| blob | a2263edadb6570ac62a165887cd7914c9056acdb |
1 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3 PR target/56540
4 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
7 PR target/56540
8 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
9 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
11 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
12 instead of -1U in last predictors element's probability member.
14 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
16 PR ipa/83983
17 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
18 arguments if they are comparable.
20 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
22 PR tree-optimization/84634
23 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
24 masks and masked_loop_p with a single loop_masks, making sure it's
25 null for bb vectorization.
27 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
29 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
30 (vect_analyze_data_ref_access): Use loop->safe_len rather than
31 loop->force_vectorize to check whether there is no alias.
33 2018-03-02 Jakub Jelinek <jakub@redhat.com>
35 PR target/84614
36 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
37 prototypes.
38 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
39 comments.
40 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
41 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
42 instead of a loop around prev_real_insn.
43 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
44 prev_real_insn.
46 PR inline-asm/84625
47 * config/i386/i386.c (ix86_print_operand): Use conditional
48 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
49 zero vector.
51 2018-03-02 Richard Biener <rguenther@suse.de>
53 PR tree-optimization/84427
54 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
55 (bitmap_set_subtract_values): Rewrite to handle multiple
56 exprs per value.
57 (clean): Likewise.
58 (prune_clobbered_mems): Likewise.
59 (phi_translate): Take edge instead of pred/phiblock.
60 (phi_translate_1): Likewise.
61 (phi_translate_set): Likewise. Insert all translated
62 exprs for a value into the set, keeping possibly multiple
63 expressions per value.
64 (compute_antic_aux): Adjust for phi_translate changes.
65 When intersecting union the expressions and prune those
66 not in the final value set, keeping possibly multiple
67 expressions per value. Do not use value-insertion
68 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
69 all expressions. Add verification that the value-sets
70 only shrink during iteration.
71 (compute_partial_antic_aux): Adjust for the phi_translate changes.
72 (do_pre_regular_insertion): Likewise.
73 (do_pre_partial_partial_insertion): Likewise.
75 2018-03-02 Richard Biener <rguenther@suse.de>
77 PR target/82005
78 * config/darwin.c (saved_debug_info_level): New static global.
79 (darwin_asm_lto_start): Disable debug info generation for LTO out.
80 (darwin_asm_lto_end): Restore debug info generation settings.
82 2018-03-01 Martin Liska <mliska@suse.cz>
84 PR sanitizer/82484
85 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
86 volatile arguments.
88 2018-03-01 Richard Biener <rguenther@suse.de>
90 PR debug/84645
91 * dwarf2out.c (gen_variable_die): Properly handle late VLA
92 type annotation with LTO when debug was disabled at compile-time.
94 2018-03-01 Matthew Fortune <mfortune@gmail.com>
96 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
97 XINT with INTVAL.
98 (mips_final_postscan_insn): Likewise.
100 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
102 PR rtl-optimization/84528
103 * alias.c (init_alias_target): Add commentary.
104 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
105 a unique base value if the frame pointer is not eliminated
106 to the stack pointer.
108 2018-03-01 Tom de Vries <tom@codesourcery.com>
110 PR rtl-optimization/83327
111 * lra-int.h (hard_regs_spilled_into): Declare.
112 * lra.c (hard_regs_spilled_into): Define.
113 (init_reg_info): Init hard_regs_spilled_into.
114 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
115 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
116 (process_bb_lives): Handle hard_regs_spilled_into.
117 (lra_create_live_ranges_1): Before doing liveness propagation, clear
118 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
120 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
122 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
123 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
124 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
125 * config/rs6000/aix72.h: New file.
127 2018-02-28 Jakub Jelinek <jakub@redhat.com>
129 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
130 instead of warning_at with conditional singular and plural messages
131 where possible.
133 PR target/52991
134 * stor-layout.c (update_alignment_for_field): For
135 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
136 && !DECL_PACKED (field), do the alignment update, just use
137 only desired_align instead of MAX (type_align, desired_align)
138 as the alignment.
139 (place_field): Don't do known_align < desired_align handling
140 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
141 is non-NULL, instead do it after rli->prev_field handling and
142 only if not within a bitfield word. For DECL_PACKED (field)
143 use type_align of BITS_PER_UNIT.
145 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
147 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
148 superfluous parentheses and trailing spaces.
150 2018-02-28 Richard Biener <rguenther@suse.de>
152 PR tree-optimization/84584
153 * graphite-scop-detection.c (scop_detection::add_scop): Discard
154 SCoPs with fake exit edge.
156 2018-02-28 Martin Liska <mliska@suse.cz>
158 PR testsuite/84597
159 * timevar.c (timer::print): Fix format to properly print 100%
160 values.
162 2018-02-28 Richard Biener <rguenther@suse.de>
164 PR middle-end/84607
165 * genmatch.c (capture_info::walk_match): Do not mark
166 captured expressions without operands as expr_p given
167 they act more like predicates and should be subject to
168 "lost tail" side-effect preserving.
170 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
172 PR rtl-optimization/81611
173 * auto-inc-dec.c (attempt_change): Move dead note from
174 mem_insn if it's the next use of regno
175 (find_address): Take address use of reg holding
176 non-incremented value. Add parm to limit search to the named
177 reg only.
178 (merge_in_block): Attempt to use a mem insn that is the next
179 use of the original regno.
181 2018-02-27 Martin Sebor <msebor@redhat.com>
183 PR c++/83871
184 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
185 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
187 2018-02-27 Martin Sebor <msebor@redhat.com>
189 PR translation/84207
190 * diagnostic-core.h (warning_n, error_n, inform_n): Change
191 n argument to unsigned HOST_WIDE_INT.
192 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
193 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
194 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
195 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
197 2018-02-27 Richard Biener <rguenther@suse.de>
199 PR tree-optimization/84512
200 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
201 Do not use the estimate returned from record_stmt_cost for
202 the scalar iteration cost but sum properly using add_stmt_cost.
204 2018-02-27 Richard Biener <rguenther@suse.de>
206 PR tree-optimization/84466
207 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
208 Adjust last change to less strictly validate use operands.
210 2018-02-27 Martin Liska <mliska@suse.cz>
212 PR gcov-profile/84548
213 * gcov.c (process_file): Allow partial overlap and consider it
214 also as group functions.
215 (output_lines): Properly calculate range of lines for a group.
217 2018-02-27 Martin Liska <mliska@suse.cz>
219 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
220 'ggc' suffixes. Change first column width.
221 (timer::print): Fix formatting of the column.
223 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
225 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
226 preserve inline entry blocks for the sake of debug inline
227 entry point markers alone.
228 (remove_unused_locals): Suggest in comments a better place to
229 force the preservation of inline entry blocks that are
230 otherwise unused, but do not preserve them.
232 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
234 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
236 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
238 PR target/84039
239 * config/i386/constraints.md (Bs): Replace
240 ix86_indirect_branch_register with
241 TARGET_INDIRECT_BRANCH_REGISTER.
242 (Bw): Likewise.
243 * config/i386/i386.md (indirect_jump): Likewise.
