to_marker): Likewise.
8702 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
8704 PR tree-optimization/84419
8705 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
8706 with the required type if its current type is compatible but
8709 2018-02-20 Jakub Jelinek <jakub@redhat.com>
8712 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
8713 after vectorization.
8715 2018-02-20 Martin Liska <mliska@suse.cz>
8718 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
8719 possible values if we don't have a hint.
8721 2018-02-20 Martin Liska <mliska@suse.cz>
8725 * final.c (shorten_branches): Build align_tab array with one
8727 * opts.c (finish_options): Add alignment option limit check.
8728 (MAX_CODE_ALIGN): Likewise.
8729 (MAX_CODE_ALIGN_VALUE): Likewise.
8730 * doc/invoke.texi: Document maximum allowed option value for
8731 all -falign-* options.
8733 2018-02-19 Jakub Jelinek <jakub@redhat.com>
8736 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
8737 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
8738 * var-tracking.c (emit_note_insn_var_location): Remove all references
8739 to NOTE_INSN_CALL_ARG_LOCATION.
8740 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
8741 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
8742 Use copy_rtx_if_shared.
8743 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
8744 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
8745 (dwarf2out_var_location): Remove handling of
8746 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
8748 * final.c (final_scan_insn): Remove all references to
8749 NOTE_INSN_CALL_ARG_LOCATION.
8750 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
8751 before dumping final insns.
8752 * except.c (emit_note_eh_region_end): Remove all references to
8753 NOTE_INSN_CALL_ARG_LOCATION.
8754 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
8755 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
8756 * config/arc/arc.c (hwloop_optimize): Likewise.
8757 * config/arm/arm.c (create_fix_barrier): Likewise.
8758 * config/s390/s390.c (s390_chunkify_start): Likewise.
8759 * config/sh/sh.c (find_barrier): Likewise.
8760 * config/i386/i386.c (rest_of_insert_endbranch,
8761 ix86_seh_fixup_eh_fallthru): Likewise.
8762 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
8763 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
8764 * config/frv/frv.c (frv_function_prologue): Likewise.
8765 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
8767 (note_outside_basic_block_p): Remove all references to
8768 NOTE_INSN_CALL_ARG_LOCATION.
8769 * gengtype.c (adjust_field_rtx_def): Likewise.
8770 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
8772 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
8773 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
8776 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
8779 PR tree-optimization/84452
8780 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
8781 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
8784 2018-02-19 Martin Liska <mliska@suse.cz>
8787 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
8789 2018-02-19 Martin Liska <mliska@suse.cz>
8790 Richard Sandiford <richard.sandiford@linaro.org>
8792 PR tree-optimization/82491
8793 * gimple-fold.c (get_base_constructor): Make earlier bail out
8796 2018-02-19 Carl Love <cel@us.ibm.com>
8798 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
8799 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
8801 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
8802 P8V_BUILTIN_VEC_NEG.
8804 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
8806 * config/rl78/rl78.md (movdf): New define expand.
8808 2018-02-19 Martin Liska <mliska@suse.cz>
8811 * doc/invoke.texi: Fix typo.
8812 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
8814 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
8816 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8817 handle rs6000_single_float and rs6000_double_float specially for
8820 2018-02-16 Jeff Law <law@redhat.com>
8822 * config/rx/rx.c (add_pop_cfi_notes): New function.;
8825 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8828 * ipa-inline.c (inline_small_functions): Fix a typo.
8830 2018-02-16 Nathan Sidwell <nathan@acm.org>
8832 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
8834 2018-02-16 Carl Love <cel@us.ibm.com>
8836 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
8837 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
8838 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
8839 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
8840 expansion to P8V_BUILTIN_VEC_FLOAT2.
8842 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
8844 PR rtl-optimization/70023
8845 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
8846 src_regno into account.
8848 2018-02-16 Carl Love <cel@us.ibm.com>
8850 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
8851 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
8852 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
8853 * config/rs6000/rs6000.c: Remove case statements for
8854 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
8855 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
8856 and P9V_BUILTIN_VEC_VINSERT4B.
8857 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
8858 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
8859 * config/rs6000/vsx.md:
8860 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
8863 2018-02-16 Carl Love <cel@us.ibm.com>
8865 * config/rs6000/altivec.h: Add builtin names vec_extract4b
8867 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
8869 * config/rs6000/rs6000-c.c: Add the definitions for
8870 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
8871 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
8872 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
8873 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
8874 definition for insert4b and define insn *insert3b_internal.
8875 * doc/extend.texi: Add documentation for vec_extract4b.
8877 2018-02-16 Nathan Sidwell <nathan@acm.org>
8879 * doc/extend.texi (Backwards Compatibility): Mention friend
8880 injection. Note for-scope is deprecated.
8881 * doc/invoke.texi (-ffriend-injection): Deprecate.
8883 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
8885 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
8886 that moved to I2, also allow destinations that are a paradoxical
8887 subreg (instead of a normal reg).
8889 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
8892 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
8895 2018-02-16 Richard Biener <rguenther@suse.de>
8897 PR tree-optimization/84037
8898 PR tree-optimization/84016
8900 * config/i386/i386.c (ix86_builtin_vectorization_cost):
8901 Adjust vec_construct for the fact we need additional higher latency
8902 128bit inserts for AVX256 and AVX512 vector builds.
8903 (ix86_add_stmt_cost): Scale vector construction cost for
8906 2018-02-16 Richard Biener <rguenther@suse.de>
8908 PR tree-optimization/84417
8909 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
8910 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
8911 (non_rewritable_lvalue_p): Likewise, use poly-ints.
8913 2018-02-16 Martin Liska <mliska@suse.cz>
8916 * internal-fn.def (ASAN_CHECK): Set proper flags.
8917 (ASAN_MARK): Likewise.
8919 2018-02-16 Julia Koval <julia.koval@intel.com>
8921 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
8922 from PTA_CANNONLAKE.
8924 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8927 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
8928 Use ++iter rather than iter++ for std::list iterators.
8929 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
8930 defer deleting them until all nodes in the forest are processed. Do
8931 free even leaf nodes. Change to_process into auto_vec.
8934 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
8935 * vec.h (vec_default_construct): Use memset instead of placement new
8936 if BROKEN_VALUE_INITIALIZATION is defined.
8937 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
8938 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
8941 PR rtl-optimization/83723
8942 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
8943 * lra.c (lra_substitute_pseudo): Likewise. If true, use
8944 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
8946 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
8948 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
8950 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
8952 PR rtl-optimization/81443
8953 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
8954 from inner REGs to paradoxical SUBREGs.
8956 2018-02-16 Richard Biener <rguenther@suse.de>
8958 PR tree-optimization/84399
8959 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8960 For operands we can analyze at their definition make sure we can
8961 analyze them at each use as well.
8963 2018-02-16 Richard Biener <rguenther@suse.de>
8965 PR tree-optimization/84190
8966 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
8967 volatile accesses if the decl isn't volatile.
8969 2018-02-15 Jason Merrill <jason@redhat.com>
8971 PR c++/84314 - ICE with templates and fastcall attribute.
8972 * attribs.c (build_type_attribute_qual_variant): Don't clobber
8973 TYPE_CANONICAL on an existing type.
8975 2018-02-15 Jakub Jelinek <jakub@redhat.com>
8977 PR tree-optimization/84383
8978 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
8979 dstoff nor call operand_equal_p if dstbase is NULL.
8981 PR tree-optimization/84334
8982 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
8983 also a CONSTANT_CLASS_P, punt.
8985 2018-02-14 Jim Wilson <jimw@sifive.com>
8987 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
8988 first SMALL_OPERAND check. New local min_second_step. Move assert
8989 to where locals are set. Add TARGET_RVC support.
8990 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
8992 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
8994 * doc/invoke.texi: Correct -Wformat-overflow code sample.
8996 2018-02-14 Martin Sebor <msebor@redhat.com>
8998 PR tree-optimization/83698
8999 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
9000 arrays constrain the offset range to their bounds.
9001 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
9002 (builtin_access::overlap): Avoid setting the size of overlap if it's
9004 (maybe_diag_overlap): Also consider arrays when deciding what values
9005 of offsets to include in diagnostics.
9007 2018-02-14 Martin Sebor <msebor@redhat.com>
9010 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
9011 that correspond to the kind of a declaration.
9013 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
9016 * config/pa/pa.md: Load address of PIC label using the linkage table
9017 if the label is nonlocal.
9019 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
9021 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
9022 warning message if user requests -maltivec=be.
9023 * doc/invoke.texi: Document deprecation of -maltivec=be.
9025 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
9028 * config/rs6000/rs6000-c.c: Update definitions for
9029 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
9030 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
9032 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9035 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
9036 add _get_ssp intrinsics. Remove argument from
9037 __builtin_ia32_rdssp[d|q].
9038 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
9039 * config/i386/i386-builtin.def: Remove argument from
9040 __builtin_ia32_rdssp[d|q].
9041 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
9042 ix86_expand_special_args_builtin for _rdssp[d|q].
9043 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
9044 Clear register before usage.
9045 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
9046 Add documentation for new _get_ssp and _inc_ssp intrinsics.
9048 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
9050 PR tree-optimization/84357
9051 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
9052 operand 1 of an ARRAY_REF too.
9054 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
9057 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
9058 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
9060 (set_of_reg): New struct.
9061 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
9062 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
9063 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
9065 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
9066 Split into bitclr, bitset, bitinvert patterns if appropriate.
9067 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
9068 use rx_fuse_in_memory_bitop.
9069 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
9070 to named insn, correct maximum insn length.
9072 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
9075 * machmode.def: Define a complex mode for PARTIAL_INT.
9076 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
9078 * doc/rtl.texi: Document CSPImode.
9079 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
9081 (msp430_hard_regno_nregs_with_padding): Likewise.
9083 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
9086 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
9088 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
9090 PR rtl-optimization/84169
9091 * combine.c (try_combine): New variable split_i2i3. Set it to true if
9092 we generated a parallel as new i3 and we split that to new i2 and i3
9093 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
9094 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
9095 those to i2, not i1. Partially rewrite this scan code.
9097 2018-02-13 Jakub Jelinek <jakub@redhat.com>
9100 * stor-layout.c (place_field): For variable length fields, adjust
9101 offset_align afterwards not just based on the field's alignment,
9102 but also on the size.
9105 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
9106 of exps and logs in the use_exp2 case.
9108 2018-02-13 Jeff Law <law@redhat.com>
9110 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
9113 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
9116 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
9120 * common.opt (gas-loc-support, gas-locview-support): New.
9121 (ginline-points, ginternal-reset-location-views): New.
9122 * doc/invoke.texi: Document them. Use @itemx where intended.
9123 (gvariable-location-views): Adjust.
9124 * target.def (reset_location_view): New.
9125 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
9126 (TARGET_RESET_LOCATION_VIEW): New.
9127 * doc/tm.texi: Rebuilt.
9128 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
9129 (dwarf2out_default_as_locview_support): New.
9130 (output_asm_line_debug_info): Use option variables.
9131 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
9132 (output_loc_list): Likewise.
9133 (add_high_low_attributes): Check option variables.
9134 Don't output entry view attribute in strict mode.
9135 (gen_inlined_subroutine_die): Check option variables.
9136 (dwarf2out_inline_entry): Likewise.
9137 (init_sections_and_labels): Likewise.
9138 (dwarf2out_early_finish): Likewise.
9139 (maybe_reset_location_view): New, from...
9140 (dwarf2out_var_location): ... here. Call it.
9141 * debug.h (dwarf2out_default_as_loc_support): Declare.
9142 (dwarf2out_default_as_locview_support): Declare.
9143 * hooks.c (hook_int_rtx_insn_0): New.
9144 * hooks.h (hook_int_rtx_insn_0): Declare.
9145 * toplev.c (process_options): Take -gas-loc-support and
9146 -gas-locview-support from dwarf2out. Enable
9147 -gvariable-location-views by default only with locview
9148 assembler support. Enable -ginternal-reset-location-views by
9149 default only if the target defines the corresponding hook.
9150 Enable -ginline-points by default if location views are
9151 enabled; force it disabled if statement frontiers are
9153 * tree-inline.c (expand_call_inline): Check option variables.
9154 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
9156 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
9158 PR tree-optimization/84321
9159 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
9160 handling. Also check whether the anti-range contains any values
9161 that satisfy the mask; switch to a VR_RANGE if not.
9163 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
9166 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
9168 2018-02-13 Martin Jambor <mjambor@suse.cz>
9171 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
9172 of call statements, also set location of a load to a temporary.
9174 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
9176 * config/rl78/rl78.c (add_vector_labels): New function.
9177 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
9178 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
9179 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
9180 which checks that no arguments are passed.
9181 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
9182 * doc/extend.texi: Documentation for the new attribute.
9184 2018-02-13 Andreas Schwab <schwab@suse.de>
9186 * config/riscv/linux.h (CPP_SPEC): Define.
9188 2018-02-13 Jakub Jelinek <jakub@redhat.com>
9191 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
9192 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
9193 OPTION_MASK_ISA_AES as first argument to def_builtin_const
9194 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
9195 instead of OPTION_MASK_ISA_PCLMUL as first argument to
9196 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
9197 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
9198 temporarily for AES and PCLMUL builtins.
9200 PR tree-optimization/84339
9201 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
9202 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
9206 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
9207 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
9208 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
9210 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
9212 * omp-simd-clone.h: New file.
9213 * omp-simd-clone.c: Include omp-simd-clone.h.
9214 (expand_simd_clones): No longer static.
9215 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
9216 cgraph.h and omp-simd-clone.h.
9217 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
9218 (vect_recog_widen_shift_pattern): Formatting fix.
9219 (vect_pattern_recog_1): Don't check optab for calls.
9222 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
9223 operands[2] into a REG before using gen_lowpart on it.
9225 2018-02-12 Jeff Law <law@redhat.com>
9228 * config/sh/sh.c (find_barrier): Consider a sibling call
9231 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
9232 successfully back substituting a reg.
9234 2018-02-12 Richard Biener <rguenther@suse.de>
9236 PR tree-optimization/84037
9237 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
9238 parameter, move visited init to caller.
9239 (vect_slp_analyze_operations): Separate cost from validity
9240 check, initialize visited once for all instances.
9241 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
9243 * tree-vect-stmts.c (vect_model_simple_cost): Make early
9245 (vect_model_promotion_demotion_cost): Likewise.
9246 (vectorizable_bswap): Guard cost modeling with !slp_node
9247 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
9249 (vectorizable_call): Likewise.
9250 (vectorizable_conversion): Likewise.
9251 (vectorizable_assignment): Likewise.
9252 (vectorizable_shift): Likewise.
9253 (vectorizable_operation): Likewise.
9254 (vectorizable_store): Likewise.
9255 (vectorizable_load): Likewise.
9256 (vectorizable_condition): Likewise.
9257 (vectorizable_comparison): Likewise.
9259 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
9262 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
9263 (ASAN_MARK): Fix fnspec to account for return value, change pointer
9264 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
9266 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
9269 * params.def (inline-min-speedup): Increase from 8 to 15.
9270 (max-inline-insns-auto): Decrease from 40 to 30.
9271 * ipa-split.c (consider_split): Add some buffer for function to
9272 be considered inlining candidate.
9273 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
9276 2018-02-12 Richard Biener <rguenther@suse.de>
9278 PR tree-optimization/84037
9279 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
9280 matched stmts if we cannot swap the non-matched ones.
9282 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
9284 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
9285 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
9286 _mm_maskz_scalef_round_ss): New intrinsics.
9287 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
9288 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
9289 __builtin_ia32_scalefss_round): Remove.
9290 (__builtin_ia32_scalefsd_mask_round,
9291 __builtin_ia32_scalefss_mask_round): New intrinsics.
9292 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
9293 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
9294 ((match_operand:VF_128 2 "<round_nimm_predicate>"
9295 "<round_constraint>")): Changed to ...
9296 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
9297 "<round_scalar_constraint>")): ... this.
9298 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
9299 %0, %1, %2<round_op3>}"): Changed to ...
9300 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
9301 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
9302 %2<round_scalar_mask_op3>}"): ... this.
9303 * config/i386/subst.md (round_scalar_nimm_predicate): New.
9305 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
9307 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
9308 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
9309 (_mm_maskz_sqrt_round_ss): New intrinsics.
9310 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
9311 (__builtin_ia32_sqrtsd_mask_round)
9312 (__builtin_ia32_sqrtss_mask_round): New builtins.
9313 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
9314 (__builtin_ia32_sqrtss_round): Remove.
9315 (__builtin_ia32_sqrtsd_mask_round)
9316 (__builtin_ia32_sqrtss_mask_round): New builtins.
9317 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
9318 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
9319 ((match_operand:VF_128 1 "vector_operand"
9320 "xBm,<round_constraint>")): Changed to ...
9321 ((match_operand:VF_128 1 "vector_operand"
9322 "xBm,<round_scalar_constraint>")): ... this.
9323 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
9324 %0, %2, %<iptr>1<round_op3>}): Changed to ...
9325 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
9326 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
9327 %<iptr>1<round_scalar_mask_op3>}): ... this.
9328 ((set_attr "prefix" "<round_prefix>")): Changed to ...
9329 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
9331 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
9334 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
9335 Cast vec_cmpeq result to correct type.
9336 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
9337 Cast vec_cmpgt result to correct type.
9339 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
9341 * final.c (final_scan_insn_1): Renamed from...
9342 (final_scan_insn): ... this. New wrapper, to recover
9343 seen from the outermost call in recursive ones.
9344 * config/sparc/sparc.c (output_return): Drop seen from call.
9345 (output_sibcall): Likewise.
9346 * config/visium/visium.c (output_branch): Likewise.
9348 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
9350 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
9353 2018-02-10 Alan Modra <amodra@gmail.com>
9356 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
9357 Specify LR as an input.
9359 2018-02-10 Jakub Jelinek <jakub@redhat.com>
9362 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
9363 remove_member_access_dummy_vars): New functions.
9364 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
9365 lower_omp_1, execute_lower_omp): Use them.
9367 PR rtl-optimization/84308
9368 * shrink-wrap.c (spread_components): Release todo vector.
9370 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
9372 PR rtl-optimization/57193
9373 * ira-color.c (struct allocno_color_data): Add member
9374 conflict_allocno_hard_prefs.
9375 (update_conflict_allocno_hard_prefs): New.
9376 (bucket_allocno_compare_func): Add a preference based on
9377 conflict_allocno_hard_prefs.
9378 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
9379 (color_allocnos): Remove a dead code. Initiate
9380 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
9382 2018-02-09 Jakub Jelinek <jakub@redhat.com>
9385 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
9386 constraint from =wa to wa. Avoid a subreg on the output operand,
9387 instead use a pseudo and subreg it in a move.
9388 (p9_xxbrd_<mode>): Changed to ...
9389 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
9390 (p9_xxbrd_v2df): New expander.
9391 (p9_xxbrw_<mode>): Changed to ...
9392 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
9393 (p9_xxbrw_v4sf): New expander.
9395 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
9397 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
9399 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
9402 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
9403 multiply in 32-bit mode.
9404 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
9405 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
9408 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
9410 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
9411 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
9412 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
9413 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
9415 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
9418 * dwarf2out.c (is_trivial_indirect_ref): New function.
9419 (dwarf2out_late_global_decl): Do not generate a location
9420 attribute for variables that have a non-trivial DECL_VALUE_EXPR
9421 and that are not defined in the current unit.
9423 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
9425 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
9426 instead of a libcall for UNORDERED.
9428 2018-02-09 Tamar Christina <tamar.christina@arm.com>
9431 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
9432 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
9434 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9437 * config/s390/s390.c (s390_set_current_function): Invoke
9438 s390_indirect_branch_settings also if fndecl didn't change.
9440 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
9442 * config/rs6000/rs6000.md (blockage): Set length to zero.
9444 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
9446 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
9448 2018-02-09 Jakub Jelinek <jakub@redhat.com>
9451 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
9452 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
9456 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
9457 PARALLEL incoming that failed vt_get_decl_and_offset check.
9460 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
9461 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
9463 (get_variable_section): For decls in named .bss* sections pass true as
9464 second argument to bss_initializer_p.
9466 2018-02-09 Marek Polacek <polacek@redhat.com>
9467 Jakub Jelinek <jakub@redhat.com>
9470 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
9471 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
9472 Sync some changes from cxx_fold_indirect_ref.
9474 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
9476 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
9478 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
9479 (BLOCK_INLINE_ENTRY_LABEL): New.
9480 (dwarf2out_var_location): Disregard inline entry markers.
9481 (inline_entry_data): New struct.
9482 (inline_entry_data_hasher): New hashtable type.
9483 (inline_entry_data_hasher::hash): New.
9484 (inline_entry_data_hasher::equal): New.
9485 (inline_entry_data_table): New variable.
9486 (add_high_low_attributes): Add DW_AT_entry_pc and
9487 DW_AT_GNU_entry_view attributes if a pending entry is found
9488 in inline_entry_data_table. Add old entry_pc attribute only
9489 if debug nonbinding markers are disabled.
9490 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
9491 markers are enabled.
9492 (block_within_block_p, dwarf2out_inline_entry): New.
9493 (dwarf2out_finish): Check that no entries remained in
9494 inline_entry_data_table.
9495 * final.c (reemit_insn_block_notes): Handle inline entry notes.
9496 (final_scan_insn, notice_source_line): Likewise.
9497 (rest_of_clean_state): Skip inline entry markers.
9498 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
9500 * gimple.c (gimple_build_debug_inline_entry): New.
9501 * gimple.h (enum gimple_debug_subcode): Add
9502 GIMPLE_DEBUG_INLINE_ENTRY.
9503 (gimple_build_debug_inline_entry): Declare.
9504 (gimple_debug_inline_entry_p): New.
9505 (gimple_debug_nonbind_marker_p): Adjust.
9506 * insn-notes.def (INLINE_ENTRY): New.
9507 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
9508 inline entry marker notes.
9509 (print_insn): Likewise.
9510 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
9511 (INSN_DEBUG_MARKER_KIND): Likewise.
9512 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
9513 * tree-inline.c (expand_call_inline): Build and insert
9514 debug_inline_entry stmt.
9515 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
9516 inline entry blocks early, if nonbind markers are enabled.
9517 (dump_scope_block): Dump fragment info.
9518 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
9519 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
9520 (gimple_build_debug_inline_entry): New.
9521 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
9522 Enable/disable inline entry points too.
9523 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
9524 (DEBUG_INSN): Describe inline entry markers.
9526 * common.opt (gvariable-location-views): New.
9527 (gvariable-location-views=incompat5): New.
9528 * config.in: Rebuilt.
9529 * configure: Rebuilt.
9530 * configure.ac: Test assembler for view support.
9531 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
9532 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
9533 * dwarf2out.c (var_loc_view): New typedef.
9534 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
9535 (dwarf2out_locviews_in_attribute): New.
9536 (dwarf2out_locviews_in_loclist): New.
9537 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
9538 (enum dw_line_info_opcode): Add LI_adv_address.
9539 (struct dw_line_info_table): Add view.
9540 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
9541 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
9542 (zero_view_p): New variable.
9543 (ZERO_VIEW_P): New macro.
9544 (output_asm_line_debug_info): New.
9545 (struct var_loc_node): Add view.
9546 (add_AT_view_list, AT_loc_list): New.
9547 (add_var_loc_to_decl): Add view param. Test it against last.
9548 (new_loc_list): Add view params. Record them.
9549 (AT_loc_list_ptr): Handle loc and view lists.
9550 (view_list_to_loc_list_val_node): New.
9551 (print_dw_val): Handle dw_val_class_view_list.
9552 (size_of_die): Likewise.
9553 (value_format): Likewise.
9554 (loc_list_has_views): New.
9555 (gen_llsym): Set vl_symbol too.
9556 (maybe_gen_llsym, skip_loc_list_entry): New.
9557 (dwarf2out_maybe_output_loclist_view_pair): New.
9558 (output_loc_list): Output view list or entries too.
9559 (output_view_list_offset): New.
9560 (output_die): Handle dw_val_class_view_list.
9561 (output_dwarf_version): New.
9562 (output_compilation_unit_header): Use it.
9563 (output_skeleton_debug_sections): Likewise.
9564 (output_rnglists, output_line_info): Likewise.
9565 (output_pubnames, output_aranges): Update version comments.
9566 (output_one_line_info_table): Output view numbers in asm comments.
9567 (dw_loc_list): Determine current endview, pass it to new_loc_list.
9568 Call maybe_gen_llsym.
9569 (loc_list_from_tree_1): Adjust.
9570 (add_AT_location_description): Create view list attribute if
9571 needed, check it's absent otherwise.
9572 (convert_cfa_to_fb_loc_list): Adjust.
9573 (maybe_emit_file): Call output_asm_line_debug_info for test.
9574 (dwarf2out_var_location): Reset views as needed. Precompute
9575 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
9576 attribute. Set view.
9577 (new_line_info_table): Reset next view.
9578 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
9579 (dwarf2out_source_line): Likewise. Output view resets and labels to
9580 the assembler, or select appropriate line info opcodes.
9581 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
9582 (optimize_string_length): Catch it. Adjust.
9583 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
9584 dw_val_class_view_list, and remove it if no longer needed.
9585 (hash_loc_list): Hash view numbers.
9586 (loc_list_hasher::equal): Compare them.
9587 (optimize_location_lists): Check whether a view list symbol is
9588 needed, and whether the locview attribute is present, and
9589 whether they match. Remove the locview attribute if no longer
9591 (index_location_lists): Call skip_loc_list_entry for test.
9592 (dwarf2out_finish): Call output_asm_line_debug_info for test.
9593 Use output_dwarf_version.
9594 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
9595 (struct dw_val_node): Add val_view_list.
9596 * final.c (SEEN_NEXT_VIEW): New.
9597 (set_next_view_needed): New.
9598 (clear_next_view_needed): New.
9599 (maybe_output_next_view): New.
9600 (final_start_function): Rename to...
9601 (final_start_function_1): ... this. Take pointer to FIRST,
9602 add SEEN parameter. Emit param bindings in the initial view.
9603 (final_start_function): Reintroduce SEEN-less interface.
9604 (final): Rename to...
9605 (final_1): ... this. Take SEEN parameter. Output final pending
9606 next view at the end.
9607 (final): Reintroduce seen-less interface.
9608 (final_scan_insn): Output pending next view before switching
9609 sections or ending a block. Mark the next view as needed when
9610 outputting variable locations. Notify debug backend of section
9611 changes, and of location view changes.
9612 (rest_of_handle_final): Adjust.
9613 * toplev.c (process_options): Autodetect value for debug variable
9614 location views option. Warn on incompat5 without -gdwarf-5.
9615 * doc/invoke.texi (gvariable-location-views): New.
9616 (gvariable-location-views=incompat5): New.
9617 (gno-variable-location-views): New.
9619 2018-02-08 David Malcolm <dmalcolm@redhat.com>
9621 PR tree-optimization/84136
9622 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
9623 that the result of find_edge is non-NULL.
9625 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
9628 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
9629 storing integer register in SImode. Fix cost of 256 and 512
9630 byte aligned SSE register store.
9632 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
9634 * config/i386/i386.c (ix86_multiplication_cost): Fix
9635 multiplication cost for TARGET_AVX512DQ.
9637 2018-02-08 Marek Polacek <polacek@redhat.com>
9639 PR tree-optimization/84238
9640 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
9643 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9645 PR tree-optimization/84265
9646 * tree-vect-stmts.c (vectorizable_store): Don't treat
9647 VMAT_CONTIGUOUS accesses as grouped.
9648 (vectorizable_load): Likewise.
9650 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9652 PR tree-optimization/81635
9653 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
9654 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
9655 (test_round_for_mask): New functions.
9656 (wide_int_cc_tests): Call test_round_for_mask.
9657 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
9658 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
9659 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
9660 range returned by get_range_info.
9662 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
9665 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
9666 * symtab.c: Include builtins.h
9667 (symtab_node::output_to_lto_symbol_table_p): Move here
9668 from lto-streamer-out.c:output_symbol_p.
9669 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
9670 (output_symbol_p): Move all logic to symtab.c
9671 (produce_symtab): Update.
9673 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9675 * config/s390/s390-opts.h (enum indirect_branch): Define.
9676 * config/s390/s390-protos.h (s390_return_addr_from_memory)
9677 (s390_indirect_branch_via_thunk)
9678 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
9679 (enum s390_indirect_branch_type): Define.
