1 2018-02-27 Martin Liska <mliska@suse.cz>
3 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
4 'ggc' suffixes. Change first column width.
5 (timer::print): Fix formatting of the column.
7 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
9 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
10 preserve inline entry blocks for the sake of debug inline
11 entry point markers alone.
12 (remove_unused_locals): Suggest in comments a better place to
13 force the preservation of inline entry blocks that are
14 otherwise unused, but do not preserve them.
16 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
18 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
20 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
23 * config/i386/constraints.md (Bs): Replace
24 ix86_indirect_branch_register with
25 TARGET_INDIRECT_BRANCH_REGISTER.
27 * config/i386/i386.md (indirect_jump): Likewise.
28 (tablejump): Likewise.
29 (*sibcall_memory): Likewise.
30 (*sibcall_value_memory): Likewise.
31 Peepholes of indirect call and jump via memory: Likewise.
32 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
33 (*sibcall_value_GOT_32): Likewise.
34 * config/i386/i386.opt: Likewise.
35 * config/i386/predicates.md (indirect_branch_operand): Likewise.
36 (GOT_memory_operand): Likewise.
37 (call_insn_operand): Likewise.
38 (sibcall_insn_operand): Likewise.
39 (GOT32_symbol_operand): Likewise.
40 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
42 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
44 PR rtl-optimization/83496
45 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
46 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
47 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
48 redundant insn, if any.
49 (relax_delay_slots): Likewise.
50 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
52 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
54 PR tree-optimization/83965
55 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
56 that grouped statements are part of a reduction chain. Return
57 true if the statement is not marked as a reduction itself but
59 (vect_recog_dot_prod_pattern): Don't check whether the statement
60 is part of a group here.
61 (vect_recog_sad_pattern): Likewise.
62 (vect_recog_widen_sum_pattern): Likewise.
64 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
67 * final.c (rest_of_clean_state): Also look for calls inside sequences.
69 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
72 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
74 (ix86_output_indirect_function_return): New prototype.
75 (ix86_split_simple_return_pop_internal): Likewise.
76 * config/i386/i386.c (indirect_return_via_cx): New.
77 (indirect_return_via_cx_bnd): Likewise.
78 (indirect_thunk_name): Handle return va CX_REG.
79 (output_indirect_thunk_function): Create alias for
80 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
81 (ix86_output_indirect_jmp): Remove the bool argument.
82 (ix86_output_indirect_function_return): New function.
83 (ix86_split_simple_return_pop_internal): Likewise.
84 * config/i386/i386.md (*indirect_jump): Don't pass false
85 to ix86_output_indirect_jmp.
86 (*tablejump_1): Likewise.
87 (simple_return_pop_internal): Change it to define_insn_and_split.
88 Call ix86_split_simple_return_pop_internal to split it for
90 (simple_return_indirect_internal): Call
91 ix86_output_indirect_function_return instead of
92 ix86_output_indirect_jmp.
94 2018-02-26 Jakub Jelinek <jakub@redhat.com>
97 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
98 memset and value initialization afterwards.
100 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
102 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
104 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
107 * common/config/aarch64/aarch64-common.c
108 (aarch_option_optimization_table[]): Switch
109 off fomit-frame-pointer
111 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
112 Chung-Ju Wu <jasonwucj@gmail.com>
114 * config/nds32/nds32-multiple.md (load_multiple): Disallow
116 (store_multiple): Ditto.
118 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
120 * config.gcc: Add --with-cpu support for nds32 target.
121 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
122 * config/nds32/nds32.opt: Add -mcpu= option.
124 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
126 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
127 isel=yes): Warn for these deprecated options.
129 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
131 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
132 ISA_2_5_MASKS_EMBEDDED.
134 2018-02-23 Jakub Jelinek <jakub@redhat.com>
136 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
137 p->max as pointers rather than using iterative_hash_expr.
139 2018-02-23 Carl Love <cel@us.ibm.com>
141 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
142 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
144 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
145 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
146 P8V_BUILTIN_VEC_VUNSIGNED2.
148 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
151 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
152 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
154 * lra-constraints.c (curr_insn_transform): Set up
155 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
157 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
158 * lra-eliminations.c (spill_pseudos): Ditto.
159 (process_insn_for_elimination): Ditto.
160 * lra-lives.c (reg_early_clobber_p): Use the new macros.
161 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
162 LRA_NON_CLOBBERED_ALT.
164 2018-02-22 Martin Sebor <msebor@redhat.com>
166 PR tree-optimization/84480
167 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
168 to maybe_diag_stxncpy_trunc. Call it.
169 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
170 from gimple_fold_builtin_strcpy. Print inlining stack.
171 (handle_builtin_stxncpy): Print inlining stack.
172 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
174 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
177 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
178 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
179 and -fcheck-pointer-bounds are used together.
180 (indirect_thunk_prefix): New enum.
181 (indirect_thunk_need_prefix): New function.
182 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
183 "_nt" instead of "_bnd" for NOTRACK prefix.
184 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
185 (output_indirect_thunk_function): Likewise.
187 (ix86_code_end): Update output_indirect_thunk_function calls.
188 (ix86_output_indirect_branch_via_reg): Replace
189 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
190 (ix86_output_indirect_branch_via_push): Likewise.
191 (ix86_output_function_return): Likewise.
192 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
193 incompatible with -fcf-protection=branch and
194 -fcheck-pointer-bounds.
196 2018-02-22 Steve Ellcey <sellcey@cavium.com>
199 * config/aarch64/aarch64.c (aarch64_print_address_internal):
200 Change gcc_assert call to output_operand_lossage.
202 2018-02-22 Steve Ellcey <sellcey@cavium.com>
204 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
206 2018-02-22 DJ Delorie <dj@redhat.com>
207 Sebastian Perta <sebastian.perta@renesas.com>
208 Oleg Endo <olegendo@gcc.gnu.org>
210 * config/rx/rx.c (rx_rtx_costs): New function.
211 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
213 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
215 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
217 2018-02-22 Martin Liska <mliska@suse.cz>
220 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
221 Add "native" as a possible value.
223 2018-02-22 Martin Liska <mliska@suse.cz>
226 * config/i386/i386.c (ix86_option_override_internal):
227 Add "native" as a possible value for -march and -mtune.
229 2018-02-22 Jakub Jelinek <jakub@redhat.com>
232 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
233 to all type variants.
235 PR tree-optimization/84503
236 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
237 width as info->bitpos + info->bitsize - start.
238 (merged_store_group::merge_overlapping): Simplify width computation.
239 (check_no_overlap): New function.
240 (imm_store_chain_info::try_coalesce_bswap): Compute expected
241 start + width and last_order of the group, fail if check_no_overlap
243 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
244 to group if check_no_overlap fails.
246 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
248 * config/rs6000/altivec.md: Delete contraint arguments to
249 define_expand, define_split, and define_peephole2, and in
250 define_insn_and_split if always unused.
251 * config/rs6000/darwin.md: Ditto.
252 * config/rs6000/dfp.md: Ditto.
253 * config/rs6000/rs6000.md: Ditto.
254 * config/rs6000/sync.md: Ditto.
255 * config/rs6000/vector.md: Ditto.
256 * config/rs6000/vsx.md: Ditto.
258 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
260 * config/rs6000/altivec.md: Write output control strings as braced
261 blocks instead of double-quoted strings.
262 * config/rs6000/darwin.md: Ditto.
263 * config/rs6000/rs6000.md: Ditto.
264 * config/rs6000/vector.md: Ditto.
265 * config/rs6000/vsx.md: Ditto.
267 2018-02-21 Jason Merrill <jason@redhat.com>
269 PR c++/84314 - ICE with templates and fastcall attribute.
270 * attribs.c (build_type_attribute_qual_variant): Remove assert.
272 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
274 * ipa-cp.c (determine_versionability): Fix comment typos.
276 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
279 * ipa-cp.c (determine_versionability): Do not version functions caling
282 2018-02-21 Martin Liska <mliska@suse.cz>
285 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
286 Add "native" as a possible value.
287 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
288 the macro when native cpu detection is available.
290 2018-02-21 Martin Liska <mliska@suse.cz>
293 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
294 Add "native" as a possible value.
295 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
296 when native cpu detection is available.
298 2018-02-21 Jakub Jelinek <jakub@redhat.com>
299 Martin Sebor <msebor@redhat.com>
301 PR tree-optimization/84478
302 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
304 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
305 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
306 support which is conservatively correct, for 2 only stay conservative
307 for maxlen. Formatting and comment capitalization fixes. Add STRICT
308 argument to the 2 argument get_range_strlen, adjust 6 arg
309 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
311 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
312 (gimple_fold_builtin_strlen): Pass true as last argument to
315 2018-02-20 Martin Sebor <msebor@redhat.com>
318 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
319 (builtin_memref::set_base_and_offset): Same. Handle inner references.
320 (builtin_memref::builtin_memref): Factor out parts into
321 set_base_and_offset and call it.
323 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
326 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
327 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
328 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
329 search at the associated MODE_INT.
331 2018-02-20 Jeff Law <law@redhat.com>
334 PR tree-optimization/81592
336 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
337 for range data rather than using global data.
338 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
339 range data rather than using global data.
340 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
341 pass it to children as needed.
342 (struct directive::fmtresult): Similarly.
343 (struct directive::set_width): Similarly.
344 (struct directive::set_precision): Similarly.
345 (format_integer, format_directive, parse_directive): Similarly.
346 (format_none): Accept unnamed vr_values parameter.
347 (format_percent, format_floating, format_character): Similarly.
348 (format_string, format_plain): Similarly.
349 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
350 the EVRP range analyzer for range data rather than using global data.
351 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
352 gimple-ssa-evrp-analyze.h
353 (class sprintf_dom_walker): Add after_dom_children member function.
354 Add evrp_range_analyzer member.
355 (sprintf_dom_walker::before_dom_children): Call into the EVRP
356 range analyzer as needed.
357 (sprintf_dom_walker::after_dom_children): New member function.
358 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
360 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
361 (evrp_range_analyzer::pop_to_marker): Likewise.
363 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
365 PR tree-optimization/84419
366 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
367 with the required type if its current type is compatible but
370 2018-02-20 Jakub Jelinek <jakub@redhat.com>
373 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
376 2018-02-20 Martin Liska <mliska@suse.cz>
379 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
380 possible values if we don't have a hint.
382 2018-02-20 Martin Liska <mliska@suse.cz>
386 * final.c (shorten_branches): Build align_tab array with one
388 * opts.c (finish_options): Add alignment option limit check.
389 (MAX_CODE_ALIGN): Likewise.
390 (MAX_CODE_ALIGN_VALUE): Likewise.
391 * doc/invoke.texi: Document maximum allowed option value for
392 all -falign-* options.
394 2018-02-19 Jakub Jelinek <jakub@redhat.com>
397 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
398 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
399 * var-tracking.c (emit_note_insn_var_location): Remove all references
400 to NOTE_INSN_CALL_ARG_LOCATION.
401 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
402 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
403 Use copy_rtx_if_shared.
404 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
405 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
406 (dwarf2out_var_location): Remove handling of
407 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
409 * final.c (final_scan_insn): Remove all references to
410 NOTE_INSN_CALL_ARG_LOCATION.
411 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
412 before dumping final insns.
413 * except.c (emit_note_eh_region_end): Remove all references to
414 NOTE_INSN_CALL_ARG_LOCATION.
415 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
416 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
417 * config/arc/arc.c (hwloop_optimize): Likewise.
418 * config/arm/arm.c (create_fix_barrier): Likewise.
419 * config/s390/s390.c (s390_chunkify_start): Likewise.
420 * config/sh/sh.c (find_barrier): Likewise.
421 * config/i386/i386.c (rest_of_insert_endbranch,
422 ix86_seh_fixup_eh_fallthru): Likewise.
423 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
424 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
425 * config/frv/frv.c (frv_function_prologue): Likewise.
426 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
428 (note_outside_basic_block_p): Remove all references to
429 NOTE_INSN_CALL_ARG_LOCATION.
430 * gengtype.c (adjust_field_rtx_def): Likewise.
431 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
433 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
434 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
437 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
440 PR tree-optimization/84452
441 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
442 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
445 2018-02-19 Martin Liska <mliska@suse.cz>
448 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
450 2018-02-19 Martin Liska <mliska@suse.cz>
451 Richard Sandiford <richard.sandiford@linaro.org>
453 PR tree-optimization/82491
454 * gimple-fold.c (get_base_constructor): Make earlier bail out
457 2018-02-19 Carl Love <cel@us.ibm.com>
459 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
460 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
462 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
465 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
467 * config/rl78/rl78.md (movdf): New define expand.
469 2018-02-19 Martin Liska <mliska@suse.cz>
472 * doc/invoke.texi: Fix typo.
473 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
475 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
477 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
478 handle rs6000_single_float and rs6000_double_float specially for
481 2018-02-16 Jeff Law <law@redhat.com>
483 * config/rx/rx.c (add_pop_cfi_notes): New function.;
486 2018-02-16 Jakub Jelinek <jakub@redhat.com>
489 * ipa-inline.c (inline_small_functions): Fix a typo.
491 2018-02-16 Nathan Sidwell <nathan@acm.org>
493 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
495 2018-02-16 Carl Love <cel@us.ibm.com>
497 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
498 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
499 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
500 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
501 expansion to P8V_BUILTIN_VEC_FLOAT2.
503 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
505 PR rtl-optimization/70023
506 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
507 src_regno into account.
509 2018-02-16 Carl Love <cel@us.ibm.com>
511 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
512 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
513 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
514 * config/rs6000/rs6000.c: Remove case statements for
515 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
516 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
517 and P9V_BUILTIN_VEC_VINSERT4B.
518 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
519 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
520 * config/rs6000/vsx.md:
521 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
524 2018-02-16 Carl Love <cel@us.ibm.com>
526 * config/rs6000/altivec.h: Add builtin names vec_extract4b
528 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
530 * config/rs6000/rs6000-c.c: Add the definitions for
531 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
532 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
533 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
534 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
535 definition for insert4b and define insn *insert3b_internal.
536 * doc/extend.texi: Add documentation for vec_extract4b.
538 2018-02-16 Nathan Sidwell <nathan@acm.org>
540 * doc/extend.texi (Backwards Compatibility): Mention friend
541 injection. Note for-scope is deprecated.
542 * doc/invoke.texi (-ffriend-injection): Deprecate.
544 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
546 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
547 that moved to I2, also allow destinations that are a paradoxical
548 subreg (instead of a normal reg).
550 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
553 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
556 2018-02-16 Richard Biener <rguenther@suse.de>
558 PR tree-optimization/84037
559 PR tree-optimization/84016
561 * config/i386/i386.c (ix86_builtin_vectorization_cost):
562 Adjust vec_construct for the fact we need additional higher latency
563 128bit inserts for AVX256 and AVX512 vector builds.
564 (ix86_add_stmt_cost): Scale vector construction cost for
567 2018-02-16 Richard Biener <rguenther@suse.de>
569 PR tree-optimization/84417
570 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
571 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
572 (non_rewritable_lvalue_p): Likewise, use poly-ints.
574 2018-02-16 Martin Liska <mliska@suse.cz>
577 * internal-fn.def (ASAN_CHECK): Set proper flags.
578 (ASAN_MARK): Likewise.
580 2018-02-16 Julia Koval <julia.koval@intel.com>
582 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
585 2018-02-16 Jakub Jelinek <jakub@redhat.com>
588 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
589 Use ++iter rather than iter++ for std::list iterators.
590 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
591 defer deleting them until all nodes in the forest are processed. Do
592 free even leaf nodes. Change to_process into auto_vec.
595 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
596 * vec.h (vec_default_construct): Use memset instead of placement new
597 if BROKEN_VALUE_INITIALIZATION is defined.
598 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
599 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
602 PR rtl-optimization/83723
603 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
604 * lra.c (lra_substitute_pseudo): Likewise. If true, use
605 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
607 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
609 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
611 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
613 PR rtl-optimization/81443
614 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
615 from inner REGs to paradoxical SUBREGs.
617 2018-02-16 Richard Biener <rguenther@suse.de>
619 PR tree-optimization/84399
620 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
621 For operands we can analyze at their definition make sure we can
622 analyze them at each use as well.
624 2018-02-16 Richard Biener <rguenther@suse.de>
626 PR tree-optimization/84190
627 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
628 volatile accesses if the decl isn't volatile.
630 2018-02-15 Jason Merrill <jason@redhat.com>
632 PR c++/84314 - ICE with templates and fastcall attribute.
633 * attribs.c (build_type_attribute_qual_variant): Don't clobber
634 TYPE_CANONICAL on an existing type.
636 2018-02-15 Jakub Jelinek <jakub@redhat.com>
638 PR tree-optimization/84383
639 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
640 dstoff nor call operand_equal_p if dstbase is NULL.
642 PR tree-optimization/84334
643 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
644 also a CONSTANT_CLASS_P, punt.
646 2018-02-14 Jim Wilson <jimw@sifive.com>
648 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
649 first SMALL_OPERAND check. New local min_second_step. Move assert
650 to where locals are set. Add TARGET_RVC support.
651 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
653 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
655 * doc/invoke.texi: Correct -Wformat-overflow code sample.
657 2018-02-14 Martin Sebor <msebor@redhat.com>
659 PR tree-optimization/83698
660 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
661 arrays constrain the offset range to their bounds.
662 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
663 (builtin_access::overlap): Avoid setting the size of overlap if it's
665 (maybe_diag_overlap): Also consider arrays when deciding what values
666 of offsets to include in diagnostics.
668 2018-02-14 Martin Sebor <msebor@redhat.com>
671 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
672 that correspond to the kind of a declaration.
674 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
677 * config/pa/pa.md: Load address of PIC label using the linkage table
678 if the label is nonlocal.
680 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
682 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
683 warning message if user requests -maltivec=be.
684 * doc/invoke.texi: Document deprecation of -maltivec=be.
686 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
689 * config/rs6000/rs6000-c.c: Update definitions for
690 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
691 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
693 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
696 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
697 add _get_ssp intrinsics. Remove argument from
698 __builtin_ia32_rdssp[d|q].
699 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
700 * config/i386/i386-builtin.def: Remove argument from
701 __builtin_ia32_rdssp[d|q].
702 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
703 ix86_expand_special_args_builtin for _rdssp[d|q].
704 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
705 Clear register before usage.
706 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
707 Add documentation for new _get_ssp and _inc_ssp intrinsics.
709 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
711 PR tree-optimization/84357
712 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
713 operand 1 of an ARRAY_REF too.
715 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
718 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
719 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
721 (set_of_reg): New struct.
722 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
723 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
724 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
726 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
727 Split into bitclr, bitset, bitinvert patterns if appropriate.
728 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
729 use rx_fuse_in_memory_bitop.
730 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
731 to named insn, correct maximum insn length.
733 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
736 * machmode.def: Define a complex mode for PARTIAL_INT.
737 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
739 * doc/rtl.texi: Document CSPImode.
740 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
742 (msp430_hard_regno_nregs_with_padding): Likewise.
744 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
747 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
749 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
751 PR rtl-optimization/84169
752 * combine.c (try_combine): New variable split_i2i3. Set it to true if
753 we generated a parallel as new i3 and we split that to new i2 and i3
754 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
755 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
756 those to i2, not i1. Partially rewrite this scan code.
758 2018-02-13 Jakub Jelinek <jakub@redhat.com>
761 * stor-layout.c (place_field): For variable length fields, adjust
762 offset_align afterwards not just based on the field's alignment,
763 but also on the size.
766 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
767 of exps and logs in the use_exp2 case.
769 2018-02-13 Jeff Law <law@redhat.com>
771 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
774 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
777 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
781 * common.opt (gas-loc-support, gas-locview-support): New.
782 (ginline-points, ginternal-reset-location-views): New.
783 * doc/invoke.texi: Document them. Use @itemx where intended.
784 (gvariable-location-views): Adjust.
785 * target.def (reset_location_view): New.
786 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
787 (TARGET_RESET_LOCATION_VIEW): New.
788 * doc/tm.texi: Rebuilt.
789 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
790 (dwarf2out_default_as_locview_support): New.
791 (output_asm_line_debug_info): Use option variables.
792 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
793 (output_loc_list): Likewise.
794 (add_high_low_attributes): Check option variables.
795 Don't output entry view attribute in strict mode.
796 (gen_inlined_subroutine_die): Check option variables.
797 (dwarf2out_inline_entry): Likewise.
798 (init_sections_and_labels): Likewise.
799 (dwarf2out_early_finish): Likewise.
800 (maybe_reset_location_view): New, from...
801 (dwarf2out_var_location): ... here. Call it.
802 * debug.h (dwarf2out_default_as_loc_support): Declare.
803 (dwarf2out_default_as_locview_support): Declare.
804 * hooks.c (hook_int_rtx_insn_0): New.
805 * hooks.h (hook_int_rtx_insn_0): Declare.
806 * toplev.c (process_options): Take -gas-loc-support and
807 -gas-locview-support from dwarf2out. Enable
808 -gvariable-location-views by default only with locview
809 assembler support. Enable -ginternal-reset-location-views by
810 default only if the target defines the corresponding hook.
811 Enable -ginline-points by default if location views are
812 enabled; force it disabled if statement frontiers are
814 * tree-inline.c (expand_call_inline): Check option variables.
815 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
817 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
819 PR tree-optimization/84321
820 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
821 handling. Also check whether the anti-range contains any values
822 that satisfy the mask; switch to a VR_RANGE if not.
824 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
827 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
829 2018-02-13 Martin Jambor <mjambor@suse.cz>
832 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
833 of call statements, also set location of a load to a temporary.
835 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
837 * config/rl78/rl78.c (add_vector_labels): New function.
838 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
839 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
840 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
841 which checks that no arguments are passed.
842 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
843 * doc/extend.texi: Documentation for the new attribute.
845 2018-02-13 Andreas Schwab <schwab@suse.de>
847 * config/riscv/linux.h (CPP_SPEC): Define.
849 2018-02-13 Jakub Jelinek <jakub@redhat.com>
852 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
853 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
854 OPTION_MASK_ISA_AES as first argument to def_builtin_const
855 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
856 instead of OPTION_MASK_ISA_PCLMUL as first argument to
857 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
858 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
859 temporarily for AES and PCLMUL builtins.
861 PR tree-optimization/84339
862 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
863 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
867 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
868 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
869 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
871 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
873 * omp-simd-clone.h: New file.
874 * omp-simd-clone.c: Include omp-simd-clone.h.
875 (expand_simd_clones): No longer static.
876 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
877 cgraph.h and omp-simd-clone.h.
878 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
879 (vect_recog_widen_shift_pattern): Formatting fix.
880 (vect_pattern_recog_1): Don't check optab for calls.
883 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
884 operands[2] into a REG before using gen_lowpart on it.
886 2018-02-12 Jeff Law <law@redhat.com>
889 * config/sh/sh.c (find_barrier): Consider a sibling call
892 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
893 successfully back substituting a reg.
895 2018-02-12 Richard Biener <rguenther@suse.de>
897 PR tree-optimization/84037
898 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
899 parameter, move visited init to caller.
900 (vect_slp_analyze_operations): Separate cost from validity
901 check, initialize visited once for all instances.
902 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
904 * tree-vect-stmts.c (vect_model_simple_cost): Make early
906 (vect_model_promotion_demotion_cost): Likewise.
907 (vectorizable_bswap): Guard cost modeling with !slp_node
908 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
910 (vectorizable_call): Likewise.
911 (vectorizable_conversion): Likewise.
912 (vectorizable_assignment): Likewise.
913 (vectorizable_shift): Likewise.
914 (vectorizable_operation): Likewise.
915 (vectorizable_store): Likewise.
916 (vectorizable_load): Likewise.
917 (vectorizable_condition): Likewise.
918 (vectorizable_comparison): Likewise.
920 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
923 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
924 (ASAN_MARK): Fix fnspec to account for return value, change pointer
925 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
927 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
930 * params.def (inline-min-speedup): Increase from 8 to 15.
931 (max-inline-insns-auto): Decrease from 40 to 30.
932 * ipa-split.c (consider_split): Add some buffer for function to
933 be considered inlining candidate.
934 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
937 2018-02-12 Richard Biener <rguenther@suse.de>
939 PR tree-optimization/84037
940 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
941 matched stmts if we cannot swap the non-matched ones.
943 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
945 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
946 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
947 _mm_maskz_scalef_round_ss): New intrinsics.
948 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
949 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
950 __builtin_ia32_scalefss_round): Remove.
951 (__builtin_ia32_scalefsd_mask_round,
952 __builtin_ia32_scalefss_mask_round): New intrinsics.
953 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
954 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
955 ((match_operand:VF_128 2 "<round_nimm_predicate>"
956 "<round_constraint>")): Changed to ...
957 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
958 "<round_scalar_constraint>")): ... this.
959 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
960 %0, %1, %2<round_op3>}"): Changed to ...
961 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
962 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
963 %2<round_scalar_mask_op3>}"): ... this.
964 * config/i386/subst.md (round_scalar_nimm_predicate): New.
966 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
968 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
969 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
970 (_mm_maskz_sqrt_round_ss): New intrinsics.
971 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
972 (__builtin_ia32_sqrtsd_mask_round)
973 (__builtin_ia32_sqrtss_mask_round): New builtins.
974 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
975 (__builtin_ia32_sqrtss_round): Remove.
976 (__builtin_ia32_sqrtsd_mask_round)
977 (__builtin_ia32_sqrtss_mask_round): New builtins.
978 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
979 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
980 ((match_operand:VF_128 1 "vector_operand"
981 "xBm,<round_constraint>")): Changed to ...
982 ((match_operand:VF_128 1 "vector_operand"
983 "xBm,<round_scalar_constraint>")): ... this.
984 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
985 %0, %2, %<iptr>1<round_op3>}): Changed to ...
986 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
987 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
988 %<iptr>1<round_scalar_mask_op3>}): ... this.
989 ((set_attr "prefix" "<round_prefix>")): Changed to ...
990 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
992 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
995 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
996 Cast vec_cmpeq result to correct type.
997 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
998 Cast vec_cmpgt result to correct type.
1000 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
1002 * final.c (final_scan_insn_1): Renamed from...
1003 (final_scan_insn): ... this. New wrapper, to recover
1004 seen from the outermost call in recursive ones.
1005 * config/sparc/sparc.c (output_return): Drop seen from call.
1006 (output_sibcall): Likewise.
1007 * config/visium/visium.c (output_branch): Likewise.
1009 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
1011 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
1014 2018-02-10 Alan Modra <amodra@gmail.com>
1017 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
1018 Specify LR as an input.
1020 2018-02-10 Jakub Jelinek <jakub@redhat.com>
1023 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
1024 remove_member_access_dummy_vars): New functions.
1025 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
1026 lower_omp_1, execute_lower_omp): Use them.
1028 PR rtl-optimization/84308
1029 * shrink-wrap.c (spread_components): Release todo vector.
1031 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
1033 PR rtl-optimization/57193
1034 * ira-color.c (struct allocno_color_data): Add member
1035 conflict_allocno_hard_prefs.
1036 (update_conflict_allocno_hard_prefs): New.
1037 (bucket_allocno_compare_func): Add a preference based on
1038 conflict_allocno_hard_prefs.
1039 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
1040 (color_allocnos): Remove a dead code. Initiate
1041 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
1043 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1046 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
1047 constraint from =wa to wa. Avoid a subreg on the output operand,
1048 instead use a pseudo and subreg it in a move.
1049 (p9_xxbrd_<mode>): Changed to ...
1050 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
1051 (p9_xxbrd_v2df): New expander.
1052 (p9_xxbrw_<mode>): Changed to ...
1053 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
1054 (p9_xxbrw_v4sf): New expander.
1056 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1058 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
1060 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
1063 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
1064 multiply in 32-bit mode.
1065 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
1066 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
1069 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1071 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
1072 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
1073 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
1074 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
1076 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
1079 * dwarf2out.c (is_trivial_indirect_ref): New function.
1080 (dwarf2out_late_global_decl): Do not generate a location
1081 attribute for variables that have a non-trivial DECL_VALUE_EXPR
1082 and that are not defined in the current unit.
1084 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1086 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
1087 instead of a libcall for UNORDERED.
1089 2018-02-09 Tamar Christina <tamar.christina@arm.com>
1092 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
1093 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
1095 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1098 * config/s390/s390.c (s390_set_current_function): Invoke
1099 s390_indirect_branch_settings also if fndecl didn't change.
1101 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1103 * config/rs6000/rs6000.md (blockage): Set length to zero.
1105 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1107 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
1109 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1112 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
1113 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
1117 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
1118 PARALLEL incoming that failed vt_get_decl_and_offset check.
1121 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
1122 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
1124 (get_variable_section): For decls in named .bss* sections pass true as
1125 second argument to bss_initializer_p.
1127 2018-02-09 Marek Polacek <polacek@redhat.com>
1128 Jakub Jelinek <jakub@redhat.com>
1131 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
1132 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
1133 Sync some changes from cxx_fold_indirect_ref.
1135 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1137 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
1139 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
1140 (BLOCK_INLINE_ENTRY_LABEL): New.
1141 (dwarf2out_var_location): Disregard inline entry markers.
1142 (inline_entry_data): New struct.
1143 (inline_entry_data_hasher): New hashtable type.
1144 (inline_entry_data_hasher::hash): New.