244 (tablejump): Likewise.
245 (*sibcall_memory): Likewise.
246 (*sibcall_value_memory): Likewise.
247 Peepholes of indirect call and jump via memory: Likewise.
248 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
249 (*sibcall_value_GOT_32): Likewise.
250 * config/i386/predicates.md (indirect_branch_operand): Likewise.
251 (GOT_memory_operand): Likewise.
252 (call_insn_operand): Likewise.
253 (sibcall_insn_operand): Likewise.
254 (GOT32_symbol_operand): Likewise.
255 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
257 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
259 PR rtl-optimization/83496
260 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
261 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
262 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
263 redundant insn, if any.
264 (relax_delay_slots): Likewise.
265 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
267 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
269 PR tree-optimization/83965
270 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
271 that grouped statements are part of a reduction chain. Return
272 true if the statement is not marked as a reduction itself but
273 is part of a group.
274 (vect_recog_dot_prod_pattern): Don't check whether the statement
275 is part of a group here.
276 (vect_recog_sad_pattern): Likewise.
277 (vect_recog_widen_sum_pattern): Likewise.
279 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
281 PR debug/84545
282 * final.c (rest_of_clean_state): Also look for calls inside sequences.
284 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
286 PR target/84530
287 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
288 the bool argument.
289 (ix86_output_indirect_function_return): New prototype.
290 (ix86_split_simple_return_pop_internal): Likewise.
291 * config/i386/i386.c (indirect_return_via_cx): New.
292 (indirect_return_via_cx_bnd): Likewise.
293 (indirect_thunk_name): Handle return va CX_REG.
294 (output_indirect_thunk_function): Create alias for
295 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
296 (ix86_output_indirect_jmp): Remove the bool argument.
297 (ix86_output_indirect_function_return): New function.
298 (ix86_split_simple_return_pop_internal): Likewise.
299 * config/i386/i386.md (*indirect_jump): Don't pass false
300 to ix86_output_indirect_jmp.
301 (*tablejump_1): Likewise.
302 (simple_return_pop_internal): Change it to define_insn_and_split.
303 Call ix86_split_simple_return_pop_internal to split it for
304 -mfunction-return=.
305 (simple_return_indirect_internal): Call
306 ix86_output_indirect_function_return instead of
307 ix86_output_indirect_jmp.
309 2018-02-26 Jakub Jelinek <jakub@redhat.com>
311 PR bootstrap/84405
312 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
313 memset and value initialization afterwards.
315 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
317 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
319 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
321 PR target/84521
322 * common/config/aarch64/aarch64-common.c
323 (aarch_option_optimization_table[]): Switch
324 off fomit-frame-pointer
326 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
327 Chung-Ju Wu <jasonwucj@gmail.com>
329 * config/nds32/nds32-multiple.md (load_multiple): Disallow
330 volatile memory.
331 (store_multiple): Ditto.
333 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
335 * config.gcc: Add --with-cpu support for nds32 target.
336 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
337 * config/nds32/nds32.opt: Add -mcpu= option.
339 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
341 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
342 isel=yes): Warn for these deprecated options.
344 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
346 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
347 ISA_2_5_MASKS_EMBEDDED.
349 2018-02-23 Jakub Jelinek <jakub@redhat.com>
351 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
352 p->max as pointers rather than using iterative_hash_expr.
354 2018-02-23 Carl Love <cel@us.ibm.com>
356 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
357 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
358 BU_P8V_OVERLOAD_2.
359 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
360 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
361 P8V_BUILTIN_VEC_VUNSIGNED2.
363 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
365 PR target/81572
366 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
367 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
368 LRA_UNKNOWN_ALT.
369 * lra-constraints.c (curr_insn_transform): Set up
370 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
371 LRA_UNKNOWN_ALT.
372 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
373 * lra-eliminations.c (spill_pseudos): Ditto.
374 (process_insn_for_elimination): Ditto.
375 * lra-lives.c (reg_early_clobber_p): Use the new macros.
376 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
377 LRA_NON_CLOBBERED_ALT.
379 2018-02-22 Martin Sebor <msebor@redhat.com>
381 PR tree-optimization/84480
382 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
383 to maybe_diag_stxncpy_trunc. Call it.
384 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
385 from gimple_fold_builtin_strcpy. Print inlining stack.
386 (handle_builtin_stxncpy): Print inlining stack.
387 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
389 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
391 PR target/84176
392 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
393 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
394 and -fcheck-pointer-bounds are used together.
395 (indirect_thunk_prefix): New enum.
396 (indirect_thunk_need_prefix): New function.
397 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
398 "_nt" instead of "_bnd" for NOTRACK prefix.
399 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
400 (output_indirect_thunk_function): Likewise.
401 (): Likewise.
402 (ix86_code_end): Update output_indirect_thunk_function calls.
403 (ix86_output_indirect_branch_via_reg): Replace
404 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
405 (ix86_output_indirect_branch_via_push): Likewise.
406 (ix86_output_function_return): Likewise.
407 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
408 incompatible with -fcf-protection=branch and
409 -fcheck-pointer-bounds.
411 2018-02-22 Steve Ellcey <sellcey@cavium.com>
413 PR target/83335
414 * config/aarch64/aarch64.c (aarch64_print_address_internal):
415 Change gcc_assert call to output_operand_lossage.
417 2018-02-22 Steve Ellcey <sellcey@cavium.com>
419 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
421 2018-02-22 DJ Delorie <dj@redhat.com>
422 Sebastian Perta <sebastian.perta@renesas.com>
423 Oleg Endo <olegendo@gcc.gnu.org>
425 * config/rx/rx.c (rx_rtx_costs): New function.
426 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
428 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
430 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
432 2018-02-22 Martin Liska <mliska@suse.cz>
434 PR driver/83193
435 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
436 Add "native" as a possible value.
438 2018-02-22 Martin Liska <mliska@suse.cz>
440 PR driver/83193
441 * config/i386/i386.c (ix86_option_override_internal):
442 Add "native" as a possible value for -march and -mtune.
444 2018-02-22 Jakub Jelinek <jakub@redhat.com>
446 PR target/84502
447 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
448 to all type variants.
450 PR tree-optimization/84503
451 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
452 width as info->bitpos + info->bitsize - start.
453 (merged_store_group::merge_overlapping): Simplify width computation.
454 (check_no_overlap): New function.
455 (imm_store_chain_info::try_coalesce_bswap): Compute expected
456 start + width and last_order of the group, fail if check_no_overlap
457 fails.
458 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
459 to group if check_no_overlap fails.
461 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
463 * config/rs6000/altivec.md: Delete contraint arguments to
464 define_expand, define_split, and define_peephole2, and in
465 define_insn_and_split if always unused.
466 * config/rs6000/darwin.md: Ditto.
467 * config/rs6000/dfp.md: Ditto.
468 * config/rs6000/rs6000.md: Ditto.
469 * config/rs6000/sync.md: Ditto.
470 * config/rs6000/vector.md: Ditto.
471 * config/rs6000/vsx.md: Ditto.