9680 * config/s390/s390.c (struct s390_frame_layout, struct
9681 machine_function): Remove.
9682 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
9683 (indirect_branch_table_label_no, indirect_branch_table_name):
9685 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
9686 (enum s390_indirect_branch_option): Define.
9687 (s390_return_addr_from_memory): New function.
9688 (s390_handle_string_attribute): New function.
9689 (s390_attribute_table): Add new attribute handler.
9690 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
9691 (s390_indirect_branch_via_thunk): New function.
9692 (s390_indirect_branch_via_inline_thunk): New function.
9693 (s390_function_ok_for_sibcall): When jumping via thunk disallow
9694 sibling call optimization for non z10 compiles.
9695 (s390_emit_call): Force indirect branch target to be a single
9696 register. Add r1 clobber for non-z10 compiles.
9697 (s390_emit_epilogue): Emit return jump via return_use expander.
9698 (s390_reorg): Handle JUMP_INSNs as execute targets.
9699 (s390_option_override_internal): Perform validity checks for the
9700 new command line options.
9701 (s390_indirect_branch_attrvalue): New function.
9702 (s390_indirect_branch_settings): New function.
9703 (s390_set_current_function): Invoke s390_indirect_branch_settings.
9704 (s390_output_indirect_thunk_function): New function.
9705 (s390_code_end): Implement target hook.
9706 (s390_case_values_threshold): Implement target hook.
9707 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
9709 * config/s390/s390.h (struct s390_frame_layout)
9710 (struct machine_function): Move here from s390.c.
9711 (TARGET_INDIRECT_BRANCH_NOBP_RET)
9712 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
9713 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
9714 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
9715 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
9716 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
9717 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
9718 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
9719 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
9720 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
9721 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
9722 (mnemonic attribute): Add values which aren't recognized
9724 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
9725 pattern for branch conversion. Fix mnemonic attribute.
9726 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
9727 indirect branch via thunk if requested.
9728 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
9729 ("*indirect_jump"): Disable for branch conversion using out of
9731 ("indirect_jump_via_thunk<mode>_z10")
9732 ("indirect_jump_via_thunk<mode>")
9733 ("indirect_jump_via_inlinethunk<mode>_z10")
9734 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
9735 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
9736 ("casesi_jump_via_inlinethunk<mode>_z10")
9737 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
9738 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
9739 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
9740 ("*indirect2_jump"): Disable for branch conversion.
9741 ("casesi_jump"): Turn into expander and expand patterns for branch
9743 ("return_use"): New expander.
9744 ("*return"): Emit return via thunk and rename it to ...
9745 ("*return<mode>"): ... this one.
9746 * config/s390/s390.opt: Add new options and and enum for the
9749 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9751 * lra-constraints.c (match_reload): Unconditionally use
9752 gen_lowpart_SUBREG, rather than selecting between that
9753 and equivalent gen_rtx_SUBREG code.
9755 2018-02-08 Richard Biener <rguenther@suse.de>
9757 PR tree-optimization/84233
9758 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
9759 changed flag instead of boguously re-using phi_inserted.
9761 2018-02-08 Martin Jambor <mjambor@suse.cz>
9763 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
9764 static local variables.
9766 2018-02-08 Richard Biener <rguenther@suse.de>
9768 PR tree-optimization/84278
9769 * tree-vect-stmts.c (vectorizable_store): When looking for
9770 smaller vector types to perform grouped strided loads/stores
9771 make sure the mode is supported by the target.
9772 (vectorizable_load): Likewise.
9774 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
9776 * config/aarch64/aarch64.c (aarch64_components_for_bb):
9777 Increase LDP/STP opportunities by adding adjacent callee-saves.
9779 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
9781 PR rtl-optimization/84068
9782 PR rtl-optimization/83459
9783 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
9785 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
9787 PR tree-optimization/84224
9788 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
9789 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
9792 2018-02-07 Iain Sandoe <iain@codesourcery.com>
9795 * config/rs6000/altivec.md (*restore_world): Remove LR use.
9796 * config/rs6000/predicates.md (restore_world_operation): Adjust op
9797 count, remove one USE.
9799 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
9801 * doc/install.texi (Configuration): Document the
9802 --with-long-double-format={ibm,ieee} PowerPC configuration
9806 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
9807 Convert from define_expand to be define_insn_and_split. Rework
9808 float/double/_Float128 conversions to QI/HI/SImode to work with
9809 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
9810 conversions to QI/HImode types did a store and then a load to
9811 truncate the value. For conversions to VSX registers, don't split
9812 the insn, instead emit the code directly. Use the code iterator
9813 any_fix to combine signed and unsigned conversions.
9814 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
9815 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
9816 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9817 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
9818 (fix_<mode>di2_hw): Likewise.
9819 (fixuns_<mode>di2_hw): Likewise.
9820 (fix_<mode>si2_hw): Likewise.
9821 (fixuns_<mode>si2_hw): Likewise.
9822 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
9823 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9824 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
9825 fix<uns>_trunc<SFDF:mode>si2_p8.
9826 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
9828 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
9829 (fix<uns>_<mode>_mem): Likewise.
9830 (fctiw<u>z_<mode>_mem): Likewise.
9831 (fix<uns>_<mode>_mem): Likewise.
9832 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
9833 the register allocator from doing a direct move to the GPRs to do
9834 a store, and instead use the ISA 3.0 store byte/half-word from
9835 vector register instruction. For IEEE 128-bit floating point,
9836 also optimize stores of 32-bit ints.
9837 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
9839 2018-02-07 Alan Hayward <alan.hayward@arm.com>
9841 * genextract.c (push_pathstr_operand): New function to support
9843 (walk_rtx): Call push_pathstr_operand.
9844 (print_path): Support [a-zA-Z].
9846 2018-02-07 Richard Biener <rguenther@suse.de>
9848 PR tree-optimization/84037
9849 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
9850 (cse_and_gimplify_to_preheader): Declare.
9851 (vect_get_place_in_interleaving_chain): Likewise.
9852 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9854 (_loop_vec_info::~_loop_vec_info): Delete it.
9855 (cse_and_gimplify_to_preheader): New function.
9856 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
9857 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
9858 (vectorizable_load): Likewise. For grouped stores always base
9859 the IV on the first element.
9860 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
9861 condition before gimplifying.
9863 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9865 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
9866 *DIV_EXPR and *MOD_EXPR.
9868 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
9871 * config/i386/i386.c (ix86_option_override_internal): Mask out
9872 the CF_SET bit when checking -fcf-protection.
9874 2018-02-07 Tom de Vries <tom@codesourcery.com>
9877 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
9880 2018-02-07 Richard Biener <rguenther@suse.de>
9882 PR tree-optimization/84204
9883 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
9886 PR tree-optimization/84205
9887 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
9888 special-case isl_ast_op_zdiv_r.
9890 PR tree-optimization/84223
9891 * graphite-scop-detection.c (gather_bbs::before_dom_children):
9892 Only add conditions from within the region.
9893 (gather_bbs::after_dom_children): Adjust.
9895 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
9898 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
9899 * config/avr/avr.md: Only post-reload split REG-REG moves if
9900 either register is GENERAL_REG_P.
9902 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9904 PR tree-optimization/84235
9905 * tree-ssa-scopedtables.c
9906 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
9907 if the subtraction is performed in floating point type where NaNs are
9908 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
9909 build 1. Formatting fix.
9911 2018-02-06 Jakub Jelinek <jakub@redhat.com>
9914 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
9915 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
9916 and skip it regardless of bb boundaries. Use CALL_P macro,
9917 don't test INSN_P (insn) together with CALL_P or JUMP_P check
9918 unnecessarily, formatting fix.
9920 2018-02-06 Michael Collison <michael.collison@arm.com>
9922 * config/arm/thumb2.md:
9923 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
9924 (*thumb_mov_notscc): Ditto.
9926 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
9929 * config/rs6000/rs6000.md (su code attribute): Use "u" for
9930 unsigned_fix, not "s".
9932 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9934 * configure.ac (gcc_fn_eh_frame_ro): New function.
9935 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
9936 correct .eh_frame permissions.
9937 * configure: Regenerate.
9939 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
9941 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
9944 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9946 * config/rs6000/rs6000.c (rs6000_option_override_internal):
9947 Display warning message for -mno-speculate-indirect-jumps.
9949 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
9951 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
9953 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
9955 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
9957 PR tree-optimization/84225
9958 * tree-eh.c (find_trapping_overflow): Only call
9959 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
9961 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9964 * config/i386/i386.c: Reimplement the check of possible options
9965 -mibt/-mshstk conbination. Change error messages.
9966 * doc/invoke.texi: Fix a typo: remove extra '='.
9968 2018-02-06 Marek Polacek <polacek@redhat.com>
9970 PR tree-optimization/84228
9971 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
9973 2018-02-06 Tamar Christina <tamar.christina@arm.com>
9976 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
9977 emitted arch directives.
9978 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
9979 __ARM_FEATURE_COPROC before changing architectures.
9981 2018-02-06 Richard Biener <rguenther@suse.de>
9983 * config/i386/i386.c (print_reg): Fix typo.
9984 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
9986 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
9988 * configure: Regenerate.
9990 2018-02-05 Martin Sebor <msebor@redhat.com>
9992 PR tree-optimization/83369
9993 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
9996 2018-02-05 Martin Liska <mliska@suse.cz>
9998 * doc/invoke.texi: Cherry-pick upstream r323995.
10000 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
10002 * ira.c (ira_init_register_move_cost): Adjust comment.
10004 2018-02-05 Martin Liska <mliska@suse.cz>
10006 PR gcov-profile/84137
10007 * doc/gcov.texi: Fix typo in documentation.
10009 2018-02-05 Martin Liska <mliska@suse.cz>
10011 PR gcov-profile/83879
10012 * doc/gcov.texi: Document necessity of --dynamic-list-data when
10013 using dlopen functionality.
10015 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
10017 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
10018 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
10019 _mm_maskz_range_ss, _mm_mask_range_round_ss,
10020 _mm_maskz_range_round_ss): New intrinsics.
10021 (__builtin_ia32_rangesd128_round)
10022 (__builtin_ia32_rangess128_round): Remove.
10023 (__builtin_ia32_rangesd128_mask_round,
10024 __builtin_ia32_rangess128_mask_round): New builtins.
10025 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
10026 __builtin_ia32_rangess128_round): Remove.
10027 (__builtin_ia32_rangesd128_mask_round,
10028 __builtin_ia32_rangess128_mask_round): New builtins.
10029 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
10030 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
10031 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
10032 "<round_saeonly_constraint>")): Changed to ...
10033 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
10034 "<round_saeonly_scalar_constraint>")): ... this.
10035 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
10036 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
10037 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
10038 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
10039 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
10041 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
10043 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
10045 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
10046 Remove all values except native, 8540 and 8548.
10048 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
10050 * config/i386/i386.c (ix86_output_function_return): Pass
10051 INVALID_REGNUM, instead of -1, as invalid register number to
10052 indirect_thunk_name and output_indirect_thunk.
10054 2018-02-02 Julia Koval <julia.koval@intel.com>
10056 * config.gcc: Add -march=icelake.
10057 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
10058 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
10059 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
10060 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
10061 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
10062 (processor_target_table): Add icelake.
10063 (ix86_option_override_internal): Handle new PTAs.
10064 (get_builtin_code_for_version): Handle icelake.
10065 (M_INTEL_COREI7_ICELAKE): New.
10066 (fold_builtin_cpu): Handle icelake.
10067 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
10068 * doc/invoke.texi: Add -march=icelake.
10070 2018-02-02 Julia Koval <julia.koval@intel.com>
10072 * config/i386/i386.c (ix86_option_override_internal): Change flags type
10073 to wide_int_bitmask.
10074 * wide-int-bitmask.h: New.
10076 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10079 * config/i386/i386.md: Replace Pmode with word_mode in
10080 builtin_setjmp_setup and builtin_longjmp to support x32.
10082 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
10086 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
10088 (rs6000_supported_cpu_names): New static variable.
10089 (linux_cpu_translation_table): Likewise.
10090 (elf_platform) <cpu>: Define new static variable and use it.
10091 Translate kernel AT_PLATFORM name to canonical name if needed.
10092 Error if platform name is unknown.
10094 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
10097 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
10099 2018-02-01 Jeff Law <law@redhat.com>
10102 * config/i386/i386.c (release_scratch_register_on_entry): Add new
10103 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
10104 the scratch if RELEASE_VIA_POP is false.
10105 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
10106 If we have to save a temporary register, decrement SIZE appropriately.
10107 Pass new arguments to release_scratch_register_on_entry.
10108 (ix86_adjust_stack_and_probe): Likewise.
10109 (ix86_emit_probe_stack_range): Pass new arguments to
10110 release_scratch_register_on_entry.
10112 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
10114 PR rtl-optimization/84157
10115 * combine.c (change_zero_ext): Use REG_P predicate in
10116 front of HARD_REGISTER_P predicate.
10118 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
10120 * config/avr/avr.c (avr_option_override): Move disabling of
10121 -fdelete-null-pointer-checks to...
10122 * common/config/avr/avr-common.c (avr_option_optimization_table):
10125 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10127 PR tree-optimization/81635
10128 * tree-data-ref.c (split_constant_offset_1): For types that
10129 wrap on overflow, try to use range info to prove that wrapping
10132 2018-02-01 Renlin Li <renlin.li@arm.com>
10135 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
10136 TAILCALL_ADDR_REGS.
10137 (aarch64_register_move_cost): Likewise.
10138 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
10139 TAILCALL_ADDR_REGS.
10140 (REG_CLASS_NAMES): Likewise.
10141 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
10142 TAILCALL_ADDR_REGS. Remove IP registers.
10143 * config/aarch64/aarch64.md (Ucs): Update register constraint.
10145 2018-02-01 Richard Biener <rguenther@suse.de>
10147 * domwalk.h (dom_walker::dom_walker): Add additional constructor
10148 for specifying RPO order and allow NULL for that.
10149 * domwalk.c (dom_walker::dom_walker): Likewise.
10150 (dom_walker::walk): Handle NULL RPO order.
10151 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
10153 (rewrite_update_dom_walker): Likewise.
10154 (mark_def_dom_walker): Likewise.
10156 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10158 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
10159 (aarch64_maybe_expand_sve_subreg_move): Declare.
10160 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
10161 * config/aarch64/predicates.md (aarch64_any_register_operand): New
10163 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
10164 that are semantically a reverse operation.
10165 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
10166 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
10167 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
10169 (aarch64_can_change_mode_class): For big-endian, forbid changes
10170 between two SVE modes if they have different element sizes.
10172 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10174 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
10175 the TImode handling for big-endian targets.
10177 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10179 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
10180 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
10182 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
10183 Remove BSWAP handing for big-endian targets and use the form of
10184 LD1RQ appropariate for the mode.
10186 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10188 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
10189 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
10190 duplicated element.
10192 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10195 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
10196 check for operands that need to go through aarch64_sve_reload_be.
10198 2018-02-01 Jakub Jelinek <jakub@redhat.com>
10200 PR tree-optimization/81661
10201 PR tree-optimization/84117
10202 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
10203 * tree-eh.c: Include gimplify.h.
10204 (find_trapping_overflow, replace_trapping_overflow,
10205 rewrite_to_non_trapping_overflow): New functions.
10206 * tree-vect-loop.c: Include tree-eh.h.
10207 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
10208 * tree-data-ref.c: Include tree-eh.h.
10209 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
10211 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
10213 PR rtl-optimization/84123
10214 * combine.c (change_zero_ext): Check if hard register satisfies
10215 can_change_dest_mode before calling gen_lowpart_SUBREG.
10217 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
10220 * ira.c (ira_init_register_move_cost): Remove assert.
10222 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10224 PR rtl-optimization/84071
10225 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
10226 * doc/tm.texi: Regenerate.
10228 2018-01-31 Richard Biener <rguenther@suse.de>
10230 PR tree-optimization/84132
10231 * tree-data-ref.c (analyze_miv_subscript): Properly
10232 check whether evolution_function_is_affine_multivariate_p
10233 before calling gcd_of_steps_may_divide_p.
10235 2018-01-31 Julia Koval <julia.koval@intel.com>
10238 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
10239 * config/i386/i386.md (rdpid_rex64) New.
10240 (rdpid): Make 32bit only.
10242 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
10245 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
10246 an IDENTIFIER_NODE for FUNCTION_TYPE's.
10248 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10251 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10253 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10255 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10257 PR rtl-optimization/84071
10258 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
10259 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
10261 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
10263 * config/arc/arc.c (arc_handle_aux_attribute): New function.
10264 (arc_attribute_table): Add 'aux' attribute.
10265 (arc_in_small_data_p): Consider aux like variables.
10266 (arc_is_aux_reg_p): New function.
10267 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
10268 (arc_get_aux_arg): New function.
10269 (prepare_move_operands): Handle aux-register access.
10270 (arc_handle_aux_attribute): New function.
10271 * doc/extend.texi (ARC Variable attributes): Add subsection.
10273 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
10275 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
10276 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
10277 (arc_attribute_table): Add 'uncached' attribute.
10278 (arc_print_operand): Print '.di' flag for uncached memory
10280 (arc_in_small_data_p): Do not consider for small data the uncached
10282 (arc_is_uncached_mem_p): New function.
10283 * config/arc/predicates.md (compact_store_memory_operand): Check
10284 for uncached memory accesses.
10285 (nonvol_nonimm_operand): Likewise.
10286 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
10288 2018-01-31 Jakub Jelinek <jakub@redhat.com>
10291 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
10292 falign-loops=): Add Optimization flag.
10294 2018-01-30 Jeff Law <law@redhat.com>
10297 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
10298 INT_REGISTERS_SAVED. Check it prior to calling
10299 get_scratch_register_on_entry.
10300 (ix86_adjust_stack_and_probe): Similarly.
10301 (ix86_emit_probe_stack_range): Similarly.
10302 (ix86_expand_prologue): Corresponding changes.
10304 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10307 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
10308 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
10310 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
10313 * lra-constraints.c (curr_insn_transform): Process AND in the
10316 2018-01-30 Jakub Jelinek <jakub@redhat.com>
10318 PR rtl-optimization/83986
10319 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
10320 dependence against last_pending_memory_flush in addition to
10321 pending_jump_insns.
10323 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
10325 PR tree-optimization/81611
10326 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
10329 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10332 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
10335 2018-01-30 Richard Biener <rguenther@suse.de>
10336 Jakub Jelinek <jakub@redhat.com>
10338 PR tree-optimization/84111
10339 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
10340 inner loops added during recursion, as they don't have up-to-date
10343 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
10346 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
10347 (can_inline_edge_by_limits_p): ... here.
10348 (can_early_inline_edge_p, check_callers,
10349 update_caller_keys, update_callee_keys, recursive_inlining,
10350 add_new_edges_to_heap, speculation_useful_p,
10351 inline_small_functions,
10352 inline_small_functions, flatten_function,
10353 inline_to_all_callers_1): Update.
10355 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
10357 * profile-count.c (profile_count::combine_with_ipa_count): Handle
10360 2018-01-30 Richard Biener <rguenther@suse.de>
10362 PR tree-optimization/83008
10363 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
10364 invariant and constant vector uses in stmts when they need
10365 more than one stmt.
10367 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10370 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
10371 * configure: Regenerate.
10373 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
10375 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
10377 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
10378 Use gen_rtx_REG rather than gen_lowpart.
10380 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
10382 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
10383 rather than 0 when creating partial subregs.
10385 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
10387 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
10390 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
10393 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
10394 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
10395 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
10396 flags. This restores the settings used before the 2017-07-24.
10397 Turning off pre increment/decrement/modify allows IVOPTS to
10398 optimize DF/SF loops where the index is an int.
10400 2018-01-29 Richard Biener <rguenther@suse.de>
10401 Kelvin Nilsen <kelvin@gcc.gnu.org>
10404 * tree-vect-stmts.c (vectorizable_call): Don't call
10405 targetm.vectorize_builtin_md_vectorized_function if callee is
10408 2018-01-22 Carl Love <cel@us.ibm.com>
10410 * doc/extend.tex: Fix typo in second arg in
10411 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
10413 2018-01-29 Richard Biener <rguenther@suse.de>
10415 PR tree-optimization/84086
10416 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
10417 (flush_ssaname_freelist): When SSA names were released reset
10418 the SCEV hash table.
10420 2018-01-29 Richard Biener <rguenther@suse.de>
10422 PR tree-optimization/84057
10423 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
10424 removed paths when removing edges.
10426 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
10428 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
10429 -mfunction-return=@var{choice}.
10431 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
10433 PR diagnostic/84034
10434 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
10435 Handle CR like TAB.
10436 (layout::print_source_line): Likewise.
10437 (test_get_line_width_without_trailing_whitespace): Add test cases.
10439 2018-01-27 Jakub Jelinek <jakub@redhat.com>
10441 PR middle-end/84040
10442 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
10445 2018-01-26 Jim Wilson <jimw@sifive.com>
10447 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
10449 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
10452 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10454 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
10455 and CMP + SUB-immediate -> SUBS.
10457 2018-01-26 Martin Sebor <msebor@redhat.com>
10459 PR tree-optimization/83896
10460 * tree-ssa-strlen.c (get_string_len): Rename...
10461 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
10462 Avoid assuming length is constant.
10463 (handle_char_store): Use HOST_WIDE_INT for string length.
10465 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
10468 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
10469 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
10471 2018-01-26 Richard Biener <rguenther@suse.de>
10473 PR rtl-optimization/84003
10474 * dse.c (record_store): Only record redundant stores when
10475 the earlier store aliases at least all accesses the later one does.
10477 2018-01-26 Jakub Jelinek <jakub@redhat.com>
10479 PR rtl-optimization/83985
10480 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
10481 REG_CFA_RESTORE insns.
10482 (delete_unmarked_insns): Don't ignore separate shrink wrapping
10483 REG_CFA_RESTORE insns here.
10486 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
10487 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
10489 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10491 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
10492 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
10493 (arc_init): Likewise.
10494 (arc_override_options): Likewise.
10495 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
10497 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
10499 * config/arc/arc.h (TARGET_DBNZ): Define.
10500 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
10501 properly set the tune attribute.
10502 (dbnz): Use TARGET_DBNZ guard.
10503 * config/arc/arc.opt (mtune): Add core3 option.
10505 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10507 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
10508 recognize new pic like addresses.
10509 (arc_delegitimize_address): Clean up.
10511 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10513 * config/arc/arc-arches.def: Option mrf16 valid for all
10515 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
10516 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
10517 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
10518 * config/arc/arc-tables.opt: Regenerate.
10519 * config/arc/arc.c (arc_conditional_register_usage): Handle
10520 reduced register file case.
10521 (arc_file_start): Set must have build attributes.
10522 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
10523 mrf16 option value.
10524 * config/arc/arc.opt (mrf16): Add new option.
10525 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
10526 * config/arc/genmultilib.awk: Handle new mrf16 option.
10527 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
10528 * config/arc/t-multilib: Regenerate.
10529 * doc/invoke.texi (ARC Options): Document mrf16 option.
10531 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10533 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
10534 * config/arc/arc.c (arc_handle_secure_attribute): New function.
10535 (arc_attribute_table): Add 'secure_call' attribute.
10536 (arc_print_operand): Print secure call operand.
10537 (arc_function_ok_for_sibcall): Don't optimize tail calls when
10539 (arc_is_secure_call_p): New function. * config/arc/arc.md
10540 (call_i): Add support for sjli instruction.
10541 (call_value_i): Likewise.
10542 * config/arc/constraints.md (Csc): New constraint.
10544 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10545 John Eric Martin <John.Martin@emmicro-us.com>
10547 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
10548 * config/arc/arc.c (_arc_jli_section): New struct.
10549 (arc_jli_section): New type.
10550 (rc_jli_sections): New static variable.
10551 (arc_handle_jli_attribute): New function.
10552 (arc_attribute_table): Add jli_always and jli_fixed attribute.
10553 (arc_file_end): New function.
10554 (TARGET_ASM_FILE_END): Define.
10555 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
10556 (arc_add_jli_section): New function.
10557 (jli_call_scan): Likewise.
10558 (arc_reorg): Call jli_call_scan.
10559 (arc_output_addsi): Remove 'S' from printing asm operand.
10560 (arc_is_jli_call_p): New function.
10561 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
10563 (movhi_insn): Likewise.
10564 (movsi_insn): Likewise.
10565 (movsi_set_cc_insn): Likewise.
10566 (loadqi_update): Likewise.
10567 (load_zeroextendqisi_update): Likewise.
10568 (load_signextendqisi_update): Likewise.
10569 (loadhi_update): Likewise.
10570 (load_zeroextendhisi_update): Likewise.
10571 (load_signextendhisi_update): Likewise.
10572 (loadsi_update): Likewise.
10573 (loadsf_update): Likewise.
10574 (movsicc_insn): Likewise.
10575 (bset_insn): Likewise.
10576 (bxor_insn): Likewise.
10577 (bclr_insn): Likewise.
10578 (bmsk_insn): Likewise.
10579 (bicsi3_insn): Likewise.
10580 (cmpsi_cc_c_insn): Likewise.
10581 (movsi_ne): Likewise.
10582 (movsi_cond_exec): Likewise.
10583 (clrsbsi2): Likewise.
10584 (norm_f): Likewise.
10594 (call_i): Remove 'S' asm letter, add jli instruction.
10595 (call_value_i): Likewise.
10596 * config/arc/arc.op (mjli-always): New option.
10597 * config/arc/constraints.md (Cji): New constraint.
10598 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
10600 (subsf3_fpx): Likewise.
10601 (mulsf3_fpx): Likewise.
10602 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
10604 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
10605 function attrbutes.
10606 * doc/invoke.texi (ARC): Document mjli-always option.
10608 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
10610 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
10611 avoid addition with 0 and use incw and decw where possible.
10613 2018-01-26 Richard Biener <rguenther@suse.de>
10615 PR tree-optimization/81082
10616 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
10617 association if it requires casting to unsigned.
10618 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
10619 from fold_plusminus_mult_expr to catch important cases late when
10620 range info is available.
10622 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10624 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
10625 * configure.ac (hidden_linkonce): New test.
10626 * configure: Regenerate.
10627 * config.in: Regenerate.
10629 2018-01-26 Julia Koval <julia.koval@intel.com>
10631 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
10632 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
10633 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
10634 _mm_mask_bitshuffle_epi64_mask): Fix type.
10635 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
10636 USI_FTYPE_V4DI_V4DI_USI): Remove.
10637 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
10638 __builtin_ia32_vpshufbitqmb256_mask,
10639 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
10640 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
10641 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
10643 2018-01-26 Alan Modra <amodra@gmail.com>
10646 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
10647 UNSPEC_VBPERMQ. Sort other unspecs.
10649 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
10651 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
10653 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
10655 PR middle-end/83055
10656 * predict.c (drop_profile): Do not push/pop cfun; update also
10658 (handle_missing_profiles): Fix logic looking for zero profiles.
10660 2018-01-25 Jakub Jelinek <jakub@redhat.com>
10662 PR middle-end/83977
10663 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
10664 on functions with #pragma omp declare simd or functions with simd
10666 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
10667 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
10668 Remove trailing \n from warning_at calls.
10670 2018-01-25 Tom de Vries <tom@codesourcery.com>
10673 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
10674 for neutered workers.
10676 2018-01-24 Joseph Myers <joseph@codesourcery.com>
10679 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
10680 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
10682 2018-01-24 Jeff Law <law@redhat.com>
10685 * i386.c (get_probe_interval): Move to earlier point.
10686 (ix86_compute_frame_layout): If -fstack-clash-protection and
10687 the frame is larger than the probe interval, then use pushes
10688 to save registers rather than reg->mem moves.
10689 (ix86_expand_prologue): Remove conditional for int_registers_saved
10692 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
10695 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
10696 min/max for never referenced object.
10698 2018-01-24 Jakub Jelinek <jakub@redhat.com>
10700 PR middle-end/83977
10701 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
10703 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
10704 attributes from DECL_ATTRIBUTES (decl) without affecting
10705 DECL_ATTRIBUTES (current_function_decl).
10706 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
10707 functions with non-NULL DECL_ABSTRACT_ORIGIN.
10709 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
10711 PR tree-optimization/83979
10712 * fold-const.c (fold_comparison): Use constant_boolean_node
10713 instead of boolean_{true,false}_node.
10715 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
10717 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
10720 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10722 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10723 Simplify the clause that sets the length attribute.