1145 (inline_entry_data_hasher::equal): New.
1146 (inline_entry_data_table): New variable.
1147 (add_high_low_attributes): Add DW_AT_entry_pc and
1148 DW_AT_GNU_entry_view attributes if a pending entry is found
1149 in inline_entry_data_table. Add old entry_pc attribute only
1150 if debug nonbinding markers are disabled.
1151 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
1152 markers are enabled.
1153 (block_within_block_p, dwarf2out_inline_entry): New.
1154 (dwarf2out_finish): Check that no entries remained in
1155 inline_entry_data_table.
1156 * final.c (reemit_insn_block_notes): Handle inline entry notes.
1157 (final_scan_insn, notice_source_line): Likewise.
1158 (rest_of_clean_state): Skip inline entry markers.
1159 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
1161 * gimple.c (gimple_build_debug_inline_entry): New.
1162 * gimple.h (enum gimple_debug_subcode): Add
1163 GIMPLE_DEBUG_INLINE_ENTRY.
1164 (gimple_build_debug_inline_entry): Declare.
1165 (gimple_debug_inline_entry_p): New.
1166 (gimple_debug_nonbind_marker_p): Adjust.
1167 * insn-notes.def (INLINE_ENTRY): New.
1168 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
1169 inline entry marker notes.
1170 (print_insn): Likewise.
1171 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
1172 (INSN_DEBUG_MARKER_KIND): Likewise.
1173 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
1174 * tree-inline.c (expand_call_inline): Build and insert
1175 debug_inline_entry stmt.
1176 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
1177 inline entry blocks early, if nonbind markers are enabled.
1178 (dump_scope_block): Dump fragment info.
1179 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
1180 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
1181 (gimple_build_debug_inline_entry): New.
1182 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
1183 Enable/disable inline entry points too.
1184 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
1185 (DEBUG_INSN): Describe inline entry markers.
1187 * common.opt (gvariable-location-views): New.
1188 (gvariable-location-views=incompat5): New.
1189 * config.in: Rebuilt.
1190 * configure: Rebuilt.
1191 * configure.ac: Test assembler for view support.
1192 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
1193 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
1194 * dwarf2out.c (var_loc_view): New typedef.
1195 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
1196 (dwarf2out_locviews_in_attribute): New.
1197 (dwarf2out_locviews_in_loclist): New.
1198 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
1199 (enum dw_line_info_opcode): Add LI_adv_address.
1200 (struct dw_line_info_table): Add view.
1201 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
1202 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
1203 (zero_view_p): New variable.
1204 (ZERO_VIEW_P): New macro.
1205 (output_asm_line_debug_info): New.
1206 (struct var_loc_node): Add view.
1207 (add_AT_view_list, AT_loc_list): New.
1208 (add_var_loc_to_decl): Add view param. Test it against last.
1209 (new_loc_list): Add view params. Record them.
1210 (AT_loc_list_ptr): Handle loc and view lists.
1211 (view_list_to_loc_list_val_node): New.
1212 (print_dw_val): Handle dw_val_class_view_list.
1213 (size_of_die): Likewise.
1214 (value_format): Likewise.
1215 (loc_list_has_views): New.
1216 (gen_llsym): Set vl_symbol too.
1217 (maybe_gen_llsym, skip_loc_list_entry): New.
1218 (dwarf2out_maybe_output_loclist_view_pair): New.
1219 (output_loc_list): Output view list or entries too.
1220 (output_view_list_offset): New.
1221 (output_die): Handle dw_val_class_view_list.
1222 (output_dwarf_version): New.
1223 (output_compilation_unit_header): Use it.
1224 (output_skeleton_debug_sections): Likewise.
1225 (output_rnglists, output_line_info): Likewise.
1226 (output_pubnames, output_aranges): Update version comments.
1227 (output_one_line_info_table): Output view numbers in asm comments.
1228 (dw_loc_list): Determine current endview, pass it to new_loc_list.
1229 Call maybe_gen_llsym.
1230 (loc_list_from_tree_1): Adjust.
1231 (add_AT_location_description): Create view list attribute if
1232 needed, check it's absent otherwise.
1233 (convert_cfa_to_fb_loc_list): Adjust.
1234 (maybe_emit_file): Call output_asm_line_debug_info for test.
1235 (dwarf2out_var_location): Reset views as needed. Precompute
1236 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
1237 attribute. Set view.
1238 (new_line_info_table): Reset next view.
1239 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
1240 (dwarf2out_source_line): Likewise. Output view resets and labels to
1241 the assembler, or select appropriate line info opcodes.
1242 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
1243 (optimize_string_length): Catch it. Adjust.
1244 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
1245 dw_val_class_view_list, and remove it if no longer needed.
1246 (hash_loc_list): Hash view numbers.
1247 (loc_list_hasher::equal): Compare them.
1248 (optimize_location_lists): Check whether a view list symbol is
1249 needed, and whether the locview attribute is present, and
1250 whether they match. Remove the locview attribute if no longer
1252 (index_location_lists): Call skip_loc_list_entry for test.
1253 (dwarf2out_finish): Call output_asm_line_debug_info for test.
1254 Use output_dwarf_version.
1255 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
1256 (struct dw_val_node): Add val_view_list.
1257 * final.c (SEEN_NEXT_VIEW): New.
1258 (set_next_view_needed): New.
1259 (clear_next_view_needed): New.
1260 (maybe_output_next_view): New.
1261 (final_start_function): Rename to...
1262 (final_start_function_1): ... this. Take pointer to FIRST,
1263 add SEEN parameter. Emit param bindings in the initial view.
1264 (final_start_function): Reintroduce SEEN-less interface.
1265 (final): Rename to...
1266 (final_1): ... this. Take SEEN parameter. Output final pending
1267 next view at the end.
1268 (final): Reintroduce seen-less interface.
1269 (final_scan_insn): Output pending next view before switching
1270 sections or ending a block. Mark the next view as needed when
1271 outputting variable locations. Notify debug backend of section
1272 changes, and of location view changes.
1273 (rest_of_handle_final): Adjust.
1274 * toplev.c (process_options): Autodetect value for debug variable
1275 location views option. Warn on incompat5 without -gdwarf-5.
1276 * doc/invoke.texi (gvariable-location-views): New.
1277 (gvariable-location-views=incompat5): New.
1278 (gno-variable-location-views): New.
1280 2018-02-08 David Malcolm <dmalcolm@redhat.com>
1282 PR tree-optimization/84136
1283 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
1284 that the result of find_edge is non-NULL.
1286 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1289 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
1290 storing integer register in SImode. Fix cost of 256 and 512
1291 byte aligned SSE register store.
1293 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1295 * config/i386/i386.c (ix86_multiplication_cost): Fix
1296 multiplication cost for TARGET_AVX512DQ.
1298 2018-02-08 Marek Polacek <polacek@redhat.com>
1300 PR tree-optimization/84238
1301 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
1304 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1306 PR tree-optimization/84265
1307 * tree-vect-stmts.c (vectorizable_store): Don't treat
1308 VMAT_CONTIGUOUS accesses as grouped.
1309 (vectorizable_load): Likewise.
1311 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1313 PR tree-optimization/81635
1314 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
1315 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
1316 (test_round_for_mask): New functions.
1317 (wide_int_cc_tests): Call test_round_for_mask.
1318 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
1319 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
1320 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
1321 range returned by get_range_info.
1323 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1326 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
1327 * symtab.c: Include builtins.h
1328 (symtab_node::output_to_lto_symbol_table_p): Move here
1329 from lto-streamer-out.c:output_symbol_p.
1330 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
1331 (output_symbol_p): Move all logic to symtab.c
1332 (produce_symtab): Update.
1334 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1336 * config/s390/s390-opts.h (enum indirect_branch): Define.
1337 * config/s390/s390-protos.h (s390_return_addr_from_memory)
1338 (s390_indirect_branch_via_thunk)
1339 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
1340 (enum s390_indirect_branch_type): Define.
1341 * config/s390/s390.c (struct s390_frame_layout, struct
1342 machine_function): Remove.
1343 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
1344 (indirect_branch_table_label_no, indirect_branch_table_name):
1346 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
1347 (enum s390_indirect_branch_option): Define.
1348 (s390_return_addr_from_memory): New function.
1349 (s390_handle_string_attribute): New function.
1350 (s390_attribute_table): Add new attribute handler.
1351 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
1352 (s390_indirect_branch_via_thunk): New function.
1353 (s390_indirect_branch_via_inline_thunk): New function.
1354 (s390_function_ok_for_sibcall): When jumping via thunk disallow
1355 sibling call optimization for non z10 compiles.
1356 (s390_emit_call): Force indirect branch target to be a single
1357 register. Add r1 clobber for non-z10 compiles.
1358 (s390_emit_epilogue): Emit return jump via return_use expander.
1359 (s390_reorg): Handle JUMP_INSNs as execute targets.
1360 (s390_option_override_internal): Perform validity checks for the
1361 new command line options.
1362 (s390_indirect_branch_attrvalue): New function.
1363 (s390_indirect_branch_settings): New function.
1364 (s390_set_current_function): Invoke s390_indirect_branch_settings.
1365 (s390_output_indirect_thunk_function): New function.
1366 (s390_code_end): Implement target hook.
1367 (s390_case_values_threshold): Implement target hook.
1368 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
1370 * config/s390/s390.h (struct s390_frame_layout)
1371 (struct machine_function): Move here from s390.c.
1372 (TARGET_INDIRECT_BRANCH_NOBP_RET)
1373 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
1374 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
1375 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
1376 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
1377 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
1378 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
1379 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
1380 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
1381 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
1382 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
1383 (mnemonic attribute): Add values which aren't recognized
1385 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
1386 pattern for branch conversion. Fix mnemonic attribute.
1387 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
1388 indirect branch via thunk if requested.
1389 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
1390 ("*indirect_jump"): Disable for branch conversion using out of
1392 ("indirect_jump_via_thunk<mode>_z10")
1393 ("indirect_jump_via_thunk<mode>")
1394 ("indirect_jump_via_inlinethunk<mode>_z10")
1395 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
1396 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
1397 ("casesi_jump_via_inlinethunk<mode>_z10")
1398 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
1399 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
1400 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
1401 ("*indirect2_jump"): Disable for branch conversion.
1402 ("casesi_jump"): Turn into expander and expand patterns for branch
1404 ("return_use"): New expander.
1405 ("*return"): Emit return via thunk and rename it to ...
1406 ("*return<mode>"): ... this one.
1407 * config/s390/s390.opt: Add new options and and enum for the
1410 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1412 * lra-constraints.c (match_reload): Unconditionally use
1413 gen_lowpart_SUBREG, rather than selecting between that
1414 and equivalent gen_rtx_SUBREG code.
1416 2018-02-08 Richard Biener <rguenther@suse.de>
1418 PR tree-optimization/84233
1419 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
1420 changed flag instead of boguously re-using phi_inserted.
1422 2018-02-08 Martin Jambor <mjambor@suse.cz>
1424 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
1425 static local variables.
1427 2018-02-08 Richard Biener <rguenther@suse.de>
1429 PR tree-optimization/84278
1430 * tree-vect-stmts.c (vectorizable_store): When looking for
1431 smaller vector types to perform grouped strided loads/stores
1432 make sure the mode is supported by the target.
1433 (vectorizable_load): Likewise.
1435 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1437 * config/aarch64/aarch64.c (aarch64_components_for_bb):
1438 Increase LDP/STP opportunities by adding adjacent callee-saves.
1440 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1442 PR rtl-optimization/84068
1443 PR rtl-optimization/83459
1444 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
1446 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
1448 PR tree-optimization/84224
1449 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
1450 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
1453 2018-02-07 Iain Sandoe <iain@codesourcery.com>
1456 * config/rs6000/altivec.md (*restore_world): Remove LR use.
1457 * config/rs6000/predicates.md (restore_world_operation): Adjust op
1458 count, remove one USE.
1460 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
1462 * doc/install.texi (Configuration): Document the
1463 --with-long-double-format={ibm,ieee} PowerPC configuration
1467 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
1468 Convert from define_expand to be define_insn_and_split. Rework
1469 float/double/_Float128 conversions to QI/HI/SImode to work with
1470 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
1471 conversions to QI/HImode types did a store and then a load to
1472 truncate the value. For conversions to VSX registers, don't split
1473 the insn, instead emit the code directly. Use the code iterator
1474 any_fix to combine signed and unsigned conversions.
1475 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
1476 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
1477 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1478 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
1479 (fix_<mode>di2_hw): Likewise.
1480 (fixuns_<mode>di2_hw): Likewise.
1481 (fix_<mode>si2_hw): Likewise.
1482 (fixuns_<mode>si2_hw): Likewise.
1483 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
1484 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1485 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
1486 fix<uns>_trunc<SFDF:mode>si2_p8.
1487 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
1489 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
1490 (fix<uns>_<mode>_mem): Likewise.
1491 (fctiw<u>z_<mode>_mem): Likewise.
1492 (fix<uns>_<mode>_mem): Likewise.
1493 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
1494 the register allocator from doing a direct move to the GPRs to do
1495 a store, and instead use the ISA 3.0 store byte/half-word from
1496 vector register instruction. For IEEE 128-bit floating point,
1497 also optimize stores of 32-bit ints.
1498 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
1500 2018-02-07 Alan Hayward <alan.hayward@arm.com>
1502 * genextract.c (push_pathstr_operand): New function to support
1504 (walk_rtx): Call push_pathstr_operand.
1505 (print_path): Support [a-zA-Z].
1507 2018-02-07 Richard Biener <rguenther@suse.de>
1509 PR tree-optimization/84037
1510 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
1511 (cse_and_gimplify_to_preheader): Declare.
1512 (vect_get_place_in_interleaving_chain): Likewise.
1513 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
1515 (_loop_vec_info::~_loop_vec_info): Delete it.
1516 (cse_and_gimplify_to_preheader): New function.
1517 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
1518 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
1519 (vectorizable_load): Likewise. For grouped stores always base
1520 the IV on the first element.
1521 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
1522 condition before gimplifying.
1524 2018-02-07 Jakub Jelinek <jakub@redhat.com>
1526 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
1527 *DIV_EXPR and *MOD_EXPR.
1529 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
1532 * config/i386/i386.c (ix86_option_override_internal): Mask out
1533 the CF_SET bit when checking -fcf-protection.
1535 2018-02-07 Tom de Vries <tom@codesourcery.com>
1538 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
1541 2018-02-07 Richard Biener <rguenther@suse.de>
1543 PR tree-optimization/84204
1544 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
1547 PR tree-optimization/84205
1548 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
1549 special-case isl_ast_op_zdiv_r.
1551 PR tree-optimization/84223
1552 * graphite-scop-detection.c (gather_bbs::before_dom_children):
1553 Only add conditions from within the region.
1554 (gather_bbs::after_dom_children): Adjust.
1556 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
1559 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
1560 * config/avr/avr.md: Only post-reload split REG-REG moves if
1561 either register is REGERAL_REG_P.
1563 2018-02-07 Jakub Jelinek <jakub@redhat.com>
1565 PR tree-optimization/84235
1566 * tree-ssa-scopedtables.c
1567 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
1568 if the subtraction is performed in floating point type where NaNs are
1569 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
1570 build 1. Formatting fix.
1572 2018-02-06 Jakub Jelinek <jakub@redhat.com>
1575 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
1576 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
1577 and skip it regardless of bb boundaries. Use CALL_P macro,
1578 don't test INSN_P (insn) together with CALL_P or JUMP_P check
1579 unnecessarily, formatting fix.
1581 2018-02-06 Michael Collison <michael.collison@arm.com>
1583 * config/arm/thumb2.md:
1584 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
1585 (*thumb_mov_notscc): Ditto.
1587 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
1590 * config/rs6000/rs6000.md (su code attribute): Use "u" for
1591 unsigned_fix, not "s".
1593 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1595 * configure.ac (gcc_fn_eh_frame_ro): New function.
1596 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
1597 correct .eh_frame permissions.
1598 * configure: Regenerate.
1600 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
1602 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
1605 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1607 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1608 Display warning message for -mno-speculate-indirect-jumps.
1610 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
1612 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
1614 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
1616 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
1618 PR tree-optimization/84225
1619 * tree-eh.c (find_trapping_overflow): Only call
1620 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
1622 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1625 * config/i386/i386.c: Reimplement the check of possible options
1626 -mibt/-mshstk conbination. Change error messages.
1627 * doc/invoke.texi: Fix a typo: remove extra '='.
1629 2018-02-06 Marek Polacek <polacek@redhat.com>
1631 PR tree-optimization/84228
1632 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
1634 2018-02-06 Tamar Christina <tamar.christina@arm.com>
1637 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
1638 emitted arch directives.
1639 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
1640 __ARM_FEATURE_COPROC before changing architectures.
1642 2018-02-06 Richard Biener <rguenther@suse.de>
1644 * config/i386/i386.c (print_reg): Fix typo.
1645 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
1647 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
1649 * configure: Regenerate.
1651 2018-02-05 Martin Sebor <msebor@redhat.com>
1653 PR tree-optimization/83369
1654 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
1657 2018-02-05 Martin Liska <mliska@suse.cz>
1659 * doc/invoke.texi: Cherry-pick upstream r323995.
1661 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
1663 * ira.c (ira_init_register_move_cost): Adjust comment.
1665 2018-02-05 Martin Liska <mliska@suse.cz>
1667 PR gcov-profile/84137
1668 * doc/gcov.texi: Fix typo in documentation.
1670 2018-02-05 Martin Liska <mliska@suse.cz>
1672 PR gcov-profile/83879
1673 * doc/gcov.texi: Document necessity of --dynamic-list-data when
1674 using dlopen functionality.
1676 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
1678 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
1679 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
1680 _mm_maskz_range_ss, _mm_mask_range_round_ss,
1681 _mm_maskz_range_round_ss): New intrinsics.
1682 (__builtin_ia32_rangesd128_round)
1683 (__builtin_ia32_rangess128_round): Remove.
1684 (__builtin_ia32_rangesd128_mask_round,
1685 __builtin_ia32_rangess128_mask_round): New builtins.
1686 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
1687 __builtin_ia32_rangess128_round): Remove.
1688 (__builtin_ia32_rangesd128_mask_round,
1689 __builtin_ia32_rangess128_mask_round): New builtins.
1690 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
1691 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
1692 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
1693 "<round_saeonly_constraint>")): Changed to ...
1694 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
1695 "<round_saeonly_scalar_constraint>")): ... this.
1696 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
1697 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
1698 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
1699 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
1700 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
1702 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
1704 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
1706 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
1707 Remove all values except native, 8540 and 8548.
1709 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
1711 * config/i386/i386.c (ix86_output_function_return): Pass
1712 INVALID_REGNUM, instead of -1, as invalid register number to
1713 indirect_thunk_name and output_indirect_thunk.
1715 2018-02-02 Julia Koval <julia.koval@intel.com>
1717 * config.gcc: Add -march=icelake.
1718 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
1719 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
1720 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
1721 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
1722 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
1723 (processor_target_table): Add icelake.
1724 (ix86_option_override_internal): Handle new PTAs.
1725 (get_builtin_code_for_version): Handle icelake.
1726 (M_INTEL_COREI7_ICELAKE): New.
1727 (fold_builtin_cpu): Handle icelake.
1728 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
1729 * doc/invoke.texi: Add -march=icelake.
1731 2018-02-02 Julia Koval <julia.koval@intel.com>
1733 * config/i386/i386.c (ix86_option_override_internal): Change flags type
1734 to wide_int_bitmask.
1735 * wide-int-bitmask.h: New.
1737 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1740 * config/i386/i386.md: Replace Pmode with word_mode in
1741 builtin_setjmp_setup and builtin_longjmp to support x32.
1743 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
1747 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
1749 (rs6000_supported_cpu_names): New static variable.
1750 (linux_cpu_translation_table): Likewise.
1751 (elf_platform) <cpu>: Define new static variable and use it.
1752 Translate kernel AT_PLATFORM name to canonical name if needed.
1753 Error if platform name is unknown.
1755 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
1758 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
1760 2018-02-01 Jeff Law <law@redhat.com>
1763 * config/i386/i386.c (release_scratch_register_on_entry): Add new
1764 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
1765 the scratch if RELEASE_VIA_POP is false.
1766 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
1767 If we have to save a temporary register, decrement SIZE appropriately.
1768 Pass new arguments to release_scratch_register_on_entry.
1769 (ix86_adjust_stack_and_probe): Likewise.
1770 (ix86_emit_probe_stack_range): Pass new arguments to
1771 release_scratch_register_on_entry.
1773 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
1775 PR rtl-optimization/84157
1776 * combine.c (change_zero_ext): Use REG_P predicate in
1777 front of HARD_REGISTER_P predicate.
1779 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
1781 * config/avr/avr.c (avr_option_override): Move disabling of
1782 -fdelete-null-pointer-checks to...
1783 * common/config/avr/avr-common.c (avr_option_optimization_table):
1786 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1788 PR tree-optimization/81635
1789 * tree-data-ref.c (split_constant_offset_1): For types that
1790 wrap on overflow, try to use range info to prove that wrapping
1793 2018-02-01 Renlin Li <renlin.li@arm.com>
1796 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
1798 (aarch64_register_move_cost): Likewise.
1799 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
1801 (REG_CLASS_NAMES): Likewise.
1802 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
1803 TAILCALL_ADDR_REGS. Remove IP registers.
1804 * config/aarch64/aarch64.md (Ucs): Update register constraint.
1806 2018-02-01 Richard Biener <rguenther@suse.de>
1808 * domwalk.h (dom_walker::dom_walker): Add additional constructor
1809 for specifying RPO order and allow NULL for that.
1810 * domwalk.c (dom_walker::dom_walker): Likewise.
1811 (dom_walker::walk): Handle NULL RPO order.
1812 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
1814 (rewrite_update_dom_walker): Likewise.
1815 (mark_def_dom_walker): Likewise.
1817 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1819 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
1820 (aarch64_maybe_expand_sve_subreg_move): Declare.
1821 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
1822 * config/aarch64/predicates.md (aarch64_any_register_operand): New
1824 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
1825 that are semantically a reverse operation.
1826 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
1827 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
1828 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
1830 (aarch64_can_change_mode_class): For big-endian, forbid changes
1831 between two SVE modes if they have different element sizes.
1833 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1835 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
1836 the TImode handling for big-endian targets.
1838 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1840 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
1841 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
1843 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
1844 Remove BSWAP handing for big-endian targets and use the form of
1845 LD1RQ appropariate for the mode.
1847 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1849 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
1850 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
1853 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1856 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
1857 check for operands that need to go through aarch64_sve_reload_be.
1859 2018-02-01 Jakub Jelinek <jakub@redhat.com>
1861 PR tree-optimization/81661
1862 PR tree-optimization/84117
1863 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
1864 * tree-eh.c: Include gimplify.h.
1865 (find_trapping_overflow, replace_trapping_overflow,
1866 rewrite_to_non_trapping_overflow): New functions.
1867 * tree-vect-loop.c: Include tree-eh.h.
1868 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
1869 * tree-data-ref.c: Include tree-eh.h.
1870 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
1872 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
1874 PR rtl-optimization/84123
1875 * combine.c (change_zero_ext): Check if hard register satisfies
1876 can_change_dest_mode before calling gen_lowpart_SUBREG.
1878 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
1881 * ira.c (ira_init_register_move_cost): Remove assert.
1883 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
1885 PR rtl-optimization/84071
1886 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
1887 * doc/tm.texi: Regenerate.
1889 2018-01-31 Richard Biener <rguenther@suse.de>
1891 PR tree-optimization/84132
1892 * tree-data-ref.c (analyze_miv_subscript): Properly
1893 check whether evolution_function_is_affine_multivariate_p
1894 before calling gcd_of_steps_may_divide_p.
1896 2018-01-31 Julia Koval <julia.koval@intel.com>
1899 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
1900 * config/i386/i386.md (rdpid_rex64) New.
1901 (rdpid): Make 32bit only.
1903 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
1906 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
1907 an IDENTIFIER_NODE for FUNCTION_TYPE's.
1909 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
1912 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
1914 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
1916 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
1918 PR rtl-optimization/84071
1919 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
1920 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
1922 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
1924 * config/arc/arc.c (arc_handle_aux_attribute): New function.
1925 (arc_attribute_table): Add 'aux' attribute.
1926 (arc_in_small_data_p): Consider aux like variables.
1927 (arc_is_aux_reg_p): New function.
1928 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
1929 (arc_get_aux_arg): New function.
1930 (prepare_move_operands): Handle aux-register access.
1931 (arc_handle_aux_attribute): New function.
1932 * doc/extend.texi (ARC Variable attributes): Add subsection.
1934 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
1936 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
1937 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
1938 (arc_attribute_table): Add 'uncached' attribute.
1939 (arc_print_operand): Print '.di' flag for uncached memory
1941 (arc_in_small_data_p): Do not consider for small data the uncached
1943 (arc_is_uncached_mem_p): New function.
1944 * config/arc/predicates.md (compact_store_memory_operand): Check
1945 for uncached memory accesses.
1946 (nonvol_nonimm_operand): Likewise.
1947 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
1949 2018-01-31 Jakub Jelinek <jakub@redhat.com>
1952 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
1953 falign-loops=): Add Optimization flag.
1955 2018-01-30 Jeff Law <law@redhat.com>
1958 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
1959 INT_REGISTERS_SAVED. Check it prior to calling
1960 get_scratch_register_on_entry.
1961 (ix86_adjust_stack_and_probe): Similarly.
1962 (ix86_emit_probe_stack_range): Similarly.
1963 (ix86_expand_prologue): Corresponding changes.
1965 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1968 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
1969 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
1971 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
1974 * lra-constraints.c (curr_insn_transform): Process AND in the
1977 2018-01-30 Jakub Jelinek <jakub@redhat.com>
1979 PR rtl-optimization/83986
1980 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
1981 dependence against last_pending_memory_flush in addition to
1984 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
1986 PR tree-optimization/81611
1987 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
1990 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1993 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
1996 2018-01-30 Richard Biener <rguenther@suse.de>
1997 Jakub Jelinek <jakub@redhat.com>
1999 PR tree-optimization/84111
2000 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
2001 inner loops added during recursion, as they don't have up-to-date
2004 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2007 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
2008 (can_inline_edge_by_limits_p): ... here.
2009 (can_early_inline_edge_p, check_callers,
2010 update_caller_keys, update_callee_keys, recursive_inlining,
2011 add_new_edges_to_heap, speculation_useful_p,
2012 inline_small_functions,
2013 inline_small_functions, flatten_function,
2014 inline_to_all_callers_1): Update.
2016 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2018 * profile-count.c (profile_count::combine_with_ipa_count): Handle
2021 2018-01-30 Richard Biener <rguenther@suse.de>
2023 PR tree-optimization/83008
2024 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
2025 invariant and constant vector uses in stmts when they need
2028 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2031 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
2032 * configure: Regenerate.
2034 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2036 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
2038 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
2039 Use gen_rtx_REG rather than gen_lowpart.
2041 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2043 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
2044 rather than 0 when creating partial subregs.
2046 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2048 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
2051 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2054 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
2055 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
2056 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
2057 flags. This restores the settings used before the 2017-07-24.
2058 Turning off pre increment/decrement/modify allows IVOPTS to
2059 optimize DF/SF loops where the index is an int.
2061 2018-01-29 Richard Biener <rguenther@suse.de>
2062 Kelvin Nilsen <kelvin@gcc.gnu.org>
2065 * tree-vect-stmts.c (vectorizable_call): Don't call
2066 targetm.vectorize_builtin_md_vectorized_function if callee is
2069 2018-01-22 Carl Love <cel@us.ibm.com>
2071 * doc/extend.tex: Fix typo in second arg in
2072 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
2074 2018-01-29 Richard Biener <rguenther@suse.de>
2076 PR tree-optimization/84086
2077 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
2078 (flush_ssaname_freelist): When SSA names were released reset
2079 the SCEV hash table.
2081 2018-01-29 Richard Biener <rguenther@suse.de>
2083 PR tree-optimization/84057
2084 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
2085 removed paths when removing edges.
2087 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
2089 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
2090 -mfunction-return=@var{choice}.
2092 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
2095 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
2097 (layout::print_source_line): Likewise.
2098 (test_get_line_width_without_trailing_whitespace): Add test cases.
2100 2018-01-27 Jakub Jelinek <jakub@redhat.com>
2103 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
2106 2018-01-26 Jim Wilson <jimw@sifive.com>
2108 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
2110 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
2113 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2115 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
2116 and CMP + SUB-immediate -> SUBS.
2118 2018-01-26 Martin Sebor <msebor@redhat.com>
2120 PR tree-optimization/83896
2121 * tree-ssa-strlen.c (get_string_len): Rename...
2122 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
2123 Avoid assuming length is constant.
2124 (handle_char_store): Use HOST_WIDE_INT for string length.
2126 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
2129 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
2130 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
2132 2018-01-26 Richard Biener <rguenther@suse.de>
2134 PR rtl-optimization/84003
2135 * dse.c (record_store): Only record redundant stores when
2136 the earlier store aliases at least all accesses the later one does.