473 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
475 * config/rs6000/altivec.md: Write output control strings as braced
476 blocks instead of double-quoted strings.
477 * config/rs6000/darwin.md: Ditto.
478 * config/rs6000/rs6000.md: Ditto.
479 * config/rs6000/vector.md: Ditto.
480 * config/rs6000/vsx.md: Ditto.
482 2018-02-21 Jason Merrill <jason@redhat.com>
484 PR c++/84314 - ICE with templates and fastcall attribute.
485 * attribs.c (build_type_attribute_qual_variant): Remove assert.
487 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
489 * ipa-cp.c (determine_versionability): Fix comment typos.
491 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
493 PR c/84229
494 * ipa-cp.c (determine_versionability): Do not version functions caling
495 va_arg_pack.
497 2018-02-21 Martin Liska <mliska@suse.cz>
499 PR driver/83193
500 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
501 Add "native" as a possible value.
502 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
503 the macro when native cpu detection is available.
505 2018-02-21 Martin Liska <mliska@suse.cz>
507 PR driver/83193
508 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
509 Add "native" as a possible value.
510 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
511 when native cpu detection is available.
513 2018-02-21 Jakub Jelinek <jakub@redhat.com>
514 Martin Sebor <msebor@redhat.com>
516 PR tree-optimization/84478
517 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
518 false.
519 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
520 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
521 support which is conservatively correct, for 2 only stay conservative
522 for maxlen. Formatting and comment capitalization fixes. Add STRICT
523 argument to the 2 argument get_range_strlen, adjust 6 arg
524 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
525 false.
526 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
527 (gimple_fold_builtin_strlen): Pass true as last argument to
528 get_range_strlen.
530 2018-02-20 Martin Sebor <msebor@redhat.com>
532 PR middle-end/84095
533 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
534 (builtin_memref::set_base_and_offset): Same. Handle inner references.
535 (builtin_memref::builtin_memref): Factor out parts into
536 set_base_and_offset and call it.
538 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
540 PR middle-end/84406
541 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
542 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
543 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
544 search at the associated MODE_INT.
546 2018-02-20 Jeff Law <law@redhat.com>
548 PR middle-end/82123
549 PR tree-optimization/81592
550 PR middle-end/79257
551 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
552 for range data rather than using global data.
553 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
554 range data rather than using global data.
555 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
556 pass it to children as needed.
557 (struct directive::fmtresult): Similarly.
558 (struct directive::set_width): Similarly.
559 (struct directive::set_precision): Similarly.
560 (format_integer, format_directive, parse_directive): Similarly.
561 (format_none): Accept unnamed vr_values parameter.
562 (format_percent, format_floating, format_character): Similarly.
563 (format_string, format_plain): Similarly.
564 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
565 the EVRP range analyzer for range data rather than using global data.
566 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
567 gimple-ssa-evrp-analyze.h
568 (class sprintf_dom_walker): Add after_dom_children member function.
569 Add evrp_range_analyzer member.
570 (sprintf_dom_walker::before_dom_children): Call into the EVRP
571 range analyzer as needed.
572 (sprintf_dom_walker::after_dom_children): New member function.
573 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
574 if not optimizing.
575 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
576 (evrp_range_analyzer::pop_to_marker): Likewise.
578 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
580 PR tree-optimization/84419
581 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
582 with the required type if its current type is compatible but
583 different.
585 2018-02-20 Jakub Jelinek <jakub@redhat.com>
587 PR middle-end/82004
588 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
589 after vectorization.
591 2018-02-20 Martin Liska <mliska@suse.cz>
593 PR driver/83193
594 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
595 possible values if we don't have a hint.
597 2018-02-20 Martin Liska <mliska@suse.cz>
599 PR c/84310
600 PR target/79747
601 * final.c (shorten_branches): Build align_tab array with one
602 more element.
603 * opts.c (finish_options): Add alignment option limit check.
604 (MAX_CODE_ALIGN): Likewise.
605 (MAX_CODE_ALIGN_VALUE): Likewise.
606 * doc/invoke.texi: Document maximum allowed option value for
607 all -falign-* options.
609 2018-02-19 Jakub Jelinek <jakub@redhat.com>
611 PR target/84146
612 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
613 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
614 * var-tracking.c (emit_note_insn_var_location): Remove all references
615 to NOTE_INSN_CALL_ARG_LOCATION.
616 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
617 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
618 Use copy_rtx_if_shared.
619 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
620 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
621 (dwarf2out_var_location): Remove handling of
622 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
623 on call_insn.
624 * final.c (final_scan_insn): Remove all references to
625 NOTE_INSN_CALL_ARG_LOCATION.
626 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
627 before dumping final insns.
628 * except.c (emit_note_eh_region_end): Remove all references to
629 NOTE_INSN_CALL_ARG_LOCATION.
630 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
631 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
632 * config/arc/arc.c (hwloop_optimize): Likewise.
633 * config/arm/arm.c (create_fix_barrier): Likewise.
634 * config/s390/s390.c (s390_chunkify_start): Likewise.
635 * config/sh/sh.c (find_barrier): Likewise.
636 * config/i386/i386.c (rest_of_insert_endbranch,
637 ix86_seh_fixup_eh_fallthru): Likewise.
638 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
639 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
640 * config/frv/frv.c (frv_function_prologue): Likewise.
641 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
642 reg note.
643 (note_outside_basic_block_p): Remove all references to
644 NOTE_INSN_CALL_ARG_LOCATION.
645 * gengtype.c (adjust_field_rtx_def): Likewise.
646 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
647 Likewise.
648 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
649 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
651 PR c++/84444
652 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
653 is ADDR_EXPR.
655 PR tree-optimization/84452
656 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
657 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
658 is NULL.
660 2018-02-19 Martin Liska <mliska@suse.cz>
662 PR sanitizer/82183
663 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
665 2018-02-19 Martin Liska <mliska@suse.cz>
666 Richard Sandiford <richard.sandiford@linaro.org>
668 PR tree-optimization/82491
669 * gimple-fold.c (get_base_constructor): Make earlier bail out
670 to prevent ubsan.
672 2018-02-19 Carl Love <cel@us.ibm.com>
674 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
675 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
676 BU_P8V_OVERLOAD_1.
677 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
678 P8V_BUILTIN_VEC_NEG.
680 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
682 * config/rl78/rl78.md (movdf): New define expand.
684 2018-02-19 Martin Liska <mliska@suse.cz>
686 PR other/80589
687 * doc/invoke.texi: Fix typo.
688 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
690 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
692 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
693 handle rs6000_single_float and rs6000_double_float specially for
694 e500 family CPUs.
696 2018-02-16 Jeff Law <law@redhat.com>
698 * config/rx/rx.c (add_pop_cfi_notes): New function.;
699 (pop_regs): Use it.
701 2018-02-16 Jakub Jelinek <jakub@redhat.com>
703 PR ipa/84425
704 * ipa-inline.c (inline_small_functions): Fix a typo.
706 2018-02-16 Nathan Sidwell <nathan@acm.org>
708 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
710 2018-02-16 Carl Love <cel@us.ibm.com>
712 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
713 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
714 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
715 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
716 expansion to P8V_BUILTIN_VEC_FLOAT2.