10724 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10725 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
10726 clause that sets the length attribute.
10727 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10729 2018-01-24 Tom de Vries <tom@codesourcery.com>
10732 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
10733 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
10734 Add strict parameter.
10735 (prevent_branch_around_nothing): Insert dummy insn between branch to
10736 label and label with no ptx insn inbetween.
10737 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
10739 2018-01-24 Tom de Vries <tom@codesourcery.com>
10742 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
10743 for neutered threads in warp.
10744 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
10746 2018-01-24 Richard Biener <rguenther@suse.de>
10748 PR tree-optimization/83176
10749 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
10752 2018-01-24 Richard Biener <rguenther@suse.de>
10754 PR tree-optimization/82819
10755 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
10756 code generating pluses that are no-ops in the target precision.
10758 2018-01-24 Richard Biener <rguenther@suse.de>
10760 PR middle-end/84000
10761 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
10763 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10765 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
10766 to merge probabilities.
10767 * predict.c (probably_never_executed): Also mark as cold functions
10768 with global 0 profile and guessed local profile.
10769 * profile-count.c (profile_probability::combine_with_count): New
10771 * profile-count.h (profile_probability::operator*,
10772 profile_probability::operator*=, profile_probability::operator/,
10773 profile_probability::operator/=): Reduce precision to adjusted
10774 and set value to guessed on contradictory divisions.
10775 (profile_probability::combine_with_freq): Remove.
10776 (profile_probability::combine_wiht_count): Declare.
10777 (profile_count::force_nonzero):: Set to adjusted.
10778 (profile_count::probability_in):: Set quality to adjusted.
10779 * tree-ssa-tail-merge.c (replace_block_by): Use
10780 combine_with_count.
10782 2018-01-23 Andrew Waterman <andrew@sifive.com>
10783 Jim Wilson <jimw@sifive.com>
10785 * config/riscv/riscv.c (riscv_stack_boundary): New.
10786 (riscv_option_override): Set riscv_stack_boundary. Handle
10787 riscv_preferred_stack_boundary_arg.
10788 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
10789 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
10790 (STACK_BOUNDARY): Set to riscv_stack_boundary.
10791 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
10792 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
10793 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
10795 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
10798 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
10799 of struct ix86_frame.
10800 (ix86_expand_epilogue): Likewise. Add a local variable for
10801 the reg_save_offset field in struct ix86_frame.
10803 2018-01-23 Bin Cheng <bin.cheng@arm.com>
10805 PR tree-optimization/82604
10806 * tree-loop-distribution.c (enum partition_kind): New enum item
10807 PKIND_PARTIAL_MEMSET.
10808 (partition_builtin_p): Support above new enum item.
10809 (generate_code_for_partition): Ditto.
10810 (compute_access_range): Differentiate cases that equality can be
10811 proven at all loops, the innermost loops or no loops.
10812 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
10813 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
10814 (finalize_partitions, distribute_loop): Don't fuse partition of
10815 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
10816 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
10817 parloop is enabled.
10819 2018-01-23 Martin Liska <mliska@suse.cz>
10821 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
10822 order to ignore the predictor.
10823 (PRED_POLYMORPHIC_CALL): Likewise.
10824 (PRED_RECURSIVE_CALL): Likewise.
10826 2018-01-23 Martin Liska <mliska@suse.cz>
10828 * tree-profile.c (tree_profiling): Print function header to
10829 aware reader which function we are working on.
10830 * value-prof.c (gimple_find_values_to_profile): Do not print
10831 not interesting value histograms.
10833 2018-01-23 Martin Liska <mliska@suse.cz>
10835 * profile-count.h (enum profile_quality): Add
10836 profile_uninitialized as the first value. Do not number values
10837 as they are zero based.
10838 (profile_count::verify): Update sanity check.
10839 (profile_probability::verify): Likewise.
10841 2018-01-23 Nathan Sidwell <nathan@acm.org>
10843 * doc/invoke.texi (ffor-scope): Deprecate.
10845 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10847 PR tree-optimization/83510
10848 * domwalk.c (set_all_edges_as_executable): New function.
10849 (dom_walker::dom_walker): Convert bool param
10850 "skip_unreachable_blocks" to enum reachability. Move setup of
10851 edge flags to set_all_edges_as_executable and only do it when
10852 reachability is REACHABLE_BLOCKS.
10853 * domwalk.h (enum dom_walker::reachability): New enum.
10854 (dom_walker::dom_walker): Convert bool param
10855 "skip_unreachable_blocks" to enum reachability.
10856 (set_all_edges_as_executable): New decl.
10857 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
10858 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
10860 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
10861 but converting true to REACHABLE_BLOCKS.
10862 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
10864 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
10865 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
10866 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
10868 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
10869 if check_all_array_refs will be called.
10871 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10873 * tree.c (selftest::test_location_wrappers): Add more test
10876 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10878 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
10879 (selftest::test_bit_in_range): Likewise.
10881 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10884 * doc/sourcebuild.texi (vect_float): Say that the selector
10885 only describes the situation when -funsafe-math-optimizations is on.
10886 (vect_float_strict): Document.
10888 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10890 PR tree-optimization/83965
10891 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
10892 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
10893 instead of checking only for a reduction.
10894 (vect_recog_widen_sum_pattern): Likewise.
10896 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10898 * predict.c (probably_never_executed): Only use precise profile info.
10899 (compute_function_frequency): Skip after inlining hack since we now
10900 have quality checking.
10902 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10904 * profile-count.h (profile_probability::very_unlikely,
10905 profile_probability::unlikely, profile_probability::even): Set
10906 precision to guessed.
10908 2018-01-23 Richard Biener <rguenther@suse.de>
10910 PR tree-optimization/83963
10911 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
10912 Properly terminate dominator walk when crossing the exit edge not
10913 when visiting its source block.
10915 2018-01-23 Jakub Jelinek <jakub@redhat.com>
10918 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
10919 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
10921 2018-01-22 Jakub Jelinek <jakub@redhat.com>
10923 PR tree-optimization/83957
10924 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
10925 semicolon after for body surrounded by braces.
10927 PR tree-optimization/83081
10928 * profile-count.h (profile_probability::split): New method.
10929 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
10930 Use profile_probability::split.
10931 (do_compare_rtx_and_jump): Fix adjustment of probabilities
10932 when splitting a single conditional jump into 2.
10934 2018-01-22 David Malcolm <dmalcolm@redhat.com>
10936 PR tree-optimization/69452
10937 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
10940 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10942 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
10943 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
10944 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
10946 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10948 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
10949 * config/rl78/rl78.md (movdi): New define_expand.
10950 * config/rl78/rl78.c (rl78_split_movdi): New function.
10952 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
10955 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
10957 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
10958 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
10959 128-bit to produce an UNSPEC move to get the double word with the
10960 signbit and then a shift directly to do signbit.
10961 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
10962 implementation with a new version that just does either a direct
10963 move or a regular move. Move memory interface to separate insns.
10964 Move insns so they are next to the expander.
10965 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
10966 with signbit move. Split big and little endian case.
10967 (signbit<mode>2_dm_mem_le): Likewise.
10968 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
10969 (signbit<mode>2_dm2): Likewise.
10971 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10973 * config/rl78/rl78.md (anddi3): New define_expand.
10975 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10977 * config/rl78/rl78.md (umindi3): New define_expand.
10979 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10981 * config/rl78/rl78.md (smindi3): New define_expand.
10983 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10985 * config/rl78/rl78.md (smaxdi3): New define_expand.
10987 2018-01-22 Carl Love <cel@us.ibm.com>
10989 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
10990 LVX_V1TI): Add macro expansion.
10991 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
10992 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
10993 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
10994 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
10995 Change check to determine if the instruction is a byte reversing
10996 entry. Fix typo in comment.
10997 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
10998 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
10999 Add def_builtin calls for new builtins.
11000 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
11001 Add define_insn expansion.
11003 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11005 * config/rl78/rl78.md (umaxdi3): New define_expand.
11007 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11009 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
11010 for non-QImode registers.
11012 2018-01-22 Richard Biener <rguenther@suse.de>
11014 PR tree-optimization/83963
11015 * graphite-scop-detection.c (scop_detection::get_sese): Delay
11016 including the loop exit block.
11017 (scop_detection::merge_sese): Likewise.
11018 (scop_detection::add_scop): Do it here instead.
11020 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11022 * doc/sourcebuild.texi (arm_softfloat): Document.
11024 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
11027 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
11028 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
11029 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
11031 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11032 David Edelsohn <dje.gcc@gmail.com>
11035 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11036 Change "crset eq" to "crset 2".
11037 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11038 (*call_indirect_aix<mode>_nospec): Likewise.
11039 (*call_value_indirect_aix<mode>_nospec): Likewise.
11040 (*call_indirect_elfv2<mode>_nospec): Likewise.
11041 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
11042 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
11043 change assembly output from . to $.
11044 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11045 (indirect_jump<mode>_nospec): Change assembly output from . to $.
11046 (*tablejump<mode>_internal1_nospec): Likewise.
11048 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
11051 * config/sh/sh_optimize_sett_clrt.cc:
11052 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
11054 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
11056 PR tree-optimization/83940
11057 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
11058 offset_dt to vect_constant_def rather than vect_unknown_def_type.
11059 (vect_check_load_store_mask): Add a mask_dt_out parameter and
11060 use it to pass back the definition type.
11061 (vect_check_store_rhs): Likewise rhs_dt_out.
11062 (vect_build_gather_load_calls): Add a mask_dt argument and use
11063 it instead of a call to vect_is_simple_use.
11064 (vectorizable_store): Update calls to vect_check_load_store_mask
11065 and vect_check_store_rhs. Use the dt returned by the latter instead
11066 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
11067 instead of calls to vect_is_simple_use. Pass the scalar rather
11068 than the vector operand to vect_is_simple_use when handling
11069 second and subsequent copies of an rhs value.
11070 (vectorizable_load): Update calls to vect_check_load_store_mask
11071 and vect_build_gather_load_calls. Use the cached mask_dt and
11072 gs_info.offset_dt instead of calls to vect_is_simple_use.
11074 2018-01-20 Jakub Jelinek <jakub@redhat.com>
11076 PR middle-end/83945
11077 * tree-emutls.c: Include gimplify.h.
11078 (lower_emutls_2): New function.
11079 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
11080 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
11081 it before further processing.
11084 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
11085 UINTVAL (trueop1) instead of INTVAL (op1).
11087 2018-01-19 Jakub Jelinek <jakub@redhat.com>
11091 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
11092 INCOMING_FRAME_SP_OFFSET if not defined.
11093 (scan_trace): Add ENTRY argument. If true and
11094 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
11095 emit a note to adjust the CFA offset.
11096 (create_cfi_notes): Adjust scan_trace callers.
11097 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
11098 INCOMING_FRAME_SP_OFFSET in the CIE.
11099 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
11100 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
11102 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
11103 * doc/tm.texi: Regenerated.
11105 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11107 PR rtl-optimization/83147
11108 * lra-constraints.c (remove_inheritance_pseudos): Use
11109 lra_substitute_pseudo_within_insn.
11111 2018-01-19 Tom de Vries <tom@codesourcery.com>
11112 Cesar Philippidis <cesar@codesourcery.com>
11115 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
11117 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
11120 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
11121 spaces for function labels.
11123 2018-01-19 Martin Liska <mliska@suse.cz>
11125 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
11126 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
11127 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
11128 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
11129 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
11130 (PRED_CONST_RETURN): Change from 69 to 65.
11131 (PRED_NULL_RETURN): Change from 91 to 71.
11132 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
11133 (PRED_LOOP_GUARD): Change from 66 to 73.
11135 2018-01-19 Martin Liska <mliska@suse.cz>
11137 * predict.c (predict_insn_def): Add new assert.
11138 (struct branch_predictor): Change type to signed integer.
11139 (test_prediction_value_range): Amend test to cover
11140 PROB_UNINITIALIZED.
11141 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
11142 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
11143 (PRED_LOOP_ITERATIONS_MAX): Likewise.
11144 (PRED_LOOP_IV_COMPARE): Likewise.
11145 * predict.h (PROB_UNINITIALIZED): Define new constant.
11147 2018-01-19 Martin Liska <mliska@suse.cz>
11149 * predict.c (dump_prediction): Add new format for
11150 analyze_brprob.py script which is enabled with -details
11152 * profile-count.h (precise_p): New function.
11154 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
11156 PR tree-optimization/83922
11157 * tree-vect-loop.c (vect_verify_full_masking): Return false if
11158 there are no statements that need masking.
11159 (vect_active_double_reduction_p): New function.
11160 (vect_analyze_loop_operations): Use it when handling phis that
11161 are not in the loop header.
11163 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
11165 PR tree-optimization/83914
11166 * tree-vect-loop.c (vectorizable_induction): Don't convert
11167 init_expr or apply the peeling adjustment for inductions
11168 that are nested within the vectorized loop.
11170 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11172 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
11175 2018-01-18 Jakub Jelinek <jakub@redhat.com>
11179 * function.h (gimplify_parameters): Add gimple_seq * argument.
11180 * function.c: Include gimple.h and options.h.
11181 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
11182 for the added local temporaries if needed.
11183 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
11184 if there are any parameter cleanups, wrap whole body into a
11185 try/finally with the cleanups.
11187 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
11190 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
11191 Use GET_MODE_CLASS for scalar floating point.
11193 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
11197 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
11198 Fix call of call_cgraph_insertion_hooks.
11200 2018-01-18 Martin Sebor <msebor@redhat.com>
11202 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
11204 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
11207 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
11210 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
11213 * common.opt: (-ffile-prefix-map): New option.
11214 * opts.c (common_handle_option): Defer it.
11215 * opts-global.c (handle_common_deferred_options): Handle it.
11216 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
11217 * file-prefix-map.h: New file.
11218 (remap_debug_filename, add_debug_prefix_map): ...here.
11219 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
11220 * final.c (debug_prefix_map, add_debug_prefix_map
11221 remap_debug_filename): Move to...
11222 * file-prefix-map.c: New file.
11223 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
11224 generalize, get rid of alloca(), use strrchr() instead of strchr().
11225 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
11226 Implement in terms of add_prefix_map().
11227 (remap_macro_filename, remap_debug_filename): Implement in term of
11229 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
11230 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
11231 * dbxout.c: Include file-prefix-map.h.
11232 * varasm.c: Likewise.
11233 * vmsdbgout.c: Likewise.
11234 * xcoffout.c: Likewise.
11235 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
11236 * doc/cppopts.texi (-fmacro-prefix-map): Document.
11237 * doc/invoke.texi (-ffile-prefix-map): Document.
11238 (-fdebug-prefix-map): Update description.
11240 2018-01-18 Martin Liska <mliska@suse.cz>
11242 * config/i386/i386.c (indirect_thunk_name): Document that also
11244 (output_indirect_thunk): Document why both instructions
11245 (pause and lfence) are generated.
11247 2018-01-18 Richard Biener <rguenther@suse.de>
11249 PR tree-optimization/83887
11250 * graphite-scop-detection.c
11251 (scop_detection::get_nearest_dom_with_single_entry): Remove.
11252 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
11253 (scop_detection::merge_sese): Re-implement with a flood-fill
11254 algorithm that properly finds a SESE region if it exists.
11256 2018-01-18 Jakub Jelinek <jakub@redhat.com>
11259 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
11260 pointer_diff optimizations use view_convert instead of convert.
11262 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11264 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11265 Generate different code for -mno-speculate-indirect-jumps.
11266 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11267 (*call_indirect_aix<mode>): Disable for
11268 -mno-speculate-indirect-jumps.
11269 (*call_indirect_aix<mode>_nospec): New define_insn.
11270 (*call_value_indirect_aix<mode>): Disable for
11271 -mno-speculate-indirect-jumps.
11272 (*call_value_indirect_aix<mode>_nospec): New define_insn.
11273 (*sibcall_nonlocal_sysv<mode>): Generate different code for
11274 -mno-speculate-indirect-jumps.
11275 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11277 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
11279 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
11280 long double type, set the flags for noting the default long double
11281 type, even if we don't pass or return a long double type.
11283 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
11286 * ipa-inline.c (flatten_function): Do not overwrite final inlining
11289 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
11291 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
11292 support for merge[hl].
11293 (fold_mergehl_helper): New helper function.
11294 (tree-vector-builder.h): New #include for tree_vector_builder usage.
11295 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
11296 (altivec_vmrglw_direct): Add xxmrglw insn.
11298 2018-01-17 Andrew Waterman <andrew@sifive.com>
11300 * config/riscv/riscv.c (riscv_conditional_register_usage): If
11301 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
11303 2018-01-17 David Malcolm <dmalcolm@redhat.com>
11306 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
11307 call the lto_location_cache before reading the
11308 DECL_SOURCE_LOCATION of the types.
11310 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
11311 Richard Sandiford <richard.sandiford@linaro.org>
11313 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
11314 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
11315 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
11316 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
11317 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
11319 * config/aarch64/constraints.md (aarch64_movti_operand):
11321 * config/aarch64/predicates.md (Uti): Add new constraint.
11323 2018-01-17 Carl Love <cel@us.ibm.com>
11325 * config/rs6000/vsx.md (define_expand xl_len_r,
11326 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
11327 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
11329 (define_expand, define_insn): Move the shift left from the
11330 define_insn to the define_expand for lxvl and stxvl instructions.
11331 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
11332 and XL_LEN_R definitions to PURE.
11334 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
11336 * config/i386/i386.c (indirect_thunk_name): Declare regno
11337 as unsigned int. Compare regno with INVALID_REGNUM.
11338 (output_indirect_thunk): Ditto.
11339 (output_indirect_thunk_function): Ditto.
11340 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
11341 in the call to output_indirect_thunk_function.
11343 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
11345 PR middle-end/83884
11346 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
11347 rather than the size of inner_type to determine the stack slot size
11348 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
11350 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
11353 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
11356 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
11358 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
11359 endian Linux systems to optionally enable multilibs for selecting
11360 the long double type if the user configured an explicit type.
11361 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
11362 have no long double multilibs if not defined.
11363 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
11364 warn if the user used -mabi={ieee,ibm}longdouble and we built
11365 multilibs for long double.
11366 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
11367 appropriate multilib option.
11368 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
11370 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
11371 for building long double multilibs.
11372 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
11374 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
11376 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
11379 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
11381 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
11384 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
11387 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
11390 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
11392 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
11393 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
11395 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
11397 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
11398 different rtl trees depending on TARGET_64BIT.
11399 (rs6000_gen_lvx): Likewise.
11401 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
11403 * config/visium/visium.md (nop): Tweak comment.
11404 (hazard_nop): Likewise.
11406 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11408 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
11409 -mspeculate-indirect-jumps.
11410 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
11411 for -mno-speculate-indirect-jumps.
11412 (*call_indirect_elfv2<mode>_nospec): New define_insn.
11413 (*call_value_indirect_elfv2<mode>): Disable for
11414 -mno-speculate-indirect-jumps.
11415 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
11416 (indirect_jump): Emit different RTL for
11417 -mno-speculate-indirect-jumps.
11418 (*indirect_jump<mode>): Disable for
11419 -mno-speculate-indirect-jumps.
11420 (*indirect_jump<mode>_nospec): New define_insn.
11421 (tablejump): Emit different RTL for
11422 -mno-speculate-indirect-jumps.
11423 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
11424 (tablejumpsi_nospec): New define_expand.
11425 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
11426 (tablejumpdi_nospec): New define_expand.
11427 (*tablejump<mode>_internal1): Disable for
11428 -mno-speculate-indirect-jumps.
11429 (*tablejump<mode>_internal1_nospec): New define_insn.
11430 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
11433 2018-01-16 Artyom Skrobov tyomitch@gmail.com
11435 * caller-save.c (insert_save): Drop unnecessary parameter. All
11438 2018-01-16 Jakub Jelinek <jakub@redhat.com>
11439 Richard Biener <rguenth@suse.de>
11442 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
11443 return early, inline manually is_gimple_sizepos. Make sure if we
11444 call gimplify_expr we don't end up with a gimple constant.
11445 * tree.c (variably_modified_type_p): Don't return true for
11446 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
11447 * gimplify.h (is_gimple_sizepos): Remove.
11449 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11451 PR tree-optimization/83857
11452 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
11453 vectorizable_live_operation for pure SLP statements.
11454 (vectorizable_live_operation): Handle PHIs.
11456 2018-01-16 Richard Biener <rguenther@suse.de>
11458 PR tree-optimization/83867
11459 * tree-vect-stmts.c (vect_transform_stmt): Precompute
11460 nested_in_vect_loop_p since the scalar stmt may get invalidated.
11462 2018-01-16 Jakub Jelinek <jakub@redhat.com>
11465 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
11466 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
11467 If off is not INTEGER_CST, issue a may not be aligned warning
11468 rather than isn't aligned. Use isn%'t rather than isn't.
11469 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
11471 <case MULT_EXPR>: Improve the case when bottom and one of the
11472 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
11473 operand, in that case check if the other operand is multiple of
11474 bottom divided by the INTEGER_CST operand.
11476 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11479 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
11480 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
11481 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
11482 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
11483 * config/pa/pa.c (pa_function_arg_advance): Likewise.
11484 (pa_function_arg, pa_arg_partial_bytes): Likewise.
11485 (pa_function_arg_size): New function.
11487 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11489 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
11490 in a separate statement.
11492 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11494 PR tree-optimization/83847
11495 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
11496 group gathers and scatters.
11498 2018-01-16 Jakub Jelinek <jakub@redhat.com>
11500 PR rtl-optimization/86620
11501 * params.def (max-sched-ready-insns): Bump minimum value to 1.
11503 PR rtl-optimization/83213
11504 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
11505 to last if both are JUMP_INSNs.
11507 PR tree-optimization/83843
11508 * gimple-ssa-store-merging.c
11509 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
11510 store_immediate_info for bswap/nop orig_stores.
11512 2018-01-15 Andrew Waterman <andrew@sifive.com>
11514 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
11516 <UDIV>: Increase cost if !TARGET_DIV.
11518 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
11520 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
11521 (define_attr "cr_logical_3op"): New.
11522 (cceq_ior_compare): Adjust.
11523 (cceq_ior_compare_complement): Adjust.
11524 (*cceq_rev_compare): Adjust.
11525 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
11526 (is_cracked_insn): Adjust.
11527 (insn_must_be_first_in_group): Adjust.
11528 * config/rs6000/40x.md: Adjust.
11529 * config/rs6000/440.md: Adjust.
11530 * config/rs6000/476.md: Adjust.
11531 * config/rs6000/601.md: Adjust.
11532 * config/rs6000/603.md: Adjust.
11533 * config/rs6000/6xx.md: Adjust.
11534 * config/rs6000/7450.md: Adjust.
11535 * config/rs6000/7xx.md: Adjust.
11536 * config/rs6000/8540.md: Adjust.
11537 * config/rs6000/cell.md: Adjust.
11538 * config/rs6000/e300c2c3.md: Adjust.
11539 * config/rs6000/e500mc.md: Adjust.
11540 * config/rs6000/e500mc64.md: Adjust.
11541 * config/rs6000/e5500.md: Adjust.
11542 * config/rs6000/e6500.md: Adjust.
11543 * config/rs6000/mpc.md: Adjust.
11544 * config/rs6000/power4.md: Adjust.
11545 * config/rs6000/power5.md: Adjust.
11546 * config/rs6000/power6.md: Adjust.
11547 * config/rs6000/power7.md: Adjust.
11548 * config/rs6000/power8.md: Adjust.
11549 * config/rs6000/power9.md: Adjust.
11550 * config/rs6000/rs64.md: Adjust.
11551 * config/rs6000/titan.md: Adjust.
11553 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11555 * config/i386/predicates.md (indirect_branch_operand): Rewrite
11556 ix86_indirect_branch_register logic.
11558 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11560 * config/i386/constraints.md (Bs): Update
11561 ix86_indirect_branch_register check. Don't check
11562 ix86_indirect_branch_register with GOT_memory_operand.
11564 * config/i386/predicates.md (GOT_memory_operand): Don't check
11565 ix86_indirect_branch_register here.
11566 (GOT32_symbol_operand): Likewise.
11568 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11570 * config/i386/predicates.md (constant_call_address_operand):
11571 Rewrite ix86_indirect_branch_register logic.
11572 (sibcall_insn_operand): Likewise.
11574 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11576 * config/i386/constraints.md (Bs): Replace
11577 ix86_indirect_branch_thunk_register with
11578 ix86_indirect_branch_register.
11580 * config/i386/i386.md (indirect_jump): Likewise.
11581 (tablejump): Likewise.
11582 (*sibcall_memory): Likewise.
11583 (*sibcall_value_memory): Likewise.
11584 Peepholes of indirect call and jump via memory: Likewise.
11585 * config/i386/i386.opt: Likewise.
11586 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11587 (GOT_memory_operand): Likewise.
11588 (call_insn_operand): Likewise.
11589 (sibcall_insn_operand): Likewise.
11590 (GOT32_symbol_operand): Likewise.
11592 2018-01-15 Jakub Jelinek <jakub@redhat.com>
11594 PR middle-end/83837
11595 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
11596 type rather than type addr's type points to.
11597 (expand_omp_atomic_mutex): Likewise.
11598 (expand_omp_atomic): Likewise.
11600 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11603 * config/i386/i386.c (output_indirect_thunk_function): Use
11604 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
11605 for __x86_return_thunk.
11607 2018-01-15 Richard Biener <rguenther@suse.de>
11609 PR middle-end/83850
11610 * expmed.c (extract_bit_field_1): Fix typo.
11612 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11615 * config/arm/iterators.md (VF): New mode iterator.
11616 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
11617 Remove integer-related logic from pattern.
11618 (neon_vabd<mode>_3): Likewise.
11620 2018-01-15 Jakub Jelinek <jakub@redhat.com>
11622 PR middle-end/82694
11623 * common.opt (fstrict-overflow): No longer an alias.
11624 (fwrapv-pointer): New option.
11625 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
11626 also for pointer types based on flag_wrapv_pointer.
11627 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
11628 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
11629 opts->x_flag_wrapv got set.
11630 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
11631 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
11632 POINTER_TYPE_OVERFLOW_UNDEFINED.
11633 * match.pd: Likewise in address comparison pattern.
11634 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
11636 2018-01-15 Richard Biener <rguenther@suse.de>
11639 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
11640 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
11641 Reset type names to their identifier if their TYPE_DECL doesn't
11642 have linkage (and thus is used for ODR and devirt).
11643 (save_debug_info_for_decl): Remove.
11644 (save_debug_info_for_type): Likewise.
11645 (add_tree_to_fld_list): Adjust.
11646 * tree-pretty-print.c (dump_generic_node): Make dumping of
11647 type names more robust.
11649 2018-01-15 Richard Biener <rguenther@suse.de>
11651 * BASE-VER: Bump to 8.0.1.
11653 2018-01-14 Martin Sebor <msebor@redhat.com>
11656 * builtins.c (check_access): Avoid warning when the no-warning bit
11659 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
11661 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
11662 * ira-color (allocno_hard_regs_compare): Likewise.
11664 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
11667 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
11668 Use .pushsection/.popsection.
11670 2018-01-14 Martin Sebor <msebor@redhat.com>
11673 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
11675 2018-01-14 Jakub Jelinek <jakub@redhat.com>
11677 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
11678 entry from extra_headers.
11679 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
11680 extra_headers, make the list bitwise identical to the i?86-*-* one.
11682 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11684 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
11685 -mcmodel=large with -mindirect-branch=thunk,
11686 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
11687 -mfunction-return=thunk-extern.
11688 * doc/invoke.texi: Document -mcmodel=large is incompatible with
11689 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
11690 -mfunction-return=thunk and -mfunction-return=thunk-extern.
11692 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11694 * config/i386/i386.c (print_reg): Print the name of the full
11695 integer register without '%'.
11696 (ix86_print_operand): Handle 'V'.
11697 * doc/extend.texi: Document 'V' modifier.
11699 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11701 * config/i386/constraints.md (Bs): Disallow memory operand for
11702 -mindirect-branch-register.
11704 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11705 (GOT_memory_operand): Likewise.
11706 (call_insn_operand): Likewise.
11707 (sibcall_insn_operand): Likewise.
11708 (GOT32_symbol_operand): Likewise.
11709 * config/i386/i386.md (indirect_jump): Call convert_memory_address
11710 for -mindirect-branch-register.
11711 (tablejump): Likewise.