2138 2018-01-26 Jakub Jelinek <jakub@redhat.com>
2140 PR rtl-optimization/83985
2141 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
2142 REG_CFA_RESTORE insns.
2143 (delete_unmarked_insns): Don't ignore separate shrink wrapping
2144 REG_CFA_RESTORE insns here.
2147 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
2148 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
2150 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2152 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
2153 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
2154 (arc_init): Likewise.
2155 (arc_override_options): Likewise.
2156 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
2158 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
2160 * config/arc/arc.h (TARGET_DBNZ): Define.
2161 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
2162 properly set the tune attribute.
2163 (dbnz): Use TARGET_DBNZ guard.
2164 * config/arc/arc.opt (mtune): Add core3 option.
2166 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2168 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
2169 recognize new pic like addresses.
2170 (arc_delegitimize_address): Clean up.
2172 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2174 * config/arc/arc-arches.def: Option mrf16 valid for all
2176 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
2177 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
2178 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
2179 * config/arc/arc-tables.opt: Regenerate.
2180 * config/arc/arc.c (arc_conditional_register_usage): Handle
2181 reduced register file case.
2182 (arc_file_start): Set must have build attributes.
2183 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
2185 * config/arc/arc.opt (mrf16): Add new option.
2186 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
2187 * config/arc/genmultilib.awk: Handle new mrf16 option.
2188 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
2189 * config/arc/t-multilib: Regenerate.
2190 * doc/invoke.texi (ARC Options): Document mrf16 option.
2192 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2194 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
2195 * config/arc/arc.c (arc_handle_secure_attribute): New function.
2196 (arc_attribute_table): Add 'secure_call' attribute.
2197 (arc_print_operand): Print secure call operand.
2198 (arc_function_ok_for_sibcall): Don't optimize tail calls when
2200 (arc_is_secure_call_p): New function. * config/arc/arc.md
2201 (call_i): Add support for sjli instruction.
2202 (call_value_i): Likewise.
2203 * config/arc/constraints.md (Csc): New constraint.
2205 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2206 John Eric Martin <John.Martin@emmicro-us.com>
2208 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
2209 * config/arc/arc.c (_arc_jli_section): New struct.
2210 (arc_jli_section): New type.
2211 (rc_jli_sections): New static variable.
2212 (arc_handle_jli_attribute): New function.
2213 (arc_attribute_table): Add jli_always and jli_fixed attribute.
2214 (arc_file_end): New function.
2215 (TARGET_ASM_FILE_END): Define.
2216 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
2217 (arc_add_jli_section): New function.
2218 (jli_call_scan): Likewise.
2219 (arc_reorg): Call jli_call_scan.
2220 (arc_output_addsi): Remove 'S' from printing asm operand.
2221 (arc_is_jli_call_p): New function.
2222 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
2224 (movhi_insn): Likewise.
2225 (movsi_insn): Likewise.
2226 (movsi_set_cc_insn): Likewise.
2227 (loadqi_update): Likewise.
2228 (load_zeroextendqisi_update): Likewise.
2229 (load_signextendqisi_update): Likewise.
2230 (loadhi_update): Likewise.
2231 (load_zeroextendhisi_update): Likewise.
2232 (load_signextendhisi_update): Likewise.
2233 (loadsi_update): Likewise.
2234 (loadsf_update): Likewise.
2235 (movsicc_insn): Likewise.
2236 (bset_insn): Likewise.
2237 (bxor_insn): Likewise.
2238 (bclr_insn): Likewise.
2239 (bmsk_insn): Likewise.
2240 (bicsi3_insn): Likewise.
2241 (cmpsi_cc_c_insn): Likewise.
2242 (movsi_ne): Likewise.
2243 (movsi_cond_exec): Likewise.
2244 (clrsbsi2): Likewise.
2255 (call_i): Remove 'S' asm letter, add jli instruction.
2256 (call_value_i): Likewise.
2257 * config/arc/arc.op (mjli-always): New option.
2258 * config/arc/constraints.md (Cji): New constraint.
2259 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
2261 (subsf3_fpx): Likewise.
2262 (mulsf3_fpx): Likewise.
2263 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
2265 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
2267 * doc/invoke.texi (ARC): Document mjli-always option.
2269 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
2271 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
2272 avoid addition with 0 and use incw and decw where possible.
2274 2018-01-26 Richard Biener <rguenther@suse.de>
2276 PR tree-optimization/81082
2277 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
2278 association if it requires casting to unsigned.
2279 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
2280 from fold_plusminus_mult_expr to catch important cases late when
2281 range info is available.
2283 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2285 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
2286 * configure.ac (hidden_linkonce): New test.
2287 * configure: Regenerate.
2288 * config.in: Regenerate.
2290 2018-01-26 Julia Koval <julia.koval@intel.com>
2292 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
2293 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
2294 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
2295 _mm_mask_bitshuffle_epi64_mask): Fix type.
2296 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
2297 USI_FTYPE_V4DI_V4DI_USI): Remove.
2298 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
2299 __builtin_ia32_vpshufbitqmb256_mask,
2300 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
2301 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
2302 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
2304 2018-01-26 Alan Modra <amodra@gmail.com>
2307 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
2308 UNSPEC_VBPERMQ. Sort other unspecs.
2310 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
2312 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
2314 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
2317 * predict.c (drop_profile): Do not push/pop cfun; update also
2319 (handle_missing_profiles): Fix logic looking for zero profiles.
2321 2018-01-25 Jakub Jelinek <jakub@redhat.com>
2324 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
2325 on functions with #pragma omp declare simd or functions with simd
2327 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
2328 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
2329 Remove trailing \n from warning_at calls.
2331 2018-01-25 Tom de Vries <tom@codesourcery.com>
2334 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2335 for neutered workers.
2337 2018-01-24 Joseph Myers <joseph@codesourcery.com>
2340 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
2341 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
2343 2018-01-24 Jeff Law <law@redhat.com>
2346 * i386.c (get_probe_interval): Move to earlier point.
2347 (ix86_compute_frame_layout): If -fstack-clash-protection and
2348 the frame is larger than the probe interval, then use pushes
2349 to save registers rather than reg->mem moves.
2350 (ix86_expand_prologue): Remove conditional for int_registers_saved
2353 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
2356 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
2357 min/max for never referenced object.
2359 2018-01-24 Jakub Jelinek <jakub@redhat.com>
2362 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
2364 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
2365 attributes from DECL_ATTRIBUTES (decl) without affecting
2366 DECL_ATTRIBUTES (current_function_decl).
2367 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
2368 functions with non-NULL DECL_ABSTRACT_ORIGIN.
2370 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
2372 PR tree-optimization/83979
2373 * fold-const.c (fold_comparison): Use constant_boolean_node
2374 instead of boolean_{true,false}_node.
2376 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
2378 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
2381 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2383 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2384 Simplify the clause that sets the length attribute.
2385 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2386 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
2387 clause that sets the length attribute.
2388 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2390 2018-01-24 Tom de Vries <tom@codesourcery.com>
2393 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
2394 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
2395 Add strict parameter.
2396 (prevent_branch_around_nothing): Insert dummy insn between branch to
2397 label and label with no ptx insn inbetween.
2398 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
2400 2018-01-24 Tom de Vries <tom@codesourcery.com>
2403 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2404 for neutered threads in warp.
2405 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
2407 2018-01-24 Richard Biener <rguenther@suse.de>
2409 PR tree-optimization/83176
2410 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
2413 2018-01-24 Richard Biener <rguenther@suse.de>
2415 PR tree-optimization/82819
2416 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
2417 code generating pluses that are no-ops in the target precision.
2419 2018-01-24 Richard Biener <rguenther@suse.de>
2422 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
2424 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2426 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
2427 to merge probabilities.
2428 * predict.c (probably_never_executed): Also mark as cold functions
2429 with global 0 profile and guessed local profile.
2430 * profile-count.c (profile_probability::combine_with_count): New
2432 * profile-count.h (profile_probability::operator*,
2433 profile_probability::operator*=, profile_probability::operator/,
2434 profile_probability::operator/=): Reduce precision to adjusted
2435 and set value to guessed on contradictory divisions.
2436 (profile_probability::combine_with_freq): Remove.
2437 (profile_probability::combine_wiht_count): Declare.
2438 (profile_count::force_nonzero):: Set to adjusted.
2439 (profile_count::probability_in):: Set quality to adjusted.
2440 * tree-ssa-tail-merge.c (replace_block_by): Use
2443 2018-01-23 Andrew Waterman <andrew@sifive.com>
2444 Jim Wilson <jimw@sifive.com>
2446 * config/riscv/riscv.c (riscv_stack_boundary): New.
2447 (riscv_option_override): Set riscv_stack_boundary. Handle
2448 riscv_preferred_stack_boundary_arg.
2449 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
2450 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
2451 (STACK_BOUNDARY): Set to riscv_stack_boundary.
2452 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
2453 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
2454 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
2456 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
2459 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
2460 of struct ix86_frame.
2461 (ix86_expand_epilogue): Likewise. Add a local variable for
2462 the reg_save_offset field in struct ix86_frame.
2464 2018-01-23 Bin Cheng <bin.cheng@arm.com>
2466 PR tree-optimization/82604
2467 * tree-loop-distribution.c (enum partition_kind): New enum item
2468 PKIND_PARTIAL_MEMSET.
2469 (partition_builtin_p): Support above new enum item.
2470 (generate_code_for_partition): Ditto.
2471 (compute_access_range): Differentiate cases that equality can be
2472 proven at all loops, the innermost loops or no loops.
2473 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
2474 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
2475 (finalize_partitions, distribute_loop): Don't fuse partition of
2476 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
2477 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
2480 2018-01-23 Martin Liska <mliska@suse.cz>
2482 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
2483 order to ignore the predictor.
2484 (PRED_POLYMORPHIC_CALL): Likewise.
2485 (PRED_RECURSIVE_CALL): Likewise.
2487 2018-01-23 Martin Liska <mliska@suse.cz>
2489 * tree-profile.c (tree_profiling): Print function header to
2490 aware reader which function we are working on.
2491 * value-prof.c (gimple_find_values_to_profile): Do not print
2492 not interesting value histograms.
2494 2018-01-23 Martin Liska <mliska@suse.cz>
2496 * profile-count.h (enum profile_quality): Add
2497 profile_uninitialized as the first value. Do not number values
2498 as they are zero based.
2499 (profile_count::verify): Update sanity check.
2500 (profile_probability::verify): Likewise.
2502 2018-01-23 Nathan Sidwell <nathan@acm.org>
2504 * doc/invoke.texi (ffor-scope): Deprecate.
2506 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2508 PR tree-optimization/83510
2509 * domwalk.c (set_all_edges_as_executable): New function.
2510 (dom_walker::dom_walker): Convert bool param
2511 "skip_unreachable_blocks" to enum reachability. Move setup of
2512 edge flags to set_all_edges_as_executable and only do it when
2513 reachability is REACHABLE_BLOCKS.
2514 * domwalk.h (enum dom_walker::reachability): New enum.
2515 (dom_walker::dom_walker): Convert bool param
2516 "skip_unreachable_blocks" to enum reachability.
2517 (set_all_edges_as_executable): New decl.
2518 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
2519 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
2521 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
2522 but converting true to REACHABLE_BLOCKS.
2523 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
2525 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
2526 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
2527 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
2529 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
2530 if check_all_array_refs will be called.
2532 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2534 * tree.c (selftest::test_location_wrappers): Add more test
2537 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2539 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
2540 (selftest::test_bit_in_range): Likewise.
2542 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
2545 * doc/sourcebuild.texi (vect_float): Say that the selector
2546 only describes the situation when -funsafe-math-optimizations is on.
2547 (vect_float_strict): Document.
2549 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
2551 PR tree-optimization/83965
2552 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
2553 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
2554 instead of checking only for a reduction.
2555 (vect_recog_widen_sum_pattern): Likewise.
2557 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2559 * predict.c (probably_never_executed): Only use precise profile info.
2560 (compute_function_frequency): Skip after inlining hack since we now
2561 have quality checking.
2563 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2565 * profile-count.h (profile_probability::very_unlikely,
2566 profile_probability::unlikely, profile_probability::even): Set
2567 precision to guessed.
2569 2018-01-23 Richard Biener <rguenther@suse.de>
2571 PR tree-optimization/83963
2572 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
2573 Properly terminate dominator walk when crossing the exit edge not
2574 when visiting its source block.
2576 2018-01-23 Jakub Jelinek <jakub@redhat.com>
2579 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
2580 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
2582 2018-01-22 Jakub Jelinek <jakub@redhat.com>
2584 PR tree-optimization/83957
2585 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
2586 semicolon after for body surrounded by braces.
2588 PR tree-optimization/83081
2589 * profile-count.h (profile_probability::split): New method.
2590 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
2591 Use profile_probability::split.
2592 (do_compare_rtx_and_jump): Fix adjustment of probabilities
2593 when splitting a single conditional jump into 2.
2595 2018-01-22 David Malcolm <dmalcolm@redhat.com>
2597 PR tree-optimization/69452
2598 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
2601 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2603 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
2604 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
2605 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
2607 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2609 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
2610 * config/rl78/rl78.md (movdi): New define_expand.
2611 * config/rl78/rl78.c (rl78_split_movdi): New function.
2613 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
2616 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
2618 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
2619 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
2620 128-bit to produce an UNSPEC move to get the double word with the
2621 signbit and then a shift directly to do signbit.
2622 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
2623 implementation with a new version that just does either a direct
2624 move or a regular move. Move memory interface to separate insns.
2625 Move insns so they are next to the expander.
2626 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
2627 with signbit move. Split big and little endian case.
2628 (signbit<mode>2_dm_mem_le): Likewise.
2629 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
2630 (signbit<mode>2_dm2): Likewise.
2632 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2634 * config/rl78/rl78.md (anddi3): New define_expand.
2636 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2638 * config/rl78/rl78.md (umindi3): New define_expand.
2640 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2642 * config/rl78/rl78.md (smindi3): New define_expand.
2644 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2646 * config/rl78/rl78.md (smaxdi3): New define_expand.
2648 2018-01-22 Carl Love <cel@us.ibm.com>
2650 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
2651 LVX_V1TI): Add macro expansion.
2652 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
2653 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
2654 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
2655 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
2656 Change check to determine if the instruction is a byte reversing
2657 entry. Fix typo in comment.
2658 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
2659 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
2660 Add def_builtin calls for new builtins.
2661 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
2662 Add define_insn expansion.
2664 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2666 * config/rl78/rl78.md (umaxdi3): New define_expand.
2668 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2670 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
2671 for non-QImode registers.
2673 2018-01-22 Richard Biener <rguenther@suse.de>
2675 PR tree-optimization/83963
2676 * graphite-scop-detection.c (scop_detection::get_sese): Delay
2677 including the loop exit block.
2678 (scop_detection::merge_sese): Likewise.
2679 (scop_detection::add_scop): Do it here instead.
2681 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2683 * doc/sourcebuild.texi (arm_softfloat): Document.
2685 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
2688 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
2689 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
2690 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
2692 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2693 David Edelsohn <dje.gcc@gmail.com>
2696 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2697 Change "crset eq" to "crset 2".
2698 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2699 (*call_indirect_aix<mode>_nospec): Likewise.
2700 (*call_value_indirect_aix<mode>_nospec): Likewise.
2701 (*call_indirect_elfv2<mode>_nospec): Likewise.
2702 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
2703 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
2704 change assembly output from . to $.
2705 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2706 (indirect_jump<mode>_nospec): Change assembly output from . to $.
2707 (*tablejump<mode>_internal1_nospec): Likewise.
2709 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
2712 * config/sh/sh_optimize_sett_clrt.cc:
2713 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
2715 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
2717 PR tree-optimization/83940
2718 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
2719 offset_dt to vect_constant_def rather than vect_unknown_def_type.
2720 (vect_check_load_store_mask): Add a mask_dt_out parameter and
2721 use it to pass back the definition type.
2722 (vect_check_store_rhs): Likewise rhs_dt_out.
2723 (vect_build_gather_load_calls): Add a mask_dt argument and use
2724 it instead of a call to vect_is_simple_use.
2725 (vectorizable_store): Update calls to vect_check_load_store_mask
2726 and vect_check_store_rhs. Use the dt returned by the latter instead
2727 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
2728 instead of calls to vect_is_simple_use. Pass the scalar rather
2729 than the vector operand to vect_is_simple_use when handling
2730 second and subsequent copies of an rhs value.
2731 (vectorizable_load): Update calls to vect_check_load_store_mask
2732 and vect_build_gather_load_calls. Use the cached mask_dt and
2733 gs_info.offset_dt instead of calls to vect_is_simple_use.
2735 2018-01-20 Jakub Jelinek <jakub@redhat.com>
2738 * tree-emutls.c: Include gimplify.h.
2739 (lower_emutls_2): New function.
2740 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
2741 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
2742 it before further processing.
2745 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
2746 UINTVAL (trueop1) instead of INTVAL (op1).
2748 2018-01-19 Jakub Jelinek <jakub@redhat.com>
2752 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
2753 INCOMING_FRAME_SP_OFFSET if not defined.
2754 (scan_trace): Add ENTRY argument. If true and
2755 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
2756 emit a note to adjust the CFA offset.
2757 (create_cfi_notes): Adjust scan_trace callers.
2758 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
2759 INCOMING_FRAME_SP_OFFSET in the CIE.
2760 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
2761 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
2763 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
2764 * doc/tm.texi: Regenerated.
2766 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2768 PR rtl-optimization/83147
2769 * lra-constraints.c (remove_inheritance_pseudos): Use
2770 lra_substitute_pseudo_within_insn.
2772 2018-01-19 Tom de Vries <tom@codesourcery.com>
2773 Cesar Philippidis <cesar@codesourcery.com>
2776 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
2778 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
2781 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
2782 spaces for function labels.
2784 2018-01-19 Martin Liska <mliska@suse.cz>
2786 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
2787 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
2788 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
2789 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
2790 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
2791 (PRED_CONST_RETURN): Change from 69 to 65.
2792 (PRED_NULL_RETURN): Change from 91 to 71.
2793 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
2794 (PRED_LOOP_GUARD): Change from 66 to 73.
2796 2018-01-19 Martin Liska <mliska@suse.cz>
2798 * predict.c (predict_insn_def): Add new assert.
2799 (struct branch_predictor): Change type to signed integer.
2800 (test_prediction_value_range): Amend test to cover
2802 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
2803 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
2804 (PRED_LOOP_ITERATIONS_MAX): Likewise.
2805 (PRED_LOOP_IV_COMPARE): Likewise.
2806 * predict.h (PROB_UNINITIALIZED): Define new constant.
2808 2018-01-19 Martin Liska <mliska@suse.cz>
2810 * predict.c (dump_prediction): Add new format for
2811 analyze_brprob.py script which is enabled with -details
2813 * profile-count.h (precise_p): New function.
2815 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
2817 PR tree-optimization/83922
2818 * tree-vect-loop.c (vect_verify_full_masking): Return false if
2819 there are no statements that need masking.
2820 (vect_active_double_reduction_p): New function.
2821 (vect_analyze_loop_operations): Use it when handling phis that
2822 are not in the loop header.
2824 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
2826 PR tree-optimization/83914
2827 * tree-vect-loop.c (vectorizable_induction): Don't convert
2828 init_expr or apply the peeling adjustment for inductions
2829 that are nested within the vectorized loop.
2831 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2833 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
2836 2018-01-18 Jakub Jelinek <jakub@redhat.com>
2840 * function.h (gimplify_parameters): Add gimple_seq * argument.
2841 * function.c: Include gimple.h and options.h.
2842 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
2843 for the added local temporaries if needed.
2844 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
2845 if there are any parameter cleanups, wrap whole body into a
2846 try/finally with the cleanups.
2848 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
2851 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
2852 Use GET_MODE_CLASS for scalar floating point.
2854 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
2858 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
2859 Fix call of call_cgraph_insertion_hooks.
2861 2018-01-18 Martin Sebor <msebor@redhat.com>
2863 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
2865 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
2868 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
2871 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
2874 * common.opt: (-ffile-prefix-map): New option.
2875 * opts.c (common_handle_option): Defer it.
2876 * opts-global.c (handle_common_deferred_options): Handle it.
2877 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
2878 * file-prefix-map.h: New file.
2879 (remap_debug_filename, add_debug_prefix_map): ...here.
2880 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
2881 * final.c (debug_prefix_map, add_debug_prefix_map
2882 remap_debug_filename): Move to...
2883 * file-prefix-map.c: New file.
2884 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
2885 generalize, get rid of alloca(), use strrchr() instead of strchr().
2886 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
2887 Implement in terms of add_prefix_map().
2888 (remap_macro_filename, remap_debug_filename): Implement in term of
2890 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
2891 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
2892 * dbxout.c: Include file-prefix-map.h.
2893 * varasm.c: Likewise.
2894 * vmsdbgout.c: Likewise.
2895 * xcoffout.c: Likewise.
2896 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
2897 * doc/cppopts.texi (-fmacro-prefix-map): Document.
2898 * doc/invoke.texi (-ffile-prefix-map): Document.
2899 (-fdebug-prefix-map): Update description.
2901 2018-01-18 Martin Liska <mliska@suse.cz>
2903 * config/i386/i386.c (indirect_thunk_name): Document that also
2905 (output_indirect_thunk): Document why both instructions
2906 (pause and lfence) are generated.
2908 2018-01-18 Richard Biener <rguenther@suse.de>
2910 PR tree-optimization/83887
2911 * graphite-scop-detection.c
2912 (scop_detection::get_nearest_dom_with_single_entry): Remove.
2913 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
2914 (scop_detection::merge_sese): Re-implement with a flood-fill
2915 algorithm that properly finds a SESE region if it exists.
2917 2018-01-18 Jakub Jelinek <jakub@redhat.com>
2920 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
2921 pointer_diff optimizations use view_convert instead of convert.
2923 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2925 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2926 Generate different code for -mno-speculate-indirect-jumps.
2927 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2928 (*call_indirect_aix<mode>): Disable for
2929 -mno-speculate-indirect-jumps.
2930 (*call_indirect_aix<mode>_nospec): New define_insn.
2931 (*call_value_indirect_aix<mode>): Disable for
2932 -mno-speculate-indirect-jumps.
2933 (*call_value_indirect_aix<mode>_nospec): New define_insn.
2934 (*sibcall_nonlocal_sysv<mode>): Generate different code for
2935 -mno-speculate-indirect-jumps.
2936 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2938 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2940 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
2941 long double type, set the flags for noting the default long double
2942 type, even if we don't pass or return a long double type.
2944 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
2947 * ipa-inline.c (flatten_function): Do not overwrite final inlining
2950 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
2952 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
2953 support for merge[hl].
2954 (fold_mergehl_helper): New helper function.
2955 (tree-vector-builder.h): New #include for tree_vector_builder usage.
2956 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
2957 (altivec_vmrglw_direct): Add xxmrglw insn.
2959 2018-01-17 Andrew Waterman <andrew@sifive.com>
2961 * config/riscv/riscv.c (riscv_conditional_register_usage): If
2962 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
2964 2018-01-17 David Malcolm <dmalcolm@redhat.com>
2967 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
2968 call the lto_location_cache before reading the
2969 DECL_SOURCE_LOCATION of the types.
2971 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2972 Richard Sandiford <richard.sandiford@linaro.org>
2974 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
2975 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
2976 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
2977 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
2978 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
2980 * config/aarch64/constraints.md (aarch64_movti_operand):
2982 * config/aarch64/predicates.md (Uti): Add new constraint.
2984 2018-01-17 Carl Love <cel@us.ibm.com>
2986 * config/rs6000/vsx.md (define_expand xl_len_r,
2987 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
2988 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
2990 (define_expand, define_insn): Move the shift left from the
2991 define_insn to the define_expand for lxvl and stxvl instructions.
2992 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
2993 and XL_LEN_R definitions to PURE.
2995 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
2997 * config/i386/i386.c (indirect_thunk_name): Declare regno
2998 as unsigned int. Compare regno with INVALID_REGNUM.
2999 (output_indirect_thunk): Ditto.
3000 (output_indirect_thunk_function): Ditto.
3001 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
3002 in the call to output_indirect_thunk_function.
3004 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
3007 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
3008 rather than the size of inner_type to determine the stack slot size
3009 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
3011 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
3014 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
3017 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3019 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
3020 endian Linux systems to optionally enable multilibs for selecting
3021 the long double type if the user configured an explicit type.
3022 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
3023 have no long double multilibs if not defined.
3024 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
3025 warn if the user used -mabi={ieee,ibm}longdouble and we built
3026 multilibs for long double.
3027 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
3028 appropriate multilib option.
3029 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
3031 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
3032 for building long double multilibs.
3033 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
3035 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
3037 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
3040 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
3042 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
3045 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
3048 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
3051 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3053 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
3054 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
3056 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3058 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
3059 different rtl trees depending on TARGET_64BIT.
3060 (rs6000_gen_lvx): Likewise.
3062 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3064 * config/visium/visium.md (nop): Tweak comment.
3065 (hazard_nop): Likewise.
3067 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3069 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
3070 -mspeculate-indirect-jumps.
3071 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
3072 for -mno-speculate-indirect-jumps.
3073 (*call_indirect_elfv2<mode>_nospec): New define_insn.
3074 (*call_value_indirect_elfv2<mode>): Disable for
3075 -mno-speculate-indirect-jumps.
3076 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
3077 (indirect_jump): Emit different RTL for
3078 -mno-speculate-indirect-jumps.
3079 (*indirect_jump<mode>): Disable for
3080 -mno-speculate-indirect-jumps.
3081 (*indirect_jump<mode>_nospec): New define_insn.
3082 (tablejump): Emit different RTL for
3083 -mno-speculate-indirect-jumps.
3084 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
3085 (tablejumpsi_nospec): New define_expand.
3086 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
3087 (tablejumpdi_nospec): New define_expand.
3088 (*tablejump<mode>_internal1): Disable for
3089 -mno-speculate-indirect-jumps.
3090 (*tablejump<mode>_internal1_nospec): New define_insn.
3091 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
3094 2018-01-16 Artyom Skrobov tyomitch@gmail.com
3096 * caller-save.c (insert_save): Drop unnecessary parameter. All
3099 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3100 Richard Biener <rguenth@suse.de>
3103 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
3104 return early, inline manually is_gimple_sizepos. Make sure if we
3105 call gimplify_expr we don't end up with a gimple constant.
3106 * tree.c (variably_modified_type_p): Don't return true for
3107 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
3108 * gimplify.h (is_gimple_sizepos): Remove.
3110 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3112 PR tree-optimization/83857
3113 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
3114 vectorizable_live_operation for pure SLP statements.
3115 (vectorizable_live_operation): Handle PHIs.
3117 2018-01-16 Richard Biener <rguenther@suse.de>
3119 PR tree-optimization/83867
3120 * tree-vect-stmts.c (vect_transform_stmt): Precompute
3121 nested_in_vect_loop_p since the scalar stmt may get invalidated.
3123 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3126 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
3127 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
3128 If off is not INTEGER_CST, issue a may not be aligned warning
3129 rather than isn't aligned. Use isn%'t rather than isn't.
3130 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
3132 <case MULT_EXPR>: Improve the case when bottom and one of the
3133 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
3134 operand, in that case check if the other operand is multiple of
3135 bottom divided by the INTEGER_CST operand.
3137 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3140 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
3141 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
3142 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
3143 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
3144 * config/pa/pa.c (pa_function_arg_advance): Likewise.
3145 (pa_function_arg, pa_arg_partial_bytes): Likewise.
3146 (pa_function_arg_size): New function.
3148 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3150 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
3151 in a separate statement.
3153 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3155 PR tree-optimization/83847
3156 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
3157 group gathers and scatters.
3159 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3161 PR rtl-optimization/86620
3162 * params.def (max-sched-ready-insns): Bump minimum value to 1.
3164 PR rtl-optimization/83213
3165 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
3166 to last if both are JUMP_INSNs.
3168 PR tree-optimization/83843
3169 * gimple-ssa-store-merging.c
3170 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
3171 store_immediate_info for bswap/nop orig_stores.
3173 2018-01-15 Andrew Waterman <andrew@sifive.com>
3175 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
3177 <UDIV>: Increase cost if !TARGET_DIV.
3179 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
3181 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
3182 (define_attr "cr_logical_3op"): New.
3183 (cceq_ior_compare): Adjust.
3184 (cceq_ior_compare_complement): Adjust.
3185 (*cceq_rev_compare): Adjust.
3186 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
3187 (is_cracked_insn): Adjust.
3188 (insn_must_be_first_in_group): Adjust.
3189 * config/rs6000/40x.md: Adjust.
3190 * config/rs6000/440.md: Adjust.
3191 * config/rs6000/476.md: Adjust.
3192 * config/rs6000/601.md: Adjust.
3193 * config/rs6000/603.md: Adjust.
3194 * config/rs6000/6xx.md: Adjust.
3195 * config/rs6000/7450.md: Adjust.
3196 * config/rs6000/7xx.md: Adjust.
3197 * config/rs6000/8540.md: Adjust.
3198 * config/rs6000/cell.md: Adjust.
3199 * config/rs6000/e300c2c3.md: Adjust.
3200 * config/rs6000/e500mc.md: Adjust.
3201 * config/rs6000/e500mc64.md: Adjust.
3202 * config/rs6000/e5500.md: Adjust.
3203 * config/rs6000/e6500.md: Adjust.
3204 * config/rs6000/mpc.md: Adjust.