718 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
720 PR rtl-optimization/70023
721 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
722 src_regno into account.
724 2018-02-16 Carl Love <cel@us.ibm.com>
726 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
727 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
728 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
729 * config/rs6000/rs6000.c: Remove case statements for
730 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
731 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
732 and P9V_BUILTIN_VEC_VINSERT4B.
733 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
734 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
735 * config/rs6000/vsx.md:
736 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
737 vec_insert4b.
739 2018-02-16 Carl Love <cel@us.ibm.com>
741 * config/rs6000/altivec.h: Add builtin names vec_extract4b
742 vec_insert4b.
743 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
744 definitions.
745 * config/rs6000/rs6000-c.c: Add the definitions for
746 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
747 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
748 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
749 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
750 definition for insert4b and define insn *insert3b_internal.
751 * doc/extend.texi: Add documentation for vec_extract4b.
753 2018-02-16 Nathan Sidwell <nathan@acm.org>
755 * doc/extend.texi (Backwards Compatibility): Mention friend
756 injection. Note for-scope is deprecated.
757 * doc/invoke.texi (-ffriend-injection): Deprecate.
759 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
761 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
762 that moved to I2, also allow destinations that are a paradoxical
763 subreg (instead of a normal reg).
765 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
767 PR target/83831
768 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
769 to QImode.
771 2018-02-16 Richard Biener <rguenther@suse.de>
773 PR tree-optimization/84037
774 PR tree-optimization/84016
775 PR target/82862
776 * config/i386/i386.c (ix86_builtin_vectorization_cost):
777 Adjust vec_construct for the fact we need additional higher latency
778 128bit inserts for AVX256 and AVX512 vector builds.
779 (ix86_add_stmt_cost): Scale vector construction cost for
780 elementwise loads.
782 2018-02-16 Richard Biener <rguenther@suse.de>
784 PR tree-optimization/84417
785 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
786 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
787 (non_rewritable_lvalue_p): Likewise, use poly-ints.
789 2018-02-16 Martin Liska <mliska@suse.cz>
791 PR sanitizer/84307
792 * internal-fn.def (ASAN_CHECK): Set proper flags.
793 (ASAN_MARK): Likewise.
795 2018-02-16 Julia Koval <julia.koval@intel.com>
797 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
798 from PTA_CANNONLAKE.
800 2018-02-16 Jakub Jelinek <jakub@redhat.com>
802 PR target/84272
803 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
804 Use ++iter rather than iter++ for std::list iterators.
805 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
806 defer deleting them until all nodes in the forest are processed. Do
807 free even leaf nodes. Change to_process into auto_vec.
809 PR bootstrap/84405
810 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
811 * vec.h (vec_default_construct): Use memset instead of placement new
812 if BROKEN_VALUE_INITIALIZATION is defined.
813 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
814 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
815 is defined.
817 PR rtl-optimization/83723
818 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
819 * lra.c (lra_substitute_pseudo): Likewise. If true, use
820 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
821 recursive calls.
822 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
823 callers.
824 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
826 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
828 PR rtl-optimization/81443
829 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
830 from inner REGs to paradoxical SUBREGs.
832 2018-02-16 Richard Biener <rguenther@suse.de>
834 PR tree-optimization/84399
835 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
836 For operands we can analyze at their definition make sure we can
837 analyze them at each use as well.
839 2018-02-16 Richard Biener <rguenther@suse.de>
841 PR tree-optimization/84190
842 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
843 volatile accesses if the decl isn't volatile.
845 2018-02-15 Jason Merrill <jason@redhat.com>
847 PR c++/84314 - ICE with templates and fastcall attribute.
848 * attribs.c (build_type_attribute_qual_variant): Don't clobber
849 TYPE_CANONICAL on an existing type.
851 2018-02-15 Jakub Jelinek <jakub@redhat.com>
853 PR tree-optimization/84383
854 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
855 dstoff nor call operand_equal_p if dstbase is NULL.
857 PR tree-optimization/84334
858 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
859 also a CONSTANT_CLASS_P, punt.
861 2018-02-14 Jim Wilson <jimw@sifive.com>
863 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
864 first SMALL_OPERAND check. New local min_second_step. Move assert
865 to where locals are set. Add TARGET_RVC support.
866 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
868 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
870 * doc/invoke.texi: Correct -Wformat-overflow code sample.
872 2018-02-14 Martin Sebor <msebor@redhat.com>
874 PR tree-optimization/83698
875 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
876 arrays constrain the offset range to their bounds.
877 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
878 (builtin_access::overlap): Avoid setting the size of overlap if it's
879 already been set.
880 (maybe_diag_overlap): Also consider arrays when deciding what values
881 of offsets to include in diagnostics.
883 2018-02-14 Martin Sebor <msebor@redhat.com>
885 PR c/84108
886 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
887 that correspond to the kind of a declaration.
889 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
891 PR target/83984
892 * config/pa/pa.md: Load address of PIC label using the linkage table
893 if the label is nonlocal.
895 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
897 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
898 warning message if user requests -maltivec=be.
899 * doc/invoke.texi: Document deprecation of -maltivec=be.
901 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
903 PR target/84220
904 * config/rs6000/rs6000-c.c: Update definitions for
905 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
906 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
908 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
910 PR target/84239
911 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
912 add _get_ssp intrinsics. Remove argument from
913 __builtin_ia32_rdssp[d|q].
914 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
915 * config/i386/i386-builtin.def: Remove argument from
916 __builtin_ia32_rdssp[d|q].
917 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
918 ix86_expand_special_args_builtin for _rdssp[d|q].
919 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
920 Clear register before usage.
921 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
922 Add documentation for new _get_ssp and _inc_ssp intrinsics.
924 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
926 PR tree-optimization/84357
927 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
928 operand 1 of an ARRAY_REF too.
930 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
932 PR target/83831
933 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
934 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
935 declarations.
936 (set_of_reg): New struct.
937 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
938 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
939 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
940 functions.
941 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
942 Split into bitclr, bitset, bitinvert patterns if appropriate.
943 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
944 use rx_fuse_in_memory_bitop.
945 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
946 to named insn, correct maximum insn length.
948 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
950 PR target/79242
951 * machmode.def: Define a complex mode for PARTIAL_INT.
952 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
953 MODE_PARTIAL_INT.
954 * doc/rtl.texi: Document CSPImode.
955 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
956 handling.
957 (msp430_hard_regno_nregs_with_padding): Likewise.
959 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
961 PR target/84279
962 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
964 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
966 PR rtl-optimization/84169
967 * combine.c (try_combine): New variable split_i2i3. Set it to true if
968 we generated a parallel as new i3 and we split that to new i2 and i3
969 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
970 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
971 those to i2, not i1. Partially rewrite this scan code.
973 2018-02-13 Jakub Jelinek <jakub@redhat.com>
975 PR c/82210
976 * stor-layout.c (place_field): For variable length fields, adjust
977 offset_align afterwards not just based on the field's alignment,
978 but also on the size.