11712 (*sibcall_memory): Likewise.
11713 (*sibcall_value_memory): Likewise.
11714 Disallow peepholes of indirect call and jump via memory for
11715 -mindirect-branch-register.
11716 (*call_pop): Replace m with Bw.
11717 (*call_value_pop): Likewise.
11718 (*sibcall_pop_memory): Replace m with Bs.
11719 * config/i386/i386.opt (mindirect-branch-register): New option.
11720 * doc/invoke.texi: Document -mindirect-branch-register option.
11722 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11724 * config/i386/i386-protos.h (ix86_output_function_return): New.
11725 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
11726 set function_return_type.
11727 (indirect_thunk_name): Add ret_p to indicate thunk for function
11729 (output_indirect_thunk_function): Pass false to
11730 indirect_thunk_name.
11731 (ix86_output_indirect_branch_via_reg): Likewise.
11732 (ix86_output_indirect_branch_via_push): Likewise.
11733 (output_indirect_thunk_function): Create alias for function
11734 return thunk if regno < 0.
11735 (ix86_output_function_return): New function.
11736 (ix86_handle_fndecl_attribute): Handle function_return.
11737 (ix86_attribute_table): Add function_return.
11738 * config/i386/i386.h (machine_function): Add
11739 function_return_type.
11740 * config/i386/i386.md (simple_return_internal): Use
11741 ix86_output_function_return.
11742 (simple_return_internal_long): Likewise.
11743 * config/i386/i386.opt (mfunction-return=): New option.
11744 (indirect_branch): Mention -mfunction-return=.
11745 * doc/extend.texi: Document function_return function attribute.
11746 * doc/invoke.texi: Document -mfunction-return= option.
11748 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11750 * config/i386/i386-opts.h (indirect_branch): New.
11751 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
11752 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
11753 with local indirect jump when converting indirect call and jump.
11754 (ix86_set_indirect_branch_type): New.
11755 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
11756 (indirectlabelno): New.
11757 (indirect_thunk_needed): Likewise.
11758 (indirect_thunk_bnd_needed): Likewise.
11759 (indirect_thunks_used): Likewise.
11760 (indirect_thunks_bnd_used): Likewise.
11761 (INDIRECT_LABEL): Likewise.
11762 (indirect_thunk_name): Likewise.
11763 (output_indirect_thunk): Likewise.
11764 (output_indirect_thunk_function): Likewise.
11765 (ix86_output_indirect_branch_via_reg): Likewise.
11766 (ix86_output_indirect_branch_via_push): Likewise.
11767 (ix86_output_indirect_branch): Likewise.
11768 (ix86_output_indirect_jmp): Likewise.
11769 (ix86_code_end): Call output_indirect_thunk_function if needed.
11770 (ix86_output_call_insn): Call ix86_output_indirect_branch if
11772 (ix86_handle_fndecl_attribute): Handle indirect_branch.
11773 (ix86_attribute_table): Add indirect_branch.
11774 * config/i386/i386.h (machine_function): Add indirect_branch_type
11775 and has_local_indirect_jump.
11776 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
11778 (tablejump): Likewise.
11779 (*indirect_jump): Use ix86_output_indirect_jmp.
11780 (*tablejump_1): Likewise.
11781 (simple_return_indirect_internal): Likewise.
11782 * config/i386/i386.opt (mindirect-branch=): New option.
11783 (indirect_branch): New.
11786 (thunk-inline): Likewise.
11787 (thunk-extern): Likewise.
11788 * doc/extend.texi: Document indirect_branch function attribute.
11789 * doc/invoke.texi: Document -mindirect-branch= option.
11791 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
11794 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
11796 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
11798 * ipa-inline.c (want_inline_small_function_p): Return false if
11799 inlining has already failed with CIF_FINAL_ERROR.
11800 (update_caller_keys): Call want_inline_small_function_p before
11802 (update_callee_keys): Likewise.
11804 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
11806 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
11808 (rs6000_quadword_masked_address_p): Likewise.
11809 (quad_aligned_load_p): Likewise.
11810 (quad_aligned_store_p): Likewise.
11811 (const_load_sequence_p): Add comment to describe the outer-most loop.
11812 (mimic_memory_attributes_and_flags): New function.
11813 (rs6000_gen_stvx): Likewise.
11814 (replace_swapped_aligned_store): Likewise.
11815 (rs6000_gen_lvx): Likewise.
11816 (replace_swapped_aligned_load): Likewise.
11817 (replace_swapped_load_constant): Capitalize argument name in
11818 comment describing this function.
11819 (rs6000_analyze_swaps): Add a third pass to search for vector loads
11820 and stores that access quad-word aligned addresses and replace
11821 with stvx or lvx instructions when appropriate.
11822 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
11823 New function prototype.
11824 (rs6000_quadword_masked_address_p): Likewise.
11825 (rs6000_gen_lvx): Likewise.
11826 (rs6000_gen_stvx): Likewise.
11827 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
11828 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
11829 when memory address is aligned.
11830 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
11831 this split to select lvx instruction when memory address is aligned.
11832 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
11833 instruction when memory address is aligned.
11834 (*vsx_le_perm_load_v16qi): Likewise.
11835 (four unnamed splitters): Modify to select the stvx instruction
11836 when memory is aligned.
11838 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
11840 * predict.c (determine_unlikely_bbs): Handle correctly BBs
11841 which appears in the queue multiple times.
11843 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11844 Alan Hayward <alan.hayward@arm.com>
11845 David Sherwood <david.sherwood@arm.com>
11847 * tree-vectorizer.h (vec_lower_bound): New structure.
11848 (_loop_vec_info): Add check_nonzero and lower_bounds.
11849 (LOOP_VINFO_CHECK_NONZERO): New macro.
11850 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
11851 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
11852 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
11853 fields. Make seg_len the distance travelled, not including the
11855 (dr_direction_indicator): Declare.
11856 (dr_zero_step_indicator): Likewise.
11857 (dr_known_forward_stride_p): Likewise.
11858 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
11860 (runtime_alias_check_p): Allow runtime alias checks with
11862 (operator ==): Compare access_size and align.
11863 (prune_runtime_alias_test_list): Rework for new distinction between
11864 the access_size and seg_len.
11865 (create_intersect_range_checks_index): Likewise. Cope with polynomial
11867 (get_segment_min_max): New function.
11868 (create_intersect_range_checks): Use it.
11869 (dr_step_indicator): New function.
11870 (dr_direction_indicator): Likewise.
11871 (dr_zero_step_indicator): Likewise.
11872 (dr_known_forward_stride_p): Likewise.
11873 * tree-loop-distribution.c (data_ref_segment_size): Return
11874 DR_STEP * (niters - 1).
11875 (compute_alias_check_pairs): Update call to the dr_with_seg_len
11877 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
11878 (vect_preserves_scalar_order_p): New function, split out from...
11879 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
11880 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
11881 (vect_vfa_access_size): New function.
11882 (vect_vfa_align): Likewise.
11883 (vect_compile_time_alias): Take access_size_a and access_b arguments.
11884 (dump_lower_bound): New function.
11885 (vect_check_lower_bound): Likewise.
11886 (vect_small_gap_p): Likewise.
11887 (vectorizable_with_step_bound_p): Likewise.
11888 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
11889 depencies if the vectorization factor is 1. Convert the checks
11890 for nonzero steps into checks on the bounds of DR_STEP. Try using
11891 a bunds check for variable steps if the minimum required step is
11892 relatively small. Update calls to the dr_with_seg_len
11893 constructor and to vect_compile_time_alias.
11894 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
11896 (vect_loop_versioning): Call it.
11897 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
11899 (vect_estimate_min_profitable_iters): Account for any bounds checks.
11901 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11902 Alan Hayward <alan.hayward@arm.com>
11903 David Sherwood <david.sherwood@arm.com>
11905 * doc/sourcebuild.texi (vect_scatter_store): Document.
11906 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
11908 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
11910 * genopinit.c (main): Add supports_vec_scatter_store and
11911 supports_vec_scatter_store_cached to target_optabs.
11912 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
11913 IFN_MASK_SCATTER_STORE.
11914 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
11916 * internal-fn.h (internal_store_fn_p): Declare.
11917 (internal_fn_stored_value_index): Likewise.
11918 * internal-fn.c (scatter_store_direct): New macro.
11919 (expand_scatter_store_optab_fn): New function.
11920 (direct_scatter_store_optab_supported_p): New macro.
11921 (internal_store_fn_p): New function.
11922 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
11923 IFN_MASK_SCATTER_STORE.
11924 (internal_fn_mask_index): Likewise.
11925 (internal_fn_stored_value_index): New function.
11926 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
11927 for scatter stores.
11928 * optabs-query.h (supports_vec_scatter_store_p): Declare.
11929 * optabs-query.c (supports_vec_scatter_store_p): New function.
11930 * tree-vectorizer.h (vect_get_store_rhs): Declare.
11931 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
11932 true for scatter stores.
11933 (vect_gather_scatter_fn_p): Handle scatter stores too.
11934 (vect_check_gather_scatter): Consider using scatter stores if
11935 supports_vec_scatter_store_p.
11936 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
11937 scatter stores too.
11938 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11939 internal_fn_stored_value_index.
11940 (check_load_store_masking): Handle scatter stores too.
11941 (vect_get_store_rhs): Make public.
11942 (vectorizable_call): Use internal_store_fn_p.
11943 (vectorizable_store): Handle scatter store internal functions.
11944 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
11945 when deciding whether the end of the group has been reached.
11946 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
11947 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
11948 (mask_scatter_store<mode>): New insns.
11950 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11951 Alan Hayward <alan.hayward@arm.com>
11952 David Sherwood <david.sherwood@arm.com>
11954 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
11955 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
11956 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
11958 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
11959 Use vect_truncate_gather_scatter_offset if we can't treat the
11960 operation as a normal gather load or scatter store.
11961 (get_group_load_store_type): Take the gather_scatter_info
11962 as argument. Try using a gather load or scatter store for
11963 single-element groups.
11964 (get_load_store_type): Update calls to get_group_load_store_type
11965 and vect_use_strided_gather_scatters_p.
11967 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11968 Alan Hayward <alan.hayward@arm.com>
11969 David Sherwood <david.sherwood@arm.com>
11971 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
11972 optional tree argument.
11973 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
11975 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
11976 but continue to use the current value as a fallback.
11977 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
11978 to compare the updates.
11979 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
11980 (get_load_store_type): Use it when handling a strided access.
11981 (vect_get_strided_load_store_ops): New function.
11982 (vect_get_data_ptr_increment): Likewise.
11983 (vectorizable_load): Handle strided gather loads. Always pass
11984 a step to vect_create_data_ref_ptr and bump_vector_ptr.
11986 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11987 Alan Hayward <alan.hayward@arm.com>
11988 David Sherwood <david.sherwood@arm.com>
11990 * doc/md.texi (gather_load@var{m}): Document.
11991 (mask_gather_load@var{m}): Likewise.
11992 * genopinit.c (main): Add supports_vec_gather_load and
11993 supports_vec_gather_load_cached to target_optabs.
11994 * optabs-tree.c (init_tree_optimization_optabs): Use
11995 ggc_cleared_alloc to allocate target_optabs.
11996 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
11997 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
11999 * internal-fn.h (internal_load_fn_p): Declare.
12000 (internal_gather_scatter_fn_p): Likewise.
12001 (internal_fn_mask_index): Likewise.
12002 (internal_gather_scatter_fn_supported_p): Likewise.
12003 * internal-fn.c (gather_load_direct): New macro.
12004 (expand_gather_load_optab_fn): New function.
12005 (direct_gather_load_optab_supported_p): New macro.
12006 (direct_internal_fn_optab): New function.
12007 (internal_load_fn_p): Likewise.
12008 (internal_gather_scatter_fn_p): Likewise.
12009 (internal_fn_mask_index): Likewise.
12010 (internal_gather_scatter_fn_supported_p): Likewise.
12011 * optabs-query.c (supports_at_least_one_mode_p): New function.
12012 (supports_vec_gather_load_p): Likewise.
12013 * optabs-query.h (supports_vec_gather_load_p): Declare.
12014 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
12015 and memory_type field.
12016 (NUM_PATTERNS): Bump to 15.
12017 * tree-vect-data-refs.c: Include internal-fn.h.
12018 (vect_gather_scatter_fn_p): New function.
12019 (vect_describe_gather_scatter_call): Likewise.
12020 (vect_check_gather_scatter): Try using internal functions for
12021 gather loads. Recognize existing calls to a gather load function.
12022 (vect_analyze_data_refs): Consider using gather loads if
12023 supports_vec_gather_load_p.
12024 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
12025 (vect_get_gather_scatter_offset_type): Likewise.
12026 (vect_convert_mask_for_vectype): Likewise.
12027 (vect_add_conversion_to_patterm): Likewise.
12028 (vect_try_gather_scatter_pattern): Likewise.
12029 (vect_recog_gather_scatter_pattern): New pattern recognizer.
12030 (vect_vect_recog_func_ptrs): Add it.
12031 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
12032 internal_fn_mask_index and internal_gather_scatter_fn_p.
12033 (check_load_store_masking): Take the gather_scatter_info as an
12034 argument and handle gather loads.
12035 (vect_get_gather_scatter_ops): New function.
12036 (vectorizable_call): Check internal_load_fn_p.
12037 (vectorizable_load): Likewise. Handle gather load internal
12039 (vectorizable_store): Update call to check_load_store_masking.
12040 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
12041 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
12042 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
12043 (aarch64_gather_scale_operand_d): New predicates.
12044 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
12045 (mask_gather_load<mode>): New insns.
12047 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12048 Alan Hayward <alan.hayward@arm.com>
12049 David Sherwood <david.sherwood@arm.com>
12051 * optabs.def (fold_left_plus_optab): New optab.
12052 * doc/md.texi (fold_left_plus_@var{m}): Document.
12053 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
12054 * internal-fn.c (fold_left_direct): Define.
12055 (expand_fold_left_optab_fn): Likewise.
12056 (direct_fold_left_optab_supported_p): Likewise.
12057 * fold-const-call.c (fold_const_fold_left): New function.
12058 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
12059 * tree-parloops.c (valid_reduction_p): New function.
12060 (gather_scalar_reductions): Use it.
12061 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
12062 (vect_finish_replace_stmt): Declare.
12063 * tree-vect-loop.c (fold_left_reduction_fn): New function.
12064 (needs_fold_left_reduction_p): New function, split out from...
12065 (vect_is_simple_reduction): ...here. Accept reductions that
12066 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
12067 (vect_force_simple_reduction): Also store the reduction type in
12068 the assignment's STMT_VINFO_REDUC_TYPE.
12069 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
12070 (merge_with_identity): New function.
12071 (vect_expand_fold_left): Likewise.
12072 (vectorize_fold_left_reduction): Likewise.
12073 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
12074 scalar phi in place for it. Check for target support and reject
12075 cases that would reassociate the operation. Defer the transform
12076 phase to vectorize_fold_left_reduction.
12077 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
12078 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
12079 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
12081 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12083 * tree-if-conv.c (predicate_mem_writes): Remove redundant
12084 call to ifc_temp_var.
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 * target.def (legitimize_address_displacement): Take the original
12091 offset as a poly_int.
12092 * targhooks.h (default_legitimize_address_displacement): Update
12094 * targhooks.c (default_legitimize_address_displacement): Likewise.
12095 * doc/tm.texi: Regenerate.
12096 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
12097 as an argument, moving assert of ad->disp == ad->disp_term to...
12098 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
12099 Try calling targetm.legitimize_address_displacement before expanding
12100 the address rather than afterwards, and adjust for the new interface.
12101 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
12102 Match the new hook interface. Handle SVE addresses.
12103 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
12104 new hook interface.
12106 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12108 * Makefile.in (OBJS): Add early-remat.o.
12109 * target.def (select_early_remat_modes): New hook.
12110 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
12111 * doc/tm.texi: Regenerate.
12112 * targhooks.h (default_select_early_remat_modes): Declare.
12113 * targhooks.c (default_select_early_remat_modes): New function.
12114 * timevar.def (TV_EARLY_REMAT): New timevar.
12115 * passes.def (pass_early_remat): New pass.
12116 * tree-pass.h (make_pass_early_remat): Declare.
12117 * early-remat.c: New file.
12118 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
12120 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
12122 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12123 Alan Hayward <alan.hayward@arm.com>
12124 David Sherwood <david.sherwood@arm.com>
12126 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
12127 vfm1 with a bound_epilog parameter.
12128 (vect_do_peeling): Update calls accordingly, and move the prologue
12129 call earlier in the function. Treat the base bound_epilog as 0 for
12130 fully-masked loops and retain vf - 1 for other loops. Add 1 to
12131 this base when peeling for gaps.
12132 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
12133 with fully-masked loops.
12134 (vect_estimate_min_profitable_iters): Handle the single peeled
12135 iteration in that case.
12137 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12138 Alan Hayward <alan.hayward@arm.com>
12139 David Sherwood <david.sherwood@arm.com>
12141 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
12142 single-element interleaving even if the size is not a power of 2.
12143 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
12144 accesses for single-element interleaving if the group size is
12147 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12148 Alan Hayward <alan.hayward@arm.com>
12149 David Sherwood <david.sherwood@arm.com>
12151 * doc/md.texi (fold_extract_last_@var{m}): Document.
12152 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
12153 * optabs.def (fold_extract_last_optab): New optab.
12154 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
12155 * internal-fn.c (fold_extract_direct): New macro.
12156 (expand_fold_extract_optab_fn): Likewise.
12157 (direct_fold_extract_optab_supported_p): Likewise.
12158 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
12159 * tree-vect-loop.c (vect_model_reduction_cost): Handle
12160 EXTRACT_LAST_REDUCTION.
12161 (get_initial_def_for_reduction): Do not create an initial vector
12162 for EXTRACT_LAST_REDUCTION reductions.
12163 (vectorizable_reduction): Leave the scalar phi in place for
12164 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
12165 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
12166 epilogue code for EXTRACT_LAST_REDUCTION and defer the
12167 transform phase to vectorizable_condition.
12168 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
12170 (vect_finish_stmt_generation): ...here.
12171 (vect_finish_replace_stmt): New function.
12172 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
12173 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
12175 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
12177 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12178 Alan Hayward <alan.hayward@arm.com>
12179 David Sherwood <david.sherwood@arm.com>
12181 * doc/md.texi (extract_last_@var{m}): Document.
12182 * optabs.def (extract_last_optab): New optab.
12183 * internal-fn.def (EXTRACT_LAST): New internal function.
12184 * internal-fn.c (cond_unary_direct): New macro.
12185 (expand_cond_unary_optab_fn): Likewise.
12186 (direct_cond_unary_optab_supported_p): Likewise.
12187 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
12188 loops using EXTRACT_LAST.
12189 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
12190 (extract_last_<mode>): ...this optab.
12191 (vec_extract<mode><Vel>): Update accordingly.
12193 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12194 Alan Hayward <alan.hayward@arm.com>
12195 David Sherwood <david.sherwood@arm.com>
12197 * target.def (empty_mask_is_expensive): New hook.
12198 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
12199 * doc/tm.texi: Regenerate.
12200 * targhooks.h (default_empty_mask_is_expensive): Declare.
12201 * targhooks.c (default_empty_mask_is_expensive): New function.
12202 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
12203 if the target says that empty masks are expensive.
12204 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
12206 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
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 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
12213 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
12214 (vect_use_loop_mask_for_alignment_p): New function.
12215 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
12216 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
12217 niters_skip argument. Make sure that the first niters_skip elements
12218 of the first iteration are inactive.
12219 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
12220 Update call to vect_set_loop_masks_directly.
12221 (get_misalign_in_elems): New function, split out from...
12222 (vect_gen_prolog_loop_niters): ...here.
12223 (vect_update_init_of_dr): Take a code argument that specifies whether
12224 the adjustment should be added or subtracted.
12225 (vect_update_init_of_drs): Likewise.
12226 (vect_prepare_for_masked_peels): New function.
12227 (vect_do_peeling): Skip prologue peeling if we're using a mask
12228 instead. Update call to vect_update_inits_of_drs.
12229 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12231 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
12232 alignment. Do not include the number of peeled iterations in
12233 the minimum threshold in that case.
12234 (vectorizable_induction): Adjust the start value down by
12235 LOOP_VINFO_MASK_SKIP_NITERS iterations.
12236 (vect_transform_loop): Call vect_prepare_for_masked_peels.
12237 Take the number of skipped iterations into account when calculating
12239 * tree-vect-stmts.c (vect_gen_while_not): New function.
12241 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12242 Alan Hayward <alan.hayward@arm.com>
12243 David Sherwood <david.sherwood@arm.com>
12245 * doc/sourcebuild.texi (vect_fully_masked): Document.
12246 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
12247 default value to 0.
12248 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
12250 (vect_analyze_loop_2): ...here. Don't check the vectorization
12251 factor against the number of loop iterations if the loop is
12254 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12255 Alan Hayward <alan.hayward@arm.com>
12256 David Sherwood <david.sherwood@arm.com>
12258 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
12259 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
12260 (dump_groups): Update accordingly.
12261 (iv_use::mem_type): New member variable.
12262 (address_p): New function.
12263 (record_use): Add a mem_type argument and initialize the new
12265 (record_group_use): Add a mem_type argument. Use address_p.
12266 Remove obsolete null checks of base_object. Update call to record_use.
12267 (find_interesting_uses_op): Update call to record_group_use.
12268 (find_interesting_uses_cond): Likewise.
12269 (find_interesting_uses_address): Likewise.
12270 (get_mem_type_for_internal_fn): New function.
12271 (find_address_like_use): Likewise.
12272 (find_interesting_uses_stmt): Try find_address_like_use before
12273 calling find_interesting_uses_op.
12274 (addr_offset_valid_p): Use the iv mem_type field as the type
12275 of the addressed memory.
12276 (add_autoinc_candidates): Likewise.
12277 (get_address_cost): Likewise.
12278 (split_small_address_groups_p): Use address_p.
12279 (split_address_groups): Likewise.
12280 (add_iv_candidate_for_use): Likewise.
12281 (autoinc_possible_for_pair): Likewise.
12282 (rewrite_groups): Likewise.
12283 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
12284 (determine_group_iv_cost): Update after split of USE_ADDRESS.
12285 (get_alias_ptr_type_for_ptr_address): New function.
12286 (rewrite_use_address): Rewrite address uses in calls that were
12287 identified by find_address_like_use.
12289 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12290 Alan Hayward <alan.hayward@arm.com>
12291 David Sherwood <david.sherwood@arm.com>
12293 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
12295 * gimple-expr.h (is_gimple_addressable: Likewise.
12296 * gimple-expr.c (is_gimple_address): Likewise.
12297 * internal-fn.c (expand_call_mem_ref): New function.
12298 (expand_mask_load_optab_fn): Use it.
12299 (expand_mask_store_optab_fn): Likewise.
12301 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12302 Alan Hayward <alan.hayward@arm.com>
12303 David Sherwood <david.sherwood@arm.com>
12305 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
12306 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
12307 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
12308 (cond_umax@var{mode}): Document.
12309 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
12310 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
12311 (cond_umin_optab, cond_umax_optab): New optabs.
12312 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
12313 (COND_IOR, COND_XOR): New internal functions.
12314 * internal-fn.h (get_conditional_internal_fn): Declare.
12315 * internal-fn.c (cond_binary_direct): New macro.
12316 (expand_cond_binary_optab_fn): Likewise.
12317 (direct_cond_binary_optab_supported_p): Likewise.
12318 (get_conditional_internal_fn): New function.
12319 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
12320 Cope with reduction statements that are vectorized as calls rather
12322 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
12323 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
12324 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
12325 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
12326 (UNSPEC_COND_EOR): New unspecs.
12327 (optab): Add mappings for them.
12328 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
12329 (sve_int_op, sve_fp_op): New int attributes.
12331 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12332 Alan Hayward <alan.hayward@arm.com>
12333 David Sherwood <david.sherwood@arm.com>
12335 * optabs.def (while_ult_optab): New optab.
12336 * doc/md.texi (while_ult@var{m}@var{n}): Document.
12337 * internal-fn.def (WHILE_ULT): New internal function.
12338 * internal-fn.h (direct_internal_fn_supported_p): New override
12339 that takes two types as argument.
12340 * internal-fn.c (while_direct): New macro.
12341 (expand_while_optab_fn): New function.
12342 (convert_optab_supported_p): Likewise.
12343 (direct_while_optab_supported_p): New macro.
12344 * wide-int.h (wi::udiv_ceil): New function.
12345 * tree-vectorizer.h (rgroup_masks): New structure.
12346 (vec_loop_masks): New typedef.
12347 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
12348 and fully_masked_p.
12349 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
12350 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
12351 (vect_max_vf): New function.
12352 (slpeel_make_loop_iterate_ntimes): Delete.
12353 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
12354 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
12355 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
12356 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
12357 internal-fn.h, stor-layout.h and optabs-query.h.
12358 (vect_set_loop_mask): New function.
12359 (add_preheader_seq): Likewise.
12360 (add_header_seq): Likewise.
12361 (interleave_supported_p): Likewise.
12362 (vect_maybe_permute_loop_masks): Likewise.
12363 (vect_set_loop_masks_directly): Likewise.
12364 (vect_set_loop_condition_masked): Likewise.
12365 (vect_set_loop_condition_unmasked): New function, split out from
12366 slpeel_make_loop_iterate_ntimes.
12367 (slpeel_make_loop_iterate_ntimes): Rename to..
12368 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
12369 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
12370 (vect_do_peeling): Update call accordingly.
12371 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
12373 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12374 mask_compare_type, can_fully_mask_p and fully_masked_p.
12375 (release_vec_loop_masks): New function.
12376 (_loop_vec_info): Use it to free the loop masks.
12377 (can_produce_all_loop_masks_p): New function.
12378 (vect_get_max_nscalars_per_iter): Likewise.
12379 (vect_verify_full_masking): Likewise.
12380 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
12381 retries, and free the mask rgroups before retrying. Check loop-wide
12382 reasons for disallowing fully-masked loops. Make the final decision
12383 about whether use a fully-masked loop or not.
12384 (vect_estimate_min_profitable_iters): Do not assume that peeling
12385 for the number of iterations will be needed for fully-masked loops.
12386 (vectorizable_reduction): Disable fully-masked loops.
12387 (vectorizable_live_operation): Likewise.
12388 (vect_halve_mask_nunits): New function.
12389 (vect_double_mask_nunits): Likewise.
12390 (vect_record_loop_mask): Likewise.
12391 (vect_get_loop_mask): Likewise.
12392 (vect_transform_loop): Handle the case in which the final loop
12393 iteration might handle a partial vector. Call vect_set_loop_condition
12394 instead of slpeel_make_loop_iterate_ntimes.
12395 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
12396 (check_load_store_masking): New function.
12397 (prepare_load_store_mask): Likewise.
12398 (vectorizable_store): Handle fully-masked loops.
12399 (vectorizable_load): Likewise.
12400 (supportable_widening_operation): Use vect_halve_mask_nunits for
12402 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
12403 (vect_gen_while): New function.
12404 * config/aarch64/aarch64.md (umax<mode>3): New expander.
12405 (aarch64_uqdec<mode>): New insn.
12407 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12408 Alan Hayward <alan.hayward@arm.com>
12409 David Sherwood <david.sherwood@arm.com>
12411 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
12412 (reduc_xor_scal_optab): New optabs.
12413 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
12414 (reduc_xor_scal_@var{m}): Document.
12415 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
12416 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
12417 internal functions.
12418 * fold-const-call.c (fold_const_call): Handle them.
12419 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
12420 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
12421 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
12422 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
12423 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
12424 (UNSPEC_XORV): New unspecs.
12425 (optab): Add entries for them.
12426 (BITWISEV): New int iterator.
12427 (bit_reduc_op): New int attributes.
12429 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12430 Alan Hayward <alan.hayward@arm.com>
12431 David Sherwood <david.sherwood@arm.com>
12433 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
12434 * internal-fn.def (VEC_SHL_INSERT): New internal function.
12435 * optabs.def (vec_shl_insert_optab): New optab.
12436 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
12437 (duplicate_and_interleave): Likewise.
12438 * tree-vect-loop.c: Include internal-fn.h.
12439 (neutral_op_for_slp_reduction): New function, split out from
12440 get_initial_defs_for_reduction.