3205 * config/rs6000/power4.md: Adjust.
3206 * config/rs6000/power5.md: Adjust.
3207 * config/rs6000/power6.md: Adjust.
3208 * config/rs6000/power7.md: Adjust.
3209 * config/rs6000/power8.md: Adjust.
3210 * config/rs6000/power9.md: Adjust.
3211 * config/rs6000/rs64.md: Adjust.
3212 * config/rs6000/titan.md: Adjust.
3214 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3216 * config/i386/predicates.md (indirect_branch_operand): Rewrite
3217 ix86_indirect_branch_register logic.
3219 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3221 * config/i386/constraints.md (Bs): Update
3222 ix86_indirect_branch_register check. Don't check
3223 ix86_indirect_branch_register with GOT_memory_operand.
3225 * config/i386/predicates.md (GOT_memory_operand): Don't check
3226 ix86_indirect_branch_register here.
3227 (GOT32_symbol_operand): Likewise.
3229 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3231 * config/i386/predicates.md (constant_call_address_operand):
3232 Rewrite ix86_indirect_branch_register logic.
3233 (sibcall_insn_operand): Likewise.
3235 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3237 * config/i386/constraints.md (Bs): Replace
3238 ix86_indirect_branch_thunk_register with
3239 ix86_indirect_branch_register.
3241 * config/i386/i386.md (indirect_jump): Likewise.
3242 (tablejump): Likewise.
3243 (*sibcall_memory): Likewise.
3244 (*sibcall_value_memory): Likewise.
3245 Peepholes of indirect call and jump via memory: Likewise.
3246 * config/i386/i386.opt: Likewise.
3247 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3248 (GOT_memory_operand): Likewise.
3249 (call_insn_operand): Likewise.
3250 (sibcall_insn_operand): Likewise.
3251 (GOT32_symbol_operand): Likewise.
3253 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3256 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
3257 type rather than type addr's type points to.
3258 (expand_omp_atomic_mutex): Likewise.
3259 (expand_omp_atomic): Likewise.
3261 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3264 * config/i386/i386.c (output_indirect_thunk_function): Use
3265 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
3266 for __x86_return_thunk.
3268 2018-01-15 Richard Biener <rguenther@suse.de>
3271 * expmed.c (extract_bit_field_1): Fix typo.
3273 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3276 * config/arm/iterators.md (VF): New mode iterator.
3277 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
3278 Remove integer-related logic from pattern.
3279 (neon_vabd<mode>_3): Likewise.
3281 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3284 * common.opt (fstrict-overflow): No longer an alias.
3285 (fwrapv-pointer): New option.
3286 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
3287 also for pointer types based on flag_wrapv_pointer.
3288 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
3289 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
3290 opts->x_flag_wrapv got set.
3291 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
3292 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
3293 POINTER_TYPE_OVERFLOW_UNDEFINED.
3294 * match.pd: Likewise in address comparison pattern.
3295 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
3297 2018-01-15 Richard Biener <rguenther@suse.de>
3300 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
3301 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
3302 Reset type names to their identifier if their TYPE_DECL doesn't
3303 have linkage (and thus is used for ODR and devirt).
3304 (save_debug_info_for_decl): Remove.
3305 (save_debug_info_for_type): Likewise.
3306 (add_tree_to_fld_list): Adjust.
3307 * tree-pretty-print.c (dump_generic_node): Make dumping of
3308 type names more robust.
3310 2018-01-15 Richard Biener <rguenther@suse.de>
3312 * BASE-VER: Bump to 8.0.1.
3314 2018-01-14 Martin Sebor <msebor@redhat.com>
3317 * builtins.c (check_access): Avoid warning when the no-warning bit
3320 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
3322 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
3323 * ira-color (allocno_hard_regs_compare): Likewise.
3325 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
3328 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
3329 Use .pushsection/.popsection.
3331 2018-01-14 Martin Sebor <msebor@redhat.com>
3334 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
3336 2018-01-14 Jakub Jelinek <jakub@redhat.com>
3338 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
3339 entry from extra_headers.
3340 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
3341 extra_headers, make the list bitwise identical to the i?86-*-* one.
3343 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3345 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
3346 -mcmodel=large with -mindirect-branch=thunk,
3347 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
3348 -mfunction-return=thunk-extern.
3349 * doc/invoke.texi: Document -mcmodel=large is incompatible with
3350 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
3351 -mfunction-return=thunk and -mfunction-return=thunk-extern.
3353 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3355 * config/i386/i386.c (print_reg): Print the name of the full
3356 integer register without '%'.
3357 (ix86_print_operand): Handle 'V'.
3358 * doc/extend.texi: Document 'V' modifier.
3360 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3362 * config/i386/constraints.md (Bs): Disallow memory operand for
3363 -mindirect-branch-register.
3365 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3366 (GOT_memory_operand): Likewise.
3367 (call_insn_operand): Likewise.
3368 (sibcall_insn_operand): Likewise.
3369 (GOT32_symbol_operand): Likewise.
3370 * config/i386/i386.md (indirect_jump): Call convert_memory_address
3371 for -mindirect-branch-register.
3372 (tablejump): Likewise.
3373 (*sibcall_memory): Likewise.
3374 (*sibcall_value_memory): Likewise.
3375 Disallow peepholes of indirect call and jump via memory for
3376 -mindirect-branch-register.
3377 (*call_pop): Replace m with Bw.
3378 (*call_value_pop): Likewise.
3379 (*sibcall_pop_memory): Replace m with Bs.
3380 * config/i386/i386.opt (mindirect-branch-register): New option.
3381 * doc/invoke.texi: Document -mindirect-branch-register option.
3383 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3385 * config/i386/i386-protos.h (ix86_output_function_return): New.
3386 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
3387 set function_return_type.
3388 (indirect_thunk_name): Add ret_p to indicate thunk for function
3390 (output_indirect_thunk_function): Pass false to
3391 indirect_thunk_name.
3392 (ix86_output_indirect_branch_via_reg): Likewise.
3393 (ix86_output_indirect_branch_via_push): Likewise.
3394 (output_indirect_thunk_function): Create alias for function
3395 return thunk if regno < 0.
3396 (ix86_output_function_return): New function.
3397 (ix86_handle_fndecl_attribute): Handle function_return.
3398 (ix86_attribute_table): Add function_return.
3399 * config/i386/i386.h (machine_function): Add
3400 function_return_type.
3401 * config/i386/i386.md (simple_return_internal): Use
3402 ix86_output_function_return.
3403 (simple_return_internal_long): Likewise.
3404 * config/i386/i386.opt (mfunction-return=): New option.
3405 (indirect_branch): Mention -mfunction-return=.
3406 * doc/extend.texi: Document function_return function attribute.
3407 * doc/invoke.texi: Document -mfunction-return= option.
3409 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3411 * config/i386/i386-opts.h (indirect_branch): New.
3412 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
3413 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
3414 with local indirect jump when converting indirect call and jump.
3415 (ix86_set_indirect_branch_type): New.
3416 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
3417 (indirectlabelno): New.
3418 (indirect_thunk_needed): Likewise.
3419 (indirect_thunk_bnd_needed): Likewise.
3420 (indirect_thunks_used): Likewise.
3421 (indirect_thunks_bnd_used): Likewise.
3422 (INDIRECT_LABEL): Likewise.
3423 (indirect_thunk_name): Likewise.
3424 (output_indirect_thunk): Likewise.
3425 (output_indirect_thunk_function): Likewise.
3426 (ix86_output_indirect_branch_via_reg): Likewise.
3427 (ix86_output_indirect_branch_via_push): Likewise.
3428 (ix86_output_indirect_branch): Likewise.
3429 (ix86_output_indirect_jmp): Likewise.
3430 (ix86_code_end): Call output_indirect_thunk_function if needed.
3431 (ix86_output_call_insn): Call ix86_output_indirect_branch if
3433 (ix86_handle_fndecl_attribute): Handle indirect_branch.
3434 (ix86_attribute_table): Add indirect_branch.
3435 * config/i386/i386.h (machine_function): Add indirect_branch_type
3436 and has_local_indirect_jump.
3437 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
3439 (tablejump): Likewise.
3440 (*indirect_jump): Use ix86_output_indirect_jmp.
3441 (*tablejump_1): Likewise.
3442 (simple_return_indirect_internal): Likewise.
3443 * config/i386/i386.opt (mindirect-branch=): New option.
3444 (indirect_branch): New.
3447 (thunk-inline): Likewise.
3448 (thunk-extern): Likewise.
3449 * doc/extend.texi: Document indirect_branch function attribute.
3450 * doc/invoke.texi: Document -mindirect-branch= option.
3452 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
3455 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
3457 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
3459 * ipa-inline.c (want_inline_small_function_p): Return false if
3460 inlining has already failed with CIF_FINAL_ERROR.
3461 (update_caller_keys): Call want_inline_small_function_p before
3463 (update_callee_keys): Likewise.
3465 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3467 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
3469 (rs6000_quadword_masked_address_p): Likewise.
3470 (quad_aligned_load_p): Likewise.
3471 (quad_aligned_store_p): Likewise.
3472 (const_load_sequence_p): Add comment to describe the outer-most loop.
3473 (mimic_memory_attributes_and_flags): New function.
3474 (rs6000_gen_stvx): Likewise.
3475 (replace_swapped_aligned_store): Likewise.
3476 (rs6000_gen_lvx): Likewise.
3477 (replace_swapped_aligned_load): Likewise.
3478 (replace_swapped_load_constant): Capitalize argument name in
3479 comment describing this function.
3480 (rs6000_analyze_swaps): Add a third pass to search for vector loads
3481 and stores that access quad-word aligned addresses and replace
3482 with stvx or lvx instructions when appropriate.
3483 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
3484 New function prototype.
3485 (rs6000_quadword_masked_address_p): Likewise.
3486 (rs6000_gen_lvx): Likewise.
3487 (rs6000_gen_stvx): Likewise.
3488 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
3489 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
3490 when memory address is aligned.
3491 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
3492 this split to select lvx instruction when memory address is aligned.
3493 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
3494 instruction when memory address is aligned.
3495 (*vsx_le_perm_load_v16qi): Likewise.
3496 (four unnamed splitters): Modify to select the stvx instruction
3497 when memory is aligned.
3499 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
3501 * predict.c (determine_unlikely_bbs): Handle correctly BBs
3502 which appears in the queue multiple times.
3504 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3505 Alan Hayward <alan.hayward@arm.com>
3506 David Sherwood <david.sherwood@arm.com>
3508 * tree-vectorizer.h (vec_lower_bound): New structure.
3509 (_loop_vec_info): Add check_nonzero and lower_bounds.
3510 (LOOP_VINFO_CHECK_NONZERO): New macro.
3511 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
3512 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
3513 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
3514 fields. Make seg_len the distance travelled, not including the
3516 (dr_direction_indicator): Declare.
3517 (dr_zero_step_indicator): Likewise.
3518 (dr_known_forward_stride_p): Likewise.
3519 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
3521 (runtime_alias_check_p): Allow runtime alias checks with
3523 (operator ==): Compare access_size and align.
3524 (prune_runtime_alias_test_list): Rework for new distinction between
3525 the access_size and seg_len.
3526 (create_intersect_range_checks_index): Likewise. Cope with polynomial
3528 (get_segment_min_max): New function.
3529 (create_intersect_range_checks): Use it.
3530 (dr_step_indicator): New function.
3531 (dr_direction_indicator): Likewise.
3532 (dr_zero_step_indicator): Likewise.
3533 (dr_known_forward_stride_p): Likewise.
3534 * tree-loop-distribution.c (data_ref_segment_size): Return
3535 DR_STEP * (niters - 1).
3536 (compute_alias_check_pairs): Update call to the dr_with_seg_len
3538 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
3539 (vect_preserves_scalar_order_p): New function, split out from...
3540 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
3541 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
3542 (vect_vfa_access_size): New function.
3543 (vect_vfa_align): Likewise.
3544 (vect_compile_time_alias): Take access_size_a and access_b arguments.
3545 (dump_lower_bound): New function.
3546 (vect_check_lower_bound): Likewise.
3547 (vect_small_gap_p): Likewise.
3548 (vectorizable_with_step_bound_p): Likewise.
3549 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
3550 depencies if the vectorization factor is 1. Convert the checks
3551 for nonzero steps into checks on the bounds of DR_STEP. Try using
3552 a bunds check for variable steps if the minimum required step is
3553 relatively small. Update calls to the dr_with_seg_len
3554 constructor and to vect_compile_time_alias.
3555 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
3557 (vect_loop_versioning): Call it.
3558 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
3560 (vect_estimate_min_profitable_iters): Account for any bounds checks.
3562 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3563 Alan Hayward <alan.hayward@arm.com>
3564 David Sherwood <david.sherwood@arm.com>
3566 * doc/sourcebuild.texi (vect_scatter_store): Document.
3567 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
3569 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
3571 * genopinit.c (main): Add supports_vec_scatter_store and
3572 supports_vec_scatter_store_cached to target_optabs.
3573 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
3574 IFN_MASK_SCATTER_STORE.
3575 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
3577 * internal-fn.h (internal_store_fn_p): Declare.
3578 (internal_fn_stored_value_index): Likewise.
3579 * internal-fn.c (scatter_store_direct): New macro.
3580 (expand_scatter_store_optab_fn): New function.
3581 (direct_scatter_store_optab_supported_p): New macro.
3582 (internal_store_fn_p): New function.
3583 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
3584 IFN_MASK_SCATTER_STORE.
3585 (internal_fn_mask_index): Likewise.
3586 (internal_fn_stored_value_index): New function.
3587 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
3589 * optabs-query.h (supports_vec_scatter_store_p): Declare.
3590 * optabs-query.c (supports_vec_scatter_store_p): New function.
3591 * tree-vectorizer.h (vect_get_store_rhs): Declare.
3592 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
3593 true for scatter stores.
3594 (vect_gather_scatter_fn_p): Handle scatter stores too.
3595 (vect_check_gather_scatter): Consider using scatter stores if
3596 supports_vec_scatter_store_p.
3597 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
3599 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
3600 internal_fn_stored_value_index.
3601 (check_load_store_masking): Handle scatter stores too.
3602 (vect_get_store_rhs): Make public.
3603 (vectorizable_call): Use internal_store_fn_p.
3604 (vectorizable_store): Handle scatter store internal functions.
3605 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
3606 when deciding whether the end of the group has been reached.
3607 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
3608 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
3609 (mask_scatter_store<mode>): New insns.
3611 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3612 Alan Hayward <alan.hayward@arm.com>
3613 David Sherwood <david.sherwood@arm.com>
3615 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
3616 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
3617 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
3619 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
3620 Use vect_truncate_gather_scatter_offset if we can't treat the
3621 operation as a normal gather load or scatter store.
3622 (get_group_load_store_type): Take the gather_scatter_info
3623 as argument. Try using a gather load or scatter store for
3624 single-element groups.
3625 (get_load_store_type): Update calls to get_group_load_store_type
3626 and vect_use_strided_gather_scatters_p.
3628 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3629 Alan Hayward <alan.hayward@arm.com>
3630 David Sherwood <david.sherwood@arm.com>
3632 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
3633 optional tree argument.
3634 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
3636 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
3637 but continue to use the current value as a fallback.
3638 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
3639 to compare the updates.
3640 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
3641 (get_load_store_type): Use it when handling a strided access.
3642 (vect_get_strided_load_store_ops): New function.
3643 (vect_get_data_ptr_increment): Likewise.
3644 (vectorizable_load): Handle strided gather loads. Always pass
3645 a step to vect_create_data_ref_ptr and bump_vector_ptr.
3647 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3648 Alan Hayward <alan.hayward@arm.com>
3649 David Sherwood <david.sherwood@arm.com>
3651 * doc/md.texi (gather_load@var{m}): Document.
3652 (mask_gather_load@var{m}): Likewise.
3653 * genopinit.c (main): Add supports_vec_gather_load and
3654 supports_vec_gather_load_cached to target_optabs.
3655 * optabs-tree.c (init_tree_optimization_optabs): Use
3656 ggc_cleared_alloc to allocate target_optabs.
3657 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
3658 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
3660 * internal-fn.h (internal_load_fn_p): Declare.
3661 (internal_gather_scatter_fn_p): Likewise.
3662 (internal_fn_mask_index): Likewise.
3663 (internal_gather_scatter_fn_supported_p): Likewise.
3664 * internal-fn.c (gather_load_direct): New macro.
3665 (expand_gather_load_optab_fn): New function.
3666 (direct_gather_load_optab_supported_p): New macro.
3667 (direct_internal_fn_optab): New function.
3668 (internal_load_fn_p): Likewise.
3669 (internal_gather_scatter_fn_p): Likewise.
3670 (internal_fn_mask_index): Likewise.
3671 (internal_gather_scatter_fn_supported_p): Likewise.
3672 * optabs-query.c (supports_at_least_one_mode_p): New function.
3673 (supports_vec_gather_load_p): Likewise.
3674 * optabs-query.h (supports_vec_gather_load_p): Declare.
3675 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
3676 and memory_type field.
3677 (NUM_PATTERNS): Bump to 15.
3678 * tree-vect-data-refs.c: Include internal-fn.h.
3679 (vect_gather_scatter_fn_p): New function.
3680 (vect_describe_gather_scatter_call): Likewise.
3681 (vect_check_gather_scatter): Try using internal functions for
3682 gather loads. Recognize existing calls to a gather load function.
3683 (vect_analyze_data_refs): Consider using gather loads if
3684 supports_vec_gather_load_p.
3685 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
3686 (vect_get_gather_scatter_offset_type): Likewise.
3687 (vect_convert_mask_for_vectype): Likewise.
3688 (vect_add_conversion_to_patterm): Likewise.
3689 (vect_try_gather_scatter_pattern): Likewise.
3690 (vect_recog_gather_scatter_pattern): New pattern recognizer.
3691 (vect_vect_recog_func_ptrs): Add it.
3692 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
3693 internal_fn_mask_index and internal_gather_scatter_fn_p.
3694 (check_load_store_masking): Take the gather_scatter_info as an
3695 argument and handle gather loads.
3696 (vect_get_gather_scatter_ops): New function.
3697 (vectorizable_call): Check internal_load_fn_p.
3698 (vectorizable_load): Likewise. Handle gather load internal
3700 (vectorizable_store): Update call to check_load_store_masking.
3701 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
3702 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
3703 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
3704 (aarch64_gather_scale_operand_d): New predicates.
3705 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
3706 (mask_gather_load<mode>): New insns.
3708 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3709 Alan Hayward <alan.hayward@arm.com>
3710 David Sherwood <david.sherwood@arm.com>
3712 * optabs.def (fold_left_plus_optab): New optab.
3713 * doc/md.texi (fold_left_plus_@var{m}): Document.
3714 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
3715 * internal-fn.c (fold_left_direct): Define.
3716 (expand_fold_left_optab_fn): Likewise.
3717 (direct_fold_left_optab_supported_p): Likewise.
3718 * fold-const-call.c (fold_const_fold_left): New function.
3719 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
3720 * tree-parloops.c (valid_reduction_p): New function.
3721 (gather_scalar_reductions): Use it.
3722 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
3723 (vect_finish_replace_stmt): Declare.
3724 * tree-vect-loop.c (fold_left_reduction_fn): New function.
3725 (needs_fold_left_reduction_p): New function, split out from...
3726 (vect_is_simple_reduction): ...here. Accept reductions that
3727 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
3728 (vect_force_simple_reduction): Also store the reduction type in
3729 the assignment's STMT_VINFO_REDUC_TYPE.
3730 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
3731 (merge_with_identity): New function.
3732 (vect_expand_fold_left): Likewise.
3733 (vectorize_fold_left_reduction): Likewise.
3734 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
3735 scalar phi in place for it. Check for target support and reject
3736 cases that would reassociate the operation. Defer the transform
3737 phase to vectorize_fold_left_reduction.
3738 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
3739 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
3740 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
3742 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3744 * tree-if-conv.c (predicate_mem_writes): Remove redundant
3745 call to ifc_temp_var.
3747 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3748 Alan Hayward <alan.hayward@arm.com>
3749 David Sherwood <david.sherwood@arm.com>
3751 * target.def (legitimize_address_displacement): Take the original
3752 offset as a poly_int.
3753 * targhooks.h (default_legitimize_address_displacement): Update
3755 * targhooks.c (default_legitimize_address_displacement): Likewise.
3756 * doc/tm.texi: Regenerate.
3757 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
3758 as an argument, moving assert of ad->disp == ad->disp_term to...
3759 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
3760 Try calling targetm.legitimize_address_displacement before expanding
3761 the address rather than afterwards, and adjust for the new interface.
3762 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
3763 Match the new hook interface. Handle SVE addresses.
3764 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
3767 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3769 * Makefile.in (OBJS): Add early-remat.o.
3770 * target.def (select_early_remat_modes): New hook.
3771 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
3772 * doc/tm.texi: Regenerate.
3773 * targhooks.h (default_select_early_remat_modes): Declare.
3774 * targhooks.c (default_select_early_remat_modes): New function.
3775 * timevar.def (TV_EARLY_REMAT): New timevar.
3776 * passes.def (pass_early_remat): New pass.
3777 * tree-pass.h (make_pass_early_remat): Declare.
3778 * early-remat.c: New file.
3779 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
3781 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
3783 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3784 Alan Hayward <alan.hayward@arm.com>
3785 David Sherwood <david.sherwood@arm.com>
3787 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
3788 vfm1 with a bound_epilog parameter.
3789 (vect_do_peeling): Update calls accordingly, and move the prologue
3790 call earlier in the function. Treat the base bound_epilog as 0 for
3791 fully-masked loops and retain vf - 1 for other loops. Add 1 to
3792 this base when peeling for gaps.
3793 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
3794 with fully-masked loops.
3795 (vect_estimate_min_profitable_iters): Handle the single peeled
3796 iteration in that case.
3798 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3799 Alan Hayward <alan.hayward@arm.com>
3800 David Sherwood <david.sherwood@arm.com>
3802 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
3803 single-element interleaving even if the size is not a power of 2.
3804 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
3805 accesses for single-element interleaving if the group size is
3808 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3809 Alan Hayward <alan.hayward@arm.com>
3810 David Sherwood <david.sherwood@arm.com>
3812 * doc/md.texi (fold_extract_last_@var{m}): Document.
3813 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
3814 * optabs.def (fold_extract_last_optab): New optab.
3815 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
3816 * internal-fn.c (fold_extract_direct): New macro.
3817 (expand_fold_extract_optab_fn): Likewise.
3818 (direct_fold_extract_optab_supported_p): Likewise.
3819 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
3820 * tree-vect-loop.c (vect_model_reduction_cost): Handle
3821 EXTRACT_LAST_REDUCTION.
3822 (get_initial_def_for_reduction): Do not create an initial vector
3823 for EXTRACT_LAST_REDUCTION reductions.
3824 (vectorizable_reduction): Leave the scalar phi in place for
3825 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
3826 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
3827 epilogue code for EXTRACT_LAST_REDUCTION and defer the
3828 transform phase to vectorizable_condition.
3829 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
3831 (vect_finish_stmt_generation): ...here.
3832 (vect_finish_replace_stmt): New function.
3833 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
3834 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
3836 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
3838 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3839 Alan Hayward <alan.hayward@arm.com>
3840 David Sherwood <david.sherwood@arm.com>
3842 * doc/md.texi (extract_last_@var{m}): Document.
3843 * optabs.def (extract_last_optab): New optab.
3844 * internal-fn.def (EXTRACT_LAST): New internal function.
3845 * internal-fn.c (cond_unary_direct): New macro.
3846 (expand_cond_unary_optab_fn): Likewise.
3847 (direct_cond_unary_optab_supported_p): Likewise.
3848 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
3849 loops using EXTRACT_LAST.
3850 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
3851 (extract_last_<mode>): ...this optab.
3852 (vec_extract<mode><Vel>): Update accordingly.
3854 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3855 Alan Hayward <alan.hayward@arm.com>
3856 David Sherwood <david.sherwood@arm.com>
3858 * target.def (empty_mask_is_expensive): New hook.
3859 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
3860 * doc/tm.texi: Regenerate.
3861 * targhooks.h (default_empty_mask_is_expensive): Declare.
3862 * targhooks.c (default_empty_mask_is_expensive): New function.
3863 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
3864 if the target says that empty masks are expensive.
3865 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
3867 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
3869 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3870 Alan Hayward <alan.hayward@arm.com>
3871 David Sherwood <david.sherwood@arm.com>
3873 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
3874 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
3875 (vect_use_loop_mask_for_alignment_p): New function.
3876 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
3877 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
3878 niters_skip argument. Make sure that the first niters_skip elements
3879 of the first iteration are inactive.
3880 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
3881 Update call to vect_set_loop_masks_directly.
3882 (get_misalign_in_elems): New function, split out from...
3883 (vect_gen_prolog_loop_niters): ...here.
3884 (vect_update_init_of_dr): Take a code argument that specifies whether
3885 the adjustment should be added or subtracted.
3886 (vect_update_init_of_drs): Likewise.
3887 (vect_prepare_for_masked_peels): New function.
3888 (vect_do_peeling): Skip prologue peeling if we're using a mask
3889 instead. Update call to vect_update_inits_of_drs.
3890 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
3892 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
3893 alignment. Do not include the number of peeled iterations in
3894 the minimum threshold in that case.
3895 (vectorizable_induction): Adjust the start value down by
3896 LOOP_VINFO_MASK_SKIP_NITERS iterations.
3897 (vect_transform_loop): Call vect_prepare_for_masked_peels.
3898 Take the number of skipped iterations into account when calculating
3900 * tree-vect-stmts.c (vect_gen_while_not): New function.
3902 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3903 Alan Hayward <alan.hayward@arm.com>
3904 David Sherwood <david.sherwood@arm.com>
3906 * doc/sourcebuild.texi (vect_fully_masked): Document.
3907 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
3909 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
3911 (vect_analyze_loop_2): ...here. Don't check the vectorization
3912 factor against the number of loop iterations if the loop is
3915 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3916 Alan Hayward <alan.hayward@arm.com>
3917 David Sherwood <david.sherwood@arm.com>
3919 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
3920 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
3921 (dump_groups): Update accordingly.
3922 (iv_use::mem_type): New member variable.
3923 (address_p): New function.
3924 (record_use): Add a mem_type argument and initialize the new
3926 (record_group_use): Add a mem_type argument. Use address_p.
3927 Remove obsolete null checks of base_object. Update call to record_use.
3928 (find_interesting_uses_op): Update call to record_group_use.
3929 (find_interesting_uses_cond): Likewise.
3930 (find_interesting_uses_address): Likewise.
3931 (get_mem_type_for_internal_fn): New function.
3932 (find_address_like_use): Likewise.
3933 (find_interesting_uses_stmt): Try find_address_like_use before
3934 calling find_interesting_uses_op.
3935 (addr_offset_valid_p): Use the iv mem_type field as the type
3936 of the addressed memory.
3937 (add_autoinc_candidates): Likewise.
3938 (get_address_cost): Likewise.
3939 (split_small_address_groups_p): Use address_p.
3940 (split_address_groups): Likewise.
3941 (add_iv_candidate_for_use): Likewise.
3942 (autoinc_possible_for_pair): Likewise.
3943 (rewrite_groups): Likewise.
3944 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
3945 (determine_group_iv_cost): Update after split of USE_ADDRESS.
3946 (get_alias_ptr_type_for_ptr_address): New function.
3947 (rewrite_use_address): Rewrite address uses in calls that were
3948 identified by find_address_like_use.
3950 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3951 Alan Hayward <alan.hayward@arm.com>
3952 David Sherwood <david.sherwood@arm.com>
3954 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
3956 * gimple-expr.h (is_gimple_addressable: Likewise.
3957 * gimple-expr.c (is_gimple_address): Likewise.
3958 * internal-fn.c (expand_call_mem_ref): New function.
3959 (expand_mask_load_optab_fn): Use it.
3960 (expand_mask_store_optab_fn): Likewise.
3962 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3963 Alan Hayward <alan.hayward@arm.com>
3964 David Sherwood <david.sherwood@arm.com>
3966 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
3967 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
3968 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
3969 (cond_umax@var{mode}): Document.
3970 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
3971 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
3972 (cond_umin_optab, cond_umax_optab): New optabs.
3973 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
3974 (COND_IOR, COND_XOR): New internal functions.
3975 * internal-fn.h (get_conditional_internal_fn): Declare.
3976 * internal-fn.c (cond_binary_direct): New macro.
3977 (expand_cond_binary_optab_fn): Likewise.
3978 (direct_cond_binary_optab_supported_p): Likewise.
3979 (get_conditional_internal_fn): New function.
3980 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
3981 Cope with reduction statements that are vectorized as calls rather
3983 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
3984 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
3985 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
3986 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
3987 (UNSPEC_COND_EOR): New unspecs.
3988 (optab): Add mappings for them.
3989 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
3990 (sve_int_op, sve_fp_op): New int attributes.
3992 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3993 Alan Hayward <alan.hayward@arm.com>
3994 David Sherwood <david.sherwood@arm.com>
3996 * optabs.def (while_ult_optab): New optab.
3997 * doc/md.texi (while_ult@var{m}@var{n}): Document.
3998 * internal-fn.def (WHILE_ULT): New internal function.