980 PR middle-end/84309
981 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
982 of exps and logs in the use_exp2 case.
984 2018-02-13 Jeff Law <law@redhat.com>
986 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
987 entry for "vector".
989 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
990 ARGS as unused.
992 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
994 PR debug/84342
995 PR debug/84319
996 * common.opt (gas-loc-support, gas-locview-support): New.
997 (ginline-points, ginternal-reset-location-views): New.
998 * doc/invoke.texi: Document them. Use @itemx where intended.
999 (gvariable-location-views): Adjust.
1000 * target.def (reset_location_view): New.
1001 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
1002 (TARGET_RESET_LOCATION_VIEW): New.
1003 * doc/tm.texi: Rebuilt.
1004 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
1005 (dwarf2out_default_as_locview_support): New.
1006 (output_asm_line_debug_info): Use option variables.
1007 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
1008 (output_loc_list): Likewise.
1009 (add_high_low_attributes): Check option variables.
1010 Don't output entry view attribute in strict mode.
1011 (gen_inlined_subroutine_die): Check option variables.
1012 (dwarf2out_inline_entry): Likewise.
1013 (init_sections_and_labels): Likewise.
1014 (dwarf2out_early_finish): Likewise.
1015 (maybe_reset_location_view): New, from...
1016 (dwarf2out_var_location): ... here. Call it.
1017 * debug.h (dwarf2out_default_as_loc_support): Declare.
1018 (dwarf2out_default_as_locview_support): Declare.
1019 * hooks.c (hook_int_rtx_insn_0): New.
1020 * hooks.h (hook_int_rtx_insn_0): Declare.
1021 * toplev.c (process_options): Take -gas-loc-support and
1022 -gas-locview-support from dwarf2out. Enable
1023 -gvariable-location-views by default only with locview
1024 assembler support. Enable -ginternal-reset-location-views by
1025 default only if the target defines the corresponding hook.
1026 Enable -ginline-points by default if location views are
1027 enabled; force it disabled if statement frontiers are
1028 disabled.
1029 * tree-inline.c (expand_call_inline): Check option variables.
1030 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1032 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
1034 PR tree-optimization/84321
1035 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
1036 handling. Also check whether the anti-range contains any values
1037 that satisfy the mask; switch to a VR_RANGE if not.
1039 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
1041 PR sanitizer/84340
1042 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
1044 2018-02-13 Martin Jambor <mjambor@suse.cz>
1046 PR c++/83990
1047 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
1048 of call statements, also set location of a load to a temporary.
1050 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
1052 * config/rl78/rl78.c (add_vector_labels): New function.
1053 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
1054 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
1055 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
1056 which checks that no arguments are passed.
1057 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
1058 * doc/extend.texi: Documentation for the new attribute.
1060 2018-02-13 Andreas Schwab <schwab@suse.de>
1062 * config/riscv/linux.h (CPP_SPEC): Define.
1064 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1066 PR target/84335
1067 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
1068 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
1069 OPTION_MASK_ISA_AES as first argument to def_builtin_const
1070 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
1071 instead of OPTION_MASK_ISA_PCLMUL as first argument to
1072 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
1073 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
1074 temporarily for AES and PCLMUL builtins.
1076 PR tree-optimization/84339
1077 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
1078 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
1079 Formatting fixes.
1081 PR middle-end/84309
1082 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
1083 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
1084 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
1085 inline function.
1086 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
1087 inline function.
1088 * omp-simd-clone.h: New file.
1089 * omp-simd-clone.c: Include omp-simd-clone.h.
1090 (expand_simd_clones): No longer static.
1091 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
1092 cgraph.h and omp-simd-clone.h.
1093 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1094 (vect_recog_widen_shift_pattern): Formatting fix.
1095 (vect_pattern_recog_1): Don't check optab for calls.
1097 PR target/84336
1098 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1099 operands[2] into a REG before using gen_lowpart on it.
1101 2018-02-12 Jeff Law <law@redhat.com>
1103 PR target/83760
1104 * config/sh/sh.c (find_barrier): Consider a sibling call
1105 a barrier as well.
1107 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1108 successfully back substituting a reg.
1110 2018-02-12 Richard Biener <rguenther@suse.de>
1112 PR tree-optimization/84037
1113 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1114 parameter, move visited init to caller.
1115 (vect_slp_analyze_operations): Separate cost from validity
1116 check, initialize visited once for all instances.
1117 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1118 for all instances.
1119 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1120 out an assert.
1121 (vect_model_promotion_demotion_cost): Likewise.
1122 (vectorizable_bswap): Guard cost modeling with !slp_node
1123 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
1124 SLP stmts.
1125 (vectorizable_call): Likewise.
1126 (vectorizable_conversion): Likewise.
1127 (vectorizable_assignment): Likewise.
1128 (vectorizable_shift): Likewise.
1129 (vectorizable_operation): Likewise.
1130 (vectorizable_store): Likewise.
1131 (vectorizable_load): Likewise.
1132 (vectorizable_condition): Likewise.
1133 (vectorizable_comparison): Likewise.
1135 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
1137 PR sanitizer/84307
1138 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
1139 (ASAN_MARK): Fix fnspec to account for return value, change pointer
1140 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
1142 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1144 PR middle-end/83665
1145 * params.def (inline-min-speedup): Increase from 8 to 15.
1146 (max-inline-insns-auto): Decrease from 40 to 30.
1147 * ipa-split.c (consider_split): Add some buffer for function to
1148 be considered inlining candidate.
1149 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
1150 default values.
1152 2018-02-12 Richard Biener <rguenther@suse.de>
1154 PR tree-optimization/84037
1155 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
1156 matched stmts if we cannot swap the non-matched ones.
1158 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1160 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
1161 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
1162 _mm_maskz_scalef_round_ss): New intrinsics.
1163 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
1164 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
1165 __builtin_ia32_scalefss_round): Remove.
1166 (__builtin_ia32_scalefsd_mask_round,
1167 __builtin_ia32_scalefss_mask_round): New intrinsics.
1168 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
1169 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
1170 ((match_operand:VF_128 2 "<round_nimm_predicate>"
1171 "<round_constraint>")): Changed to ...
1172 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
1173 "<round_scalar_constraint>")): ... this.
1174 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
1175 %0, %1, %2<round_op3>}"): Changed to ...
1176 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
1177 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
1178 %2<round_scalar_mask_op3>}"): ... this.
1179 * config/i386/subst.md (round_scalar_nimm_predicate): New.
1181 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1183 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
1184 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
1185 (_mm_maskz_sqrt_round_ss): New intrinsics.
1186 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
1187 (__builtin_ia32_sqrtsd_mask_round)
1188 (__builtin_ia32_sqrtss_mask_round): New builtins.
1189 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
1190 (__builtin_ia32_sqrtss_round): Remove.
1191 (__builtin_ia32_sqrtsd_mask_round)
1192 (__builtin_ia32_sqrtss_mask_round): New builtins.
1193 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
1194 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
1195 ((match_operand:VF_128 1 "vector_operand"
1196 "xBm,<round_constraint>")): Changed to ...