12441 (get_initial_def_for_reduction): Handle option 2 for variable-length
12442 vectors by loading the neutral value into a vector and then shifting
12443 the initial value into element 0.
12444 (get_initial_defs_for_reduction): Replace the code argument with
12445 the neutral value calculated by neutral_op_for_slp_reduction.
12446 Use gimple_build_vector for constant-length vectors.
12447 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
12448 but the first group_size elements have a neutral value.
12449 Use duplicate_and_interleave otherwise.
12450 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
12451 Update call to get_initial_defs_for_reduction. Handle SLP
12452 reductions for variable-length vectors by creating one vector
12453 result for each scalar result, with the elements associated
12454 with other scalar results stubbed out with the neutral value.
12455 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
12456 Require IFN_VEC_SHL_INSERT for double reductions on
12457 variable-length vectors, or SLP reductions that have
12458 a neutral value. Require can_duplicate_and_interleave_p
12459 support for variable-length unchained SLP reductions if there
12460 is no neutral value, such as for MIN/MAX reductions. Also require
12461 the number of vector elements to be a multiple of the number of
12462 SLP statements when doing variable-length unchained SLP reductions.
12463 Update call to vect_create_epilog_for_reduction.
12464 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
12465 and remove initial values.
12466 (duplicate_and_interleave): Make public.
12467 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
12468 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
12470 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12471 Alan Hayward <alan.hayward@arm.com>
12472 David Sherwood <david.sherwood@arm.com>
12474 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
12475 (can_duplicate_and_interleave_p): New function.
12476 (vect_get_and_check_slp_defs): Take the vector of statements
12477 rather than just the current one. Remove excess parentheses.
12478 Restriction rejectinon of vect_constant_def and vect_external_def
12479 for variable-length vectors to boolean types, or types for which
12480 can_duplicate_and_interleave_p is false.
12481 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
12482 (duplicate_and_interleave): New function.
12483 (vect_get_constant_vectors): Use gimple_build_vector for
12484 constant-length vectors and suitable variable-length constant
12485 vectors. Use duplicate_and_interleave for other variable-length
12486 vectors. Don't defer the update when inserting new statements.
12488 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12489 Alan Hayward <alan.hayward@arm.com>
12490 David Sherwood <david.sherwood@arm.com>
12492 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
12493 min_profitable_iters doesn't go negative.
12495 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12496 Alan Hayward <alan.hayward@arm.com>
12497 David Sherwood <david.sherwood@arm.com>
12499 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
12500 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
12501 * optabs.def (vec_mask_load_lanes_optab): New optab.
12502 (vec_mask_store_lanes_optab): Likewise.
12503 * internal-fn.def (MASK_LOAD_LANES): New internal function.
12504 (MASK_STORE_LANES): Likewise.
12505 * internal-fn.c (mask_load_lanes_direct): New macro.
12506 (mask_store_lanes_direct): Likewise.
12507 (expand_mask_load_optab_fn): Handle masked operations.
12508 (expand_mask_load_lanes_optab_fn): New macro.
12509 (expand_mask_store_optab_fn): Handle masked operations.
12510 (expand_mask_store_lanes_optab_fn): New macro.
12511 (direct_mask_load_lanes_optab_supported_p): Likewise.
12512 (direct_mask_store_lanes_optab_supported_p): Likewise.
12513 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
12515 (vect_load_lanes_supported): Likewise.
12516 * tree-vect-data-refs.c (strip_conversion): New function.
12517 (can_group_stmts_p): Likewise.
12518 (vect_analyze_data_ref_accesses): Use it instead of checking
12519 for a pair of assignments.
12520 (vect_store_lanes_supported): Take a masked_p parameter.
12521 (vect_load_lanes_supported): Likewise.
12522 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
12523 vect_store_lanes_supported and vect_load_lanes_supported.
12524 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
12525 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
12526 parameter. Don't allow gaps for masked accesses.
12527 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
12528 and vect_load_lanes_supported.
12529 (get_load_store_type): Take a masked_p parameter and update
12530 call to get_group_load_store_type.
12531 (vectorizable_store): Update call to get_load_store_type.
12532 Handle IFN_MASK_STORE_LANES.
12533 (vectorizable_load): Update call to get_load_store_type.
12534 Handle IFN_MASK_LOAD_LANES.
12536 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12537 Alan Hayward <alan.hayward@arm.com>
12538 David Sherwood <david.sherwood@arm.com>
12540 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
12542 * config/aarch64/aarch64-protos.h
12543 (aarch64_sve_struct_memory_operand_p): Declare.
12544 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
12545 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
12546 (VPRED, vpred): Handle SVE structure modes.
12547 * config/aarch64/constraints.md (Utx): New constraint.
12548 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
12549 (aarch64_sve_struct_nonimmediate_operand): New predicates.
12550 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
12551 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
12552 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
12553 structure modes. Split into pieces after RA.
12554 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
12555 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
12557 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
12558 SVE structure modes.
12559 (aarch64_classify_address): Likewise.
12560 (sizetochar): Move earlier in file.
12561 (aarch64_print_operand): Handle SVE register lists.
12562 (aarch64_array_mode): New function.
12563 (aarch64_sve_struct_memory_operand_p): Likewise.
12564 (TARGET_ARRAY_MODE): Redefine.
12566 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12567 Alan Hayward <alan.hayward@arm.com>
12568 David Sherwood <david.sherwood@arm.com>
12570 * target.def (array_mode): New target hook.
12571 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
12572 * doc/tm.texi: Regenerate.
12573 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
12574 * hooks.c (hook_optmode_mode_uhwi_none): New function.
12575 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
12576 targetm.array_mode.
12577 * stor-layout.c (mode_for_array): Likewise. Support polynomial
12580 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12581 Alan Hayward <alan.hayward@arm.com>
12582 David Sherwood <david.sherwood@arm.com>
12584 * fold-const.c (fold_binary_loc): Check the argument types
12585 rather than the result type when testing for a vector operation.
12587 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12589 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
12590 * doc/tm.texi: Regenerate.
12592 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12593 Alan Hayward <alan.hayward@arm.com>
12594 David Sherwood <david.sherwood@arm.com>
12596 * doc/invoke.texi (-msve-vector-bits=): Document new option.
12597 (sve): Document new AArch64 extension.
12598 * doc/md.texi (w): Extend the description of the AArch64
12599 constraint to include SVE vectors.
12600 (Upl, Upa): Document new AArch64 predicate constraints.
12601 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
12603 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
12604 (msve-vector-bits=): New option.
12605 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
12606 SVE when these are disabled.
12607 (sve): New extension.
12608 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
12609 modes. Adjust their number of units based on aarch64_sve_vg.
12610 (MAX_BITSIZE_MODE_ANY_MODE): Define.
12611 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
12612 aarch64_addr_query_type.
12613 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
12614 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
12615 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
12616 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
12617 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
12618 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
12619 (aarch64_simd_imm_zero_p): Delete.
12620 (aarch64_check_zero_based_sve_index_immediate): Declare.
12621 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
12622 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
12623 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
12624 (aarch64_sve_float_mul_immediate_p): Likewise.
12625 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
12626 rather than an rtx.
12627 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
12628 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
12629 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
12630 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
12631 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
12632 (aarch64_regmode_natural_size): Likewise.
12633 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
12634 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
12636 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
12637 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
12638 for VG and the SVE predicate registers.
12639 (V_ALIASES): Add a "z"-prefixed alias.
12640 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
12641 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
12642 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
12643 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
12644 (REG_CLASS_NAMES): Add entries for them.
12645 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
12646 and the predicate registers.
12647 (aarch64_sve_vg): Declare.
12648 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
12649 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
12650 (REGMODE_NATURAL_SIZE): Define.
12651 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
12653 * config/aarch64/aarch64.c: Include cfgrtl.h.
12654 (simd_immediate_info): Add a constructor for series vectors,
12655 and an associated step field.
12656 (aarch64_sve_vg): New variable.
12657 (aarch64_dbx_register_number): Handle VG and the predicate registers.
12658 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
12659 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
12660 (VEC_ANY_DATA, VEC_STRUCT): New constants.
12661 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
12662 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
12663 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
12664 (aarch64_get_mask_mode): New functions.
12665 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
12666 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
12667 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
12668 predicate modes and predicate registers. Explicitly restrict
12669 GPRs to modes of 16 bytes or smaller. Only allow FP registers
12670 to store a vector mode if it is recognized by
12671 aarch64_classify_vector_mode.
12672 (aarch64_regmode_natural_size): New function.
12673 (aarch64_hard_regno_caller_save_mode): Return the original mode
12675 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
12676 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
12677 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
12678 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
12680 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
12681 does not overlap dest if the function is frame-related. Handle
12683 (aarch64_split_add_offset): New function.
12684 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
12685 them aarch64_add_offset.
12686 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
12687 and update call to aarch64_sub_sp.
12688 (aarch64_add_cfa_expression): New function.
12689 (aarch64_expand_prologue): Pass extra temporary registers to the
12690 functions above. Handle the case in which we need to emit new
12691 DW_CFA_expressions for registers that were originally saved
12692 relative to the stack pointer, but now have to be expressed
12693 relative to the frame pointer.
12694 (aarch64_output_mi_thunk): Pass extra temporary registers to the
12696 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
12697 IP0 and IP1 values for SVE frames.
12698 (aarch64_expand_vec_series): New function.
12699 (aarch64_expand_sve_widened_duplicate): Likewise.
12700 (aarch64_expand_sve_const_vector): Likewise.
12701 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
12702 Handle SVE constants. Use emit_move_insn to move a force_const_mem
12703 into the register, rather than emitting a SET directly.
12704 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
12705 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
12706 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
12707 (offset_9bit_signed_scaled_p): New functions.
12708 (aarch64_replicate_bitmask_imm): New function.
12709 (aarch64_bitmask_imm): Use it.
12710 (aarch64_cannot_force_const_mem): Reject expressions involving
12711 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
12712 (aarch64_classify_index): Handle SVE indices, by requiring
12713 a plain register index with a scale that matches the element size.
12714 (aarch64_classify_address): Handle SVE addresses. Assert that
12715 the mode of the address is VOIDmode or an integer mode.
12716 Update call to aarch64_classify_symbol.
12717 (aarch64_classify_symbolic_expression): Update call to
12718 aarch64_classify_symbol.
12719 (aarch64_const_vec_all_in_range_p): New function.
12720 (aarch64_print_vector_float_operand): Likewise.
12721 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
12722 "vN" for FP registers with SVE modes. Handle (const ...) vectors
12723 and the FP immediates 1.0 and 0.5.
12724 (aarch64_print_address_internal): Handle SVE addresses.
12725 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
12726 (aarch64_regno_regclass): Handle predicate registers.
12727 (aarch64_secondary_reload): Handle big-endian reloads of SVE
12729 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
12730 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
12731 (aarch64_convert_sve_vector_bits): New function.
12732 (aarch64_override_options): Use it to handle -msve-vector-bits=.
12733 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
12734 rather than an rtx.
12735 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
12736 Handle SVE vector and predicate modes. Accept VL-based constants
12737 that need only one temporary register, and VL offsets that require
12738 no temporary registers.
12739 (aarch64_conditional_register_usage): Mark the predicate registers
12740 as fixed if SVE isn't available.
12741 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
12742 Return true for SVE vector and predicate modes.
12743 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
12744 rather than an unsigned int. Handle SVE modes.
12745 (aarch64_preferred_simd_mode): Update call accordingly. Handle
12747 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
12749 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
12750 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
12751 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
12752 (aarch64_sve_float_mul_immediate_p): New functions.
12753 (aarch64_sve_valid_immediate): New function.
12754 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
12755 Explicitly reject structure modes. Check for INDEX constants.
12756 Handle PTRUE and PFALSE constants.
12757 (aarch64_check_zero_based_sve_index_immediate): New function.
12758 (aarch64_simd_imm_zero_p): Delete.
12759 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
12760 vector modes. Accept constants in the range of CNT[BHWD].
12761 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
12762 ask for an Advanced SIMD mode.
12763 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
12764 (aarch64_simd_vector_alignment): Handle SVE predicates.
12765 (aarch64_vectorize_preferred_vector_alignment): New function.
12766 (aarch64_simd_vector_alignment_reachable): Use it instead of
12768 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
12769 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
12771 (MAX_VECT_LEN): Delete.
12772 (expand_vec_perm_d): Add a vec_flags field.
12773 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
12774 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
12775 (aarch64_evpc_ext): Don't apply a big-endian lane correction
12777 (aarch64_evpc_rev): Rename to...
12778 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
12779 (aarch64_evpc_rev_global): New function.
12780 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
12781 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
12783 (aarch64_evpc_sve_tbl): New function.
12784 (aarch64_expand_vec_perm_const_1): Update after rename of
12785 aarch64_evpc_rev. Handle SVE permutes too, trying
12786 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
12787 than aarch64_evpc_tbl.
12788 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
12789 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
12790 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
12791 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
12792 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
12793 (aarch64_expand_sve_vcond): New functions.
12794 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
12795 of aarch64_vector_mode_p.
12796 (aarch64_dwarf_poly_indeterminate_value): New function.
12797 (aarch64_compute_pressure_classes): Likewise.
12798 (aarch64_can_change_mode_class): Likewise.
12799 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
12800 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
12801 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
12802 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
12803 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
12804 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
12805 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
12806 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
12808 (Dn, Dl, Dr): Accept const as well as const_vector.
12809 (Dz): Likewise. Compare against CONST0_RTX.
12810 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
12811 of "vector" where appropriate.
12812 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
12813 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
12814 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
12815 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
12816 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
12817 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
12818 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
12819 (v_int_equiv): Extend to SVE modes.
12820 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
12822 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
12823 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
12824 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
12825 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
12826 (SVE_COND_FP_CMP): New int iterators.
12827 (perm_hilo): Handle the new unpack unspecs.
12828 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
12830 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
12831 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
12832 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
12833 (aarch64_equality_operator, aarch64_constant_vector_operand)
12834 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
12835 (aarch64_sve_nonimmediate_operand): Likewise.
12836 (aarch64_sve_general_operand): Likewise.
12837 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
12838 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
12839 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
12840 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
12841 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
12842 (aarch64_sve_float_arith_immediate): Likewise.
12843 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
12844 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
12845 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
12846 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
12847 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
12848 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
12849 (aarch64_sve_float_arith_operand): Likewise.
12850 (aarch64_sve_float_arith_with_sub_operand): Likewise.
12851 (aarch64_sve_float_mul_operand): Likewise.
12852 (aarch64_sve_vec_perm_operand): Likewise.
12853 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
12854 (aarch64_mov_operand): Accept const_poly_int and const_vector.
12855 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
12856 as well as const_vector.
12857 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
12858 in file. Use CONST0_RTX and CONSTM1_RTX.
12859 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
12860 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
12861 Use aarch64_simd_imm_zero.
12862 * config/aarch64/aarch64-sve.md: New file.
12863 * config/aarch64/aarch64.md: Include it.
12864 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
12865 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
12866 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
12867 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
12868 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
12869 (sve): New attribute.
12870 (enabled): Disable instructions with the sve attribute unless
12872 (movqi, movhi): Pass CONST_POLY_INT operaneds through
12873 aarch64_expand_mov_immediate.
12874 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
12875 CNT[BHSD] immediates.
12876 (movti): Split CONST_POLY_INT moves into two halves.
12877 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
12878 Split additions that need a temporary here if the destination
12879 is the stack pointer.
12880 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
12881 (*add<mode>3_poly_1): New instruction.
12882 (set_clobber_cc): New expander.
12884 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12886 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
12887 parameter and use it instead of GET_MODE_SIZE (innermode). Use
12888 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
12889 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
12890 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
12891 Change innermode from fixed_mode_size to machine_mode.
12892 (simplify_subreg): Update call accordingly. Handle a constant-sized
12893 subreg of a variable-length CONST_VECTOR.
12895 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12896 Alan Hayward <alan.hayward@arm.com>
12897 David Sherwood <david.sherwood@arm.com>
12899 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
12900 (add_offset_to_base): New function, split out from...
12901 (create_mem_ref): ...here. When handling a scale other than 1,
12902 check first whether the address is valid without the offset.
12903 Add it into the base if so, leaving the index and scale as-is.
12905 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12908 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
12909 fold_for_warn before checking if arg2 is INTEGER_CST.
12911 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
12913 * config/rs6000/predicates.md (load_multiple_operation): Delete.
12914 (store_multiple_operation): Delete.
12915 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
12916 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
12917 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
12918 guarded by TARGET_STRING.
12919 (rs6000_output_load_multiple): Delete.
12920 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
12921 OPTION_MASK_STRING / TARGET_STRING handling.
12922 (print_operand) <'N', 'O'>: Add comment that these are unused now.
12923 (const rs6000_opt_masks) <"string">: Change mask to 0.
12924 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
12925 (MASK_STRING): Delete.
12926 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
12928 (load_multiple): Delete.
12935 (store_multiple): Delete.
12942 (movmemsi_8reg): Delete.
12943 (corresponding unnamed define_insn): Delete.
12944 (movmemsi_6reg): Delete.
12945 (corresponding unnamed define_insn): Delete.
12946 (movmemsi_4reg): Delete.
12947 (corresponding unnamed define_insn): Delete.
12948 (movmemsi_2reg): Delete.
12949 (corresponding unnamed define_insn): Delete.
12950 (movmemsi_1reg): Delete.
12951 (corresponding unnamed define_insn): Delete.
12952 * config/rs6000/rs6000.opt (mno-string): New.
12953 (mstring): Replace by deprecation warning stub.
12954 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
12956 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12958 * regrename.c (regrename_do_replace): If replacing the same
12959 reg multiple times, try to reuse last created gen_raw_REG.
12962 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
12963 main to workaround a bug in GDB.
12965 2018-01-12 Tom de Vries <tom@codesourcery.com>
12968 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
12970 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
12972 PR rtl-optimization/80481
12973 * ira-color.c (get_cap_member): New function.
12974 (allocnos_conflict_by_live_ranges_p): Use it.
12975 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
12976 (setup_slot_coalesced_allocno_live_ranges): Ditto.
12978 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
12981 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
12982 (*saddl_se_1): Ditto.
12983 (*ssubsi_1): Ditto.
12984 (*ssubl_se_1): Ditto.
12986 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12988 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
12989 rather than wi::to_widest for DR_INITs.
12990 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
12991 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
12992 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
12994 (vect_analyze_group_access_1): Note that here.
12996 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12998 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
12999 polynomial type sizes.
13001 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13003 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
13004 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
13005 (gimple_add_tmp_var): Likewise.
13007 2018-01-12 Martin Liska <mliska@suse.cz>
13009 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
13010 (gimple_alloc_sizes): Likewise.
13011 (dump_gimple_statistics): Use PRIu64 in printf format.
13012 * gimple.h: Change uint64_t to int.
13014 2018-01-12 Martin Liska <mliska@suse.cz>
13016 * tree-core.h: Use uint64_t instead of int.
13017 * tree.c (tree_node_counts): Likewise.
13018 (tree_node_sizes): Likewise.
13019 (dump_tree_statistics): Use PRIu64 in printf format.
13021 2018-01-12 Martin Liska <mliska@suse.cz>
13023 * Makefile.in: As qsort_chk is implemented in vec.c, add
13024 vec.o to linkage of gencfn-macros.
13025 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
13026 passing the info to record_node_allocation_statistics.
13027 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
13029 * ggc-common.c (struct ggc_usage): Add operator== and use
13030 it in operator< and compare function.
13031 * mem-stats.h (struct mem_usage): Likewise.
13032 * vec.c (struct vec_usage): Remove operator< and compare
13033 function. Can be simply inherited.
13035 2018-01-12 Martin Jambor <mjambor@suse.cz>
13038 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
13039 * tree-ssa-math-opts.c: Include domwalk.h.
13040 (convert_mult_to_fma_1): New function.
13041 (fma_transformation_info): New type.
13042 (fma_deferring_state): Likewise.
13043 (cancel_fma_deferring): New function.
13044 (result_of_phi): Likewise.
13045 (last_fma_candidate_feeds_initial_phi): Likewise.
13046 (convert_mult_to_fma): Added deferring logic, split actual
13047 transformation to convert_mult_to_fma_1.
13048 (math_opts_dom_walker): New type.
13049 (math_opts_dom_walker::after_dom_children): New method, body moved
13050 here from pass_optimize_widening_mul::execute, added deferring logic
13052 (pass_optimize_widening_mul::execute): Moved most of code to
13053 math_opts_dom_walker::after_dom_children.
13054 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
13055 * config/i386/i386.c (ix86_option_override_internal): Added
13056 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
13058 2018-01-12 Richard Biener <rguenther@suse.de>
13061 * dwarf2out.c (gen_variable_die): Do not reset old_die for
13062 inline instance vars.
13064 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
13067 * config/rx/rx.c (rx_is_restricted_memory_address):
13068 Handle SUBREG case.
13070 2018-01-12 Richard Biener <rguenther@suse.de>
13072 PR tree-optimization/80846
13073 * target.def (split_reduction): New target hook.
13074 * targhooks.c (default_split_reduction): New function.
13075 * targhooks.h (default_split_reduction): Declare.
13076 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
13077 target requests first reduce vectors by combining low and high
13079 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
13080 (get_vectype_for_scalar_type_and_size): Export.
13081 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
13082 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
13083 * doc/tm.texi: Regenerate.
13084 * config/i386/i386.c (ix86_split_reduction): Implement
13085 TARGET_VECTORIZE_SPLIT_REDUCTION.
13087 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13090 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
13091 in PIC mode except for TARGET_VXWORKS_RTP.
13092 * config/sparc/sparc.c: Include cfgrtl.h.
13093 (TARGET_INIT_PIC_REG): Define.
13094 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
13095 (sparc_pic_register_p): New predicate.
13096 (sparc_legitimate_address_p): Use it.
13097 (sparc_legitimize_pic_address): Likewise.
13098 (sparc_delegitimize_address): Likewise.
13099 (sparc_mode_dependent_address_p): Likewise.
13100 (gen_load_pcrel_sym): Remove 4th parameter.
13101 (load_got_register): Adjust call to above. Remove obsolete stuff.
13102 (sparc_expand_prologue): Do not call load_got_register here.
13103 (sparc_flat_expand_prologue): Likewise.
13104 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
13105 (sparc_use_pseudo_pic_reg): New function.
13106 (sparc_init_pic_reg): Likewise.
13107 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
13108 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
13110 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
13112 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
13113 Add item for branch_cost.
13115 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13117 PR rtl-optimization/83565
13118 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
13119 not extend the result to a larger mode for rotate operations.
13120 (num_sign_bit_copies1): Likewise.
13122 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13125 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
13127 Use values-Xc.o for -pedantic.
13128 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
13130 2018-01-12 Martin Liska <mliska@suse.cz>
13133 * ipa-devirt.c (final_warning_record::grow_type_warnings):
13135 (possible_polymorphic_call_targets): Use it.
13136 (ipa_devirt): Likewise.
13138 2018-01-12 Martin Liska <mliska@suse.cz>
13140 * profile-count.h (enum profile_quality): Use 0 as invalid
13141 enum value of profile_quality.
13143 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
13145 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
13146 -mext-string options.
13148 2018-01-12 Richard Biener <rguenther@suse.de>
13150 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
13151 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
13152 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
13154 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
13156 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
13158 * configure.ac (--with-long-double-format): Add support for the
13159 configuration option to change the default long double format on
13161 * config.gcc (powerpc*-linux*-*): Likewise.
13162 * configure: Regenerate.
13163 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
13164 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
13165 used without modification.
13167 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13169 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
13170 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
13171 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
13172 MISC_BUILTIN_SPEC_BARRIER.
13173 (rs6000_init_builtins): Likewise.
13174 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
13176 (speculation_barrier): New define_insn.
13177 * doc/extend.texi: Document __builtin_speculation_barrier.
13179 2018-01-11 Jakub Jelinek <jakub@redhat.com>
13182 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
13183 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
13184 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
13186 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
13187 integral modes instead of "ss" and "sd".
13188 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
13189 vectors with 32-bit and 64-bit elements.
13190 (vecdupssescalarmodesuffix): New mode attribute.
13191 (vec_dup<mode>): Use it.
13193 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
13196 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
13197 frame if argument is passed on stack.
13199 2018-01-11 Jakub Jelinek <jakub@redhat.com>
13202 * ree.c (combine_reaching_defs): Optimize also
13203 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
13204 reg2=any_extend(exp); reg1=reg2;, formatting fix.
13206 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
13208 PR middle-end/83189
13209 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
13211 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
13213 PR middle-end/83718
13214 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
13215 after they are computed.
13217 2018-01-11 Bin Cheng <bin.cheng@arm.com>
13219 PR tree-optimization/83695
13220 * gimple-loop-linterchange.cc
13221 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
13222 reset cached scev information after interchange.
13223 (pass_linterchange::execute): Remove call to scev_reset_htab.
13225 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13227 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
13228 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
13229 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
13230 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
13231 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
13232 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
13233 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
13234 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
13235 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
13236 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
13237 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
13238 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
13239 (V_lane_reg): Likewise.
13240 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
13242 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
13243 (vfmal_lane_low<mode>_intrinsic,
13244 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
13245 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
13246 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
13247 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
13248 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
13249 vfmsl_lane_high<mode>_intrinsic): New define_insns.
13251 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13253 * config/arm/arm-cpus.in (fp16fml): New feature.
13254 (ALL_SIMD): Add fp16fml.
13255 (armv8.2-a): Add fp16fml as an option.
13256 (armv8.3-a): Likewise.
13257 (armv8.4-a): Add fp16fml as part of fp16.
13258 * config/arm/arm.h (TARGET_FP16FML): Define.
13259 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
13261 * config/arm/arm-modes.def (V2HF): Define.
13262 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
13263 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
13264 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
13265 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
13266 vfmsl_low, vfmsl_high): New set of builtins.
13267 * config/arm/iterators.md (PLUSMINUS): New code iterator.
13268 (vfml_op): New code attribute.
13269 (VFMLHALVES): New int iterator.
13270 (VFML, VFMLSEL): New mode attributes.
13271 (V_reg): Define mapping for V2HF.
13272 (V_hi, V_lo): New mode attributes.
13273 (VF_constraint): Likewise.
13274 (vfml_half, vfml_half_selector): New int attributes.
13275 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
13277 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
13278 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
13280 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
13281 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
13282 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
13283 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
13285 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
13286 Document new effective target and option set.
13288 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13290 * config/arm/arm-cpus.in (armv8_4): New feature.
13291 (ARMv8_4a): New fgroup.
13292 (armv8.4-a): New arch.
13293 * config/arm/arm-tables.opt: Regenerate.
13294 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
13295 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
13296 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
13297 Add matching rules for -march=armv8.4-a and extensions.
13298 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
13300 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
13303 * config/rx/rx.md (BW): New mode attribute.
13304 (sync_lock_test_and_setsi): Add mode suffix to insn output.
13306 2018-01-11 Richard Biener <rguenther@suse.de>
13308 PR tree-optimization/83435
13309 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
13310 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
13311 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
13313 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13314 Alan Hayward <alan.hayward@arm.com>
13315 David Sherwood <david.sherwood@arm.com>
13317 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
13319 (aarch64_classify_address): Initialize it. Track polynomial offsets.
13320 (aarch64_print_address_internal): Use it to check for a zero offset.
13322 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13323 Alan Hayward <alan.hayward@arm.com>
13324 David Sherwood <david.sherwood@arm.com>
13326 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
13327 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
13328 Return a poly_int64 rather than a HOST_WIDE_INT.
13329 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
13330 rather than a HOST_WIDE_INT.
13331 * config/aarch64/aarch64.h (aarch64_frame): Protect with
13332 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
13333 hard_fp_offset, frame_size, initial_adjust, callee_offset and
13334 final_offset from HOST_WIDE_INT to poly_int64.
13335 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
13336 to_constant when getting the number of units in an Advanced SIMD
13338 (aarch64_builtin_vectorized_function): Check for a constant number
13340 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
13342 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
13343 attribute instead of GET_MODE_NUNITS.
13344 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
13345 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
13346 GET_MODE_SIZE for fixed-size registers.
13347 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
13348 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
13349 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
13350 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
13351 (aarch64_print_operand, aarch64_print_address_internal)
13352 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
13353 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
13354 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
13355 Handle polynomial GET_MODE_SIZE.
13356 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
13357 wider than SImode without modification.
13358 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
13359 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
13360 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
13361 passing and returning SVE modes.
13362 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
13363 rather than GEN_INT.