3999 * internal-fn.h (direct_internal_fn_supported_p): New override
4000 that takes two types as argument.
4001 * internal-fn.c (while_direct): New macro.
4002 (expand_while_optab_fn): New function.
4003 (convert_optab_supported_p): Likewise.
4004 (direct_while_optab_supported_p): New macro.
4005 * wide-int.h (wi::udiv_ceil): New function.
4006 * tree-vectorizer.h (rgroup_masks): New structure.
4007 (vec_loop_masks): New typedef.
4008 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
4010 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
4011 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
4012 (vect_max_vf): New function.
4013 (slpeel_make_loop_iterate_ntimes): Delete.
4014 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
4015 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
4016 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
4017 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
4018 internal-fn.h, stor-layout.h and optabs-query.h.
4019 (vect_set_loop_mask): New function.
4020 (add_preheader_seq): Likewise.
4021 (add_header_seq): Likewise.
4022 (interleave_supported_p): Likewise.
4023 (vect_maybe_permute_loop_masks): Likewise.
4024 (vect_set_loop_masks_directly): Likewise.
4025 (vect_set_loop_condition_masked): Likewise.
4026 (vect_set_loop_condition_unmasked): New function, split out from
4027 slpeel_make_loop_iterate_ntimes.
4028 (slpeel_make_loop_iterate_ntimes): Rename to..
4029 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
4030 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
4031 (vect_do_peeling): Update call accordingly.
4032 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
4034 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4035 mask_compare_type, can_fully_mask_p and fully_masked_p.
4036 (release_vec_loop_masks): New function.
4037 (_loop_vec_info): Use it to free the loop masks.
4038 (can_produce_all_loop_masks_p): New function.
4039 (vect_get_max_nscalars_per_iter): Likewise.
4040 (vect_verify_full_masking): Likewise.
4041 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
4042 retries, and free the mask rgroups before retrying. Check loop-wide
4043 reasons for disallowing fully-masked loops. Make the final decision
4044 about whether use a fully-masked loop or not.
4045 (vect_estimate_min_profitable_iters): Do not assume that peeling
4046 for the number of iterations will be needed for fully-masked loops.
4047 (vectorizable_reduction): Disable fully-masked loops.
4048 (vectorizable_live_operation): Likewise.
4049 (vect_halve_mask_nunits): New function.
4050 (vect_double_mask_nunits): Likewise.
4051 (vect_record_loop_mask): Likewise.
4052 (vect_get_loop_mask): Likewise.
4053 (vect_transform_loop): Handle the case in which the final loop
4054 iteration might handle a partial vector. Call vect_set_loop_condition
4055 instead of slpeel_make_loop_iterate_ntimes.
4056 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
4057 (check_load_store_masking): New function.
4058 (prepare_load_store_mask): Likewise.
4059 (vectorizable_store): Handle fully-masked loops.
4060 (vectorizable_load): Likewise.
4061 (supportable_widening_operation): Use vect_halve_mask_nunits for
4063 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
4064 (vect_gen_while): New function.
4065 * config/aarch64/aarch64.md (umax<mode>3): New expander.
4066 (aarch64_uqdec<mode>): New insn.
4068 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4069 Alan Hayward <alan.hayward@arm.com>
4070 David Sherwood <david.sherwood@arm.com>
4072 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
4073 (reduc_xor_scal_optab): New optabs.
4074 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
4075 (reduc_xor_scal_@var{m}): Document.
4076 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
4077 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
4079 * fold-const-call.c (fold_const_call): Handle them.
4080 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
4081 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
4082 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
4083 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
4084 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
4085 (UNSPEC_XORV): New unspecs.
4086 (optab): Add entries for them.
4087 (BITWISEV): New int iterator.
4088 (bit_reduc_op): New int attributes.
4090 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4091 Alan Hayward <alan.hayward@arm.com>
4092 David Sherwood <david.sherwood@arm.com>
4094 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
4095 * internal-fn.def (VEC_SHL_INSERT): New internal function.
4096 * optabs.def (vec_shl_insert_optab): New optab.
4097 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
4098 (duplicate_and_interleave): Likewise.
4099 * tree-vect-loop.c: Include internal-fn.h.
4100 (neutral_op_for_slp_reduction): New function, split out from
4101 get_initial_defs_for_reduction.
4102 (get_initial_def_for_reduction): Handle option 2 for variable-length
4103 vectors by loading the neutral value into a vector and then shifting
4104 the initial value into element 0.
4105 (get_initial_defs_for_reduction): Replace the code argument with
4106 the neutral value calculated by neutral_op_for_slp_reduction.
4107 Use gimple_build_vector for constant-length vectors.
4108 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
4109 but the first group_size elements have a neutral value.
4110 Use duplicate_and_interleave otherwise.
4111 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
4112 Update call to get_initial_defs_for_reduction. Handle SLP
4113 reductions for variable-length vectors by creating one vector
4114 result for each scalar result, with the elements associated
4115 with other scalar results stubbed out with the neutral value.
4116 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
4117 Require IFN_VEC_SHL_INSERT for double reductions on
4118 variable-length vectors, or SLP reductions that have
4119 a neutral value. Require can_duplicate_and_interleave_p
4120 support for variable-length unchained SLP reductions if there
4121 is no neutral value, such as for MIN/MAX reductions. Also require
4122 the number of vector elements to be a multiple of the number of
4123 SLP statements when doing variable-length unchained SLP reductions.
4124 Update call to vect_create_epilog_for_reduction.
4125 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
4126 and remove initial values.
4127 (duplicate_and_interleave): Make public.
4128 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
4129 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
4131 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4132 Alan Hayward <alan.hayward@arm.com>
4133 David Sherwood <david.sherwood@arm.com>
4135 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
4136 (can_duplicate_and_interleave_p): New function.
4137 (vect_get_and_check_slp_defs): Take the vector of statements
4138 rather than just the current one. Remove excess parentheses.
4139 Restriction rejectinon of vect_constant_def and vect_external_def
4140 for variable-length vectors to boolean types, or types for which
4141 can_duplicate_and_interleave_p is false.
4142 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
4143 (duplicate_and_interleave): New function.
4144 (vect_get_constant_vectors): Use gimple_build_vector for
4145 constant-length vectors and suitable variable-length constant
4146 vectors. Use duplicate_and_interleave for other variable-length
4147 vectors. Don't defer the update when inserting new statements.
4149 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4150 Alan Hayward <alan.hayward@arm.com>
4151 David Sherwood <david.sherwood@arm.com>
4153 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
4154 min_profitable_iters doesn't go negative.
4156 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4157 Alan Hayward <alan.hayward@arm.com>
4158 David Sherwood <david.sherwood@arm.com>
4160 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
4161 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
4162 * optabs.def (vec_mask_load_lanes_optab): New optab.
4163 (vec_mask_store_lanes_optab): Likewise.
4164 * internal-fn.def (MASK_LOAD_LANES): New internal function.
4165 (MASK_STORE_LANES): Likewise.
4166 * internal-fn.c (mask_load_lanes_direct): New macro.
4167 (mask_store_lanes_direct): Likewise.
4168 (expand_mask_load_optab_fn): Handle masked operations.
4169 (expand_mask_load_lanes_optab_fn): New macro.
4170 (expand_mask_store_optab_fn): Handle masked operations.
4171 (expand_mask_store_lanes_optab_fn): New macro.
4172 (direct_mask_load_lanes_optab_supported_p): Likewise.
4173 (direct_mask_store_lanes_optab_supported_p): Likewise.
4174 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
4176 (vect_load_lanes_supported): Likewise.
4177 * tree-vect-data-refs.c (strip_conversion): New function.
4178 (can_group_stmts_p): Likewise.
4179 (vect_analyze_data_ref_accesses): Use it instead of checking
4180 for a pair of assignments.
4181 (vect_store_lanes_supported): Take a masked_p parameter.
4182 (vect_load_lanes_supported): Likewise.
4183 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
4184 vect_store_lanes_supported and vect_load_lanes_supported.
4185 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4186 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
4187 parameter. Don't allow gaps for masked accesses.
4188 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
4189 and vect_load_lanes_supported.
4190 (get_load_store_type): Take a masked_p parameter and update
4191 call to get_group_load_store_type.
4192 (vectorizable_store): Update call to get_load_store_type.
4193 Handle IFN_MASK_STORE_LANES.
4194 (vectorizable_load): Update call to get_load_store_type.
4195 Handle IFN_MASK_LOAD_LANES.
4197 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4198 Alan Hayward <alan.hayward@arm.com>
4199 David Sherwood <david.sherwood@arm.com>
4201 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
4203 * config/aarch64/aarch64-protos.h
4204 (aarch64_sve_struct_memory_operand_p): Declare.
4205 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
4206 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
4207 (VPRED, vpred): Handle SVE structure modes.
4208 * config/aarch64/constraints.md (Utx): New constraint.
4209 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
4210 (aarch64_sve_struct_nonimmediate_operand): New predicates.
4211 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
4212 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
4213 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
4214 structure modes. Split into pieces after RA.
4215 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
4216 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
4218 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
4219 SVE structure modes.
4220 (aarch64_classify_address): Likewise.
4221 (sizetochar): Move earlier in file.
4222 (aarch64_print_operand): Handle SVE register lists.
4223 (aarch64_array_mode): New function.
4224 (aarch64_sve_struct_memory_operand_p): Likewise.
4225 (TARGET_ARRAY_MODE): Redefine.
4227 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4228 Alan Hayward <alan.hayward@arm.com>
4229 David Sherwood <david.sherwood@arm.com>
4231 * target.def (array_mode): New target hook.
4232 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
4233 * doc/tm.texi: Regenerate.
4234 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
4235 * hooks.c (hook_optmode_mode_uhwi_none): New function.
4236 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
4238 * stor-layout.c (mode_for_array): Likewise. Support polynomial
4241 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4242 Alan Hayward <alan.hayward@arm.com>
4243 David Sherwood <david.sherwood@arm.com>
4245 * fold-const.c (fold_binary_loc): Check the argument types
4246 rather than the result type when testing for a vector operation.
4248 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4250 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
4251 * doc/tm.texi: Regenerate.
4253 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4254 Alan Hayward <alan.hayward@arm.com>
4255 David Sherwood <david.sherwood@arm.com>
4257 * doc/invoke.texi (-msve-vector-bits=): Document new option.
4258 (sve): Document new AArch64 extension.
4259 * doc/md.texi (w): Extend the description of the AArch64
4260 constraint to include SVE vectors.
4261 (Upl, Upa): Document new AArch64 predicate constraints.
4262 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
4264 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
4265 (msve-vector-bits=): New option.
4266 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
4267 SVE when these are disabled.
4268 (sve): New extension.
4269 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
4270 modes. Adjust their number of units based on aarch64_sve_vg.
4271 (MAX_BITSIZE_MODE_ANY_MODE): Define.
4272 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
4273 aarch64_addr_query_type.
4274 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
4275 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
4276 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
4277 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
4278 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
4279 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
4280 (aarch64_simd_imm_zero_p): Delete.
4281 (aarch64_check_zero_based_sve_index_immediate): Declare.
4282 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4283 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4284 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4285 (aarch64_sve_float_mul_immediate_p): Likewise.
4286 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4288 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
4289 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
4290 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
4291 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
4292 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
4293 (aarch64_regmode_natural_size): Likewise.
4294 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
4295 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
4297 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
4298 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
4299 for VG and the SVE predicate registers.
4300 (V_ALIASES): Add a "z"-prefixed alias.
4301 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
4302 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
4303 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
4304 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
4305 (REG_CLASS_NAMES): Add entries for them.
4306 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
4307 and the predicate registers.
4308 (aarch64_sve_vg): Declare.
4309 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
4310 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
4311 (REGMODE_NATURAL_SIZE): Define.
4312 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4314 * config/aarch64/aarch64.c: Include cfgrtl.h.
4315 (simd_immediate_info): Add a constructor for series vectors,
4316 and an associated step field.
4317 (aarch64_sve_vg): New variable.
4318 (aarch64_dbx_register_number): Handle VG and the predicate registers.
4319 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
4320 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
4321 (VEC_ANY_DATA, VEC_STRUCT): New constants.
4322 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
4323 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
4324 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
4325 (aarch64_get_mask_mode): New functions.
4326 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
4327 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
4328 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
4329 predicate modes and predicate registers. Explicitly restrict
4330 GPRs to modes of 16 bytes or smaller. Only allow FP registers
4331 to store a vector mode if it is recognized by
4332 aarch64_classify_vector_mode.
4333 (aarch64_regmode_natural_size): New function.
4334 (aarch64_hard_regno_caller_save_mode): Return the original mode
4336 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
4337 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
4338 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
4339 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
4341 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
4342 does not overlap dest if the function is frame-related. Handle
4344 (aarch64_split_add_offset): New function.
4345 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
4346 them aarch64_add_offset.
4347 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
4348 and update call to aarch64_sub_sp.
4349 (aarch64_add_cfa_expression): New function.
4350 (aarch64_expand_prologue): Pass extra temporary registers to the
4351 functions above. Handle the case in which we need to emit new
4352 DW_CFA_expressions for registers that were originally saved
4353 relative to the stack pointer, but now have to be expressed
4354 relative to the frame pointer.
4355 (aarch64_output_mi_thunk): Pass extra temporary registers to the
4357 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
4358 IP0 and IP1 values for SVE frames.
4359 (aarch64_expand_vec_series): New function.
4360 (aarch64_expand_sve_widened_duplicate): Likewise.
4361 (aarch64_expand_sve_const_vector): Likewise.
4362 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
4363 Handle SVE constants. Use emit_move_insn to move a force_const_mem
4364 into the register, rather than emitting a SET directly.
4365 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
4366 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
4367 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
4368 (offset_9bit_signed_scaled_p): New functions.
4369 (aarch64_replicate_bitmask_imm): New function.
4370 (aarch64_bitmask_imm): Use it.
4371 (aarch64_cannot_force_const_mem): Reject expressions involving
4372 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
4373 (aarch64_classify_index): Handle SVE indices, by requiring
4374 a plain register index with a scale that matches the element size.
4375 (aarch64_classify_address): Handle SVE addresses. Assert that
4376 the mode of the address is VOIDmode or an integer mode.
4377 Update call to aarch64_classify_symbol.
4378 (aarch64_classify_symbolic_expression): Update call to
4379 aarch64_classify_symbol.
4380 (aarch64_const_vec_all_in_range_p): New function.
4381 (aarch64_print_vector_float_operand): Likewise.
4382 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
4383 "vN" for FP registers with SVE modes. Handle (const ...) vectors
4384 and the FP immediates 1.0 and 0.5.
4385 (aarch64_print_address_internal): Handle SVE addresses.
4386 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
4387 (aarch64_regno_regclass): Handle predicate registers.
4388 (aarch64_secondary_reload): Handle big-endian reloads of SVE
4390 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
4391 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
4392 (aarch64_convert_sve_vector_bits): New function.
4393 (aarch64_override_options): Use it to handle -msve-vector-bits=.
4394 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4396 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
4397 Handle SVE vector and predicate modes. Accept VL-based constants
4398 that need only one temporary register, and VL offsets that require
4399 no temporary registers.
4400 (aarch64_conditional_register_usage): Mark the predicate registers
4401 as fixed if SVE isn't available.
4402 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
4403 Return true for SVE vector and predicate modes.
4404 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
4405 rather than an unsigned int. Handle SVE modes.
4406 (aarch64_preferred_simd_mode): Update call accordingly. Handle
4408 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
4410 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4411 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4412 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4413 (aarch64_sve_float_mul_immediate_p): New functions.
4414 (aarch64_sve_valid_immediate): New function.
4415 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
4416 Explicitly reject structure modes. Check for INDEX constants.
4417 Handle PTRUE and PFALSE constants.
4418 (aarch64_check_zero_based_sve_index_immediate): New function.
4419 (aarch64_simd_imm_zero_p): Delete.
4420 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
4421 vector modes. Accept constants in the range of CNT[BHWD].
4422 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
4423 ask for an Advanced SIMD mode.
4424 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
4425 (aarch64_simd_vector_alignment): Handle SVE predicates.
4426 (aarch64_vectorize_preferred_vector_alignment): New function.
4427 (aarch64_simd_vector_alignment_reachable): Use it instead of
4429 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
4430 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
4432 (MAX_VECT_LEN): Delete.
4433 (expand_vec_perm_d): Add a vec_flags field.
4434 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
4435 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
4436 (aarch64_evpc_ext): Don't apply a big-endian lane correction
4438 (aarch64_evpc_rev): Rename to...
4439 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
4440 (aarch64_evpc_rev_global): New function.
4441 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
4442 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
4444 (aarch64_evpc_sve_tbl): New function.
4445 (aarch64_expand_vec_perm_const_1): Update after rename of
4446 aarch64_evpc_rev. Handle SVE permutes too, trying
4447 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
4448 than aarch64_evpc_tbl.
4449 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
4450 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
4451 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
4452 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
4453 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
4454 (aarch64_expand_sve_vcond): New functions.
4455 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
4456 of aarch64_vector_mode_p.
4457 (aarch64_dwarf_poly_indeterminate_value): New function.
4458 (aarch64_compute_pressure_classes): Likewise.
4459 (aarch64_can_change_mode_class): Likewise.
4460 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
4461 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
4462 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
4463 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
4464 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
4465 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
4466 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
4467 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
4469 (Dn, Dl, Dr): Accept const as well as const_vector.
4470 (Dz): Likewise. Compare against CONST0_RTX.
4471 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
4472 of "vector" where appropriate.
4473 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
4474 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
4475 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
4476 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
4477 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
4478 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
4479 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
4480 (v_int_equiv): Extend to SVE modes.
4481 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
4483 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
4484 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
4485 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
4486 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
4487 (SVE_COND_FP_CMP): New int iterators.
4488 (perm_hilo): Handle the new unpack unspecs.
4489 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
4491 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
4492 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
4493 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
4494 (aarch64_equality_operator, aarch64_constant_vector_operand)
4495 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
4496 (aarch64_sve_nonimmediate_operand): Likewise.
4497 (aarch64_sve_general_operand): Likewise.
4498 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
4499 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
4500 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
4501 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
4502 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
4503 (aarch64_sve_float_arith_immediate): Likewise.
4504 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
4505 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
4506 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
4507 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
4508 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
4509 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
4510 (aarch64_sve_float_arith_operand): Likewise.
4511 (aarch64_sve_float_arith_with_sub_operand): Likewise.
4512 (aarch64_sve_float_mul_operand): Likewise.
4513 (aarch64_sve_vec_perm_operand): Likewise.
4514 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
4515 (aarch64_mov_operand): Accept const_poly_int and const_vector.
4516 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
4517 as well as const_vector.
4518 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
4519 in file. Use CONST0_RTX and CONSTM1_RTX.
4520 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
4521 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
4522 Use aarch64_simd_imm_zero.
4523 * config/aarch64/aarch64-sve.md: New file.
4524 * config/aarch64/aarch64.md: Include it.
4525 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
4526 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
4527 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
4528 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
4529 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
4530 (sve): New attribute.
4531 (enabled): Disable instructions with the sve attribute unless
4533 (movqi, movhi): Pass CONST_POLY_INT operaneds through
4534 aarch64_expand_mov_immediate.
4535 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
4536 CNT[BHSD] immediates.
4537 (movti): Split CONST_POLY_INT moves into two halves.
4538 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
4539 Split additions that need a temporary here if the destination
4540 is the stack pointer.
4541 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
4542 (*add<mode>3_poly_1): New instruction.
4543 (set_clobber_cc): New expander.
4545 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4547 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
4548 parameter and use it instead of GET_MODE_SIZE (innermode). Use
4549 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
4550 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
4551 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
4552 Change innermode from fixed_mode_size to machine_mode.
4553 (simplify_subreg): Update call accordingly. Handle a constant-sized
4554 subreg of a variable-length CONST_VECTOR.
4556 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4557 Alan Hayward <alan.hayward@arm.com>
4558 David Sherwood <david.sherwood@arm.com>
4560 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
4561 (add_offset_to_base): New function, split out from...
4562 (create_mem_ref): ...here. When handling a scale other than 1,
4563 check first whether the address is valid without the offset.
4564 Add it into the base if so, leaving the index and scale as-is.
4566 2018-01-12 Jakub Jelinek <jakub@redhat.com>
4569 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
4570 fold_for_warn before checking if arg2 is INTEGER_CST.
4572 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
4574 * config/rs6000/predicates.md (load_multiple_operation): Delete.
4575 (store_multiple_operation): Delete.
4576 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
4577 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
4578 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
4579 guarded by TARGET_STRING.
4580 (rs6000_output_load_multiple): Delete.
4581 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
4582 OPTION_MASK_STRING / TARGET_STRING handling.
4583 (print_operand) <'N', 'O'>: Add comment that these are unused now.
4584 (const rs6000_opt_masks) <"string">: Change mask to 0.
4585 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
4586 (MASK_STRING): Delete.
4587 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
4589 (load_multiple): Delete.
4596 (store_multiple): Delete.
4603 (movmemsi_8reg): Delete.
4604 (corresponding unnamed define_insn): Delete.
4605 (movmemsi_6reg): Delete.
4606 (corresponding unnamed define_insn): Delete.
4607 (movmemsi_4reg): Delete.
4608 (corresponding unnamed define_insn): Delete.
4609 (movmemsi_2reg): Delete.
4610 (corresponding unnamed define_insn): Delete.
4611 (movmemsi_1reg): Delete.
4612 (corresponding unnamed define_insn): Delete.
4613 * config/rs6000/rs6000.opt (mno-string): New.
4614 (mstring): Replace by deprecation warning stub.
4615 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
4617 2018-01-12 Jakub Jelinek <jakub@redhat.com>
4619 * regrename.c (regrename_do_replace): If replacing the same
4620 reg multiple times, try to reuse last created gen_raw_REG.
4623 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
4624 main to workaround a bug in GDB.
4626 2018-01-12 Tom de Vries <tom@codesourcery.com>
4629 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
4631 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
4633 PR rtl-optimization/80481
4634 * ira-color.c (get_cap_member): New function.
4635 (allocnos_conflict_by_live_ranges_p): Use it.
4636 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
4637 (setup_slot_coalesced_allocno_live_ranges): Ditto.
4639 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
4642 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
4643 (*saddl_se_1): Ditto.
4645 (*ssubl_se_1): Ditto.
4647 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4649 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
4650 rather than wi::to_widest for DR_INITs.
4651 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
4652 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
4653 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
4655 (vect_analyze_group_access_1): Note that here.
4657 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4659 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
4660 polynomial type sizes.
4662 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4664 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
4665 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
4666 (gimple_add_tmp_var): Likewise.
4668 2018-01-12 Martin Liska <mliska@suse.cz>
4670 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
4671 (gimple_alloc_sizes): Likewise.
4672 (dump_gimple_statistics): Use PRIu64 in printf format.
4673 * gimple.h: Change uint64_t to int.
4675 2018-01-12 Martin Liska <mliska@suse.cz>
4677 * tree-core.h: Use uint64_t instead of int.
4678 * tree.c (tree_node_counts): Likewise.
4679 (tree_node_sizes): Likewise.
4680 (dump_tree_statistics): Use PRIu64 in printf format.
4682 2018-01-12 Martin Liska <mliska@suse.cz>
4684 * Makefile.in: As qsort_chk is implemented in vec.c, add
4685 vec.o to linkage of gencfn-macros.
4686 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
4687 passing the info to record_node_allocation_statistics.
4688 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
4690 * ggc-common.c (struct ggc_usage): Add operator== and use
4691 it in operator< and compare function.
4692 * mem-stats.h (struct mem_usage): Likewise.
4693 * vec.c (struct vec_usage): Remove operator< and compare
4694 function. Can be simply inherited.
4696 2018-01-12 Martin Jambor <mjambor@suse.cz>
4699 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
4700 * tree-ssa-math-opts.c: Include domwalk.h.
4701 (convert_mult_to_fma_1): New function.
4702 (fma_transformation_info): New type.
4703 (fma_deferring_state): Likewise.
4704 (cancel_fma_deferring): New function.
4705 (result_of_phi): Likewise.
4706 (last_fma_candidate_feeds_initial_phi): Likewise.
4707 (convert_mult_to_fma): Added deferring logic, split actual
4708 transformation to convert_mult_to_fma_1.
4709 (math_opts_dom_walker): New type.
4710 (math_opts_dom_walker::after_dom_children): New method, body moved
4711 here from pass_optimize_widening_mul::execute, added deferring logic
4713 (pass_optimize_widening_mul::execute): Moved most of code to
4714 math_opts_dom_walker::after_dom_children.
4715 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
4716 * config/i386/i386.c (ix86_option_override_internal): Added
4717 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
4719 2018-01-12 Richard Biener <rguenther@suse.de>
4722 * dwarf2out.c (gen_variable_die): Do not reset old_die for
4723 inline instance vars.
4725 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
4728 * config/rx/rx.c (rx_is_restricted_memory_address):
4731 2018-01-12 Richard Biener <rguenther@suse.de>
4733 PR tree-optimization/80846
4734 * target.def (split_reduction): New target hook.
4735 * targhooks.c (default_split_reduction): New function.
4736 * targhooks.h (default_split_reduction): Declare.
4737 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
4738 target requests first reduce vectors by combining low and high
4740 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
4741 (get_vectype_for_scalar_type_and_size): Export.
4742 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
4743 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
4744 * doc/tm.texi: Regenerate.
4745 * config/i386/i386.c (ix86_split_reduction): Implement
4746 TARGET_VECTORIZE_SPLIT_REDUCTION.
4748 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4751 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
4752 in PIC mode except for TARGET_VXWORKS_RTP.
4753 * config/sparc/sparc.c: Include cfgrtl.h.
4754 (TARGET_INIT_PIC_REG): Define.
4755 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
4756 (sparc_pic_register_p): New predicate.
4757 (sparc_legitimate_address_p): Use it.
4758 (sparc_legitimize_pic_address): Likewise.
4759 (sparc_delegitimize_address): Likewise.
4760 (sparc_mode_dependent_address_p): Likewise.
4761 (gen_load_pcrel_sym): Remove 4th parameter.
4762 (load_got_register): Adjust call to above. Remove obsolete stuff.
4763 (sparc_expand_prologue): Do not call load_got_register here.
4764 (sparc_flat_expand_prologue): Likewise.
4765 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
4766 (sparc_use_pseudo_pic_reg): New function.
4767 (sparc_init_pic_reg): Likewise.
4768 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
4769 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
4771 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
4773 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
4774 Add item for branch_cost.
4776 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4778 PR rtl-optimization/83565
4779 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
4780 not extend the result to a larger mode for rotate operations.
4781 (num_sign_bit_copies1): Likewise.
4783 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4786 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
4788 Use values-Xc.o for -pedantic.
4789 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
4791 2018-01-12 Martin Liska <mliska@suse.cz>
4794 * ipa-devirt.c (final_warning_record::grow_type_warnings):
4796 (possible_polymorphic_call_targets): Use it.
4797 (ipa_devirt): Likewise.
4799 2018-01-12 Martin Liska <mliska@suse.cz>
4801 * profile-count.h (enum profile_quality): Use 0 as invalid
4802 enum value of profile_quality.
4804 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
4806 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
4807 -mext-string options.
4809 2018-01-12 Richard Biener <rguenther@suse.de>
4811 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
4812 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
4813 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
4815 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
4817 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
4819 * configure.ac (--with-long-double-format): Add support for the
4820 configuration option to change the default long double format on
4822 * config.gcc (powerpc*-linux*-*): Likewise.
4823 * configure: Regenerate.
4824 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
4825 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
4826 used without modification.
4828 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4830 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
4831 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
4832 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
4833 MISC_BUILTIN_SPEC_BARRIER.
4834 (rs6000_init_builtins): Likewise.
4835 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
4837 (speculation_barrier): New define_insn.
4838 * doc/extend.texi: Document __builtin_speculation_barrier.
4840 2018-01-11 Jakub Jelinek <jakub@redhat.com>
4843 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
4844 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
4845 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
4847 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
4848 integral modes instead of "ss" and "sd".
4849 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
4850 vectors with 32-bit and 64-bit elements.
4851 (vecdupssescalarmodesuffix): New mode attribute.
4852 (vec_dup<mode>): Use it.
4854 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
4857 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
4858 frame if argument is passed on stack.
4860 2018-01-11 Jakub Jelinek <jakub@redhat.com>
4863 * ree.c (combine_reaching_defs): Optimize also
4864 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
4865 reg2=any_extend(exp); reg1=reg2;, formatting fix.
4867 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
4870 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
4872 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
4875 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
4876 after they are computed.
4878 2018-01-11 Bin Cheng <bin.cheng@arm.com>
4880 PR tree-optimization/83695
4881 * gimple-loop-linterchange.cc
4882 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
4883 reset cached scev information after interchange.
4884 (pass_linterchange::execute): Remove call to scev_reset_htab.
4886 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4888 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
4889 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
4890 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
4891 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
4892 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
4893 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
4894 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
4895 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
4896 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
4897 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
4898 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
4899 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
4900 (V_lane_reg): Likewise.
4901 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
4903 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
4904 (vfmal_lane_low<mode>_intrinsic,
4905 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
4906 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
4907 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
4908 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
4909 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
4910 vfmsl_lane_high<mode>_intrinsic): New define_insns.
4912 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4914 * config/arm/arm-cpus.in (fp16fml): New feature.