1197 ((match_operand:VF_128 1 "vector_operand"
1198 "xBm,<round_scalar_constraint>")): ... this.
1199 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
1200 %0, %2, %<iptr>1<round_op3>}): Changed to ...
1201 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
1202 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
1203 %<iptr>1<round_scalar_mask_op3>}): ... this.
1204 ((set_attr "prefix" "<round_prefix>")): Changed to ...
1205 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
1207 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
1209 PR target/84266
1210 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
1211 Cast vec_cmpeq result to correct type.
1212 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
1213 Cast vec_cmpgt result to correct type.
1215 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
1217 * final.c (final_scan_insn_1): Renamed from...
1218 (final_scan_insn): ... this. New wrapper, to recover
1219 seen from the outermost call in recursive ones.
1220 * config/sparc/sparc.c (output_return): Drop seen from call.
1221 (output_sibcall): Likewise.
1222 * config/visium/visium.c (output_branch): Likewise.
1224 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
1226 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
1227 function label.
1229 2018-02-10 Alan Modra <amodra@gmail.com>
1231 PR target/84300
1232 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
1233 Specify LR as an input.
1235 2018-02-10 Jakub Jelinek <jakub@redhat.com>
1237 PR sanitizer/83987
1238 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
1239 remove_member_access_dummy_vars): New functions.
1240 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
1241 lower_omp_1, execute_lower_omp): Use them.
1243 PR rtl-optimization/84308
1244 * shrink-wrap.c (spread_components): Release todo vector.
1246 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
1248 PR rtl-optimization/57193
1249 * ira-color.c (struct allocno_color_data): Add member
1250 conflict_allocno_hard_prefs.
1251 (update_conflict_allocno_hard_prefs): New.
1252 (bucket_allocno_compare_func): Add a preference based on
1253 conflict_allocno_hard_prefs.
1254 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
1255 (color_allocnos): Remove a dead code. Initiate
1256 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
1258 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1260 PR target/84226
1261 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
1262 constraint from =wa to wa. Avoid a subreg on the output operand,
1263 instead use a pseudo and subreg it in a move.
1264 (p9_xxbrd_<mode>): Changed to ...
1265 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
1266 (p9_xxbrd_v2df): New expander.
1267 (p9_xxbrw_<mode>): Changed to ...
1268 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
1269 (p9_xxbrw_v4sf): New expander.
1271 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1273 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
1275 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
1277 PR target/83926
1278 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
1279 multiply in 32-bit mode.
1280 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
1281 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
1282 mode.
1284 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1286 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
1287 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
1288 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
1289 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
1291 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
1293 PR lto/84213
1294 * dwarf2out.c (is_trivial_indirect_ref): New function.
1295 (dwarf2out_late_global_decl): Do not generate a location
1296 attribute for variables that have a non-trivial DECL_VALUE_EXPR
1297 and that are not defined in the current unit.
1299 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1301 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
1302 instead of a libcall for UNORDERED.
1304 2018-02-09 Tamar Christina <tamar.christina@arm.com>
1306 PR target/82641
1307 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
1308 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
1310 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1312 PR target/PR84295
1313 * config/s390/s390.c (s390_set_current_function): Invoke
1314 s390_indirect_branch_settings also if fndecl didn't change.
1316 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1318 * config/rs6000/rs6000.md (blockage): Set length to zero.
1320 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1322 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
1324 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1326 PR sanitizer/84285
1327 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
1328 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
1329 -static-lib*san.
1331 PR debug/84252
1332 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
1333 PARALLEL incoming that failed vt_get_decl_and_offset check.
1335 PR middle-end/84237
1336 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
1337 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
1338 TREE_READONLY bit.
1339 (get_variable_section): For decls in named .bss* sections pass true as
1340 second argument to bss_initializer_p.
1342 2018-02-09 Marek Polacek <polacek@redhat.com>
1343 Jakub Jelinek <jakub@redhat.com>
1345 PR c++/83659
1346 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
1347 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
1348 Sync some changes from cxx_fold_indirect_ref.
1350 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1352 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
1353 markers.
1354 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
1355 (BLOCK_INLINE_ENTRY_LABEL): New.
1356 (dwarf2out_var_location): Disregard inline entry markers.
1357 (inline_entry_data): New struct.
1358 (inline_entry_data_hasher): New hashtable type.
1359 (inline_entry_data_hasher::hash): New.
1360 (inline_entry_data_hasher::equal): New.
1361 (inline_entry_data_table): New variable.
1362 (add_high_low_attributes): Add DW_AT_entry_pc and
1363 DW_AT_GNU_entry_view attributes if a pending entry is found
1364 in inline_entry_data_table. Add old entry_pc attribute only
1365 if debug nonbinding markers are disabled.
1366 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
1367 markers are enabled.
1368 (block_within_block_p, dwarf2out_inline_entry): New.
1369 (dwarf2out_finish): Check that no entries remained in
1370 inline_entry_data_table.
1371 * final.c (reemit_insn_block_notes): Handle inline entry notes.
1372 (final_scan_insn, notice_source_line): Likewise.
1373 (rest_of_clean_state): Skip inline entry markers.
1374 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
1375 markers.
1376 * gimple.c (gimple_build_debug_inline_entry): New.
1377 * gimple.h (enum gimple_debug_subcode): Add
1378 GIMPLE_DEBUG_INLINE_ENTRY.
1379 (gimple_build_debug_inline_entry): Declare.
1380 (gimple_debug_inline_entry_p): New.
1381 (gimple_debug_nonbind_marker_p): Adjust.
1382 * insn-notes.def (INLINE_ENTRY): New.
1383 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
1384 inline entry marker notes.
1385 (print_insn): Likewise.
1386 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
1387 (INSN_DEBUG_MARKER_KIND): Likewise.
1388 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
1389 * tree-inline.c (expand_call_inline): Build and insert
1390 debug_inline_entry stmt.
1391 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
1392 inline entry blocks early, if nonbind markers are enabled.
1393 (dump_scope_block): Dump fragment info.
1394 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
1395 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
1396 (gimple_build_debug_inline_entry): New.
1397 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
1398 Enable/disable inline entry points too.
1399 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
1400 (DEBUG_INSN): Describe inline entry markers.
1402 * common.opt (gvariable-location-views): New.
1403 (gvariable-location-views=incompat5): New.
1404 * config.in: Rebuilt.
1405 * configure: Rebuilt.
1406 * configure.ac: Test assembler for view support.
1407 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
1408 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
1409 * dwarf2out.c (var_loc_view): New typedef.
1410 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
1411 (dwarf2out_locviews_in_attribute): New.
1412 (dwarf2out_locviews_in_loclist): New.
1413 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
1414 (enum dw_line_info_opcode): Add LI_adv_address.
1415 (struct dw_line_info_table): Add view.
1416 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
1417 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
1418 (zero_view_p): New variable.
1419 (ZERO_VIEW_P): New macro.
1420 (output_asm_line_debug_info): New.
1421 (struct var_loc_node): Add view.