13364 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
13365 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
13366 (aarch64_allocate_and_probe_stack_space): Likewise.
13367 (aarch64_layout_frame): Cope with polynomial offsets.
13368 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
13369 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
13370 polynomial offsets.
13371 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
13372 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
13373 poly_int64 rather than a HOST_WIDE_INT.
13374 (aarch64_get_separate_components, aarch64_process_components)
13375 (aarch64_expand_prologue, aarch64_expand_epilogue)
13376 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
13377 (aarch64_anchor_offset): New function, split out from...
13378 (aarch64_legitimize_address): ...here.
13379 (aarch64_builtin_vectorization_cost): Handle polynomial
13380 TYPE_VECTOR_SUBPARTS.
13381 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
13383 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
13384 number of elements from the PARALLEL rather than the mode.
13385 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
13386 rather than GET_MODE_BITSIZE.
13387 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
13388 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
13389 (aarch64_expand_vec_perm_const_1): Handle polynomial
13390 d->perm.length () and d->perm elements.
13391 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
13392 Apply to_constant to d->perm elements.
13393 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
13394 polynomial CONST_VECTOR_NUNITS.
13395 (aarch64_move_pointer): Take amount as a poly_int64 rather
13397 (aarch64_progress_pointer): Avoid temporary variable.
13398 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
13399 the mode attribute instead of GET_MODE.
13401 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13402 Alan Hayward <alan.hayward@arm.com>
13403 David Sherwood <david.sherwood@arm.com>
13405 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
13406 x exists before using it.
13407 (aarch64_add_constant_internal): Rename to...
13408 (aarch64_add_offset_1): ...this. Replace regnum with separate
13409 src and dest rtxes. Handle the case in which they're different,
13410 including when the offset is zero. Replace scratchreg with an rtx.
13411 Use 2 additions if there is no spare register into which we can
13412 move a 16-bit constant.
13413 (aarch64_add_constant): Delete.
13414 (aarch64_add_offset): Replace reg with separate src and dest
13415 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
13416 Use aarch64_add_offset_1.
13417 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
13418 an rtx rather than an int. Take the delta as a poly_int64
13419 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
13420 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
13421 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
13422 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
13423 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
13424 and aarch64_add_sp.
13425 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
13426 aarch64_add_constant.
13428 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13430 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
13431 Use scalar_float_mode.
13433 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13435 * config/aarch64/aarch64-simd.md
13436 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
13437 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
13438 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
13439 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
13440 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
13441 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
13442 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
13443 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
13444 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
13445 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
13447 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13450 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
13451 targ_options->x_arm_arch_string is non NULL.
13453 2018-01-11 Tamar Christina <tamar.christina@arm.com>
13455 * config/aarch64/aarch64.h
13456 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
13458 2018-01-11 Sudakshina Das <sudi.das@arm.com>
13461 * expmed.c (emit_store_flag_force): Swap if const op0
13462 and change VOIDmode to mode of op0.
13464 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13466 PR rtl-optimization/83761
13467 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
13468 than bytes to mode_for_size.
13470 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13472 PR middle-end/83189
13473 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
13474 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
13477 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13479 PR middle-end/83575
13480 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
13481 when in layout mode.
13482 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
13483 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
13486 2018-01-10 Michael Collison <michael.collison@arm.com>
13488 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
13489 * config/aarch64/aarch64-option-extension.def: Add
13490 AARCH64_OPT_EXTENSION of 'fp16fml'.
13491 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13492 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
13493 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
13494 * config/aarch64/constraints.md (Ui7): New constraint.
13495 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
13496 (VFMLA_SEL_W): Ditto.
13499 (VFMLA16_LOW): New int iterator.
13500 (VFMLA16_HIGH): Ditto.
13501 (UNSPEC_FMLAL): New unspec.
13502 (UNSPEC_FMLSL): Ditto.
13503 (UNSPEC_FMLAL2): Ditto.
13504 (UNSPEC_FMLSL2): Ditto.
13505 (f16mac): New code attribute.
13506 * config/aarch64/aarch64-simd-builtins.def
13507 (aarch64_fmlal_lowv2sf): Ditto.
13508 (aarch64_fmlsl_lowv2sf): Ditto.
13509 (aarch64_fmlalq_lowv4sf): Ditto.
13510 (aarch64_fmlslq_lowv4sf): Ditto.
13511 (aarch64_fmlal_highv2sf): Ditto.
13512 (aarch64_fmlsl_highv2sf): Ditto.
13513 (aarch64_fmlalq_highv4sf): Ditto.
13514 (aarch64_fmlslq_highv4sf): Ditto.
13515 (aarch64_fmlal_lane_lowv2sf): Ditto.
13516 (aarch64_fmlsl_lane_lowv2sf): Ditto.
13517 (aarch64_fmlal_laneq_lowv2sf): Ditto.
13518 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
13519 (aarch64_fmlalq_lane_lowv4sf): Ditto.
13520 (aarch64_fmlsl_lane_lowv4sf): Ditto.
13521 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
13522 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
13523 (aarch64_fmlal_lane_highv2sf): Ditto.
13524 (aarch64_fmlsl_lane_highv2sf): Ditto.
13525 (aarch64_fmlal_laneq_highv2sf): Ditto.
13526 (aarch64_fmlsl_laneq_highv2sf): Ditto.
13527 (aarch64_fmlalq_lane_highv4sf): Ditto.
13528 (aarch64_fmlsl_lane_highv4sf): Ditto.
13529 (aarch64_fmlalq_laneq_highv4sf): Ditto.
13530 (aarch64_fmlsl_laneq_highv4sf): Ditto.
13531 * config/aarch64/aarch64-simd.md:
13532 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
13533 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
13534 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
13535 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
13536 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
13537 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
13538 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
13539 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
13540 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
13541 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
13542 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
13543 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
13544 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
13545 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
13546 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
13547 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
13548 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
13549 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
13550 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
13551 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
13552 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
13553 (vfmlsl_low_u32): Ditto.
13554 (vfmlalq_low_u32): Ditto.
13555 (vfmlslq_low_u32): Ditto.
13556 (vfmlal_high_u32): Ditto.
13557 (vfmlsl_high_u32): Ditto.
13558 (vfmlalq_high_u32): Ditto.
13559 (vfmlslq_high_u32): Ditto.
13560 (vfmlal_lane_low_u32): Ditto.
13561 (vfmlsl_lane_low_u32): Ditto.
13562 (vfmlal_laneq_low_u32): Ditto.
13563 (vfmlsl_laneq_low_u32): Ditto.
13564 (vfmlalq_lane_low_u32): Ditto.
13565 (vfmlslq_lane_low_u32): Ditto.
13566 (vfmlalq_laneq_low_u32): Ditto.
13567 (vfmlslq_laneq_low_u32): Ditto.
13568 (vfmlal_lane_high_u32): Ditto.
13569 (vfmlsl_lane_high_u32): Ditto.
13570 (vfmlal_laneq_high_u32): Ditto.
13571 (vfmlsl_laneq_high_u32): Ditto.
13572 (vfmlalq_lane_high_u32): Ditto.
13573 (vfmlslq_lane_high_u32): Ditto.
13574 (vfmlalq_laneq_high_u32): Ditto.
13575 (vfmlslq_laneq_high_u32): Ditto.
13576 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
13577 (AARCH64_FL_FOR_ARCH8_4): New.
13578 (AARCH64_ISA_F16FML): New ISA flag.
13579 (TARGET_F16FML): New feature flag for fp16fml.
13580 (doc/invoke.texi): Document new fp16fml option.
13582 2018-01-10 Michael Collison <michael.collison@arm.com>
13584 * config/aarch64/aarch64-builtins.c:
13585 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
13586 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13587 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
13588 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
13589 (AARCH64_ISA_SHA3): New ISA flag.
13590 (TARGET_SHA3): New feature flag for sha3.
13591 * config/aarch64/iterators.md (sha512_op): New int attribute.
13592 (CRYPTO_SHA512): New int iterator.
13593 (UNSPEC_SHA512H): New unspec.
13594 (UNSPEC_SHA512H2): Ditto.
13595 (UNSPEC_SHA512SU0): Ditto.
13596 (UNSPEC_SHA512SU1): Ditto.
13597 * config/aarch64/aarch64-simd-builtins.def
13598 (aarch64_crypto_sha512hqv2di): New builtin.
13599 (aarch64_crypto_sha512h2qv2di): Ditto.
13600 (aarch64_crypto_sha512su0qv2di): Ditto.
13601 (aarch64_crypto_sha512su1qv2di): Ditto.
13602 (aarch64_eor3qv8hi): Ditto.
13603 (aarch64_rax1qv2di): Ditto.
13604 (aarch64_xarqv2di): Ditto.
13605 (aarch64_bcaxqv8hi): Ditto.
13606 * config/aarch64/aarch64-simd.md:
13607 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
13608 (aarch64_crypto_sha512su0qv2di): Ditto.
13609 (aarch64_crypto_sha512su1qv2di): Ditto.
13610 (aarch64_eor3qv8hi): Ditto.
13611 (aarch64_rax1qv2di): Ditto.
13612 (aarch64_xarqv2di): Ditto.
13613 (aarch64_bcaxqv8hi): Ditto.
13614 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
13615 (vsha512h2q_u64): Ditto.
13616 (vsha512su0q_u64): Ditto.
13617 (vsha512su1q_u64): Ditto.
13618 (veor3q_u16): Ditto.
13619 (vrax1q_u64): Ditto.
13620 (vxarq_u64): Ditto.
13621 (vbcaxq_u16): Ditto.
13622 * config/arm/types.md (crypto_sha512): New type attribute.
13623 (crypto_sha3): Ditto.
13624 (doc/invoke.texi): Document new sha3 option.
13626 2018-01-10 Michael Collison <michael.collison@arm.com>
13628 * config/aarch64/aarch64-builtins.c:
13629 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
13630 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13631 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
13632 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
13633 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
13634 (AARCH64_ISA_SM4): New ISA flag.
13635 (TARGET_SM4): New feature flag for sm4.
13636 * config/aarch64/aarch64-simd-builtins.def
13637 (aarch64_sm3ss1qv4si): Ditto.
13638 (aarch64_sm3tt1aq4si): Ditto.
13639 (aarch64_sm3tt1bq4si): Ditto.
13640 (aarch64_sm3tt2aq4si): Ditto.
13641 (aarch64_sm3tt2bq4si): Ditto.
13642 (aarch64_sm3partw1qv4si): Ditto.
13643 (aarch64_sm3partw2qv4si): Ditto.
13644 (aarch64_sm4eqv4si): Ditto.
13645 (aarch64_sm4ekeyqv4si): Ditto.
13646 * config/aarch64/aarch64-simd.md:
13647 (aarch64_sm3ss1qv4si): Ditto.
13648 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
13649 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
13650 (aarch64_sm4eqv4si): Ditto.
13651 (aarch64_sm4ekeyqv4si): Ditto.
13652 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
13653 (sm3part_op): Ditto.
13654 (CRYPTO_SM3TT): Ditto.
13655 (CRYPTO_SM3PART): Ditto.
13656 (UNSPEC_SM3SS1): New unspec.
13657 (UNSPEC_SM3TT1A): Ditto.
13658 (UNSPEC_SM3TT1B): Ditto.
13659 (UNSPEC_SM3TT2A): Ditto.
13660 (UNSPEC_SM3TT2B): Ditto.
13661 (UNSPEC_SM3PARTW1): Ditto.
13662 (UNSPEC_SM3PARTW2): Ditto.
13663 (UNSPEC_SM4E): Ditto.
13664 (UNSPEC_SM4EKEY): Ditto.
13665 * config/aarch64/constraints.md (Ui2): New constraint.
13666 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
13667 * config/arm/types.md (crypto_sm3): New type attribute.
13668 (crypto_sm4): Ditto.
13669 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
13670 (vsm3tt1aq_u32): Ditto.
13671 (vsm3tt1bq_u32): Ditto.
13672 (vsm3tt2aq_u32): Ditto.
13673 (vsm3tt2bq_u32): Ditto.
13674 (vsm3partw1q_u32): Ditto.
13675 (vsm3partw2q_u32): Ditto.
13676 (vsm4eq_u32): Ditto.
13677 (vsm4ekeyq_u32): Ditto.
13678 (doc/invoke.texi): Document new sm4 option.
13680 2018-01-10 Michael Collison <michael.collison@arm.com>
13682 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
13683 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
13684 (AARCH64_FL_FOR_ARCH8_4): New.
13685 (AARCH64_FL_V8_4): New flag.
13686 (doc/invoke.texi): Document new armv8.4-a option.
13688 2018-01-10 Michael Collison <michael.collison@arm.com>
13690 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13691 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
13692 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
13693 * config/aarch64/aarch64-option-extension.def: Add
13694 AARCH64_OPT_EXTENSION of 'sha2'.
13695 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
13696 (crypto): Disable sha2 and aes if crypto disabled.
13697 (crypto): Enable aes and sha2 if enabled.
13698 (simd): Disable sha2 and aes if simd disabled.
13699 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
13701 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
13702 (TARGET_SHA2): New feature flag for sha2.
13703 (TARGET_AES): New feature flag for aes.
13704 * config/aarch64/aarch64-simd.md:
13705 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
13706 conditional on TARGET_AES.
13707 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
13708 (aarch64_crypto_sha1hsi): Make pattern conditional
13710 (aarch64_crypto_sha1hv4si): Ditto.
13711 (aarch64_be_crypto_sha1hv4si): Ditto.
13712 (aarch64_crypto_sha1su1v4si): Ditto.
13713 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
13714 (aarch64_crypto_sha1su0v4si): Ditto.
13715 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
13716 (aarch64_crypto_sha256su0v4si): Ditto.
13717 (aarch64_crypto_sha256su1v4si): Ditto.
13718 (doc/invoke.texi): Document new aes and sha2 options.
13720 2018-01-10 Martin Sebor <msebor@redhat.com>
13722 PR tree-optimization/83781
13723 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
13726 2018-01-11 Martin Sebor <msebor@gmail.com>
13727 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13729 PR tree-optimization/83501
13730 PR tree-optimization/81703
13732 * tree-ssa-strlen.c (get_string_cst): Rename...
13733 (get_string_len): ...to this. Handle global constants.
13734 (handle_char_store): Adjust.
13736 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
13737 Jim Wilson <jimw@sifive.com>
13739 * config/riscv/riscv-protos.h (riscv_output_return): New.
13740 * config/riscv/riscv.c (struct machine_function): New naked_p field.
13741 (riscv_attribute_table, riscv_output_return),
13742 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
13743 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
13744 (riscv_compute_frame_info): Only compute frame->mask if not a naked
13746 (riscv_expand_prologue): Add early return for naked function.
13747 (riscv_expand_epilogue): Likewise.
13748 (riscv_function_ok_for_sibcall): Return false for naked function.
13749 (riscv_set_current_function): New.
13750 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
13751 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
13752 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
13753 * doc/extend.texi (RISC-V Function Attributes): New.
13755 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
13757 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
13758 check for 128-bit long double before checking TCmode.
13759 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
13760 128-bit long doubles before checking TFmode or TCmode.
13761 (FLOAT128_IBM_P): Likewise.
13763 2018-01-10 Martin Sebor <msebor@redhat.com>
13765 PR tree-optimization/83671
13766 * builtins.c (c_strlen): Unconditionally return zero for the empty
13768 Use -Warray-bounds for warnings.
13769 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
13770 for non-constant array indices with COMPONENT_REF, arrays of
13771 arrays, and pointers to arrays.
13772 (gimple_fold_builtin_strlen): Determine and set length range for
13773 non-constant character arrays.
13775 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
13777 PR middle-end/81897
13778 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
13781 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
13783 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
13785 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
13788 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
13789 VECTOR_MEM_ALTIVEC_OR_VSX_P.
13790 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
13791 indexed_or_indirect_operand predicate.
13792 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
13793 (*vsx_le_perm_load_v8hi): Likewise.
13794 (*vsx_le_perm_load_v16qi): Likewise.
13795 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
13796 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
13797 (*vsx_le_perm_store_v8hi): Likewise.
13798 (*vsx_le_perm_store_v16qi): Likewise.
13799 (eight unnamed splitters): Likewise.
13801 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
13803 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
13804 * config/rs6000/emmintrin.h: Likewise.
13805 * config/rs6000/mmintrin.h: Likewise.
13806 * config/rs6000/xmmintrin.h: Likewise.
13808 2018-01-10 David Malcolm <dmalcolm@redhat.com>
13811 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
13813 * tree.c (tree_nop_conversion): Return true for location wrapper
13815 (maybe_wrap_with_location): New function.
13816 (selftest::check_strip_nops): New function.
13817 (selftest::test_location_wrappers): New function.
13818 (selftest::tree_c_tests): Call it.
13819 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
13820 (maybe_wrap_with_location): New decl.
13821 (EXPR_LOCATION_WRAPPER_P): New macro.
13822 (location_wrapper_p): New inline function.
13823 (tree_strip_any_location_wrapper): New inline function.
13825 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
13828 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
13829 stack_realign_offset for the largest alignment of stack slot
13831 (ix86_find_max_used_stack_alignment): New function.
13832 (ix86_finalize_stack_frame_flags): Use it. Set
13833 max_used_stack_alignment if we don't realign stack.
13834 * config/i386/i386.h (machine_function): Add
13835 max_used_stack_alignment.
13837 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
13839 * config/arm/arm.opt (-mbranch-cost): New option.
13840 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
13843 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
13846 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
13847 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
13849 2018-01-10 Richard Biener <rguenther@suse.de>
13852 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
13853 early out so it also covers the case where we have a non-NULL
13856 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13858 PR tree-optimization/83753
13859 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
13860 for non-strided grouped accesses if the number of elements is 1.
13862 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13865 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
13866 * i386.h (TARGET_USE_GATHER): Define.
13867 * x86-tune.def (X86_TUNE_USE_GATHER): New.
13869 2018-01-10 Martin Liska <mliska@suse.cz>
13872 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
13873 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
13875 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
13876 CLEANUP_NO_PARTITIONING is not set.
13878 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13880 * doc/rtl.texi: Remove documentation of (const ...) wrappers
13881 for vectors, as a partial revert of r254296.
13882 * rtl.h (const_vec_p): Delete.
13883 (const_vec_duplicate_p): Don't test for vector CONSTs.
13884 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
13885 * expmed.c (make_tree): Likewise.
13888 * common.md (E, F): Use CONSTANT_P instead of checking for
13890 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
13891 checking for CONST_VECTOR.
13893 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13895 PR middle-end/83575
13896 * predict.c (force_edge_cold): Handle in more sane way edges
13897 with no prediction.
13899 2018-01-09 Carl Love <cel@us.ibm.com>
13901 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
13903 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
13904 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
13905 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
13906 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
13907 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
13908 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
13909 * config/rs6000/rs6000-protos.h: Add extern defition for
13910 rs6000_generate_float2_double_code.
13911 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
13913 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
13914 (float2_v2df): Add define_expand.
13916 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
13919 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
13920 op_mode in the force_to_mode call.
13922 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13924 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
13925 instead of checking each element individually.
13926 (aarch64_evpc_uzp): Likewise.
13927 (aarch64_evpc_zip): Likewise.
13928 (aarch64_evpc_ext): Likewise.
13929 (aarch64_evpc_rev): Likewise.
13930 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
13931 instead of checking each element individually. Return true without
13933 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
13934 whether all selected elements come from the same input, instead of
13935 checking each element individually. Remove calls to gen_rtx_REG,
13936 start_sequence and end_sequence and instead assert that no rtl is
13939 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13941 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
13942 order of HIGH and CONST checks.
13944 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13946 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
13947 if the destination isn't an SSA_NAME.
13949 2018-01-09 Richard Biener <rguenther@suse.de>
13951 PR tree-optimization/83668
13952 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
13954 (canonicalize_loop_form): ... here, renamed from ...
13955 (canonicalize_loop_closed_ssa_form): ... this and amended to
13956 swap successor edges for loop exit blocks to make us use
13957 the RPO order we need for initial schedule generation.
13959 2018-01-09 Joseph Myers <joseph@codesourcery.com>
13961 PR tree-optimization/64811
13962 * match.pd: When optimizing comparisons with Inf, avoid
13963 introducing or losing exceptions from comparisons with NaN.
13965 2018-01-09 Martin Liska <mliska@suse.cz>
13968 * asan.c (shadow_mem_size): Add gcc_assert.
13970 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
13972 Don't save registers in main().
13975 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
13976 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
13977 * config/avr/avr.c (avr_set_current_function): Don't error if
13978 naked, OS_task or OS_main are specified at the same time.
13979 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
13981 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
13983 * common/config/avr/avr-common.c (avr_option_optimization_table):
13984 Switch on -mmain-is-OS_task for optimizing compilations.
13986 2018-01-09 Richard Biener <rguenther@suse.de>
13988 PR tree-optimization/83572
13989 * graphite.c: Include cfganal.h.
13990 (graphite_transform_loops): Connect infinite loops to exit
13991 and remove fake edges at the end.
13993 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13995 * ipa-inline.c (edge_badness): Revert accidental checkin.
13997 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14000 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
14001 symbols; not inline clones.
14003 2018-01-09 Jakub Jelinek <jakub@redhat.com>
14006 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
14007 hard registers. Formatting fixes.
14009 PR preprocessor/83722
14010 * gcc.c (try_generate_repro): Pass
14011 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
14012 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
14015 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
14016 Kito Cheng <kito.cheng@gmail.com>
14018 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
14019 (riscv_leaf_function_p): Delete.
14020 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
14022 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14024 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
14026 (do_ifelse): New function.
14027 (do_isel): New function.
14028 (do_sub3): New function.
14029 (do_add3): New function.
14030 (do_load_mask_compare): New function.
14031 (do_overlap_load_compare): New function.
14032 (expand_compare_loop): New function.
14033 (expand_block_compare): Call expand_compare_loop() when appropriate.
14034 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
14035 option description.
14036 (-mblock-compare-inline-loop-limit): New option.
14038 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14041 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
14042 Reverse order of second and third operands in first alternative.
14043 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
14044 of first and second elements in UNSPEC_VPERMR vector.
14045 (altivec_expand_vec_perm_le): Likewise.
14047 2018-01-08 Jeff Law <law@redhat.com>
14049 PR rtl-optimizatin/81308
14050 * tree-switch-conversion.c (cfg_altered): New file scoped static.
14051 (process_switch): If group_case_labels makes a change, then set
14053 (pass_convert_switch::execute): If a switch is converted, then
14054 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
14056 PR rtl-optimization/81308
14057 * recog.c (split_all_insns): Conditionally cleanup the CFG after
14060 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
14062 PR target/83663 - Revert r255946
14063 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
14064 generation for cases where splatting a value is not useful.
14065 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
14066 across a vec_duplicate and a paradoxical subreg forming a vector
14067 mode to a vec_concat.
14069 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14071 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
14072 -march=armv8.3-a variants.
14073 * config/arm/t-multilib: Likewise.
14074 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
14076 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14078 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
14080 (cceq_ior_compare_complement): Give it a name so I can use it, and
14081 change boolean_or_operator predicate to boolean_operator so it can
14082 be used to generate a crand.
14083 (eqne): New code iterator.
14084 (bd/bd_neg): New code_attrs.
14085 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
14086 a single define_insn.
14087 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
14088 decrement (bdnzt/bdnzf/bdzt/bdzf).
14089 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
14090 with the new names of the branch decrement patterns, and added the
14091 names of the branch decrement conditional patterns.
14093 2018-01-08 Richard Biener <rguenther@suse.de>
14095 PR tree-optimization/83563
14096 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
14099 2018-01-08 Richard Biener <rguenther@suse.de>
14101 PR middle-end/83713
14102 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
14104 2018-01-08 Richard Biener <rguenther@suse.de>
14106 PR tree-optimization/83685
14107 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
14108 references to abnormals.
14110 2018-01-08 Richard Biener <rguenther@suse.de>
14113 * dwarf2out.c (output_indirect_strings): Handle empty
14114 skeleton_debug_str_hash.
14115 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
14117 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
14119 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
14120 (emit_store_direct): Likewise.
14121 (arc_trampoline_adjust_address): Likewise.
14122 (arc_asm_trampoline_template): New function.
14123 (arc_initialize_trampoline): Use asm_trampoline_template.
14124 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
14125 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
14126 * config/arc/arc.md (flush_icache): Delete pattern.
14128 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
14130 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
14131 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
14134 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
14137 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
14138 by not USED_FOR_TARGET.
14139 (make_pass_resolve_sw_modes): Likewise.
14141 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
14143 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
14146 2018-01-08 Richard Biener <rguenther@suse.de>
14148 PR middle-end/83580
14149 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
14151 2018-01-08 Richard Biener <rguenther@suse.de>
14153 PR middle-end/83517
14154 * match.pd ((t * 2) / 2) -> t): Add missing :c.
14156 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
14158 PR middle-end/81897
14159 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
14160 basic blocks with a small number of successors.
14161 (convert_control_dep_chain_into_preds): Improve handling of
14163 (dump_predicates): Split apart into...
14164 (dump_pred_chain): ...here...
14165 (dump_pred_info): ...and here.
14166 (can_one_predicate_be_invalidated_p): Add debugging printfs.
14167 (can_chain_union_be_invalidated_p): Improve check for invalidation
14169 (uninit_uses_cannot_happen): Avoid unnecessary if
14170 convert_control_dep_chain_into_preds yielded nothing.
14172 2018-01-06 Martin Sebor <msebor@redhat.com>
14174 PR tree-optimization/83640
14175 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
14176 subtracting negative offset from size.
14177 (builtin_access::overlap): Adjust offset bounds of the access to fall
14178 within the size of the object if possible.
14180 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
14182 PR rtl-optimization/83699
14183 * expmed.c (extract_bit_field_1): Restrict the vector usage of
14184 extract_bit_field_as_subreg to cases in which the extracted
14185 value is also a vector.
14187 * lra-constraints.c (process_alt_operands): Test for the equivalence
14188 substitutions when detecting a possible reload cycle.
14190 2018-01-06 Jakub Jelinek <jakub@redhat.com>
14193 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
14194 by default if flag_selective_schedling{,2}. Formatting fixes.
14196 PR rtl-optimization/83682
14197 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
14198 if it has non-VECTOR_MODE element mode.
14199 (vec_duplicate_p): Likewise.
14201 PR middle-end/83694
14202 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
14203 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
14205 2018-01-05 Jakub Jelinek <jakub@redhat.com>
14208 * config/i386/i386-builtin.def
14209 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
14210 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
14211 Require also OPTION_MASK_ISA_AVX512F in addition to
14212 OPTION_MASK_ISA_GFNI.
14213 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
14214 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
14215 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
14216 to OPTION_MASK_ISA_GFNI.
14217 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
14218 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
14219 OPTION_MASK_ISA_AVX512BW.
14220 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
14221 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
14222 addition to OPTION_MASK_ISA_GFNI.
14223 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
14224 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
14225 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
14226 to OPTION_MASK_ISA_GFNI.
14227 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
14228 a requirement for all ISAs rather than any of them with a few
14230 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
14232 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
14233 bitmasks to be enabled with 3 exceptions, instead of requiring any
14234 enabled ISA with lots of exceptions.
14235 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
14236 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
14237 Change avx512bw in isa attribute to avx512f.
14238 * config/i386/sgxintrin.h: Add license boilerplate.
14239 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
14240 to __AVX512F__ and __AVX512VL to __AVX512VL__.
14241 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
14242 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
14244 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
14245 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
14246 temporarily sse2 rather than sse if not enabled already.
14249 * config/i386/sse.md (VI248_VLBW): Rename to ...
14250 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
14251 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
14252 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
14253 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
14254 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
14255 mode iterator instead of VI248_VLBW.
14257 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
14259 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
14260 (record_modified): Skip clobbers; add debug output.
14261 (param_change_prob): Use sreal frequencies.
14263 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
14265 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
14266 punt for user-aligned variables.
14268 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
14270 * tree-chrec.c (chrec_contains_symbols): Return true for
14273 2018-01-05 Sudakshina Das <sudi.das@arm.com>
14276 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
14277 of (x|y) == x for BICS pattern.
14279 2018-01-05 Jakub Jelinek <jakub@redhat.com>
14281 PR tree-optimization/83605
14282 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
14283 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
14286 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
14288 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
14289 * config/epiphany/rtems.h: New file.