4915 (ALL_SIMD): Add fp16fml.
4916 (armv8.2-a): Add fp16fml as an option.
4917 (armv8.3-a): Likewise.
4918 (armv8.4-a): Add fp16fml as part of fp16.
4919 * config/arm/arm.h (TARGET_FP16FML): Define.
4920 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
4922 * config/arm/arm-modes.def (V2HF): Define.
4923 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
4924 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
4925 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
4926 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
4927 vfmsl_low, vfmsl_high): New set of builtins.
4928 * config/arm/iterators.md (PLUSMINUS): New code iterator.
4929 (vfml_op): New code attribute.
4930 (VFMLHALVES): New int iterator.
4931 (VFML, VFMLSEL): New mode attributes.
4932 (V_reg): Define mapping for V2HF.
4933 (V_hi, V_lo): New mode attributes.
4934 (VF_constraint): Likewise.
4935 (vfml_half, vfml_half_selector): New int attributes.
4936 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
4938 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
4939 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
4941 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
4942 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
4943 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
4944 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
4946 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
4947 Document new effective target and option set.
4949 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4951 * config/arm/arm-cpus.in (armv8_4): New feature.
4952 (ARMv8_4a): New fgroup.
4953 (armv8.4-a): New arch.
4954 * config/arm/arm-tables.opt: Regenerate.
4955 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
4956 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
4957 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
4958 Add matching rules for -march=armv8.4-a and extensions.
4959 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
4961 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
4964 * config/rx/rx.md (BW): New mode attribute.
4965 (sync_lock_test_and_setsi): Add mode suffix to insn output.
4967 2018-01-11 Richard Biener <rguenther@suse.de>
4969 PR tree-optimization/83435
4970 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
4971 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
4972 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
4974 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
4975 Alan Hayward <alan.hayward@arm.com>
4976 David Sherwood <david.sherwood@arm.com>
4978 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
4980 (aarch64_classify_address): Initialize it. Track polynomial offsets.
4981 (aarch64_print_address_internal): Use it to check for a zero offset.
4983 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
4984 Alan Hayward <alan.hayward@arm.com>
4985 David Sherwood <david.sherwood@arm.com>
4987 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
4988 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
4989 Return a poly_int64 rather than a HOST_WIDE_INT.
4990 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
4991 rather than a HOST_WIDE_INT.
4992 * config/aarch64/aarch64.h (aarch64_frame): Protect with
4993 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
4994 hard_fp_offset, frame_size, initial_adjust, callee_offset and
4995 final_offset from HOST_WIDE_INT to poly_int64.
4996 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
4997 to_constant when getting the number of units in an Advanced SIMD
4999 (aarch64_builtin_vectorized_function): Check for a constant number
5001 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
5003 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
5004 attribute instead of GET_MODE_NUNITS.
5005 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
5006 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
5007 GET_MODE_SIZE for fixed-size registers.
5008 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
5009 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
5010 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
5011 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
5012 (aarch64_print_operand, aarch64_print_address_internal)
5013 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
5014 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
5015 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
5016 Handle polynomial GET_MODE_SIZE.
5017 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
5018 wider than SImode without modification.
5019 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
5020 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
5021 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
5022 passing and returning SVE modes.
5023 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
5024 rather than GEN_INT.
5025 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
5026 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
5027 (aarch64_allocate_and_probe_stack_space): Likewise.
5028 (aarch64_layout_frame): Cope with polynomial offsets.
5029 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
5030 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
5032 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
5033 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
5034 poly_int64 rather than a HOST_WIDE_INT.
5035 (aarch64_get_separate_components, aarch64_process_components)
5036 (aarch64_expand_prologue, aarch64_expand_epilogue)
5037 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
5038 (aarch64_anchor_offset): New function, split out from...
5039 (aarch64_legitimize_address): ...here.
5040 (aarch64_builtin_vectorization_cost): Handle polynomial
5041 TYPE_VECTOR_SUBPARTS.
5042 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
5044 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
5045 number of elements from the PARALLEL rather than the mode.
5046 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
5047 rather than GET_MODE_BITSIZE.
5048 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
5049 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
5050 (aarch64_expand_vec_perm_const_1): Handle polynomial
5051 d->perm.length () and d->perm elements.
5052 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
5053 Apply to_constant to d->perm elements.
5054 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
5055 polynomial CONST_VECTOR_NUNITS.
5056 (aarch64_move_pointer): Take amount as a poly_int64 rather
5058 (aarch64_progress_pointer): Avoid temporary variable.
5059 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
5060 the mode attribute instead of GET_MODE.
5062 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5063 Alan Hayward <alan.hayward@arm.com>
5064 David Sherwood <david.sherwood@arm.com>
5066 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
5067 x exists before using it.
5068 (aarch64_add_constant_internal): Rename to...
5069 (aarch64_add_offset_1): ...this. Replace regnum with separate
5070 src and dest rtxes. Handle the case in which they're different,
5071 including when the offset is zero. Replace scratchreg with an rtx.
5072 Use 2 additions if there is no spare register into which we can
5073 move a 16-bit constant.
5074 (aarch64_add_constant): Delete.
5075 (aarch64_add_offset): Replace reg with separate src and dest
5076 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
5077 Use aarch64_add_offset_1.
5078 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
5079 an rtx rather than an int. Take the delta as a poly_int64
5080 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
5081 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
5082 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
5083 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
5084 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
5086 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
5087 aarch64_add_constant.
5089 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5091 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
5092 Use scalar_float_mode.
5094 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5096 * config/aarch64/aarch64-simd.md
5097 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
5098 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
5099 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
5100 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
5101 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
5102 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
5103 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
5104 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
5105 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
5106 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
5108 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5111 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
5112 targ_options->x_arm_arch_string is non NULL.
5114 2018-01-11 Tamar Christina <tamar.christina@arm.com>
5116 * config/aarch64/aarch64.h
5117 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
5119 2018-01-11 Sudakshina Das <sudi.das@arm.com>
5122 * expmed.c (emit_store_flag_force): Swap if const op0
5123 and change VOIDmode to mode of op0.
5125 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5127 PR rtl-optimization/83761
5128 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
5129 than bytes to mode_for_size.
5131 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5134 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
5135 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
5138 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5141 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
5142 when in layout mode.
5143 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
5144 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
5147 2018-01-10 Michael Collison <michael.collison@arm.com>
5149 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
5150 * config/aarch64/aarch64-option-extension.def: Add
5151 AARCH64_OPT_EXTENSION of 'fp16fml'.
5152 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5153 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
5154 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
5155 * config/aarch64/constraints.md (Ui7): New constraint.
5156 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
5157 (VFMLA_SEL_W): Ditto.
5160 (VFMLA16_LOW): New int iterator.
5161 (VFMLA16_HIGH): Ditto.
5162 (UNSPEC_FMLAL): New unspec.
5163 (UNSPEC_FMLSL): Ditto.
5164 (UNSPEC_FMLAL2): Ditto.
5165 (UNSPEC_FMLSL2): Ditto.
5166 (f16mac): New code attribute.
5167 * config/aarch64/aarch64-simd-builtins.def
5168 (aarch64_fmlal_lowv2sf): Ditto.
5169 (aarch64_fmlsl_lowv2sf): Ditto.
5170 (aarch64_fmlalq_lowv4sf): Ditto.
5171 (aarch64_fmlslq_lowv4sf): Ditto.
5172 (aarch64_fmlal_highv2sf): Ditto.
5173 (aarch64_fmlsl_highv2sf): Ditto.
5174 (aarch64_fmlalq_highv4sf): Ditto.
5175 (aarch64_fmlslq_highv4sf): Ditto.
5176 (aarch64_fmlal_lane_lowv2sf): Ditto.
5177 (aarch64_fmlsl_lane_lowv2sf): Ditto.
5178 (aarch64_fmlal_laneq_lowv2sf): Ditto.
5179 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
5180 (aarch64_fmlalq_lane_lowv4sf): Ditto.
5181 (aarch64_fmlsl_lane_lowv4sf): Ditto.
5182 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
5183 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
5184 (aarch64_fmlal_lane_highv2sf): Ditto.
5185 (aarch64_fmlsl_lane_highv2sf): Ditto.
5186 (aarch64_fmlal_laneq_highv2sf): Ditto.
5187 (aarch64_fmlsl_laneq_highv2sf): Ditto.
5188 (aarch64_fmlalq_lane_highv4sf): Ditto.
5189 (aarch64_fmlsl_lane_highv4sf): Ditto.
5190 (aarch64_fmlalq_laneq_highv4sf): Ditto.
5191 (aarch64_fmlsl_laneq_highv4sf): Ditto.
5192 * config/aarch64/aarch64-simd.md:
5193 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
5194 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5195 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
5196 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5197 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
5198 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
5199 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
5200 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
5201 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
5202 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
5203 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
5204 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
5205 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
5206 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
5207 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
5208 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
5209 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
5210 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
5211 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
5212 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
5213 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
5214 (vfmlsl_low_u32): Ditto.
5215 (vfmlalq_low_u32): Ditto.
5216 (vfmlslq_low_u32): Ditto.
5217 (vfmlal_high_u32): Ditto.
5218 (vfmlsl_high_u32): Ditto.
5219 (vfmlalq_high_u32): Ditto.
5220 (vfmlslq_high_u32): Ditto.
5221 (vfmlal_lane_low_u32): Ditto.
5222 (vfmlsl_lane_low_u32): Ditto.
5223 (vfmlal_laneq_low_u32): Ditto.
5224 (vfmlsl_laneq_low_u32): Ditto.
5225 (vfmlalq_lane_low_u32): Ditto.
5226 (vfmlslq_lane_low_u32): Ditto.
5227 (vfmlalq_laneq_low_u32): Ditto.
5228 (vfmlslq_laneq_low_u32): Ditto.
5229 (vfmlal_lane_high_u32): Ditto.
5230 (vfmlsl_lane_high_u32): Ditto.
5231 (vfmlal_laneq_high_u32): Ditto.
5232 (vfmlsl_laneq_high_u32): Ditto.
5233 (vfmlalq_lane_high_u32): Ditto.
5234 (vfmlslq_lane_high_u32): Ditto.
5235 (vfmlalq_laneq_high_u32): Ditto.
5236 (vfmlslq_laneq_high_u32): Ditto.
5237 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
5238 (AARCH64_FL_FOR_ARCH8_4): New.
5239 (AARCH64_ISA_F16FML): New ISA flag.
5240 (TARGET_F16FML): New feature flag for fp16fml.
5241 (doc/invoke.texi): Document new fp16fml option.
5243 2018-01-10 Michael Collison <michael.collison@arm.com>
5245 * config/aarch64/aarch64-builtins.c:
5246 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
5247 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5248 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
5249 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
5250 (AARCH64_ISA_SHA3): New ISA flag.
5251 (TARGET_SHA3): New feature flag for sha3.
5252 * config/aarch64/iterators.md (sha512_op): New int attribute.
5253 (CRYPTO_SHA512): New int iterator.
5254 (UNSPEC_SHA512H): New unspec.
5255 (UNSPEC_SHA512H2): Ditto.
5256 (UNSPEC_SHA512SU0): Ditto.
5257 (UNSPEC_SHA512SU1): Ditto.
5258 * config/aarch64/aarch64-simd-builtins.def
5259 (aarch64_crypto_sha512hqv2di): New builtin.
5260 (aarch64_crypto_sha512h2qv2di): Ditto.
5261 (aarch64_crypto_sha512su0qv2di): Ditto.
5262 (aarch64_crypto_sha512su1qv2di): Ditto.
5263 (aarch64_eor3qv8hi): Ditto.
5264 (aarch64_rax1qv2di): Ditto.
5265 (aarch64_xarqv2di): Ditto.
5266 (aarch64_bcaxqv8hi): Ditto.
5267 * config/aarch64/aarch64-simd.md:
5268 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
5269 (aarch64_crypto_sha512su0qv2di): Ditto.
5270 (aarch64_crypto_sha512su1qv2di): Ditto.
5271 (aarch64_eor3qv8hi): Ditto.
5272 (aarch64_rax1qv2di): Ditto.
5273 (aarch64_xarqv2di): Ditto.
5274 (aarch64_bcaxqv8hi): Ditto.
5275 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
5276 (vsha512h2q_u64): Ditto.
5277 (vsha512su0q_u64): Ditto.
5278 (vsha512su1q_u64): Ditto.
5279 (veor3q_u16): Ditto.
5280 (vrax1q_u64): Ditto.
5282 (vbcaxq_u16): Ditto.
5283 * config/arm/types.md (crypto_sha512): New type attribute.
5284 (crypto_sha3): Ditto.
5285 (doc/invoke.texi): Document new sha3 option.
5287 2018-01-10 Michael Collison <michael.collison@arm.com>
5289 * config/aarch64/aarch64-builtins.c:
5290 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
5291 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5292 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
5293 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
5294 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
5295 (AARCH64_ISA_SM4): New ISA flag.
5296 (TARGET_SM4): New feature flag for sm4.
5297 * config/aarch64/aarch64-simd-builtins.def
5298 (aarch64_sm3ss1qv4si): Ditto.
5299 (aarch64_sm3tt1aq4si): Ditto.
5300 (aarch64_sm3tt1bq4si): Ditto.
5301 (aarch64_sm3tt2aq4si): Ditto.
5302 (aarch64_sm3tt2bq4si): Ditto.
5303 (aarch64_sm3partw1qv4si): Ditto.
5304 (aarch64_sm3partw2qv4si): Ditto.
5305 (aarch64_sm4eqv4si): Ditto.
5306 (aarch64_sm4ekeyqv4si): Ditto.
5307 * config/aarch64/aarch64-simd.md:
5308 (aarch64_sm3ss1qv4si): Ditto.
5309 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
5310 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
5311 (aarch64_sm4eqv4si): Ditto.
5312 (aarch64_sm4ekeyqv4si): Ditto.
5313 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
5314 (sm3part_op): Ditto.
5315 (CRYPTO_SM3TT): Ditto.
5316 (CRYPTO_SM3PART): Ditto.
5317 (UNSPEC_SM3SS1): New unspec.
5318 (UNSPEC_SM3TT1A): Ditto.
5319 (UNSPEC_SM3TT1B): Ditto.
5320 (UNSPEC_SM3TT2A): Ditto.
5321 (UNSPEC_SM3TT2B): Ditto.
5322 (UNSPEC_SM3PARTW1): Ditto.
5323 (UNSPEC_SM3PARTW2): Ditto.
5324 (UNSPEC_SM4E): Ditto.
5325 (UNSPEC_SM4EKEY): Ditto.
5326 * config/aarch64/constraints.md (Ui2): New constraint.
5327 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
5328 * config/arm/types.md (crypto_sm3): New type attribute.
5329 (crypto_sm4): Ditto.
5330 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
5331 (vsm3tt1aq_u32): Ditto.
5332 (vsm3tt1bq_u32): Ditto.
5333 (vsm3tt2aq_u32): Ditto.
5334 (vsm3tt2bq_u32): Ditto.
5335 (vsm3partw1q_u32): Ditto.
5336 (vsm3partw2q_u32): Ditto.
5337 (vsm4eq_u32): Ditto.
5338 (vsm4ekeyq_u32): Ditto.
5339 (doc/invoke.texi): Document new sm4 option.
5341 2018-01-10 Michael Collison <michael.collison@arm.com>
5343 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
5344 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
5345 (AARCH64_FL_FOR_ARCH8_4): New.
5346 (AARCH64_FL_V8_4): New flag.
5347 (doc/invoke.texi): Document new armv8.4-a option.
5349 2018-01-10 Michael Collison <michael.collison@arm.com>
5351 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5352 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
5353 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
5354 * config/aarch64/aarch64-option-extension.def: Add
5355 AARCH64_OPT_EXTENSION of 'sha2'.
5356 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
5357 (crypto): Disable sha2 and aes if crypto disabled.
5358 (crypto): Enable aes and sha2 if enabled.
5359 (simd): Disable sha2 and aes if simd disabled.
5360 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
5362 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
5363 (TARGET_SHA2): New feature flag for sha2.
5364 (TARGET_AES): New feature flag for aes.
5365 * config/aarch64/aarch64-simd.md:
5366 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
5367 conditional on TARGET_AES.
5368 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
5369 (aarch64_crypto_sha1hsi): Make pattern conditional
5371 (aarch64_crypto_sha1hv4si): Ditto.
5372 (aarch64_be_crypto_sha1hv4si): Ditto.
5373 (aarch64_crypto_sha1su1v4si): Ditto.
5374 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
5375 (aarch64_crypto_sha1su0v4si): Ditto.
5376 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
5377 (aarch64_crypto_sha256su0v4si): Ditto.
5378 (aarch64_crypto_sha256su1v4si): Ditto.
5379 (doc/invoke.texi): Document new aes and sha2 options.
5381 2018-01-10 Martin Sebor <msebor@redhat.com>
5383 PR tree-optimization/83781
5384 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
5387 2018-01-11 Martin Sebor <msebor@gmail.com>
5388 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5390 PR tree-optimization/83501
5391 PR tree-optimization/81703
5393 * tree-ssa-strlen.c (get_string_cst): Rename...
5394 (get_string_len): ...to this. Handle global constants.
5395 (handle_char_store): Adjust.
5397 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
5398 Jim Wilson <jimw@sifive.com>
5400 * config/riscv/riscv-protos.h (riscv_output_return): New.
5401 * config/riscv/riscv.c (struct machine_function): New naked_p field.
5402 (riscv_attribute_table, riscv_output_return),
5403 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
5404 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
5405 (riscv_compute_frame_info): Only compute frame->mask if not a naked
5407 (riscv_expand_prologue): Add early return for naked function.
5408 (riscv_expand_epilogue): Likewise.
5409 (riscv_function_ok_for_sibcall): Return false for naked function.
5410 (riscv_set_current_function): New.
5411 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
5412 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
5413 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
5414 * doc/extend.texi (RISC-V Function Attributes): New.
5416 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
5418 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
5419 check for 128-bit long double before checking TCmode.
5420 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
5421 128-bit long doubles before checking TFmode or TCmode.
5422 (FLOAT128_IBM_P): Likewise.
5424 2018-01-10 Martin Sebor <msebor@redhat.com>
5426 PR tree-optimization/83671
5427 * builtins.c (c_strlen): Unconditionally return zero for the empty
5429 Use -Warray-bounds for warnings.
5430 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
5431 for non-constant array indices with COMPONENT_REF, arrays of
5432 arrays, and pointers to arrays.
5433 (gimple_fold_builtin_strlen): Determine and set length range for
5434 non-constant character arrays.
5436 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
5439 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
5442 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
5444 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
5446 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5449 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
5450 VECTOR_MEM_ALTIVEC_OR_VSX_P.
5451 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
5452 indexed_or_indirect_operand predicate.
5453 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
5454 (*vsx_le_perm_load_v8hi): Likewise.
5455 (*vsx_le_perm_load_v16qi): Likewise.
5456 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
5457 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
5458 (*vsx_le_perm_store_v8hi): Likewise.
5459 (*vsx_le_perm_store_v16qi): Likewise.
5460 (eight unnamed splitters): Likewise.
5462 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5464 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
5465 * config/rs6000/emmintrin.h: Likewise.
5466 * config/rs6000/mmintrin.h: Likewise.
5467 * config/rs6000/xmmintrin.h: Likewise.
5469 2018-01-10 David Malcolm <dmalcolm@redhat.com>
5472 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
5474 * tree.c (tree_nop_conversion): Return true for location wrapper
5476 (maybe_wrap_with_location): New function.
5477 (selftest::check_strip_nops): New function.
5478 (selftest::test_location_wrappers): New function.
5479 (selftest::tree_c_tests): Call it.
5480 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
5481 (maybe_wrap_with_location): New decl.
5482 (EXPR_LOCATION_WRAPPER_P): New macro.
5483 (location_wrapper_p): New inline function.
5484 (tree_strip_any_location_wrapper): New inline function.
5486 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
5489 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
5490 stack_realign_offset for the largest alignment of stack slot
5492 (ix86_find_max_used_stack_alignment): New function.
5493 (ix86_finalize_stack_frame_flags): Use it. Set
5494 max_used_stack_alignment if we don't realign stack.
5495 * config/i386/i386.h (machine_function): Add
5496 max_used_stack_alignment.
5498 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
5500 * config/arm/arm.opt (-mbranch-cost): New option.
5501 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
5504 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
5507 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
5508 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
5510 2018-01-10 Richard Biener <rguenther@suse.de>
5513 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
5514 early out so it also covers the case where we have a non-NULL
5517 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
5519 PR tree-optimization/83753
5520 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
5521 for non-strided grouped accesses if the number of elements is 1.
5523 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5526 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
5527 * i386.h (TARGET_USE_GATHER): Define.
5528 * x86-tune.def (X86_TUNE_USE_GATHER): New.
5530 2018-01-10 Martin Liska <mliska@suse.cz>
5533 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
5534 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
5536 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
5537 CLEANUP_NO_PARTITIONING is not set.
5539 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
5541 * doc/rtl.texi: Remove documentation of (const ...) wrappers
5542 for vectors, as a partial revert of r254296.
5543 * rtl.h (const_vec_p): Delete.
5544 (const_vec_duplicate_p): Don't test for vector CONSTs.
5545 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
5546 * expmed.c (make_tree): Likewise.
5549 * common.md (E, F): Use CONSTANT_P instead of checking for
5551 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
5552 checking for CONST_VECTOR.
5554 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5557 * predict.c (force_edge_cold): Handle in more sane way edges
5560 2018-01-09 Carl Love <cel@us.ibm.com>
5562 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
5564 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
5565 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
5566 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
5567 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
5568 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
5569 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
5570 * config/rs6000/rs6000-protos.h: Add extern defition for
5571 rs6000_generate_float2_double_code.
5572 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
5574 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
5575 (float2_v2df): Add define_expand.
5577 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
5580 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
5581 op_mode in the force_to_mode call.
5583 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5585 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
5586 instead of checking each element individually.
5587 (aarch64_evpc_uzp): Likewise.
5588 (aarch64_evpc_zip): Likewise.
5589 (aarch64_evpc_ext): Likewise.
5590 (aarch64_evpc_rev): Likewise.
5591 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
5592 instead of checking each element individually. Return true without
5594 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
5595 whether all selected elements come from the same input, instead of
5596 checking each element individually. Remove calls to gen_rtx_REG,
5597 start_sequence and end_sequence and instead assert that no rtl is
5600 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5602 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
5603 order of HIGH and CONST checks.
5605 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5607 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
5608 if the destination isn't an SSA_NAME.
5610 2018-01-09 Richard Biener <rguenther@suse.de>
5612 PR tree-optimization/83668
5613 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
5615 (canonicalize_loop_form): ... here, renamed from ...
5616 (canonicalize_loop_closed_ssa_form): ... this and amended to
5617 swap successor edges for loop exit blocks to make us use
5618 the RPO order we need for initial schedule generation.
5620 2018-01-09 Joseph Myers <joseph@codesourcery.com>
5622 PR tree-optimization/64811
5623 * match.pd: When optimizing comparisons with Inf, avoid
5624 introducing or losing exceptions from comparisons with NaN.
5626 2018-01-09 Martin Liska <mliska@suse.cz>
5629 * asan.c (shadow_mem_size): Add gcc_assert.
5631 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
5633 Don't save registers in main().
5636 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
5637 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
5638 * config/avr/avr.c (avr_set_current_function): Don't error if
5639 naked, OS_task or OS_main are specified at the same time.
5640 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
5642 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
5644 * common/config/avr/avr-common.c (avr_option_optimization_table):
5645 Switch on -mmain-is-OS_task for optimizing compilations.
5647 2018-01-09 Richard Biener <rguenther@suse.de>
5649 PR tree-optimization/83572
5650 * graphite.c: Include cfganal.h.
5651 (graphite_transform_loops): Connect infinite loops to exit
5652 and remove fake edges at the end.
5654 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5656 * ipa-inline.c (edge_badness): Revert accidental checkin.
5658 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5661 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
5662 symbols; not inline clones.
5664 2018-01-09 Jakub Jelinek <jakub@redhat.com>
5667 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
5668 hard registers. Formatting fixes.
5670 PR preprocessor/83722
5671 * gcc.c (try_generate_repro): Pass
5672 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
5673 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
5676 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
5677 Kito Cheng <kito.cheng@gmail.com>
5679 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
5680 (riscv_leaf_function_p): Delete.
5681 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
5683 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5685 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
5687 (do_ifelse): New function.
5688 (do_isel): New function.
5689 (do_sub3): New function.
5690 (do_add3): New function.
5691 (do_load_mask_compare): New function.
5692 (do_overlap_load_compare): New function.
5693 (expand_compare_loop): New function.
5694 (expand_block_compare): Call expand_compare_loop() when appropriate.
5695 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
5697 (-mblock-compare-inline-loop-limit): New option.
5699 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5702 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
5703 Reverse order of second and third operands in first alternative.
5704 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
5705 of first and second elements in UNSPEC_VPERMR vector.
5706 (altivec_expand_vec_perm_le): Likewise.
5708 2018-01-08 Jeff Law <law@redhat.com>
5710 PR rtl-optimizatin/81308
5711 * tree-switch-conversion.c (cfg_altered): New file scoped static.
5712 (process_switch): If group_case_labels makes a change, then set
5714 (pass_convert_switch::execute): If a switch is converted, then
5715 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
5717 PR rtl-optimization/81308
5718 * recog.c (split_all_insns): Conditionally cleanup the CFG after
5721 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
5723 PR target/83663 - Revert r255946
5724 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
5725 generation for cases where splatting a value is not useful.
5726 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
5727 across a vec_duplicate and a paradoxical subreg forming a vector
5728 mode to a vec_concat.
5730 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5732 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
5733 -march=armv8.3-a variants.
5734 * config/arm/t-multilib: Likewise.
5735 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
5737 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5739 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
5741 (cceq_ior_compare_complement): Give it a name so I can use it, and
5742 change boolean_or_operator predicate to boolean_operator so it can
5743 be used to generate a crand.
5744 (eqne): New code iterator.
5745 (bd/bd_neg): New code_attrs.
5746 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
5747 a single define_insn.
5748 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
5749 decrement (bdnzt/bdnzf/bdzt/bdzf).
5750 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
5751 with the new names of the branch decrement patterns, and added the
5752 names of the branch decrement conditional patterns.
5754 2018-01-08 Richard Biener <rguenther@suse.de>
5756 PR tree-optimization/83563
5757 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
5760 2018-01-08 Richard Biener <rguenther@suse.de>
5763 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
5765 2018-01-08 Richard Biener <rguenther@suse.de>
5767 PR tree-optimization/83685
5768 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
5769 references to abnormals.
5771 2018-01-08 Richard Biener <rguenther@suse.de>
5774 * dwarf2out.c (output_indirect_strings): Handle empty
5775 skeleton_debug_str_hash.
5776 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
5778 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
5780 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
5781 (emit_store_direct): Likewise.
5782 (arc_trampoline_adjust_address): Likewise.
5783 (arc_asm_trampoline_template): New function.
5784 (arc_initialize_trampoline): Use asm_trampoline_template.
5785 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
5786 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
5787 * config/arc/arc.md (flush_icache): Delete pattern.
5789 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
5791 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
5792 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
5795 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
5798 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
5799 by not USED_FOR_TARGET.
5800 (make_pass_resolve_sw_modes): Likewise.
5802 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
5804 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
5807 2018-01-08 Richard Biener <rguenther@suse.de>
5810 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
5812 2018-01-08 Richard Biener <rguenther@suse.de>
5815 * match.pd ((t * 2) / 2) -> t): Add missing :c.
5817 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
5820 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
5821 basic blocks with a small number of successors.
5822 (convert_control_dep_chain_into_preds): Improve handling of
5824 (dump_predicates): Split apart into...
5825 (dump_pred_chain): ...here...
5826 (dump_pred_info): ...and here.
5827 (can_one_predicate_be_invalidated_p): Add debugging printfs.
5828 (can_chain_union_be_invalidated_p): Improve check for invalidation
5830 (uninit_uses_cannot_happen): Avoid unnecessary if
5831 convert_control_dep_chain_into_preds yielded nothing.
5833 2018-01-06 Martin Sebor <msebor@redhat.com>
5835 PR tree-optimization/83640
5836 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
5837 subtracting negative offset from size.
5838 (builtin_access::overlap): Adjust offset bounds of the access to fall
5839 within the size of the object if possible.
5841 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
5843 PR rtl-optimization/83699
5844 * expmed.c (extract_bit_field_1): Restrict the vector usage of
5845 extract_bit_field_as_subreg to cases in which the extracted
5846 value is also a vector.
5848 * lra-constraints.c (process_alt_operands): Test for the equivalence
5849 substitutions when detecting a possible reload cycle.
5851 2018-01-06 Jakub Jelinek <jakub@redhat.com>
5854 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
5855 by default if flag_selective_schedling{,2}. Formatting fixes.
5857 PR rtl-optimization/83682
5858 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
5859 if it has non-VECTOR_MODE element mode.
5860 (vec_duplicate_p): Likewise.
5863 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
5864 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
5866 2018-01-05 Jakub Jelinek <jakub@redhat.com>
5869 * config/i386/i386-builtin.def
5870 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
5871 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
5872 Require also OPTION_MASK_ISA_AVX512F in addition to
5873 OPTION_MASK_ISA_GFNI.
5874 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
5875 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
5876 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
5877 to OPTION_MASK_ISA_GFNI.