1422 (add_AT_view_list, AT_loc_list): New.
1423 (add_var_loc_to_decl): Add view param. Test it against last.
1424 (new_loc_list): Add view params. Record them.
1425 (AT_loc_list_ptr): Handle loc and view lists.
1426 (view_list_to_loc_list_val_node): New.
1427 (print_dw_val): Handle dw_val_class_view_list.
1428 (size_of_die): Likewise.
1429 (value_format): Likewise.
1430 (loc_list_has_views): New.
1431 (gen_llsym): Set vl_symbol too.
1432 (maybe_gen_llsym, skip_loc_list_entry): New.
1433 (dwarf2out_maybe_output_loclist_view_pair): New.
1434 (output_loc_list): Output view list or entries too.
1435 (output_view_list_offset): New.
1436 (output_die): Handle dw_val_class_view_list.
1437 (output_dwarf_version): New.
1438 (output_compilation_unit_header): Use it.
1439 (output_skeleton_debug_sections): Likewise.
1440 (output_rnglists, output_line_info): Likewise.
1441 (output_pubnames, output_aranges): Update version comments.
1442 (output_one_line_info_table): Output view numbers in asm comments.
1443 (dw_loc_list): Determine current endview, pass it to new_loc_list.
1444 Call maybe_gen_llsym.
1445 (loc_list_from_tree_1): Adjust.
1446 (add_AT_location_description): Create view list attribute if
1447 needed, check it's absent otherwise.
1448 (convert_cfa_to_fb_loc_list): Adjust.
1449 (maybe_emit_file): Call output_asm_line_debug_info for test.
1450 (dwarf2out_var_location): Reset views as needed. Precompute
1451 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
1452 attribute. Set view.
1453 (new_line_info_table): Reset next view.
1454 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
1455 (dwarf2out_source_line): Likewise. Output view resets and labels to
1456 the assembler, or select appropriate line info opcodes.
1457 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
1458 (optimize_string_length): Catch it. Adjust.
1459 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
1460 dw_val_class_view_list, and remove it if no longer needed.
1461 (hash_loc_list): Hash view numbers.
1462 (loc_list_hasher::equal): Compare them.
1463 (optimize_location_lists): Check whether a view list symbol is
1464 needed, and whether the locview attribute is present, and
1465 whether they match. Remove the locview attribute if no longer
1466 needed.
1467 (index_location_lists): Call skip_loc_list_entry for test.
1468 (dwarf2out_finish): Call output_asm_line_debug_info for test.
1469 Use output_dwarf_version.
1470 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
1471 (struct dw_val_node): Add val_view_list.
1472 * final.c (SEEN_NEXT_VIEW): New.
1473 (set_next_view_needed): New.
1474 (clear_next_view_needed): New.
1475 (maybe_output_next_view): New.
1476 (final_start_function): Rename to...
1477 (final_start_function_1): ... this. Take pointer to FIRST,
1478 add SEEN parameter. Emit param bindings in the initial view.
1479 (final_start_function): Reintroduce SEEN-less interface.
1480 (final): Rename to...
1481 (final_1): ... this. Take SEEN parameter. Output final pending
1482 next view at the end.
1483 (final): Reintroduce seen-less interface.
1484 (final_scan_insn): Output pending next view before switching
1485 sections or ending a block. Mark the next view as needed when
1486 outputting variable locations. Notify debug backend of section
1487 changes, and of location view changes.
1488 (rest_of_handle_final): Adjust.
1489 * toplev.c (process_options): Autodetect value for debug variable
1490 location views option. Warn on incompat5 without -gdwarf-5.
1491 * doc/invoke.texi (gvariable-location-views): New.
1492 (gvariable-location-views=incompat5): New.
1493 (gno-variable-location-views): New.
1495 2018-02-08 David Malcolm <dmalcolm@redhat.com>
1497 PR tree-optimization/84136
1498 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
1499 that the result of find_edge is non-NULL.
1501 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1503 PR target/83008
1504 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
1505 storing integer register in SImode. Fix cost of 256 and 512
1506 byte aligned SSE register store.
1508 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1510 * config/i386/i386.c (ix86_multiplication_cost): Fix
1511 multiplication cost for TARGET_AVX512DQ.
1513 2018-02-08 Marek Polacek <polacek@redhat.com>
1515 PR tree-optimization/84238
1516 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
1517 get_range_strlen.
1519 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1521 PR tree-optimization/84265
1522 * tree-vect-stmts.c (vectorizable_store): Don't treat
1523 VMAT_CONTIGUOUS accesses as grouped.
1524 (vectorizable_load): Likewise.
1526 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1528 PR tree-optimization/81635
1529 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
1530 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
1531 (test_round_for_mask): New functions.
1532 (wide_int_cc_tests): Call test_round_for_mask.
1533 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
1534 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
1535 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
1536 range returned by get_range_info.
1538 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1540 PR ipa/81360
1541 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
1542 * symtab.c: Include builtins.h
1543 (symtab_node::output_to_lto_symbol_table_p): Move here
1544 from lto-streamer-out.c:output_symbol_p.
1545 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
1546 (output_symbol_p): Move all logic to symtab.c
1547 (produce_symtab): Update.
1549 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1551 * config/s390/s390-opts.h (enum indirect_branch): Define.
1552 * config/s390/s390-protos.h (s390_return_addr_from_memory)
1553 (s390_indirect_branch_via_thunk)
1554 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
1555 (enum s390_indirect_branch_type): Define.
1556 * config/s390/s390.c (struct s390_frame_layout, struct
1557 machine_function): Remove.
1558 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
1559 (indirect_branch_table_label_no, indirect_branch_table_name):
1560 Define variables.
1561 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
1562 (enum s390_indirect_branch_option): Define.
1563 (s390_return_addr_from_memory): New function.
1564 (s390_handle_string_attribute): New function.
1565 (s390_attribute_table): Add new attribute handler.
1566 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
1567 (s390_indirect_branch_via_thunk): New function.
1568 (s390_indirect_branch_via_inline_thunk): New function.
1569 (s390_function_ok_for_sibcall): When jumping via thunk disallow
1570 sibling call optimization for non z10 compiles.
1571 (s390_emit_call): Force indirect branch target to be a single
1572 register. Add r1 clobber for non-z10 compiles.
1573 (s390_emit_epilogue): Emit return jump via return_use expander.
1574 (s390_reorg): Handle JUMP_INSNs as execute targets.
1575 (s390_option_override_internal): Perform validity checks for the
1576 new command line options.
1577 (s390_indirect_branch_attrvalue): New function.
1578 (s390_indirect_branch_settings): New function.
1579 (s390_set_current_function): Invoke s390_indirect_branch_settings.
1580 (s390_output_indirect_thunk_function): New function.
1581 (s390_code_end): Implement target hook.
1582 (s390_case_values_threshold): Implement target hook.
1583 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
1584 macros.
1585 * config/s390/s390.h (struct s390_frame_layout)
1586 (struct machine_function): Move here from s390.c.