14291 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14292 Uros Bizjak <ubizjak@gmail.com>
14295 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
14296 QIreg_operand instead of register_operand predicate.
14297 * config/i386/i386.c (ix86_rop_should_change_byte_p,
14298 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
14299 comments instead of -fmitigate[-_]rop.
14301 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14304 * cgraphunit.c (symbol_table::compile): Switch to text_section
14305 before calling assembly_start debug hook.
14306 * run-rtl-passes.c (run_rtl_passes): Likewise.
14309 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14311 * tree-vrp.c (extract_range_from_binary_expr_1): Check
14312 range_int_cst_p rather than !symbolic_range_p before calling
14313 extract_range_from_multiplicative_op_1.
14315 2018-01-04 Jeff Law <law@redhat.com>
14317 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
14318 redundant test in assertion.
14320 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14322 * doc/rtl.texi: Document machine_mode wrapper classes.
14324 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14326 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
14327 using tree_to_uhwi.
14329 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14331 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
14332 the VEC_PERM_EXPR fold to fail.
14334 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14337 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
14338 to switched_sections.
14340 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14343 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
14344 test for d.testing.
14346 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
14349 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
14350 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
14352 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14355 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
14356 is BLKmode and bitpos not zero or mode change is needed.
14358 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14361 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
14364 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
14367 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
14368 instead of MULT rtx. Update all corresponding splitters.
14369 (*saddl_se): Ditto.
14370 (*ssub<modesuffix>): Ditto.
14371 (*ssubl_se): Ditto.
14372 (*cmp_sadd_di): Update split patterns.
14373 (*cmp_sadd_si): Ditto.
14374 (*cmp_sadd_sidi): Ditto.
14375 (*cmp_ssub_di): Ditto.
14376 (*cmp_ssub_si): Ditto.
14377 (*cmp_ssub_sidi): Ditto.
14378 * config/alpha/predicates.md (const23_operand): New predicate.
14379 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
14380 Look for ASHIFT, not MULT inner operand.
14381 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
14383 2018-01-04 Martin Liska <mliska@suse.cz>
14385 PR gcov-profile/83669
14386 * gcov.c (output_intermediate_file): Add version to intermediate
14388 * doc/gcov.texi: Document new field 'version' in intermediate
14389 file format. Fix location of '-k' option of gcov command.
14391 2018-01-04 Martin Liska <mliska@suse.cz>
14394 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
14396 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14398 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
14400 2018-01-03 Martin Sebor <msebor@redhat.com>
14402 PR tree-optimization/83655
14403 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
14404 checking calls with invalid arguments.
14406 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14408 * tree-vect-stmts.c (vect_get_store_rhs): New function.
14409 (vectorizable_mask_load_store): Delete.
14410 (vectorizable_call): Return false for masked loads and stores.
14411 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
14412 instead of gimple_assign_rhs1.
14413 (vectorizable_load): Handle IFN_MASK_LOAD.
14414 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
14416 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14418 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
14420 (vectorizable_mask_load_store): ...here.
14421 (vectorizable_load): ...and here.
14423 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14425 * tree-vect-stmts.c (vect_build_all_ones_mask)
14426 (vect_build_zero_merge_argument): New functions, split out from...
14427 (vectorizable_load): ...here.
14429 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14431 * tree-vect-stmts.c (vect_check_store_rhs): New function,
14433 (vectorizable_mask_load_store): ...here.
14434 (vectorizable_store): ...and here.
14436 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14438 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
14440 (vectorizable_mask_load_store): ...here.
14442 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14444 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
14445 (vect_model_store_cost): Take a vec_load_store_type instead of a
14447 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
14448 (vect_model_store_cost): Take a vec_load_store_type instead of a
14450 (vectorizable_mask_load_store): Update accordingly.
14451 (vectorizable_store): Likewise.
14452 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
14454 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14456 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
14457 IFN_MASK_LOAD calls here rather than...
14458 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
14460 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14461 Alan Hayward <alan.hayward@arm.com>
14462 David Sherwood <david.sherwood@arm.com>
14464 * expmed.c (extract_bit_field_1): For vector extracts,
14465 fall back to extract_bit_field_as_subreg if vec_extract
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 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
14473 they are variable or constant sized.
14474 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
14475 slots for constant-sized data.
14477 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14478 Alan Hayward <alan.hayward@arm.com>
14479 David Sherwood <david.sherwood@arm.com>
14481 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
14482 handling COND_EXPRs with boolean comparisons, try to find a better
14483 basis for the mask type than the boolean itself.
14485 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14487 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
14488 is calculated and how it can be overridden.
14489 * genmodes.c (max_bitsize_mode_any_mode): New variable.
14490 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
14492 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
14495 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14496 Alan Hayward <alan.hayward@arm.com>
14497 David Sherwood <david.sherwood@arm.com>
14499 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
14500 Remove the mode argument.
14501 (aarch64_simd_valid_immediate): Remove the mode and inverse
14503 * config/aarch64/iterators.md (bitsize): New iterator.
14504 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
14505 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
14506 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
14507 aarch64_simd_valid_immediate.
14508 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
14509 (aarch64_reg_or_bic_imm): Likewise.
14510 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
14511 with an insn_type enum and msl with a modifier_type enum.
14512 Replace element_width with a scalar_mode. Change the shift
14513 to unsigned int. Add constructors for scalar_float_mode and
14514 scalar_int_mode elements.
14515 (aarch64_vect_float_const_representable_p): Delete.
14516 (aarch64_can_const_movi_rtx_p)
14517 (aarch64_simd_scalar_immediate_valid_for_move)
14518 (aarch64_simd_make_constant): Update call to
14519 aarch64_simd_valid_immediate.
14520 (aarch64_advsimd_valid_immediate_hs): New function.
14521 (aarch64_advsimd_valid_immediate): Likewise.
14522 (aarch64_simd_valid_immediate): Remove mode and inverse
14523 arguments. Rewrite to use the above. Use const_vec_duplicate_p
14524 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
14525 and aarch64_float_const_representable_p on the result.
14526 (aarch64_output_simd_mov_immediate): Remove mode argument.
14527 Update call to aarch64_simd_valid_immediate and use of
14528 simd_immediate_info.
14529 (aarch64_output_scalar_simd_mov_immediate): Update call
14532 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14533 Alan Hayward <alan.hayward@arm.com>
14534 David Sherwood <david.sherwood@arm.com>
14536 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
14537 (mode_nunits): Likewise CONST_MODE_NUNITS.
14538 * machmode.def (ADJUST_NUNITS): Document.
14539 * genmodes.c (mode_data::need_nunits_adj): New field.
14540 (blank_mode): Update accordingly.
14541 (adj_nunits): New variable.
14542 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
14544 (emit_mode_size_inline): Set need_bytesize_adj for all modes
14545 listed in adj_nunits.
14546 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
14547 listed in adj_nunits. Don't emit case statements for such modes.
14548 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
14549 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
14550 nothing if adj_nunits is nonnull.
14551 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
14552 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
14553 (emit_mode_fbit): Update use of print_maybe_const_decl.
14554 (emit_move_size): Likewise. Treat the array as non-const
14556 (emit_mode_adjustments): Handle adj_nunits.
14558 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14560 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
14561 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
14562 (VECTOR_MODES): Use it.
14563 (make_vector_modes): Take the prefix as an argument.
14565 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14566 Alan Hayward <alan.hayward@arm.com>
14567 David Sherwood <david.sherwood@arm.com>
14569 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
14570 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
14571 for MODE_VECTOR_BOOL.
14572 * machmode.def (VECTOR_BOOL_MODE): Document.
14573 * genmodes.c (VECTOR_BOOL_MODE): New macro.
14574 (make_vector_bool_mode): New function.
14575 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
14577 * lto-streamer-in.c (lto_input_mode_table): Likewise.
14578 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
14580 * stor-layout.c (int_mode_for_mode): Likewise.
14581 * tree.c (build_vector_type_for_mode): Likewise.
14582 * varasm.c (output_constant_pool_2): Likewise.
14583 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
14584 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
14585 for MODE_VECTOR_BOOL.
14586 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
14587 of mode class checks.
14588 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
14589 instead of a list of mode class checks.
14590 (expand_vector_scalar_condition): Likewise.
14591 (type_for_widest_vector_mode): Handle BImode as an inner mode.
14593 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14594 Alan Hayward <alan.hayward@arm.com>
14595 David Sherwood <david.sherwood@arm.com>
14597 * machmode.h (mode_size): Change from unsigned short to
14599 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
14600 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
14601 or if measurement_type is not polynomial.
14602 (fixed_size_mode::includes_p): Check for constant-sized modes.
14603 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
14604 return a poly_uint16 rather than an unsigned short.
14605 (emit_mode_size): Change the type of mode_size from unsigned short
14606 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
14607 (emit_mode_adjustments): Cope with polynomial vector sizes.
14608 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14610 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14612 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
14613 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
14614 * caller-save.c (setup_save_areas): Likewise.
14615 (replace_reg_with_saved_mem): Likewise.
14616 * calls.c (emit_library_call_value_1): Likewise.
14617 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
14618 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
14619 (gen_lowpart_for_combine): Likewise.
14620 * convert.c (convert_to_integer_1): Likewise.
14621 * cse.c (equiv_constant, cse_insn): Likewise.
14622 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
14623 (cselib_subst_to_values): Likewise.
14624 * dce.c (word_dce_process_block): Likewise.
14625 * df-problems.c (df_word_lr_mark_ref): Likewise.
14626 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
14627 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
14628 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
14629 (rtl_for_decl_location): Likewise.
14630 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
14631 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
14632 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
14633 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
14634 (expand_expr_real_1): Likewise.
14635 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
14636 (pad_below): Likewise.
14637 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
14638 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
14639 * ira.c (get_subreg_tracking_sizes): Likewise.
14640 * ira-build.c (ira_create_allocno_objects): Likewise.
14641 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
14642 (ira_sort_regnos_for_alter_reg): Likewise.
14643 * ira-costs.c (record_operand_costs): Likewise.
14644 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
14645 (resolve_simple_move): Likewise.
14646 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
14647 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
14648 (lra_constraints): Likewise.
14649 (CONST_POOL_OK_P): Reject variable-sized modes.
14650 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
14651 (add_pseudo_to_slot, lra_spill): Likewise.
14652 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14653 * optabs-query.c (get_best_extraction_insn): Likewise.
14654 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14655 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
14656 (expand_mult_highpart, valid_multiword_target_p): Likewise.
14657 * recog.c (offsettable_address_addr_space_p): Likewise.
14658 * regcprop.c (maybe_mode_change): Likewise.
14659 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
14660 * regrename.c (build_def_use): Likewise.
14661 * regstat.c (dump_reg_info): Likewise.
14662 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
14663 (find_reloads, find_reloads_subreg_address): Likewise.
14664 * reload1.c (eliminate_regs_1): Likewise.
14665 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
14666 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
14667 (simplify_binary_operation_1, simplify_subreg): Likewise.
14668 * targhooks.c (default_function_arg_padding): Likewise.
14669 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
14670 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
14671 (verify_gimple_assign_ternary): Likewise.
14672 * tree-inline.c (estimate_move_cost): Likewise.
14673 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14674 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
14675 (get_address_cost_ainc): Likewise.
14676 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
14677 (vect_supportable_dr_alignment): Likewise.
14678 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14679 (vectorizable_reduction): Likewise.
14680 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
14681 (vectorizable_operation, vectorizable_load): Likewise.
14682 * tree.c (build_same_sized_truth_vector_type): Likewise.
14683 * valtrack.c (cleanup_auto_inc_dec): Likewise.
14684 * var-tracking.c (emit_note_insn_var_location): Likewise.
14685 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
14686 (ADDR_VEC_ALIGN): Likewise.
14688 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14689 Alan Hayward <alan.hayward@arm.com>
14690 David Sherwood <david.sherwood@arm.com>
14692 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
14694 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
14695 or if measurement_type is polynomial.
14696 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
14697 * combine.c (make_extraction): Likewise.
14698 * dse.c (find_shift_sequence): Likewise.
14699 * dwarf2out.c (mem_loc_descriptor): Likewise.
14700 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
14701 (extract_bit_field, extract_low_bits): Likewise.
14702 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
14703 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
14704 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
14705 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
14706 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
14707 * reload.c (find_reloads): Likewise.
14708 * reload1.c (alter_reg): Likewise.
14709 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
14710 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
14711 * tree-if-conv.c (predicate_mem_writes): Likewise.
14712 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
14713 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
14714 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
14715 * valtrack.c (dead_debug_insert_temp): Likewise.
14716 * varasm.c (mergeable_constant_section): Likewise.
14717 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
14719 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14720 Alan Hayward <alan.hayward@arm.com>
14721 David Sherwood <david.sherwood@arm.com>
14723 * expr.c (expand_assignment): Cope with polynomial mode sizes
14724 when assigning to a CONCAT.
14726 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14727 Alan Hayward <alan.hayward@arm.com>
14728 David Sherwood <david.sherwood@arm.com>
14730 * machmode.h (mode_precision): Change from unsigned short to
14732 (mode_to_precision): Return a poly_uint16 rather than an unsigned
14734 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
14735 or if measurement_type is not polynomial.
14736 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
14737 in which the mode is already known to be a scalar_int_mode.
14738 * genmodes.c (emit_mode_precision): Change the type of mode_precision
14739 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
14741 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14742 for GET_MODE_PRECISION.
14743 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14744 for GET_MODE_PRECISION.
14745 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
14747 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
14748 (expand_field_assignment, make_extraction): Likewise.
14749 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
14750 (get_last_value): Likewise.
14751 * convert.c (convert_to_integer_1): Likewise.
14752 * cse.c (cse_insn): Likewise.
14753 * expr.c (expand_expr_real_1): Likewise.
14754 * lra-constraints.c (simplify_operand_subreg): Likewise.
14755 * optabs-query.c (can_atomic_load_p): Likewise.
14756 * optabs.c (expand_atomic_load): Likewise.
14757 (expand_atomic_store): Likewise.
14758 * ree.c (combine_reaching_defs): Likewise.
14759 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
14760 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
14761 * tree.h (type_has_mode_precision_p): Likewise.
14762 * ubsan.c (instrument_si_overflow): Likewise.
14764 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14765 Alan Hayward <alan.hayward@arm.com>
14766 David Sherwood <david.sherwood@arm.com>
14768 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
14769 polynomial numbers of units.
14770 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
14771 (valid_vector_subparts_p): New function.
14772 (build_vector_type): Remove temporary shim and take the number
14773 of units as a poly_uint64 rather than an int.
14774 (build_opaque_vector_type): Take the number of units as a
14775 poly_uint64 rather than an int.
14776 * tree.c (build_vector_from_ctor): Handle polynomial
14777 TYPE_VECTOR_SUBPARTS.
14778 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
14779 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
14780 (build_vector_from_val): If the number of units is variable,
14781 use build_vec_duplicate_cst for constant operands and
14782 VEC_DUPLICATE_EXPR otherwise.
14783 (make_vector_type): Remove temporary is_constant ().
14784 (build_vector_type, build_opaque_vector_type): Take the number of
14785 units as a poly_uint64 rather than an int.
14786 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
14788 * cfgexpand.c (expand_debug_expr): Likewise.
14789 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
14790 (store_constructor, expand_expr_real_1): Likewise.
14791 (const_scalar_mask_from_tree): Likewise.
14792 * fold-const-call.c (fold_const_reduction): Likewise.
14793 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
14794 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
14795 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
14796 (fold_relational_const): Likewise.
14797 (native_interpret_vector): Likewise. Change the size from an
14798 int to an unsigned int.
14799 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
14800 TYPE_VECTOR_SUBPARTS.
14801 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
14802 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
14803 duplicating a non-constant operand into a variable-length vector.
14804 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
14805 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
14806 * ipa-icf.c (sem_variable::equals): Likewise.
14807 * match.pd: Likewise.
14808 * omp-simd-clone.c (simd_clone_subparts): Likewise.
14809 * print-tree.c (print_node): Likewise.
14810 * stor-layout.c (layout_type): Likewise.
14811 * targhooks.c (default_builtin_vectorization_cost): Likewise.
14812 * tree-cfg.c (verify_gimple_comparison): Likewise.
14813 (verify_gimple_assign_binary): Likewise.
14814 (verify_gimple_assign_ternary): Likewise.
14815 (verify_gimple_assign_single): Likewise.
14816 * tree-pretty-print.c (dump_generic_node): Likewise.
14817 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14818 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
14819 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
14820 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
14821 (vect_shift_permute_load_chain): Likewise.
14822 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
14823 (expand_vector_condition, optimize_vector_constructor): Likewise.
14824 (lower_vec_perm, get_compute_type): Likewise.
14825 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14826 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
14827 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
14828 (vect_recog_mask_conversion_pattern): Likewise.
14829 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
14830 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
14831 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14832 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
14833 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
14834 (vectorizable_shift, vectorizable_operation, vectorizable_store)
14835 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
14836 (supportable_widening_operation): Likewise.
14837 (supportable_narrowing_operation): Likewise.
14838 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
14840 * varasm.c (output_constant): Likewise.
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 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
14847 so that both the length == 3 and length != 3 cases set up their
14848 own permute vectors. Add comments explaining why we know the
14849 number of elements is constant.
14850 (vect_permute_load_chain): Likewise.
14852 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14853 Alan Hayward <alan.hayward@arm.com>
14854 David Sherwood <david.sherwood@arm.com>
14856 * machmode.h (mode_nunits): Change from unsigned char to
14858 (ONLY_FIXED_SIZE_MODES): New macro.
14859 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
14860 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
14861 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
14863 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
14864 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
14865 or if measurement_type is not polynomial.
14866 * genmodes.c (ZERO_COEFFS): New macro.
14867 (emit_mode_nunits_inline): Make mode_nunits_inline return a
14869 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
14870 Use ZERO_COEFFS when emitting initializers.
14871 * data-streamer.h (bp_pack_poly_value): New function.
14872 (bp_unpack_poly_value): Likewise.
14873 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14874 for GET_MODE_NUNITS.
14875 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14876 for GET_MODE_NUNITS.
14877 * tree.c (make_vector_type): Remove temporary shim and make
14878 the real function take the number of units as a poly_uint64
14879 rather than an int.
14880 (build_vector_type_for_mode): Handle polynomial nunits.
14881 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
14882 * emit-rtl.c (const_vec_series_p_1): Likewise.
14883 (gen_rtx_CONST_VECTOR): Likewise.
14884 * fold-const.c (test_vec_duplicate_folding): Likewise.
14885 * genrecog.c (validate_pattern): Likewise.
14886 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
14887 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14888 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
14889 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
14890 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
14891 * rtlanal.c (subreg_get_info): Likewise.
14892 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14893 (vect_grouped_load_supported): Likewise.
14894 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
14895 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
14896 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
14897 (simplify_const_unary_operation, simplify_binary_operation_1)
14898 (simplify_const_binary_operation, simplify_ternary_operation)
14899 (test_vector_ops_duplicate, test_vector_ops): Likewise.
14900 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
14901 instead of CONST_VECTOR_NUNITS.
14902 * varasm.c (output_constant_pool_2): Likewise.
14903 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
14904 explicit-encoded elements in the XVEC for variable-length vectors.
14906 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14908 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
14910 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14911 Alan Hayward <alan.hayward@arm.com>
14912 David Sherwood <david.sherwood@arm.com>
14914 * coretypes.h (fixed_size_mode): Declare.
14915 (fixed_size_mode_pod): New typedef.
14916 * builtins.h (target_builtins::x_apply_args_mode)
14917 (target_builtins::x_apply_result_mode): Change type to
14918 fixed_size_mode_pod.
14919 * builtins.c (apply_args_size, apply_result_size, result_vector)
14920 (expand_builtin_apply_args_1, expand_builtin_apply)
14921 (expand_builtin_return): Update accordingly.
14923 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14925 * cse.c (hash_rtx_cb): Hash only the encoded elements.
14926 * cselib.c (cselib_hash_rtx): Likewise.
14927 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
14928 CONST_VECTOR encoding.
14930 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14931 Jeff Law <law@redhat.com>
14934 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
14935 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
14936 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
14937 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
14940 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
14941 explicitly probe *sp in a noreturn function if there were any callee
14942 register saves or frame pointer is needed.
14944 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14947 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
14948 BLKmode for ternary, binary or unary expressions.
14951 * var-tracking.c (delete_vta_debug_insn): New inline function.
14952 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
14953 insns from get_insns () to NULL instead of each bb separately.
14954 Use delete_vta_debug_insn. No longer static.
14955 (vt_debug_insns_local, variable_tracking_main_1): Adjust
14956 delete_vta_debug_insns callers.
14957 * rtl.h (delete_vta_debug_insns): Declare.
14958 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
14959 instead of variable_tracking_main.
14961 2018-01-03 Martin Sebor <msebor@redhat.com>
14963 PR tree-optimization/83603
14964 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
14965 arguments past the endof the argument list in functions declared
14966 without a prototype.
14967 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
14968 Avoid checking when arguments are null.
14970 2018-01-03 Martin Sebor <msebor@redhat.com>
14973 * doc/extend.texi (attribute const): Fix a typo.
14974 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
14975 issuing -Wsuggest-attribute for void functions.
14977 2018-01-03 Martin Sebor <msebor@redhat.com>
14979 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
14980 offset_int::from instead of wide_int::to_shwi.
14981 (maybe_diag_overlap): Remove assertion.
14982 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
14983 * gimple-ssa-sprintf.c (format_directive): Same.
14984 (parse_directive): Same.
14985 (sprintf_dom_walker::compute_format_length): Same.
14986 (try_substitute_return_value): Same.
14988 2018-01-03 Jeff Law <law@redhat.com>
14990 PR middle-end/83654
14991 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
14992 non-constant residual for zero at runtime and avoid probing in
14993 that case. Reorganize code for trailing problem to mirror handling
14996 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14998 PR tree-optimization/83501
14999 * tree-ssa-strlen.c (get_string_cst): New.
15000 (handle_char_store): Call get_string_cst.
15002 2018-01-03 Martin Liska <mliska@suse.cz>
15004 PR tree-optimization/83593
15005 * tree-ssa-strlen.c: Include tree-cfg.h.
15006 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
15007 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
15008 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
15010 (strlen_dom_walker::before_dom_children): Call
15011 gimple_purge_dead_eh_edges. Dump tranformation with details
15013 (strlen_dom_walker::before_dom_children): Update call by adding
15014 new argument cleanup_eh.
15015 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
15017 2018-01-03 Martin Liska <mliska@suse.cz>
15020 * cif-code.def (VARIADIC_THUNK): New enum value.
15021 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
15024 2018-01-03 Jan Beulich <jbeulich@suse.com>
15026 * sse.md (mov<mode>_internal): Tighten condition for when to use
15027 vmovdqu<ssescalarsize> for TI and OI modes.
15029 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15031 Update copyright years.
15033 2018-01-03 Martin Liska <mliska@suse.cz>
15036 * ipa-visibility.c (function_and_variable_visibility): Skip
15037 functions with noipa attribure.
15039 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15041 * gcc.c (process_command): Update copyright notice dates.
15042 * gcov-dump.c (print_version): Ditto.
15043 * gcov.c (print_version): Ditto.
15044 * gcov-tool.c (print_version): Ditto.
15045 * gengtype.c (create_file): Ditto.
15046 * doc/cpp.texi: Bump @copying's copyright year.
15047 * doc/cppinternals.texi: Ditto.
15048 * doc/gcc.texi: Ditto.
15049 * doc/gccint.texi: Ditto.
15050 * doc/gcov.texi: Ditto.
15051 * doc/install.texi: Ditto.
15052 * doc/invoke.texi: Ditto.
15054 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15056 * vector-builder.h (vector_builder::m_full_nelts): Change from
15057 unsigned int to poly_uint64.
15058 (vector_builder::full_nelts): Update prototype accordingly.
15059 (vector_builder::new_vector): Likewise.
15060 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
15061 (vector_builder::operator ==): Likewise.
15062 (vector_builder::finalize): Likewise.
15063 * int-vector-builder.h (int_vector_builder::int_vector_builder):
15064 Take the number of elements as a poly_uint64 rather than an
15066 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
15067 from unsigned int to poly_uint64.
15068 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
15069 (vec_perm_indices::new_vector): Likewise.
15070 (vec_perm_indices::length): Likewise.
15071 (vec_perm_indices::nelts_per_input): Likewise.
15072 (vec_perm_indices::input_nelts): Likewise.
15073 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
15074 number of elements per input as a poly_uint64 rather than an
15075 unsigned int. Use the original encoding for variable-length
15076 vectors, rather than clamping each individual element.
15077 For the second and subsequent elements in each pattern,
15078 clamp the step and base before clamping their sum.
15079 (vec_perm_indices::series_p): Handle polynomial element counts.
15080 (vec_perm_indices::all_in_range_p): Likewise.
15081 (vec_perm_indices_to_tree): Likewise.
15082 (vec_perm_indices_to_rtx): Likewise.
15083 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
15084 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
15085 (tree_vector_builder::new_binary_operation): Handle polynomial
15086 element counts. Return false if we need to know the number
15087 of elements at compile time.
15088 * fold-const.c (fold_vec_perm): Punt if the number of elements
15089 isn't known at compile time.
15091 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15093 * vec-perm-indices.h (vec_perm_builder): Change element type
15094 from HOST_WIDE_INT to poly_int64.
15095 (vec_perm_indices::element_type): Update accordingly.
15096 (vec_perm_indices::clamp): Handle polynomial element_types.
15097 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
15098 (vec_perm_indices::all_in_range_p): Likewise.
15099 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
15101 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
15102 polynomial vec_perm_indices element types.
15103 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
15104 * fold-const.c (fold_vec_perm): Likewise.
15105 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
15106 * tree-vect-generic.c (lower_vec_perm): Likewise.
15107 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
15108 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
15109 element type to HOST_WIDE_INT.
15111 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15112 Alan Hayward <alan.hayward@arm.com>
15113 David Sherwood <david.sherwood@arm.com>
15115 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
15116 rather than an int. Use plus_constant.
15117 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
15118 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
15120 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15121 Alan Hayward <alan.hayward@arm.com>
15122 David Sherwood <david.sherwood@arm.com>
15124 * calls.c (emit_call_1, expand_call): Change struct_value_size from
15125 a HOST_WIDE_INT to a poly_int64.
15127 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15128 Alan Hayward <alan.hayward@arm.com>
15129 David Sherwood <david.sherwood@arm.com>
15131 * calls.c (load_register_parameters): Cope with polynomial
15132 mode sizes. Require a constant size for BLKmode parameters
15133 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
15134 forces a parameter to be padded at the lsb end in order to
15135 fill a complete number of words, require the parameter size
15136 to be ordered wrt UNITS_PER_WORD.
15138 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15139 Alan Hayward <alan.hayward@arm.com>
15140 David Sherwood <david.sherwood@arm.com>
15142 * reload1.c (spill_stack_slot_width): Change element type
15143 from unsigned int to poly_uint64_pod.
15144 (alter_reg): Treat mode sizes as polynomial.
15146 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15147 Alan Hayward <alan.hayward@arm.com>
15148 David Sherwood <david.sherwood@arm.com>
15150 * reload.c (complex_word_subreg_p): New function.
15151 (reload_inner_reg_of_subreg, push_reload): Use it.
15153 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15154 Alan Hayward <alan.hayward@arm.com>
15155 David Sherwood <david.sherwood@arm.com>
15157 * lra-constraints.c (process_alt_operands): Reject matched
15158 operands whose sizes aren't ordered.
15159 (match_reload): Refer to this check here.
15161 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15162 Alan Hayward <alan.hayward@arm.com>
15163 David Sherwood <david.sherwood@arm.com>
15165 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
15166 that the mode size is in the set {1, 2, 4, 8, 16}.
15168 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15169 Alan Hayward <alan.hayward@arm.com>
15170 David Sherwood <david.sherwood@arm.com>
15172 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
15173 Use plus_constant instead of gen_rtx_PLUS.
15175 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15176 Alan Hayward <alan.hayward@arm.com>
15177 David Sherwood <david.sherwood@arm.com>
15179 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
15180 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
15181 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
15182 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
15183 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
15184 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
15185 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
15186 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
15187 * config/i386/i386.c (ix86_push_rounding): ...this new function.
15188 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
15190 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
15191 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
15192 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
15193 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
15194 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
15195 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
15196 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
15197 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
15198 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
15199 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
15201 * expr.c (emit_move_resolve_push): Treat the input and result
15202 of PUSH_ROUNDING as a poly_int64.