5878 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
5879 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
5880 OPTION_MASK_ISA_AVX512BW.
5881 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
5882 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
5883 addition to OPTION_MASK_ISA_GFNI.
5884 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
5885 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
5886 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
5887 to OPTION_MASK_ISA_GFNI.
5888 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
5889 a requirement for all ISAs rather than any of them with a few
5891 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
5893 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
5894 bitmasks to be enabled with 3 exceptions, instead of requiring any
5895 enabled ISA with lots of exceptions.
5896 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
5897 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
5898 Change avx512bw in isa attribute to avx512f.
5899 * config/i386/sgxintrin.h: Add license boilerplate.
5900 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
5901 to __AVX512F__ and __AVX512VL to __AVX512VL__.
5902 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
5903 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
5905 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
5906 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
5907 temporarily sse2 rather than sse if not enabled already.
5910 * config/i386/sse.md (VI248_VLBW): Rename to ...
5911 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
5912 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
5913 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
5914 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
5915 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
5916 mode iterator instead of VI248_VLBW.
5918 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
5920 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
5921 (record_modified): Skip clobbers; add debug output.
5922 (param_change_prob): Use sreal frequencies.
5924 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
5926 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
5927 punt for user-aligned variables.
5929 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
5931 * tree-chrec.c (chrec_contains_symbols): Return true for
5934 2018-01-05 Sudakshina Das <sudi.das@arm.com>
5937 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
5938 of (x|y) == x for BICS pattern.
5940 2018-01-05 Jakub Jelinek <jakub@redhat.com>
5942 PR tree-optimization/83605
5943 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
5944 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
5947 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
5949 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
5950 * config/epiphany/rtems.h: New file.
5952 2018-01-04 Jakub Jelinek <jakub@redhat.com>
5953 Uros Bizjak <ubizjak@gmail.com>
5956 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
5957 QIreg_operand instead of register_operand predicate.
5958 * config/i386/i386.c (ix86_rop_should_change_byte_p,
5959 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
5960 comments instead of -fmitigate[-_]rop.
5962 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5965 * cgraphunit.c (symbol_table::compile): Switch to text_section
5966 before calling assembly_start debug hook.
5967 * run-rtl-passes.c (run_rtl_passes): Likewise.
5970 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
5972 * tree-vrp.c (extract_range_from_binary_expr_1): Check
5973 range_int_cst_p rather than !symbolic_range_p before calling
5974 extract_range_from_multiplicative_op_1.
5976 2018-01-04 Jeff Law <law@redhat.com>
5978 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
5979 redundant test in assertion.
5981 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
5983 * doc/rtl.texi: Document machine_mode wrapper classes.
5985 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
5987 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
5990 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
5992 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
5993 the VEC_PERM_EXPR fold to fail.
5995 2018-01-04 Jakub Jelinek <jakub@redhat.com>
5998 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
5999 to switched_sections.
6001 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6004 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
6007 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
6010 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
6011 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
6013 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6016 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
6017 is BLKmode and bitpos not zero or mode change is needed.
6019 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6022 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
6025 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
6028 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
6029 instead of MULT rtx. Update all corresponding splitters.
6031 (*ssub<modesuffix>): Ditto.
6033 (*cmp_sadd_di): Update split patterns.
6034 (*cmp_sadd_si): Ditto.
6035 (*cmp_sadd_sidi): Ditto.
6036 (*cmp_ssub_di): Ditto.
6037 (*cmp_ssub_si): Ditto.
6038 (*cmp_ssub_sidi): Ditto.
6039 * config/alpha/predicates.md (const23_operand): New predicate.
6040 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
6041 Look for ASHIFT, not MULT inner operand.
6042 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
6044 2018-01-04 Martin Liska <mliska@suse.cz>
6046 PR gcov-profile/83669
6047 * gcov.c (output_intermediate_file): Add version to intermediate
6049 * doc/gcov.texi: Document new field 'version' in intermediate
6050 file format. Fix location of '-k' option of gcov command.
6052 2018-01-04 Martin Liska <mliska@suse.cz>
6055 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
6057 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6059 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
6061 2018-01-03 Martin Sebor <msebor@redhat.com>
6063 PR tree-optimization/83655
6064 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
6065 checking calls with invalid arguments.
6067 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6069 * tree-vect-stmts.c (vect_get_store_rhs): New function.
6070 (vectorizable_mask_load_store): Delete.
6071 (vectorizable_call): Return false for masked loads and stores.
6072 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
6073 instead of gimple_assign_rhs1.
6074 (vectorizable_load): Handle IFN_MASK_LOAD.
6075 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
6077 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6079 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
6081 (vectorizable_mask_load_store): ...here.
6082 (vectorizable_load): ...and here.
6084 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6086 * tree-vect-stmts.c (vect_build_all_ones_mask)
6087 (vect_build_zero_merge_argument): New functions, split out from...
6088 (vectorizable_load): ...here.
6090 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6092 * tree-vect-stmts.c (vect_check_store_rhs): New function,
6094 (vectorizable_mask_load_store): ...here.
6095 (vectorizable_store): ...and here.
6097 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6099 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
6101 (vectorizable_mask_load_store): ...here.
6103 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6105 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
6106 (vect_model_store_cost): Take a vec_load_store_type instead of a
6108 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
6109 (vect_model_store_cost): Take a vec_load_store_type instead of a
6111 (vectorizable_mask_load_store): Update accordingly.
6112 (vectorizable_store): Likewise.
6113 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
6115 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6117 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
6118 IFN_MASK_LOAD calls here rather than...
6119 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
6121 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6122 Alan Hayward <alan.hayward@arm.com>
6123 David Sherwood <david.sherwood@arm.com>
6125 * expmed.c (extract_bit_field_1): For vector extracts,
6126 fall back to extract_bit_field_as_subreg if vec_extract
6129 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6130 Alan Hayward <alan.hayward@arm.com>
6131 David Sherwood <david.sherwood@arm.com>
6133 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
6134 they are variable or constant sized.
6135 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
6136 slots for constant-sized data.
6138 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6139 Alan Hayward <alan.hayward@arm.com>
6140 David Sherwood <david.sherwood@arm.com>
6142 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
6143 handling COND_EXPRs with boolean comparisons, try to find a better
6144 basis for the mask type than the boolean itself.
6146 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6148 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
6149 is calculated and how it can be overridden.
6150 * genmodes.c (max_bitsize_mode_any_mode): New variable.
6151 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
6153 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
6156 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6157 Alan Hayward <alan.hayward@arm.com>
6158 David Sherwood <david.sherwood@arm.com>
6160 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
6161 Remove the mode argument.
6162 (aarch64_simd_valid_immediate): Remove the mode and inverse
6164 * config/aarch64/iterators.md (bitsize): New iterator.
6165 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
6166 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
6167 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
6168 aarch64_simd_valid_immediate.
6169 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
6170 (aarch64_reg_or_bic_imm): Likewise.
6171 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
6172 with an insn_type enum and msl with a modifier_type enum.
6173 Replace element_width with a scalar_mode. Change the shift
6174 to unsigned int. Add constructors for scalar_float_mode and
6175 scalar_int_mode elements.
6176 (aarch64_vect_float_const_representable_p): Delete.
6177 (aarch64_can_const_movi_rtx_p)
6178 (aarch64_simd_scalar_immediate_valid_for_move)
6179 (aarch64_simd_make_constant): Update call to
6180 aarch64_simd_valid_immediate.
6181 (aarch64_advsimd_valid_immediate_hs): New function.
6182 (aarch64_advsimd_valid_immediate): Likewise.
6183 (aarch64_simd_valid_immediate): Remove mode and inverse
6184 arguments. Rewrite to use the above. Use const_vec_duplicate_p
6185 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
6186 and aarch64_float_const_representable_p on the result.
6187 (aarch64_output_simd_mov_immediate): Remove mode argument.
6188 Update call to aarch64_simd_valid_immediate and use of
6189 simd_immediate_info.
6190 (aarch64_output_scalar_simd_mov_immediate): Update call
6193 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6194 Alan Hayward <alan.hayward@arm.com>
6195 David Sherwood <david.sherwood@arm.com>
6197 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
6198 (mode_nunits): Likewise CONST_MODE_NUNITS.
6199 * machmode.def (ADJUST_NUNITS): Document.
6200 * genmodes.c (mode_data::need_nunits_adj): New field.
6201 (blank_mode): Update accordingly.
6202 (adj_nunits): New variable.
6203 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
6205 (emit_mode_size_inline): Set need_bytesize_adj for all modes
6206 listed in adj_nunits.
6207 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
6208 listed in adj_nunits. Don't emit case statements for such modes.
6209 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
6210 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
6211 nothing if adj_nunits is nonnull.
6212 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
6213 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
6214 (emit_mode_fbit): Update use of print_maybe_const_decl.
6215 (emit_move_size): Likewise. Treat the array as non-const
6217 (emit_mode_adjustments): Handle adj_nunits.
6219 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6221 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
6222 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
6223 (VECTOR_MODES): Use it.
6224 (make_vector_modes): Take the prefix as an argument.
6226 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6227 Alan Hayward <alan.hayward@arm.com>
6228 David Sherwood <david.sherwood@arm.com>
6230 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
6231 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
6232 for MODE_VECTOR_BOOL.
6233 * machmode.def (VECTOR_BOOL_MODE): Document.
6234 * genmodes.c (VECTOR_BOOL_MODE): New macro.
6235 (make_vector_bool_mode): New function.
6236 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
6238 * lto-streamer-in.c (lto_input_mode_table): Likewise.
6239 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
6241 * stor-layout.c (int_mode_for_mode): Likewise.
6242 * tree.c (build_vector_type_for_mode): Likewise.
6243 * varasm.c (output_constant_pool_2): Likewise.
6244 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
6245 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
6246 for MODE_VECTOR_BOOL.
6247 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
6248 of mode class checks.
6249 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
6250 instead of a list of mode class checks.
6251 (expand_vector_scalar_condition): Likewise.
6252 (type_for_widest_vector_mode): Handle BImode as an inner mode.
6254 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6255 Alan Hayward <alan.hayward@arm.com>
6256 David Sherwood <david.sherwood@arm.com>
6258 * machmode.h (mode_size): Change from unsigned short to
6260 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
6261 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6262 or if measurement_type is not polynomial.
6263 (fixed_size_mode::includes_p): Check for constant-sized modes.
6264 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
6265 return a poly_uint16 rather than an unsigned short.
6266 (emit_mode_size): Change the type of mode_size from unsigned short
6267 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
6268 (emit_mode_adjustments): Cope with polynomial vector sizes.
6269 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6271 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6273 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
6274 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
6275 * caller-save.c (setup_save_areas): Likewise.
6276 (replace_reg_with_saved_mem): Likewise.
6277 * calls.c (emit_library_call_value_1): Likewise.
6278 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
6279 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
6280 (gen_lowpart_for_combine): Likewise.
6281 * convert.c (convert_to_integer_1): Likewise.
6282 * cse.c (equiv_constant, cse_insn): Likewise.
6283 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
6284 (cselib_subst_to_values): Likewise.
6285 * dce.c (word_dce_process_block): Likewise.
6286 * df-problems.c (df_word_lr_mark_ref): Likewise.
6287 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
6288 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
6289 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
6290 (rtl_for_decl_location): Likewise.
6291 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
6292 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
6293 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
6294 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
6295 (expand_expr_real_1): Likewise.
6296 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
6297 (pad_below): Likewise.
6298 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6299 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
6300 * ira.c (get_subreg_tracking_sizes): Likewise.
6301 * ira-build.c (ira_create_allocno_objects): Likewise.
6302 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
6303 (ira_sort_regnos_for_alter_reg): Likewise.
6304 * ira-costs.c (record_operand_costs): Likewise.
6305 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
6306 (resolve_simple_move): Likewise.
6307 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
6308 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
6309 (lra_constraints): Likewise.
6310 (CONST_POOL_OK_P): Reject variable-sized modes.
6311 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
6312 (add_pseudo_to_slot, lra_spill): Likewise.
6313 * omp-low.c (omp_clause_aligned_alignment): Likewise.
6314 * optabs-query.c (get_best_extraction_insn): Likewise.
6315 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6316 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
6317 (expand_mult_highpart, valid_multiword_target_p): Likewise.
6318 * recog.c (offsettable_address_addr_space_p): Likewise.
6319 * regcprop.c (maybe_mode_change): Likewise.
6320 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
6321 * regrename.c (build_def_use): Likewise.
6322 * regstat.c (dump_reg_info): Likewise.
6323 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
6324 (find_reloads, find_reloads_subreg_address): Likewise.
6325 * reload1.c (eliminate_regs_1): Likewise.
6326 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
6327 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
6328 (simplify_binary_operation_1, simplify_subreg): Likewise.
6329 * targhooks.c (default_function_arg_padding): Likewise.
6330 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
6331 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
6332 (verify_gimple_assign_ternary): Likewise.
6333 * tree-inline.c (estimate_move_cost): Likewise.
6334 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6335 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
6336 (get_address_cost_ainc): Likewise.
6337 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
6338 (vect_supportable_dr_alignment): Likewise.
6339 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6340 (vectorizable_reduction): Likewise.
6341 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
6342 (vectorizable_operation, vectorizable_load): Likewise.
6343 * tree.c (build_same_sized_truth_vector_type): Likewise.
6344 * valtrack.c (cleanup_auto_inc_dec): Likewise.
6345 * var-tracking.c (emit_note_insn_var_location): Likewise.
6346 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
6347 (ADDR_VEC_ALIGN): Likewise.
6349 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6350 Alan Hayward <alan.hayward@arm.com>
6351 David Sherwood <david.sherwood@arm.com>
6353 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
6355 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6356 or if measurement_type is polynomial.
6357 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
6358 * combine.c (make_extraction): Likewise.
6359 * dse.c (find_shift_sequence): Likewise.
6360 * dwarf2out.c (mem_loc_descriptor): Likewise.
6361 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
6362 (extract_bit_field, extract_low_bits): Likewise.
6363 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
6364 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
6365 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
6366 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
6367 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6368 * reload.c (find_reloads): Likewise.
6369 * reload1.c (alter_reg): Likewise.
6370 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
6371 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
6372 * tree-if-conv.c (predicate_mem_writes): Likewise.
6373 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
6374 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
6375 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
6376 * valtrack.c (dead_debug_insert_temp): Likewise.
6377 * varasm.c (mergeable_constant_section): Likewise.
6378 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
6380 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6381 Alan Hayward <alan.hayward@arm.com>
6382 David Sherwood <david.sherwood@arm.com>
6384 * expr.c (expand_assignment): Cope with polynomial mode sizes
6385 when assigning to a CONCAT.
6387 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6388 Alan Hayward <alan.hayward@arm.com>
6389 David Sherwood <david.sherwood@arm.com>
6391 * machmode.h (mode_precision): Change from unsigned short to
6393 (mode_to_precision): Return a poly_uint16 rather than an unsigned
6395 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
6396 or if measurement_type is not polynomial.
6397 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
6398 in which the mode is already known to be a scalar_int_mode.
6399 * genmodes.c (emit_mode_precision): Change the type of mode_precision
6400 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
6402 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6403 for GET_MODE_PRECISION.
6404 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6405 for GET_MODE_PRECISION.
6406 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
6408 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
6409 (expand_field_assignment, make_extraction): Likewise.
6410 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
6411 (get_last_value): Likewise.
6412 * convert.c (convert_to_integer_1): Likewise.
6413 * cse.c (cse_insn): Likewise.
6414 * expr.c (expand_expr_real_1): Likewise.
6415 * lra-constraints.c (simplify_operand_subreg): Likewise.
6416 * optabs-query.c (can_atomic_load_p): Likewise.
6417 * optabs.c (expand_atomic_load): Likewise.
6418 (expand_atomic_store): Likewise.
6419 * ree.c (combine_reaching_defs): Likewise.
6420 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
6421 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
6422 * tree.h (type_has_mode_precision_p): Likewise.
6423 * ubsan.c (instrument_si_overflow): Likewise.
6425 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6426 Alan Hayward <alan.hayward@arm.com>
6427 David Sherwood <david.sherwood@arm.com>
6429 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
6430 polynomial numbers of units.
6431 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
6432 (valid_vector_subparts_p): New function.
6433 (build_vector_type): Remove temporary shim and take the number
6434 of units as a poly_uint64 rather than an int.
6435 (build_opaque_vector_type): Take the number of units as a
6436 poly_uint64 rather than an int.
6437 * tree.c (build_vector_from_ctor): Handle polynomial
6438 TYPE_VECTOR_SUBPARTS.
6439 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
6440 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
6441 (build_vector_from_val): If the number of units is variable,
6442 use build_vec_duplicate_cst for constant operands and
6443 VEC_DUPLICATE_EXPR otherwise.
6444 (make_vector_type): Remove temporary is_constant ().
6445 (build_vector_type, build_opaque_vector_type): Take the number of
6446 units as a poly_uint64 rather than an int.
6447 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
6449 * cfgexpand.c (expand_debug_expr): Likewise.
6450 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
6451 (store_constructor, expand_expr_real_1): Likewise.
6452 (const_scalar_mask_from_tree): Likewise.
6453 * fold-const-call.c (fold_const_reduction): Likewise.
6454 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
6455 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
6456 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
6457 (fold_relational_const): Likewise.
6458 (native_interpret_vector): Likewise. Change the size from an
6459 int to an unsigned int.
6460 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
6461 TYPE_VECTOR_SUBPARTS.
6462 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
6463 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
6464 duplicating a non-constant operand into a variable-length vector.
6465 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
6466 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
6467 * ipa-icf.c (sem_variable::equals): Likewise.
6468 * match.pd: Likewise.
6469 * omp-simd-clone.c (simd_clone_subparts): Likewise.
6470 * print-tree.c (print_node): Likewise.
6471 * stor-layout.c (layout_type): Likewise.
6472 * targhooks.c (default_builtin_vectorization_cost): Likewise.
6473 * tree-cfg.c (verify_gimple_comparison): Likewise.
6474 (verify_gimple_assign_binary): Likewise.
6475 (verify_gimple_assign_ternary): Likewise.
6476 (verify_gimple_assign_single): Likewise.
6477 * tree-pretty-print.c (dump_generic_node): Likewise.
6478 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6479 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
6480 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
6481 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
6482 (vect_shift_permute_load_chain): Likewise.
6483 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
6484 (expand_vector_condition, optimize_vector_constructor): Likewise.
6485 (lower_vec_perm, get_compute_type): Likewise.
6486 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6487 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
6488 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
6489 (vect_recog_mask_conversion_pattern): Likewise.
6490 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
6491 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
6492 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
6493 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
6494 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
6495 (vectorizable_shift, vectorizable_operation, vectorizable_store)
6496 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
6497 (supportable_widening_operation): Likewise.
6498 (supportable_narrowing_operation): Likewise.
6499 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
6501 * varasm.c (output_constant): Likewise.
6503 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6504 Alan Hayward <alan.hayward@arm.com>
6505 David Sherwood <david.sherwood@arm.com>
6507 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
6508 so that both the length == 3 and length != 3 cases set up their
6509 own permute vectors. Add comments explaining why we know the
6510 number of elements is constant.
6511 (vect_permute_load_chain): Likewise.
6513 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6514 Alan Hayward <alan.hayward@arm.com>
6515 David Sherwood <david.sherwood@arm.com>
6517 * machmode.h (mode_nunits): Change from unsigned char to
6519 (ONLY_FIXED_SIZE_MODES): New macro.
6520 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
6521 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
6522 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
6524 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
6525 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
6526 or if measurement_type is not polynomial.
6527 * genmodes.c (ZERO_COEFFS): New macro.
6528 (emit_mode_nunits_inline): Make mode_nunits_inline return a
6530 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
6531 Use ZERO_COEFFS when emitting initializers.
6532 * data-streamer.h (bp_pack_poly_value): New function.
6533 (bp_unpack_poly_value): Likewise.
6534 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6535 for GET_MODE_NUNITS.
6536 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6537 for GET_MODE_NUNITS.
6538 * tree.c (make_vector_type): Remove temporary shim and make
6539 the real function take the number of units as a poly_uint64
6541 (build_vector_type_for_mode): Handle polynomial nunits.
6542 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
6543 * emit-rtl.c (const_vec_series_p_1): Likewise.
6544 (gen_rtx_CONST_VECTOR): Likewise.
6545 * fold-const.c (test_vec_duplicate_folding): Likewise.
6546 * genrecog.c (validate_pattern): Likewise.
6547 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
6548 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6549 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
6550 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
6551 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
6552 * rtlanal.c (subreg_get_info): Likewise.
6553 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
6554 (vect_grouped_load_supported): Likewise.
6555 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
6556 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
6557 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
6558 (simplify_const_unary_operation, simplify_binary_operation_1)
6559 (simplify_const_binary_operation, simplify_ternary_operation)
6560 (test_vector_ops_duplicate, test_vector_ops): Likewise.
6561 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
6562 instead of CONST_VECTOR_NUNITS.
6563 * varasm.c (output_constant_pool_2): Likewise.
6564 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
6565 explicit-encoded elements in the XVEC for variable-length vectors.
6567 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6569 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
6571 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6572 Alan Hayward <alan.hayward@arm.com>
6573 David Sherwood <david.sherwood@arm.com>
6575 * coretypes.h (fixed_size_mode): Declare.
6576 (fixed_size_mode_pod): New typedef.
6577 * builtins.h (target_builtins::x_apply_args_mode)
6578 (target_builtins::x_apply_result_mode): Change type to
6579 fixed_size_mode_pod.
6580 * builtins.c (apply_args_size, apply_result_size, result_vector)
6581 (expand_builtin_apply_args_1, expand_builtin_apply)
6582 (expand_builtin_return): Update accordingly.
6584 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6586 * cse.c (hash_rtx_cb): Hash only the encoded elements.
6587 * cselib.c (cselib_hash_rtx): Likewise.
6588 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
6589 CONST_VECTOR encoding.
6591 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6592 Jeff Law <law@redhat.com>
6595 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
6596 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
6597 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
6598 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
6601 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
6602 explicitly probe *sp in a noreturn function if there were any callee
6603 register saves or frame pointer is needed.
6605 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6608 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
6609 BLKmode for ternary, binary or unary expressions.
6612 * var-tracking.c (delete_vta_debug_insn): New inline function.
6613 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
6614 insns from get_insns () to NULL instead of each bb separately.
6615 Use delete_vta_debug_insn. No longer static.
6616 (vt_debug_insns_local, variable_tracking_main_1): Adjust
6617 delete_vta_debug_insns callers.
6618 * rtl.h (delete_vta_debug_insns): Declare.
6619 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
6620 instead of variable_tracking_main.
6622 2018-01-03 Martin Sebor <msebor@redhat.com>
6624 PR tree-optimization/83603
6625 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
6626 arguments past the endof the argument list in functions declared
6627 without a prototype.
6628 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
6629 Avoid checking when arguments are null.
6631 2018-01-03 Martin Sebor <msebor@redhat.com>
6634 * doc/extend.texi (attribute const): Fix a typo.
6635 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
6636 issuing -Wsuggest-attribute for void functions.
6638 2018-01-03 Martin Sebor <msebor@redhat.com>
6640 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
6641 offset_int::from instead of wide_int::to_shwi.
6642 (maybe_diag_overlap): Remove assertion.
6643 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
6644 * gimple-ssa-sprintf.c (format_directive): Same.
6645 (parse_directive): Same.
6646 (sprintf_dom_walker::compute_format_length): Same.
6647 (try_substitute_return_value): Same.
6649 2018-01-03 Jeff Law <law@redhat.com>
6652 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
6653 non-constant residual for zero at runtime and avoid probing in
6654 that case. Reorganize code for trailing problem to mirror handling
6657 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6659 PR tree-optimization/83501
6660 * tree-ssa-strlen.c (get_string_cst): New.
6661 (handle_char_store): Call get_string_cst.
6663 2018-01-03 Martin Liska <mliska@suse.cz>
6665 PR tree-optimization/83593
6666 * tree-ssa-strlen.c: Include tree-cfg.h.
6667 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
6668 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
6669 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
6671 (strlen_dom_walker::before_dom_children): Call
6672 gimple_purge_dead_eh_edges. Dump tranformation with details
6674 (strlen_dom_walker::before_dom_children): Update call by adding
6675 new argument cleanup_eh.
6676 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
6678 2018-01-03 Martin Liska <mliska@suse.cz>
6681 * cif-code.def (VARIADIC_THUNK): New enum value.
6682 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
6685 2018-01-03 Jan Beulich <jbeulich@suse.com>
6687 * sse.md (mov<mode>_internal): Tighten condition for when to use
6688 vmovdqu<ssescalarsize> for TI and OI modes.
6690 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6692 Update copyright years.
6694 2018-01-03 Martin Liska <mliska@suse.cz>
6697 * ipa-visibility.c (function_and_variable_visibility): Skip
6698 functions with noipa attribure.
6700 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6702 * gcc.c (process_command): Update copyright notice dates.
6703 * gcov-dump.c (print_version): Ditto.
6704 * gcov.c (print_version): Ditto.
6705 * gcov-tool.c (print_version): Ditto.
6706 * gengtype.c (create_file): Ditto.
6707 * doc/cpp.texi: Bump @copying's copyright year.
6708 * doc/cppinternals.texi: Ditto.
6709 * doc/gcc.texi: Ditto.
6710 * doc/gccint.texi: Ditto.
6711 * doc/gcov.texi: Ditto.
6712 * doc/install.texi: Ditto.
6713 * doc/invoke.texi: Ditto.
6715 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6717 * vector-builder.h (vector_builder::m_full_nelts): Change from
6718 unsigned int to poly_uint64.
6719 (vector_builder::full_nelts): Update prototype accordingly.
6720 (vector_builder::new_vector): Likewise.
6721 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
6722 (vector_builder::operator ==): Likewise.
6723 (vector_builder::finalize): Likewise.
6724 * int-vector-builder.h (int_vector_builder::int_vector_builder):
6725 Take the number of elements as a poly_uint64 rather than an
6727 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
6728 from unsigned int to poly_uint64.
6729 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
6730 (vec_perm_indices::new_vector): Likewise.
6731 (vec_perm_indices::length): Likewise.
6732 (vec_perm_indices::nelts_per_input): Likewise.
6733 (vec_perm_indices::input_nelts): Likewise.
6734 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
6735 number of elements per input as a poly_uint64 rather than an
6736 unsigned int. Use the original encoding for variable-length
6737 vectors, rather than clamping each individual element.
6738 For the second and subsequent elements in each pattern,
6739 clamp the step and base before clamping their sum.
6740 (vec_perm_indices::series_p): Handle polynomial element counts.
6741 (vec_perm_indices::all_in_range_p): Likewise.
6742 (vec_perm_indices_to_tree): Likewise.
6743 (vec_perm_indices_to_rtx): Likewise.
6744 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
6745 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
6746 (tree_vector_builder::new_binary_operation): Handle polynomial
6747 element counts. Return false if we need to know the number
6748 of elements at compile time.
6749 * fold-const.c (fold_vec_perm): Punt if the number of elements
6750 isn't known at compile time.
6752 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6754 * vec-perm-indices.h (vec_perm_builder): Change element type
6755 from HOST_WIDE_INT to poly_int64.
6756 (vec_perm_indices::element_type): Update accordingly.
6757 (vec_perm_indices::clamp): Handle polynomial element_types.
6758 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
6759 (vec_perm_indices::all_in_range_p): Likewise.
6760 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
6762 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
6763 polynomial vec_perm_indices element types.
6764 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
6765 * fold-const.c (fold_vec_perm): Likewise.
6766 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
6767 * tree-vect-generic.c (lower_vec_perm): Likewise.
6768 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
6769 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
6770 element type to HOST_WIDE_INT.
6772 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6773 Alan Hayward <alan.hayward@arm.com>
6774 David Sherwood <david.sherwood@arm.com>
6776 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
6777 rather than an int. Use plus_constant.
6778 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
6779 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
6781 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6782 Alan Hayward <alan.hayward@arm.com>
6783 David Sherwood <david.sherwood@arm.com>
6785 * calls.c (emit_call_1, expand_call): Change struct_value_size from
6786 a HOST_WIDE_INT to a poly_int64.
6788 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6789 Alan Hayward <alan.hayward@arm.com>
6790 David Sherwood <david.sherwood@arm.com>
6792 * calls.c (load_register_parameters): Cope with polynomial
6793 mode sizes. Require a constant size for BLKmode parameters
6794 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
6795 forces a parameter to be padded at the lsb end in order to
6796 fill a complete number of words, require the parameter size
6797 to be ordered wrt UNITS_PER_WORD.
6799 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6800 Alan Hayward <alan.hayward@arm.com>
6801 David Sherwood <david.sherwood@arm.com>
6803 * reload1.c (spill_stack_slot_width): Change element type
6804 from unsigned int to poly_uint64_pod.
6805 (alter_reg): Treat mode sizes as polynomial.
6807 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6808 Alan Hayward <alan.hayward@arm.com>
6809 David Sherwood <david.sherwood@arm.com>
6811 * reload.c (complex_word_subreg_p): New function.
6812 (reload_inner_reg_of_subreg, push_reload): Use it.