15203 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
15204 (emit_push_insn): Likewise.
15205 * lra-eliminations.c (mark_not_eliminable): Likewise.
15206 * recog.c (push_operand): Likewise.
15207 * reload1.c (elimination_effects): Likewise.
15208 * rtlanal.c (nonzero_bits1): Likewise.
15209 * calls.c (store_one_arg): Likewise. Require the padding to be
15210 known at compile time.
15212 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15213 Alan Hayward <alan.hayward@arm.com>
15214 David Sherwood <david.sherwood@arm.com>
15216 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
15217 Use plus_constant instead of gen_rtx_PLUS.
15219 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15220 Alan Hayward <alan.hayward@arm.com>
15221 David Sherwood <david.sherwood@arm.com>
15223 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
15224 rather than an int.
15226 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15227 Alan Hayward <alan.hayward@arm.com>
15228 David Sherwood <david.sherwood@arm.com>
15230 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
15231 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
15232 via stack temporaries. Treat the mode size as polynomial too.
15234 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15235 Alan Hayward <alan.hayward@arm.com>
15236 David Sherwood <david.sherwood@arm.com>
15238 * expr.c (expand_expr_real_2): When handling conversions involving
15239 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
15240 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
15241 as a poly_uint64 too.
15243 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15244 Alan Hayward <alan.hayward@arm.com>
15245 David Sherwood <david.sherwood@arm.com>
15247 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
15249 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15250 Alan Hayward <alan.hayward@arm.com>
15251 David Sherwood <david.sherwood@arm.com>
15253 * combine.c (can_change_dest_mode): Handle polynomial
15254 REGMODE_NATURAL_SIZE.
15255 * expmed.c (store_bit_field_1): Likewise.
15256 * expr.c (store_constructor): Likewise.
15257 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
15258 and polynomial REGMODE_NATURAL_SIZE.
15259 (gen_lowpart_common): Likewise.
15260 * reginfo.c (record_subregs_of_mode): Likewise.
15261 * rtlanal.c (read_modify_subreg_p): Likewise.
15263 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15264 Alan Hayward <alan.hayward@arm.com>
15265 David Sherwood <david.sherwood@arm.com>
15267 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
15268 numbers of elements.
15270 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15271 Alan Hayward <alan.hayward@arm.com>
15272 David Sherwood <david.sherwood@arm.com>
15274 * match.pd: Cope with polynomial numbers of vector elements.
15276 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15277 Alan Hayward <alan.hayward@arm.com>
15278 David Sherwood <david.sherwood@arm.com>
15280 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
15281 in a POINTER_PLUS_EXPR.
15283 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15284 Alan Hayward <alan.hayward@arm.com>
15285 David Sherwood <david.sherwood@arm.com>
15287 * omp-simd-clone.c (simd_clone_subparts): New function.
15288 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
15289 (ipa_simd_modify_function_body): Likewise.
15291 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15292 Alan Hayward <alan.hayward@arm.com>
15293 David Sherwood <david.sherwood@arm.com>
15295 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
15296 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
15297 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
15298 (expand_vector_condition, vector_element): Likewise.
15299 (subparts_gt): New function.
15300 (get_compute_type): Use subparts_gt.
15301 (count_type_subparts): Delete.
15302 (expand_vector_operations_1): Use subparts_gt instead of
15303 count_type_subparts.
15305 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15306 Alan Hayward <alan.hayward@arm.com>
15307 David Sherwood <david.sherwood@arm.com>
15309 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
15310 (vect_compile_time_alias): ...this new function. Do the calculation
15311 on poly_ints rather than trees.
15312 (vect_prune_runtime_alias_test_list): Update call accordingly.
15314 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15315 Alan Hayward <alan.hayward@arm.com>
15316 David Sherwood <david.sherwood@arm.com>
15318 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
15320 (vect_schedule_slp_instance): Likewise.
15322 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15323 Alan Hayward <alan.hayward@arm.com>
15324 David Sherwood <david.sherwood@arm.com>
15326 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
15327 constant and extern definitions for variable-length vectors.
15328 (vect_get_constant_vectors): Note that the number of units
15329 is known to be constant.
15331 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15332 Alan Hayward <alan.hayward@arm.com>
15333 David Sherwood <david.sherwood@arm.com>
15335 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
15336 of units as polynomial. Choose between WIDE and NARROW based
15339 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15340 Alan Hayward <alan.hayward@arm.com>
15341 David Sherwood <david.sherwood@arm.com>
15343 * tree-vect-stmts.c (simd_clone_subparts): New function.
15344 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
15346 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15347 Alan Hayward <alan.hayward@arm.com>
15348 David Sherwood <david.sherwood@arm.com>
15350 * tree-vect-stmts.c (vectorizable_call): Treat the number of
15351 vectors as polynomial. Use build_index_vector for
15352 IFN_GOMP_SIMD_LANE.
15354 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15355 Alan Hayward <alan.hayward@arm.com>
15356 David Sherwood <david.sherwood@arm.com>
15358 * tree-vect-stmts.c (get_load_store_type): Treat the number of
15359 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
15360 for variable-length vectors.
15361 (vectorizable_mask_load_store): Treat the number of units as
15362 polynomial, asserting that it is constant if the condition has
15363 already been enforced.
15364 (vectorizable_store, vectorizable_load): Likewise.
15366 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15367 Alan Hayward <alan.hayward@arm.com>
15368 David Sherwood <david.sherwood@arm.com>
15370 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
15371 of units as polynomial. Punt if we can't tell at compile time
15372 which vector contains the final result.
15374 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15375 Alan Hayward <alan.hayward@arm.com>
15376 David Sherwood <david.sherwood@arm.com>
15378 * tree-vect-loop.c (vectorizable_induction): Treat the number
15379 of units as polynomial. Punt on SLP inductions. Use an integer
15380 VEC_SERIES_EXPR for variable-length integer reductions. Use a
15381 cast of such a series for variable-length floating-point
15384 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15385 Alan Hayward <alan.hayward@arm.com>
15386 David Sherwood <david.sherwood@arm.com>
15388 * tree.h (build_index_vector): Declare.
15389 * tree.c (build_index_vector): New function.
15390 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
15391 of units as polynomial, forcibly converting it to a constant if
15392 vectorizable_reduction has already enforced the condition.
15393 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
15394 to create a {1,2,3,...} vector.
15395 (vectorizable_reduction): Treat the number of units as polynomial.
15396 Choose vectype_in based on the largest scalar element size rather
15397 than the smallest number of units. Enforce the restrictions
15400 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15401 Alan Hayward <alan.hayward@arm.com>
15402 David Sherwood <david.sherwood@arm.com>
15404 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
15405 number of units as polynomial.
15407 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15408 Alan Hayward <alan.hayward@arm.com>
15409 David Sherwood <david.sherwood@arm.com>
15411 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
15412 * target.def (autovectorize_vector_sizes): Return the vector sizes
15413 by pointer, using vector_sizes rather than a bitmask.
15414 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
15415 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
15416 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
15418 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
15419 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
15420 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
15421 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
15422 * omp-general.c (omp_max_vf): Likewise.
15423 * omp-low.c (omp_clause_aligned_alignment): Likewise.
15424 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
15425 * tree-vect-loop.c (vect_analyze_loop): Likewise.
15426 * tree-vect-slp.c (vect_slp_bb): Likewise.
15427 * doc/tm.texi: Regenerate.
15428 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
15430 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
15431 the vector size as a poly_uint64 rather than an unsigned int.
15432 (current_vector_size): Change from an unsigned int to a poly_uint64.
15433 (get_vectype_for_scalar_type): Update accordingly.
15434 * tree.h (build_truth_vector_type): Take the size and number of
15435 units as a poly_uint64 rather than an unsigned int.
15436 (build_vector_type): Add a temporary overload that takes
15437 the number of units as a poly_uint64 rather than an unsigned int.
15438 * tree.c (make_vector_type): Likewise.
15439 (build_truth_vector_type): Take the number of units as a poly_uint64
15440 rather than an unsigned int.
15442 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15443 Alan Hayward <alan.hayward@arm.com>
15444 David Sherwood <david.sherwood@arm.com>
15446 * target.def (get_mask_mode): Take the number of units and length
15447 as poly_uint64s rather than unsigned ints.
15448 * targhooks.h (default_get_mask_mode): Update accordingly.
15449 * targhooks.c (default_get_mask_mode): Likewise.
15450 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
15451 * doc/tm.texi: Regenerate.
15453 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15454 Alan Hayward <alan.hayward@arm.com>
15455 David Sherwood <david.sherwood@arm.com>
15457 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
15458 * omp-general.c (omp_max_vf): Likewise.
15459 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
15460 (expand_omp_simd): Handle polynomial safelen.
15461 * omp-low.c (omplow_simd_context): Add a default constructor.
15462 (omplow_simd_context::max_vf): Change from int to poly_uint64.
15463 (lower_rec_simd_input_clauses): Update accordingly.
15464 (lower_rec_input_clauses): Likewise.
15466 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15467 Alan Hayward <alan.hayward@arm.com>
15468 David Sherwood <david.sherwood@arm.com>
15470 * tree-vectorizer.h (vect_nunits_for_cost): New function.
15471 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
15472 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
15473 (vect_analyze_slp_cost): Likewise.
15474 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
15475 (vect_model_load_cost): Likewise.
15477 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15478 Alan Hayward <alan.hayward@arm.com>
15479 David Sherwood <david.sherwood@arm.com>
15481 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
15482 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
15483 from an unsigned int * to a poly_uint64_pod *.
15484 (calculate_unrolling_factor): New function.
15485 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
15487 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15488 Alan Hayward <alan.hayward@arm.com>
15489 David Sherwood <david.sherwood@arm.com>
15491 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
15492 from an unsigned int to a poly_uint64.
15493 (_loop_vec_info::slp_unrolling_factor): Likewise.
15494 (_loop_vec_info::vectorization_factor): Change from an int
15496 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
15497 (vect_get_num_vectors): New function.
15498 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
15499 (vect_get_num_copies): Use vect_get_num_vectors.
15500 (vect_analyze_data_ref_dependences): Change max_vf from an int *
15501 to an unsigned int *.
15502 (vect_analyze_data_refs): Change min_vf from an int * to a
15504 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
15505 than an unsigned HOST_WIDE_INT.
15506 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
15507 (vect_analyze_data_ref_dependence): Change max_vf from an int *
15508 to an unsigned int *.
15509 (vect_analyze_data_ref_dependences): Likewise.
15510 (vect_compute_data_ref_alignment): Handle polynomial vf.
15511 (vect_enhance_data_refs_alignment): Likewise.
15512 (vect_prune_runtime_alias_test_list): Likewise.
15513 (vect_shift_permute_load_chain): Likewise.
15514 (vect_supportable_dr_alignment): Likewise.
15515 (dependence_distance_ge_vf): Take the vectorization factor as a
15516 poly_uint64 rather than an unsigned HOST_WIDE_INT.
15517 (vect_analyze_data_refs): Change min_vf from an int * to a
15519 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
15520 vfm1 as a poly_uint64 rather than an int. Make the same change
15521 for the returned bound_scalar.
15522 (vect_gen_vector_loop_niters): Handle polynomial vf.
15523 (vect_do_peeling): Likewise. Update call to
15524 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
15525 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
15527 * tree-vect-loop.c (vect_determine_vectorization_factor)
15528 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
15529 (vect_get_known_peeling_cost): Likewise.
15530 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
15531 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
15532 (vect_transform_loop): Likewise. Use the lowest possible VF when
15533 updating the upper bounds of the loop.
15534 (vect_min_worthwhile_factor): Make static. Return an unsigned int
15535 rather than an int.
15536 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
15537 polynomial unroll factors.
15538 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
15539 (vect_make_slp_decision): Likewise.
15540 (vect_supported_load_permutation_p): Likewise, and polynomial
15542 (vect_analyze_slp_cost): Handle polynomial vf.
15543 (vect_slp_analyze_node_operations): Likewise.
15544 (vect_slp_analyze_bb_1): Likewise.
15545 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
15546 than an unsigned HOST_WIDE_INT.
15547 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
15548 (vectorizable_load): Handle polynomial vf.
15549 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
15551 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
15553 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15554 Alan Hayward <alan.hayward@arm.com>
15555 David Sherwood <david.sherwood@arm.com>
15557 * match.pd: Handle bit operations involving three constants
15558 and try to fold one pair.
15560 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15562 * tree-vect-loop-manip.c: Include gimple-fold.h.
15563 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
15564 niters_maybe_zero parameters. Handle other cases besides a step of 1.
15565 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
15566 Add a path that uses a step of VF instead of 1, but disable it
15568 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
15569 and niters_no_overflow parameters. Update calls to
15570 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
15571 Create a new SSA name if the latter choses to use a ste other
15572 than zero, and return it via niters_vector_mult_vf_var.
15573 * tree-vect-loop.c (vect_transform_loop): Update calls to
15574 vect_do_peeling, vect_gen_vector_loop_niters and
15575 slpeel_make_loop_iterate_ntimes.
15576 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
15577 (vect_gen_vector_loop_niters): Update declarations after above changes.
15579 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
15581 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
15582 128-bit round to integer instructions.
15583 (ceil<mode>2): Likewise.
15584 (btrunc<mode>2): Likewise.
15585 (round<mode>2): Likewise.
15587 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15589 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
15590 unaligned VSX load/store on P8/P9.
15591 (expand_block_clear): Allow the use of unaligned VSX
15592 load/store on P8/P9.
15594 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15596 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
15598 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
15599 swap associated with both a load and a store.
15601 2018-01-02 Andrew Waterman <andrew@sifive.com>
15603 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
15604 * config/riscv/riscv.md (clear_cache): Use it.
15606 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
15608 * web.c: Remove out-of-date comment.
15610 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15612 * expr.c (fixup_args_size_notes): Check that any existing
15613 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
15614 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
15615 (emit_single_push_insn): ...here.
15617 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15619 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
15620 (const_vector_encoded_nelts): New function.
15621 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
15622 (const_vector_int_elt, const_vector_elt): Declare.
15623 * emit-rtl.c (const_vector_int_elt_1): New function.
15624 (const_vector_elt): Likewise.
15625 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
15626 of CONST_VECTOR_ELT.
15628 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15630 * expr.c: Include rtx-vector-builder.h.
15631 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
15632 directly on the tree encoding.
15633 (const_vector_from_tree): Likewise.
15634 * optabs.c: Include rtx-vector-builder.h.
15635 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
15636 sequence of "u" values.
15637 * vec-perm-indices.c: Include rtx-vector-builder.h.
15638 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
15639 directly on the vec_perm_indices encoding.
15641 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15643 * doc/rtl.texi (const_vector): Describe new encoding scheme.
15644 * Makefile.in (OBJS): Add rtx-vector-builder.o.
15645 * rtx-vector-builder.h: New file.
15646 * rtx-vector-builder.c: Likewise.
15647 * rtl.h (rtx_def::u2): Add a const_vector field.
15648 (CONST_VECTOR_NPATTERNS): New macro.
15649 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
15650 (CONST_VECTOR_DUPLICATE_P): Likewise.
15651 (CONST_VECTOR_STEPPED_P): Likewise.
15652 (CONST_VECTOR_ENCODED_ELT): Likewise.
15653 (const_vec_duplicate_p): Check for a duplicated vector encoding.
15654 (unwrap_const_vec_duplicate): Likewise.
15655 (const_vec_series_p): Check for a non-duplicated vector encoding.
15656 Say that the function only returns true for integer vectors.
15657 * emit-rtl.c: Include rtx-vector-builder.h.
15658 (gen_const_vec_duplicate_1): Delete.
15659 (gen_const_vector): Call gen_const_vec_duplicate instead of
15660 gen_const_vec_duplicate_1.
15661 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
15662 (gen_const_vec_duplicate): Use rtx_vector_builder.
15663 (gen_const_vec_series): Likewise.
15664 (gen_rtx_CONST_VECTOR): Likewise.
15665 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
15666 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
15667 Build a new vector rather than modifying a CONST_VECTOR in-place.
15668 (handle_special_swappables): Update call accordingly.
15669 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
15670 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
15671 Build a new vector rather than modifying a CONST_VECTOR in-place.
15672 (handle_special_swappables): Update call accordingly.
15674 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15676 * simplify-rtx.c (simplify_const_binary_operation): Use
15677 CONST_VECTOR_ELT instead of XVECEXP.
15679 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15681 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
15682 the selector elements to be different from the data elements
15683 if the selector is a VECTOR_CST.
15684 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
15685 ssizetype for the selector.
15687 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15689 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
15690 before testing each element individually.
15691 * tree-vect-generic.c (lower_vec_perm): Likewise.
15693 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15695 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
15696 * selftest-run-tests.c (selftest::run_tests): Call it.
15697 * vector-builder.h (vector_builder::operator ==): New function.
15698 (vector_builder::operator !=): Likewise.
15699 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
15700 (vec_perm_indices::all_from_input_p): New function.
15701 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
15702 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
15703 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
15704 instead of reading the VECTOR_CST directly. Detect whether both
15705 vector inputs are the same before constructing the vec_perm_indices,
15706 and update the number of inputs argument accordingly. Use the
15707 utility functions added above. Only construct sel2 if we need to.
15709 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15711 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
15712 the broadcast of the low byte.
15713 (expand_mult_highpart): Use an explicit encoding for the permutes.
15714 * optabs-query.c (can_mult_highpart_p): Likewise.
15715 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
15716 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15717 (vectorizable_bswap): Likewise.
15718 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
15719 explicit encoding for the power-of-2 permutes.
15720 (vect_permute_store_chain): Likewise.
15721 (vect_grouped_load_supported): Likewise.
15722 (vect_permute_load_chain): Likewise.
15724 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15726 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
15727 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
15728 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
15729 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
15730 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
15731 (vect_gen_perm_mask_any): Likewise.
15733 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15735 * int-vector-builder.h: New file.
15736 * vec-perm-indices.h: Include int-vector-builder.h.
15737 (vec_perm_indices): Redefine as an int_vector_builder.
15738 (auto_vec_perm_indices): Delete.
15739 (vec_perm_builder): Redefine as a stand-alone class.
15740 (vec_perm_indices::vec_perm_indices): New function.
15741 (vec_perm_indices::clamp): Likewise.
15742 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
15743 (vec_perm_indices::new_vector): New function.
15744 (vec_perm_indices::new_expanded_vector): Update for new
15745 vec_perm_indices class.
15746 (vec_perm_indices::rotate_inputs): New function.
15747 (vec_perm_indices::all_in_range_p): Operate directly on the
15748 encoded form, without computing elided elements.
15749 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
15750 encoding. Update for new vec_perm_indices class.
15751 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
15752 the given vec_perm_builder.
15753 (expand_vec_perm_var): Update vec_perm_builder constructor.
15754 (expand_mult_highpart): Use vec_perm_builder instead of
15755 auto_vec_perm_indices.
15756 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
15757 vec_perm_indices instead of auto_vec_perm_indices. Use a single
15758 or double series encoding as appropriate.
15759 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
15760 vec_perm_indices instead of auto_vec_perm_indices.
15761 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15762 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15763 (vect_permute_store_chain): Likewise.
15764 (vect_grouped_load_supported): Likewise.
15765 (vect_permute_load_chain): Likewise.
15766 (vect_shift_permute_load_chain): Likewise.
15767 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15768 (vect_transform_slp_perm_load): Likewise.
15769 (vect_schedule_slp_instance): Likewise.
15770 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15771 (vectorizable_mask_load_store): Likewise.
15772 (vectorizable_bswap): Likewise.
15773 (vectorizable_store): Likewise.
15774 (vectorizable_load): Likewise.
15775 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
15776 vec_perm_indices instead of auto_vec_perm_indices. Use
15777 tree_to_vec_perm_builder to read the vector from a tree.
15778 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
15779 vec_perm_builder instead of a vec_perm_indices.
15780 (have_whole_vector_shift): Use vec_perm_builder and
15781 vec_perm_indices instead of auto_vec_perm_indices. Leave the
15782 truncation to calc_vec_perm_mask_for_shift.
15783 (vect_create_epilog_for_reduction): Likewise.
15784 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
15785 from auto_vec_perm_indices to vec_perm_indices.
15786 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15787 instead of changing individual elements.
15788 (aarch64_vectorize_vec_perm_const): Use new_vector to install
15789 the vector in d.perm.
15790 * config/arm/arm.c (expand_vec_perm_d::perm): Change
15791 from auto_vec_perm_indices to vec_perm_indices.
15792 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15793 instead of changing individual elements.
15794 (arm_vectorize_vec_perm_const): Use new_vector to install
15795 the vector in d.perm.
15796 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
15797 Update vec_perm_builder constructor.
15798 (rs6000_expand_interleave): Likewise.
15799 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
15800 (rs6000_expand_interleave): Likewise.
15802 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15804 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
15805 to qimode could truncate the indices.
15806 * optabs.c (expand_vec_perm_var): Likewise.
15808 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15810 * Makefile.in (OBJS): Add vec-perm-indices.o.
15811 * vec-perm-indices.h: New file.
15812 * vec-perm-indices.c: Likewise.
15813 * target.h (vec_perm_indices): Replace with a forward class
15815 (auto_vec_perm_indices): Move to vec-perm-indices.h.
15816 * optabs.h: Include vec-perm-indices.h.
15817 (expand_vec_perm): Delete.
15818 (selector_fits_mode_p, expand_vec_perm_var): Declare.
15819 (expand_vec_perm_const): Declare.
15820 * target.def (vec_perm_const_ok): Replace with...
15821 (vec_perm_const): ...this new hook.
15822 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
15823 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
15824 * doc/tm.texi: Regenerate.
15825 * optabs.def (vec_perm_const): Delete.
15826 * doc/md.texi (vec_perm_const): Likewise.
15827 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
15828 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
15829 expand_vec_perm for constant permutation vectors. Assert that
15830 the mode of variable permutation vectors is the integer equivalent
15831 of the mode that is being permuted.
15832 * optabs-query.h (selector_fits_mode_p): Declare.
15833 * optabs-query.c: Include vec-perm-indices.h.
15834 (selector_fits_mode_p): New function.
15835 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
15836 is defined, instead of checking whether the vec_perm_const_optab
15837 exists. Use targetm.vectorize.vec_perm_const instead of
15838 targetm.vectorize.vec_perm_const_ok. Check whether the indices
15839 fit in the vector mode before using a variable permute.
15840 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
15841 vec_perm_indices instead of an rtx.
15842 (expand_vec_perm): Replace with...
15843 (expand_vec_perm_const): ...this new function. Take the selector
15844 as a vec_perm_indices rather than an rtx. Also take the mode of
15845 the selector. Update call to shift_amt_for_vec_perm_mask.
15846 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
15847 Use vec_perm_indices::new_expanded_vector to expand the original
15848 selector into bytes. Check whether the indices fit in the vector
15849 mode before using a variable permute.
15850 (expand_vec_perm_var): Make global.
15851 (expand_mult_highpart): Use expand_vec_perm_const.
15852 * fold-const.c: Includes vec-perm-indices.h.
15853 * tree-ssa-forwprop.c: Likewise.
15854 * tree-vect-data-refs.c: Likewise.
15855 * tree-vect-generic.c: Likewise.
15856 * tree-vect-loop.c: Likewise.
15857 * tree-vect-slp.c: Likewise.
15858 * tree-vect-stmts.c: Likewise.
15859 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
15861 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
15862 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
15863 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
15864 (aarch64_vectorize_vec_perm_const): ...this new function.
15865 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15866 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15867 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
15868 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
15869 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15870 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15871 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
15873 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
15874 check for NEON modes.
15875 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
15876 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
15877 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
15878 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
15880 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
15881 the old VEC_PERM_CONST conditions.
15882 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
15883 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
15884 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
15885 (ia64_vectorize_vec_perm_const_ok): Merge into...
15886 (ia64_vectorize_vec_perm_const): ...this new function.
15887 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
15888 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
15889 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
15890 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
15891 * config/mips/mips.c (mips_expand_vec_perm_const)
15892 (mips_vectorize_vec_perm_const_ok): Merge into...
15893 (mips_vectorize_vec_perm_const): ...this new function.
15894 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
15895 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
15896 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
15897 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
15898 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
15899 (rs6000_expand_vec_perm_const): Delete.
15900 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
15902 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15903 (altivec_expand_vec_perm_const_le): Take each operand individually.
15904 Operate on constant selectors rather than rtxes.
15905 (altivec_expand_vec_perm_const): Likewise. Update call to
15906 altivec_expand_vec_perm_const_le.
15907 (rs6000_expand_vec_perm_const): Delete.
15908 (rs6000_vectorize_vec_perm_const_ok): Delete.
15909 (rs6000_vectorize_vec_perm_const): New function.
15910 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15911 an element count and rtx array.
15912 (rs6000_expand_extract_even): Update call accordingly.
15913 (rs6000_expand_interleave): Likewise.
15914 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
15915 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
15916 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
15917 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
15918 (rs6000_expand_vec_perm_const): Delete.
15919 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15920 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15921 (altivec_expand_vec_perm_const_le): Take each operand individually.
15922 Operate on constant selectors rather than rtxes.
15923 (altivec_expand_vec_perm_const): Likewise. Update call to
15924 altivec_expand_vec_perm_const_le.
15925 (rs6000_expand_vec_perm_const): Delete.
15926 (rs6000_vectorize_vec_perm_const_ok): Delete.
15927 (rs6000_vectorize_vec_perm_const): New function. Remove stray
15928 reference to the SPE evmerge intructions.
15929 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15930 an element count and rtx array.
15931 (rs6000_expand_extract_even): Update call accordingly.
15932 (rs6000_expand_interleave): Likewise.
15933 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
15934 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
15936 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15938 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15940 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
15941 vector mode and that that mode matches the mode of the data
15943 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
15944 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
15945 directly using expand_vec_perm_1 when forcing selectors into
15947 (expand_vec_perm_var): New function, split out from expand_vec_perm.
15949 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15951 * optabs-query.h (can_vec_perm_p): Delete.
15952 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
15953 * optabs-query.c (can_vec_perm_p): Split into...
15954 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
15955 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
15956 particular selector is valid.
15957 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15958 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15959 (vect_grouped_load_supported): Likewise.
15960 (vect_shift_permute_load_chain): Likewise.
15961 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15962 (vect_transform_slp_perm_load): Likewise.
15963 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15964 (vectorizable_bswap): Likewise.
15965 (vect_gen_perm_mask_checked): Likewise.
15966 * fold-const.c (fold_ternary_loc): Likewise. Don't take
15967 implementations of variable permutation vectors into account
15968 when deciding which selector to use.
15969 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
15970 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
15971 with a false third argument.
15972 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
15973 to test whether the constant selector is valid and can_vec_perm_var_p
15974 to test whether a variable selector is valid.
15976 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15978 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
15979 * optabs-query.c (can_vec_perm_p): Likewise.
15980 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
15981 instead of vec_perm_indices.
15982 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
15983 (vect_gen_perm_mask_checked): Likewise,
15984 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
15985 (vect_gen_perm_mask_checked): Likewise,
15987 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15989 * optabs-query.h (qimode_for_vec_perm): Declare.
15990 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
15991 (qimode_for_vec_perm): ...this new function.
15992 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
15994 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15996 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
15997 does not have a conditional at the top.
15999 2018-01-02 Richard Biener <rguenther@suse.de>
16001 * ipa-inline.c (big_speedup_p): Fix expression.
16003 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
16006 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
16009 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
16013 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
16014 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
16015 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
16016 cond_taken_branch_cost 3->4.
16018 2018-01-01 Jakub Jelinek <jakub@redhat.com>
16020 PR tree-optimization/83581
16021 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
16022 TODO_cleanup_cfg if any changes have been made.
16024 PR middle-end/83608
16025 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
16026 convert_modes if target mode has the right side, but different mode
16029 PR middle-end/83609
16030 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
16031 last argument when extracting from CONCAT. If either from_real or
16032 from_imag is NULL, use expansion through memory. If result is not
16033 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
16034 the parts directly to inner mode, if even that fails, use expansion
16037 PR middle-end/83623
16038 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
16039 check for bswap in mode rather than HImode and use that in expand_unop
16042 Copyright (C) 2018 Free Software Foundation, Inc.
16044 Copying and distribution of this file, with or without modification,
16045 are permitted in any medium without royalty provided the copyright
16046 notice and this notice are preserved.