6814 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6815 Alan Hayward <alan.hayward@arm.com>
6816 David Sherwood <david.sherwood@arm.com>
6818 * lra-constraints.c (process_alt_operands): Reject matched
6819 operands whose sizes aren't ordered.
6820 (match_reload): Refer to this check here.
6822 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6823 Alan Hayward <alan.hayward@arm.com>
6824 David Sherwood <david.sherwood@arm.com>
6826 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
6827 that the mode size is in the set {1, 2, 4, 8, 16}.
6829 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6830 Alan Hayward <alan.hayward@arm.com>
6831 David Sherwood <david.sherwood@arm.com>
6833 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
6834 Use plus_constant instead of gen_rtx_PLUS.
6836 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6837 Alan Hayward <alan.hayward@arm.com>
6838 David Sherwood <david.sherwood@arm.com>
6840 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
6841 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
6842 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
6843 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
6844 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
6845 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
6846 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
6847 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
6848 * config/i386/i386.c (ix86_push_rounding): ...this new function.
6849 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
6851 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
6852 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
6853 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
6854 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
6855 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
6856 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
6857 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
6858 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
6859 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
6860 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
6862 * expr.c (emit_move_resolve_push): Treat the input and result
6863 of PUSH_ROUNDING as a poly_int64.
6864 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
6865 (emit_push_insn): Likewise.
6866 * lra-eliminations.c (mark_not_eliminable): Likewise.
6867 * recog.c (push_operand): Likewise.
6868 * reload1.c (elimination_effects): Likewise.
6869 * rtlanal.c (nonzero_bits1): Likewise.
6870 * calls.c (store_one_arg): Likewise. Require the padding to be
6871 known at compile time.
6873 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6874 Alan Hayward <alan.hayward@arm.com>
6875 David Sherwood <david.sherwood@arm.com>
6877 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
6878 Use plus_constant instead of gen_rtx_PLUS.
6880 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6881 Alan Hayward <alan.hayward@arm.com>
6882 David Sherwood <david.sherwood@arm.com>
6884 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
6887 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6888 Alan Hayward <alan.hayward@arm.com>
6889 David Sherwood <david.sherwood@arm.com>
6891 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
6892 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
6893 via stack temporaries. Treat the mode size as polynomial too.
6895 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6896 Alan Hayward <alan.hayward@arm.com>
6897 David Sherwood <david.sherwood@arm.com>
6899 * expr.c (expand_expr_real_2): When handling conversions involving
6900 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
6901 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
6902 as a poly_uint64 too.
6904 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6905 Alan Hayward <alan.hayward@arm.com>
6906 David Sherwood <david.sherwood@arm.com>
6908 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
6910 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6911 Alan Hayward <alan.hayward@arm.com>
6912 David Sherwood <david.sherwood@arm.com>
6914 * combine.c (can_change_dest_mode): Handle polynomial
6915 REGMODE_NATURAL_SIZE.
6916 * expmed.c (store_bit_field_1): Likewise.
6917 * expr.c (store_constructor): Likewise.
6918 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
6919 and polynomial REGMODE_NATURAL_SIZE.
6920 (gen_lowpart_common): Likewise.
6921 * reginfo.c (record_subregs_of_mode): Likewise.
6922 * rtlanal.c (read_modify_subreg_p): Likewise.
6924 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6925 Alan Hayward <alan.hayward@arm.com>
6926 David Sherwood <david.sherwood@arm.com>
6928 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
6929 numbers of elements.
6931 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6932 Alan Hayward <alan.hayward@arm.com>
6933 David Sherwood <david.sherwood@arm.com>
6935 * match.pd: Cope with polynomial numbers of vector elements.
6937 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6938 Alan Hayward <alan.hayward@arm.com>
6939 David Sherwood <david.sherwood@arm.com>
6941 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
6942 in a POINTER_PLUS_EXPR.
6944 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6945 Alan Hayward <alan.hayward@arm.com>
6946 David Sherwood <david.sherwood@arm.com>
6948 * omp-simd-clone.c (simd_clone_subparts): New function.
6949 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
6950 (ipa_simd_modify_function_body): Likewise.
6952 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6953 Alan Hayward <alan.hayward@arm.com>
6954 David Sherwood <david.sherwood@arm.com>
6956 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
6957 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
6958 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
6959 (expand_vector_condition, vector_element): Likewise.
6960 (subparts_gt): New function.
6961 (get_compute_type): Use subparts_gt.
6962 (count_type_subparts): Delete.
6963 (expand_vector_operations_1): Use subparts_gt instead of
6964 count_type_subparts.
6966 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6967 Alan Hayward <alan.hayward@arm.com>
6968 David Sherwood <david.sherwood@arm.com>
6970 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
6971 (vect_compile_time_alias): ...this new function. Do the calculation
6972 on poly_ints rather than trees.
6973 (vect_prune_runtime_alias_test_list): Update call accordingly.
6975 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6976 Alan Hayward <alan.hayward@arm.com>
6977 David Sherwood <david.sherwood@arm.com>
6979 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
6981 (vect_schedule_slp_instance): Likewise.
6983 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6984 Alan Hayward <alan.hayward@arm.com>
6985 David Sherwood <david.sherwood@arm.com>
6987 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
6988 constant and extern definitions for variable-length vectors.
6989 (vect_get_constant_vectors): Note that the number of units
6990 is known to be constant.
6992 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6993 Alan Hayward <alan.hayward@arm.com>
6994 David Sherwood <david.sherwood@arm.com>
6996 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
6997 of units as polynomial. Choose between WIDE and NARROW based
7000 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7001 Alan Hayward <alan.hayward@arm.com>
7002 David Sherwood <david.sherwood@arm.com>
7004 * tree-vect-stmts.c (simd_clone_subparts): New function.
7005 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
7007 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7008 Alan Hayward <alan.hayward@arm.com>
7009 David Sherwood <david.sherwood@arm.com>
7011 * tree-vect-stmts.c (vectorizable_call): Treat the number of
7012 vectors as polynomial. Use build_index_vector for
7015 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7016 Alan Hayward <alan.hayward@arm.com>
7017 David Sherwood <david.sherwood@arm.com>
7019 * tree-vect-stmts.c (get_load_store_type): Treat the number of
7020 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
7021 for variable-length vectors.
7022 (vectorizable_mask_load_store): Treat the number of units as
7023 polynomial, asserting that it is constant if the condition has
7024 already been enforced.
7025 (vectorizable_store, vectorizable_load): Likewise.
7027 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7028 Alan Hayward <alan.hayward@arm.com>
7029 David Sherwood <david.sherwood@arm.com>
7031 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
7032 of units as polynomial. Punt if we can't tell at compile time
7033 which vector contains the final result.
7035 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7036 Alan Hayward <alan.hayward@arm.com>
7037 David Sherwood <david.sherwood@arm.com>
7039 * tree-vect-loop.c (vectorizable_induction): Treat the number
7040 of units as polynomial. Punt on SLP inductions. Use an integer
7041 VEC_SERIES_EXPR for variable-length integer reductions. Use a
7042 cast of such a series for variable-length floating-point
7045 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7046 Alan Hayward <alan.hayward@arm.com>
7047 David Sherwood <david.sherwood@arm.com>
7049 * tree.h (build_index_vector): Declare.
7050 * tree.c (build_index_vector): New function.
7051 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
7052 of units as polynomial, forcibly converting it to a constant if
7053 vectorizable_reduction has already enforced the condition.
7054 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
7055 to create a {1,2,3,...} vector.
7056 (vectorizable_reduction): Treat the number of units as polynomial.
7057 Choose vectype_in based on the largest scalar element size rather
7058 than the smallest number of units. Enforce the restrictions
7061 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7062 Alan Hayward <alan.hayward@arm.com>
7063 David Sherwood <david.sherwood@arm.com>
7065 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
7066 number of units as polynomial.
7068 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7069 Alan Hayward <alan.hayward@arm.com>
7070 David Sherwood <david.sherwood@arm.com>
7072 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
7073 * target.def (autovectorize_vector_sizes): Return the vector sizes
7074 by pointer, using vector_sizes rather than a bitmask.
7075 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
7076 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
7077 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
7079 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
7080 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
7081 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
7082 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
7083 * omp-general.c (omp_max_vf): Likewise.
7084 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7085 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
7086 * tree-vect-loop.c (vect_analyze_loop): Likewise.
7087 * tree-vect-slp.c (vect_slp_bb): Likewise.
7088 * doc/tm.texi: Regenerate.
7089 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
7091 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
7092 the vector size as a poly_uint64 rather than an unsigned int.
7093 (current_vector_size): Change from an unsigned int to a poly_uint64.
7094 (get_vectype_for_scalar_type): Update accordingly.
7095 * tree.h (build_truth_vector_type): Take the size and number of
7096 units as a poly_uint64 rather than an unsigned int.
7097 (build_vector_type): Add a temporary overload that takes
7098 the number of units as a poly_uint64 rather than an unsigned int.
7099 * tree.c (make_vector_type): Likewise.
7100 (build_truth_vector_type): Take the number of units as a poly_uint64
7101 rather than an unsigned int.
7103 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7104 Alan Hayward <alan.hayward@arm.com>
7105 David Sherwood <david.sherwood@arm.com>
7107 * target.def (get_mask_mode): Take the number of units and length
7108 as poly_uint64s rather than unsigned ints.
7109 * targhooks.h (default_get_mask_mode): Update accordingly.
7110 * targhooks.c (default_get_mask_mode): Likewise.
7111 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
7112 * doc/tm.texi: Regenerate.
7114 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7115 Alan Hayward <alan.hayward@arm.com>
7116 David Sherwood <david.sherwood@arm.com>
7118 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
7119 * omp-general.c (omp_max_vf): Likewise.
7120 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
7121 (expand_omp_simd): Handle polynomial safelen.
7122 * omp-low.c (omplow_simd_context): Add a default constructor.
7123 (omplow_simd_context::max_vf): Change from int to poly_uint64.
7124 (lower_rec_simd_input_clauses): Update accordingly.
7125 (lower_rec_input_clauses): Likewise.
7127 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7128 Alan Hayward <alan.hayward@arm.com>
7129 David Sherwood <david.sherwood@arm.com>
7131 * tree-vectorizer.h (vect_nunits_for_cost): New function.
7132 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
7133 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
7134 (vect_analyze_slp_cost): Likewise.
7135 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
7136 (vect_model_load_cost): Likewise.
7138 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7139 Alan Hayward <alan.hayward@arm.com>
7140 David Sherwood <david.sherwood@arm.com>
7142 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
7143 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
7144 from an unsigned int * to a poly_uint64_pod *.
7145 (calculate_unrolling_factor): New function.
7146 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
7148 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7149 Alan Hayward <alan.hayward@arm.com>
7150 David Sherwood <david.sherwood@arm.com>
7152 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
7153 from an unsigned int to a poly_uint64.
7154 (_loop_vec_info::slp_unrolling_factor): Likewise.
7155 (_loop_vec_info::vectorization_factor): Change from an int
7157 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
7158 (vect_get_num_vectors): New function.
7159 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
7160 (vect_get_num_copies): Use vect_get_num_vectors.
7161 (vect_analyze_data_ref_dependences): Change max_vf from an int *
7162 to an unsigned int *.
7163 (vect_analyze_data_refs): Change min_vf from an int * to a
7165 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7166 than an unsigned HOST_WIDE_INT.
7167 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7168 (vect_analyze_data_ref_dependence): Change max_vf from an int *
7169 to an unsigned int *.
7170 (vect_analyze_data_ref_dependences): Likewise.
7171 (vect_compute_data_ref_alignment): Handle polynomial vf.
7172 (vect_enhance_data_refs_alignment): Likewise.
7173 (vect_prune_runtime_alias_test_list): Likewise.
7174 (vect_shift_permute_load_chain): Likewise.
7175 (vect_supportable_dr_alignment): Likewise.
7176 (dependence_distance_ge_vf): Take the vectorization factor as a
7177 poly_uint64 rather than an unsigned HOST_WIDE_INT.
7178 (vect_analyze_data_refs): Change min_vf from an int * to a
7180 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
7181 vfm1 as a poly_uint64 rather than an int. Make the same change
7182 for the returned bound_scalar.
7183 (vect_gen_vector_loop_niters): Handle polynomial vf.
7184 (vect_do_peeling): Likewise. Update call to
7185 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
7186 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
7188 * tree-vect-loop.c (vect_determine_vectorization_factor)
7189 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
7190 (vect_get_known_peeling_cost): Likewise.
7191 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
7192 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
7193 (vect_transform_loop): Likewise. Use the lowest possible VF when
7194 updating the upper bounds of the loop.
7195 (vect_min_worthwhile_factor): Make static. Return an unsigned int
7197 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
7198 polynomial unroll factors.
7199 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
7200 (vect_make_slp_decision): Likewise.
7201 (vect_supported_load_permutation_p): Likewise, and polynomial
7203 (vect_analyze_slp_cost): Handle polynomial vf.
7204 (vect_slp_analyze_node_operations): Likewise.
7205 (vect_slp_analyze_bb_1): Likewise.
7206 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7207 than an unsigned HOST_WIDE_INT.
7208 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
7209 (vectorizable_load): Handle polynomial vf.
7210 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
7212 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
7214 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7215 Alan Hayward <alan.hayward@arm.com>
7216 David Sherwood <david.sherwood@arm.com>
7218 * match.pd: Handle bit operations involving three constants
7219 and try to fold one pair.
7221 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7223 * tree-vect-loop-manip.c: Include gimple-fold.h.
7224 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
7225 niters_maybe_zero parameters. Handle other cases besides a step of 1.
7226 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
7227 Add a path that uses a step of VF instead of 1, but disable it
7229 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
7230 and niters_no_overflow parameters. Update calls to
7231 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
7232 Create a new SSA name if the latter choses to use a ste other
7233 than zero, and return it via niters_vector_mult_vf_var.
7234 * tree-vect-loop.c (vect_transform_loop): Update calls to
7235 vect_do_peeling, vect_gen_vector_loop_niters and
7236 slpeel_make_loop_iterate_ntimes.
7237 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
7238 (vect_gen_vector_loop_niters): Update declarations after above changes.
7240 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
7242 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
7243 128-bit round to integer instructions.
7244 (ceil<mode>2): Likewise.
7245 (btrunc<mode>2): Likewise.
7246 (round<mode>2): Likewise.
7248 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7250 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
7251 unaligned VSX load/store on P8/P9.
7252 (expand_block_clear): Allow the use of unaligned VSX
7253 load/store on P8/P9.
7255 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7257 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
7259 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
7260 swap associated with both a load and a store.
7262 2018-01-02 Andrew Waterman <andrew@sifive.com>
7264 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
7265 * config/riscv/riscv.md (clear_cache): Use it.
7267 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
7269 * web.c: Remove out-of-date comment.
7271 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7273 * expr.c (fixup_args_size_notes): Check that any existing
7274 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
7275 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
7276 (emit_single_push_insn): ...here.
7278 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7280 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
7281 (const_vector_encoded_nelts): New function.
7282 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
7283 (const_vector_int_elt, const_vector_elt): Declare.
7284 * emit-rtl.c (const_vector_int_elt_1): New function.
7285 (const_vector_elt): Likewise.
7286 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
7287 of CONST_VECTOR_ELT.
7289 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7291 * expr.c: Include rtx-vector-builder.h.
7292 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
7293 directly on the tree encoding.
7294 (const_vector_from_tree): Likewise.
7295 * optabs.c: Include rtx-vector-builder.h.
7296 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
7297 sequence of "u" values.
7298 * vec-perm-indices.c: Include rtx-vector-builder.h.
7299 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
7300 directly on the vec_perm_indices encoding.
7302 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7304 * doc/rtl.texi (const_vector): Describe new encoding scheme.
7305 * Makefile.in (OBJS): Add rtx-vector-builder.o.
7306 * rtx-vector-builder.h: New file.
7307 * rtx-vector-builder.c: Likewise.
7308 * rtl.h (rtx_def::u2): Add a const_vector field.
7309 (CONST_VECTOR_NPATTERNS): New macro.
7310 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
7311 (CONST_VECTOR_DUPLICATE_P): Likewise.
7312 (CONST_VECTOR_STEPPED_P): Likewise.
7313 (CONST_VECTOR_ENCODED_ELT): Likewise.
7314 (const_vec_duplicate_p): Check for a duplicated vector encoding.
7315 (unwrap_const_vec_duplicate): Likewise.
7316 (const_vec_series_p): Check for a non-duplicated vector encoding.
7317 Say that the function only returns true for integer vectors.
7318 * emit-rtl.c: Include rtx-vector-builder.h.
7319 (gen_const_vec_duplicate_1): Delete.
7320 (gen_const_vector): Call gen_const_vec_duplicate instead of
7321 gen_const_vec_duplicate_1.
7322 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
7323 (gen_const_vec_duplicate): Use rtx_vector_builder.
7324 (gen_const_vec_series): Likewise.
7325 (gen_rtx_CONST_VECTOR): Likewise.
7326 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
7327 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7328 Build a new vector rather than modifying a CONST_VECTOR in-place.
7329 (handle_special_swappables): Update call accordingly.
7330 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
7331 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7332 Build a new vector rather than modifying a CONST_VECTOR in-place.
7333 (handle_special_swappables): Update call accordingly.
7335 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7337 * simplify-rtx.c (simplify_const_binary_operation): Use
7338 CONST_VECTOR_ELT instead of XVECEXP.
7340 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7342 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
7343 the selector elements to be different from the data elements
7344 if the selector is a VECTOR_CST.
7345 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
7346 ssizetype for the selector.
7348 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7350 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
7351 before testing each element individually.
7352 * tree-vect-generic.c (lower_vec_perm): Likewise.
7354 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7356 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
7357 * selftest-run-tests.c (selftest::run_tests): Call it.
7358 * vector-builder.h (vector_builder::operator ==): New function.
7359 (vector_builder::operator !=): Likewise.
7360 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
7361 (vec_perm_indices::all_from_input_p): New function.
7362 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7363 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
7364 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
7365 instead of reading the VECTOR_CST directly. Detect whether both
7366 vector inputs are the same before constructing the vec_perm_indices,
7367 and update the number of inputs argument accordingly. Use the
7368 utility functions added above. Only construct sel2 if we need to.
7370 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7372 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
7373 the broadcast of the low byte.
7374 (expand_mult_highpart): Use an explicit encoding for the permutes.
7375 * optabs-query.c (can_mult_highpart_p): Likewise.
7376 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
7377 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7378 (vectorizable_bswap): Likewise.
7379 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
7380 explicit encoding for the power-of-2 permutes.
7381 (vect_permute_store_chain): Likewise.
7382 (vect_grouped_load_supported): Likewise.
7383 (vect_permute_load_chain): Likewise.
7385 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7387 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
7388 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
7389 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
7390 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7391 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
7392 (vect_gen_perm_mask_any): Likewise.
7394 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7396 * int-vector-builder.h: New file.
7397 * vec-perm-indices.h: Include int-vector-builder.h.
7398 (vec_perm_indices): Redefine as an int_vector_builder.
7399 (auto_vec_perm_indices): Delete.
7400 (vec_perm_builder): Redefine as a stand-alone class.
7401 (vec_perm_indices::vec_perm_indices): New function.
7402 (vec_perm_indices::clamp): Likewise.
7403 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
7404 (vec_perm_indices::new_vector): New function.
7405 (vec_perm_indices::new_expanded_vector): Update for new
7406 vec_perm_indices class.
7407 (vec_perm_indices::rotate_inputs): New function.
7408 (vec_perm_indices::all_in_range_p): Operate directly on the
7409 encoded form, without computing elided elements.
7410 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
7411 encoding. Update for new vec_perm_indices class.
7412 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
7413 the given vec_perm_builder.
7414 (expand_vec_perm_var): Update vec_perm_builder constructor.
7415 (expand_mult_highpart): Use vec_perm_builder instead of
7416 auto_vec_perm_indices.
7417 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
7418 vec_perm_indices instead of auto_vec_perm_indices. Use a single
7419 or double series encoding as appropriate.
7420 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
7421 vec_perm_indices instead of auto_vec_perm_indices.
7422 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7423 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7424 (vect_permute_store_chain): Likewise.
7425 (vect_grouped_load_supported): Likewise.
7426 (vect_permute_load_chain): Likewise.
7427 (vect_shift_permute_load_chain): Likewise.
7428 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
7429 (vect_transform_slp_perm_load): Likewise.
7430 (vect_schedule_slp_instance): Likewise.
7431 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7432 (vectorizable_mask_load_store): Likewise.
7433 (vectorizable_bswap): Likewise.
7434 (vectorizable_store): Likewise.
7435 (vectorizable_load): Likewise.
7436 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
7437 vec_perm_indices instead of auto_vec_perm_indices. Use
7438 tree_to_vec_perm_builder to read the vector from a tree.
7439 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
7440 vec_perm_builder instead of a vec_perm_indices.
7441 (have_whole_vector_shift): Use vec_perm_builder and
7442 vec_perm_indices instead of auto_vec_perm_indices. Leave the
7443 truncation to calc_vec_perm_mask_for_shift.
7444 (vect_create_epilog_for_reduction): Likewise.
7445 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
7446 from auto_vec_perm_indices to vec_perm_indices.
7447 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7448 instead of changing individual elements.
7449 (aarch64_vectorize_vec_perm_const): Use new_vector to install
7450 the vector in d.perm.
7451 * config/arm/arm.c (expand_vec_perm_d::perm): Change
7452 from auto_vec_perm_indices to vec_perm_indices.
7453 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7454 instead of changing individual elements.
7455 (arm_vectorize_vec_perm_const): Use new_vector to install
7456 the vector in d.perm.
7457 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
7458 Update vec_perm_builder constructor.
7459 (rs6000_expand_interleave): Likewise.
7460 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
7461 (rs6000_expand_interleave): Likewise.
7463 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7465 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
7466 to qimode could truncate the indices.
7467 * optabs.c (expand_vec_perm_var): Likewise.
7469 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7471 * Makefile.in (OBJS): Add vec-perm-indices.o.
7472 * vec-perm-indices.h: New file.
7473 * vec-perm-indices.c: Likewise.
7474 * target.h (vec_perm_indices): Replace with a forward class
7476 (auto_vec_perm_indices): Move to vec-perm-indices.h.
7477 * optabs.h: Include vec-perm-indices.h.
7478 (expand_vec_perm): Delete.
7479 (selector_fits_mode_p, expand_vec_perm_var): Declare.
7480 (expand_vec_perm_const): Declare.
7481 * target.def (vec_perm_const_ok): Replace with...
7482 (vec_perm_const): ...this new hook.
7483 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
7484 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
7485 * doc/tm.texi: Regenerate.
7486 * optabs.def (vec_perm_const): Delete.
7487 * doc/md.texi (vec_perm_const): Likewise.
7488 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
7489 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
7490 expand_vec_perm for constant permutation vectors. Assert that
7491 the mode of variable permutation vectors is the integer equivalent
7492 of the mode that is being permuted.
7493 * optabs-query.h (selector_fits_mode_p): Declare.
7494 * optabs-query.c: Include vec-perm-indices.h.
7495 (selector_fits_mode_p): New function.
7496 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
7497 is defined, instead of checking whether the vec_perm_const_optab
7498 exists. Use targetm.vectorize.vec_perm_const instead of
7499 targetm.vectorize.vec_perm_const_ok. Check whether the indices
7500 fit in the vector mode before using a variable permute.
7501 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
7502 vec_perm_indices instead of an rtx.
7503 (expand_vec_perm): Replace with...
7504 (expand_vec_perm_const): ...this new function. Take the selector
7505 as a vec_perm_indices rather than an rtx. Also take the mode of
7506 the selector. Update call to shift_amt_for_vec_perm_mask.
7507 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
7508 Use vec_perm_indices::new_expanded_vector to expand the original
7509 selector into bytes. Check whether the indices fit in the vector
7510 mode before using a variable permute.
7511 (expand_vec_perm_var): Make global.
7512 (expand_mult_highpart): Use expand_vec_perm_const.
7513 * fold-const.c: Includes vec-perm-indices.h.
7514 * tree-ssa-forwprop.c: Likewise.
7515 * tree-vect-data-refs.c: Likewise.
7516 * tree-vect-generic.c: Likewise.
7517 * tree-vect-loop.c: Likewise.
7518 * tree-vect-slp.c: Likewise.
7519 * tree-vect-stmts.c: Likewise.
7520 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
7522 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
7523 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
7524 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
7525 (aarch64_vectorize_vec_perm_const): ...this new function.
7526 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7527 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7528 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
7529 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
7530 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7531 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7532 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
7534 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
7535 check for NEON modes.
7536 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
7537 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
7538 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
7539 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
7541 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
7542 the old VEC_PERM_CONST conditions.
7543 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
7544 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
7545 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
7546 (ia64_vectorize_vec_perm_const_ok): Merge into...
7547 (ia64_vectorize_vec_perm_const): ...this new function.
7548 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
7549 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
7550 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
7551 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
7552 * config/mips/mips.c (mips_expand_vec_perm_const)
7553 (mips_vectorize_vec_perm_const_ok): Merge into...
7554 (mips_vectorize_vec_perm_const): ...this new function.
7555 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
7556 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
7557 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
7558 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
7559 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
7560 (rs6000_expand_vec_perm_const): Delete.
7561 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
7563 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7564 (altivec_expand_vec_perm_const_le): Take each operand individually.
7565 Operate on constant selectors rather than rtxes.
7566 (altivec_expand_vec_perm_const): Likewise. Update call to
7567 altivec_expand_vec_perm_const_le.
7568 (rs6000_expand_vec_perm_const): Delete.
7569 (rs6000_vectorize_vec_perm_const_ok): Delete.
7570 (rs6000_vectorize_vec_perm_const): New function.
7571 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
7572 an element count and rtx array.
7573 (rs6000_expand_extract_even): Update call accordingly.
7574 (rs6000_expand_interleave): Likewise.
7575 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
7576 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
7577 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
7578 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
7579 (rs6000_expand_vec_perm_const): Delete.
7580 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7581 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7582 (altivec_expand_vec_perm_const_le): Take each operand individually.
7583 Operate on constant selectors rather than rtxes.
7584 (altivec_expand_vec_perm_const): Likewise. Update call to
7585 altivec_expand_vec_perm_const_le.
7586 (rs6000_expand_vec_perm_const): Delete.
7587 (rs6000_vectorize_vec_perm_const_ok): Delete.
7588 (rs6000_vectorize_vec_perm_const): New function. Remove stray
7589 reference to the SPE evmerge intructions.
7590 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
7591 an element count and rtx array.
7592 (rs6000_expand_extract_even): Update call accordingly.
7593 (rs6000_expand_interleave): Likewise.
7594 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
7595 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
7597 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7599 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7601 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
7602 vector mode and that that mode matches the mode of the data
7604 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
7605 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
7606 directly using expand_vec_perm_1 when forcing selectors into
7608 (expand_vec_perm_var): New function, split out from expand_vec_perm.
7610 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7612 * optabs-query.h (can_vec_perm_p): Delete.
7613 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
7614 * optabs-query.c (can_vec_perm_p): Split into...
7615 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
7616 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
7617 particular selector is valid.
7618 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7619 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7620 (vect_grouped_load_supported): Likewise.
7621 (vect_shift_permute_load_chain): Likewise.
7622 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
7623 (vect_transform_slp_perm_load): Likewise.
7624 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7625 (vectorizable_bswap): Likewise.
7626 (vect_gen_perm_mask_checked): Likewise.
7627 * fold-const.c (fold_ternary_loc): Likewise. Don't take
7628 implementations of variable permutation vectors into account
7629 when deciding which selector to use.
7630 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
7631 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
7632 with a false third argument.
7633 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
7634 to test whether the constant selector is valid and can_vec_perm_var_p
7635 to test whether a variable selector is valid.
7637 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7639 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
7640 * optabs-query.c (can_vec_perm_p): Likewise.
7641 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
7642 instead of vec_perm_indices.
7643 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
7644 (vect_gen_perm_mask_checked): Likewise,
7645 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
7646 (vect_gen_perm_mask_checked): Likewise,
7648 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7650 * optabs-query.h (qimode_for_vec_perm): Declare.
7651 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
7652 (qimode_for_vec_perm): ...this new function.
7653 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
7655 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7657 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
7658 does not have a conditional at the top.
7660 2018-01-02 Richard Biener <rguenther@suse.de>
7662 * ipa-inline.c (big_speedup_p): Fix expression.
7664 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
7667 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
7670 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
7674 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
7675 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
7676 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
7677 cond_taken_branch_cost 3->4.
7679 2018-01-01 Jakub Jelinek <jakub@redhat.com>
7681 PR tree-optimization/83581
7682 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
7683 TODO_cleanup_cfg if any changes have been made.
7686 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
7687 convert_modes if target mode has the right side, but different mode
7691 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
7692 last argument when extracting from CONCAT. If either from_real or
7693 from_imag is NULL, use expansion through memory. If result is not
7694 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
7695 the parts directly to inner mode, if even that fails, use expansion
7699 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
7700 check for bswap in mode rather than HImode and use that in expand_unop
7703 Copyright (C) 2018 Free Software Foundation, Inc.
7705 Copying and distribution of this file, with or without modification,
7706 are permitted in any medium without royalty provided the copyright
7707 notice and this notice are preserved.