1587 (TARGET_INDIRECT_BRANCH_NOBP_RET)
1588 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
1589 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
1590 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
1591 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
1592 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
1593 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
1594 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
1595 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
1596 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
1597 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
1598 (mnemonic attribute): Add values which aren't recognized
1599 automatically.
1600 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
1601 pattern for branch conversion. Fix mnemonic attribute.
1602 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
1603 indirect branch via thunk if requested.
1604 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
1605 ("*indirect_jump"): Disable for branch conversion using out of
1606 line thunks.
1607 ("indirect_jump_via_thunk<mode>_z10")
1608 ("indirect_jump_via_thunk<mode>")
1609 ("indirect_jump_via_inlinethunk<mode>_z10")
1610 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
1611 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
1612 ("casesi_jump_via_inlinethunk<mode>_z10")
1613 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
1614 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
1615 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
1616 ("*indirect2_jump"): Disable for branch conversion.
1617 ("casesi_jump"): Turn into expander and expand patterns for branch
1618 conversion.
1619 ("return_use"): New expander.
1620 ("*return"): Emit return via thunk and rename it to ...
1621 ("*return<mode>"): ... this one.
1622 * config/s390/s390.opt: Add new options and and enum for the
1623 option values.
1625 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1627 * lra-constraints.c (match_reload): Unconditionally use
1628 gen_lowpart_SUBREG, rather than selecting between that
1629 and equivalent gen_rtx_SUBREG code.
1631 2018-02-08 Richard Biener <rguenther@suse.de>
1633 PR tree-optimization/84233
1634 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
1635 changed flag instead of boguously re-using phi_inserted.
1637 2018-02-08 Martin Jambor <mjambor@suse.cz>
1639 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
1640 static local variables.
1642 2018-02-08 Richard Biener <rguenther@suse.de>
1644 PR tree-optimization/84278
1645 * tree-vect-stmts.c (vectorizable_store): When looking for
1646 smaller vector types to perform grouped strided loads/stores
1647 make sure the mode is supported by the target.
1648 (vectorizable_load): Likewise.
1650 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1652 * config/aarch64/aarch64.c (aarch64_components_for_bb):
1653 Increase LDP/STP opportunities by adding adjacent callee-saves.
1655 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1657 PR rtl-optimization/84068
1658 PR rtl-optimization/83459
1659 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
1661 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
1663 PR tree-optimization/84224
1664 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
1665 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
1666 non-zero arguments.
1668 2018-02-07 Iain Sandoe <iain@codesourcery.com>
1670 PR target/84113
1671 * config/rs6000/altivec.md (*restore_world): Remove LR use.
1672 * config/rs6000/predicates.md (restore_world_operation): Adjust op
1673 count, remove one USE.
1675 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
1677 * doc/install.texi (Configuration): Document the
1678 --with-long-double-format={ibm,ieee} PowerPC configuration
1679 options.
1681 PR target/84154
1682 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
1683 Convert from define_expand to be define_insn_and_split. Rework
1684 float/double/_Float128 conversions to QI/HI/SImode to work with
1685 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
1686 conversions to QI/HImode types did a store and then a load to
1687 truncate the value. For conversions to VSX registers, don't split
1688 the insn, instead emit the code directly. Use the code iterator
1689 any_fix to combine signed and unsigned conversions.
1690 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
1691 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
1692 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1693 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
1694 (fix_<mode>di2_hw): Likewise.
1695 (fixuns_<mode>di2_hw): Likewise.
1696 (fix_<mode>si2_hw): Likewise.
1697 (fixuns_<mode>si2_hw): Likewise.
1698 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
1699 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1700 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
1701 fix<uns>_trunc<SFDF:mode>si2_p8.
1702 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
1703 used.
1704 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
1705 (fix<uns>_<mode>_mem): Likewise.
1706 (fctiw<u>z_<mode>_mem): Likewise.
1707 (fix<uns>_<mode>_mem): Likewise.
1708 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
1709 the register allocator from doing a direct move to the GPRs to do
1710 a store, and instead use the ISA 3.0 store byte/half-word from
1711 vector register instruction. For IEEE 128-bit floating point,
1712 also optimize stores of 32-bit ints.
1713 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
1715 2018-02-07 Alan Hayward <alan.hayward@arm.com>
1717 * genextract.c (push_pathstr_operand): New function to support
1718 [a-zA-Z].
1719 (walk_rtx): Call push_pathstr_operand.
1720 (print_path): Support [a-zA-Z].
1722 2018-02-07 Richard Biener <rguenther@suse.de>
1724 PR tree-optimization/84037
1725 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
1726 (cse_and_gimplify_to_preheader): Declare.
1727 (vect_get_place_in_interleaving_chain): Likewise.
1728 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
1729 ivexpr_map.
1730 (_loop_vec_info::~_loop_vec_info): Delete it.
1731 (cse_and_gimplify_to_preheader): New function.
1732 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
1733 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
1734 (vectorizable_load): Likewise. For grouped stores always base
1735 the IV on the first element.
1736 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
1737 condition before gimplifying.
1739 2018-02-07 Jakub Jelinek <jakub@redhat.com>
1741 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
1742 *DIV_EXPR and *MOD_EXPR.
1744 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
1746 PR target/84248
1747 * config/i386/i386.c (ix86_option_override_internal): Mask out
1748 the CF_SET bit when checking -fcf-protection.
1750 2018-02-07 Tom de Vries <tom@codesourcery.com>
1752 PR libgomp/84217
1753 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
1754 enough.
1756 2018-02-07 Richard Biener <rguenther@suse.de>
1758 PR tree-optimization/84204
1759 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
1760 this place.
1762 PR tree-optimization/84205
1763 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
1764 special-case isl_ast_op_zdiv_r.
1766 PR tree-optimization/84223
1767 * graphite-scop-detection.c (gather_bbs::before_dom_children):
1768 Only add conditions from within the region.
1769 (gather_bbs::after_dom_children): Adjust.
1771 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
1773 PR target/84209
1774 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
1775 * config/avr/avr.md: Only post-reload split REG-REG moves if
1776 either register is REGERAL_REG_P.
1778 2018-02-07 Jakub Jelinek <jakub@redhat.com>
1780 PR tree-optimization/84235
1781 * tree-ssa-scopedtables.c
1782 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
1783 if the subtraction is performed in floating point type where NaNs are
1784 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
1785 build 1. Formatting fix.
1787 2018-02-06 Jakub Jelinek <jakub@redhat.com>
1789 PR target/84146
1790 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
1791 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
1792 and skip it regardless of bb boundaries. Use CALL_P macro,
1793 don't test INSN_P (insn) together with CALL_P or JUMP_P check
1794 unnecessarily, formatting fix.
1796 2018-02-06 Michael Collison <michael.collison@arm.com>
1798 * config/arm/thumb2.md:
1799 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
1800 (*thumb_mov_notscc): Ditto.
1802 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
1804 PR target/84154
1805 * config/rs6000/rs6000.md (su code attribute): Use "u" for
1806 unsigned_fix, not "s".
1808 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1810 * configure.ac (gcc_fn_eh_frame_ro): New function.
1811 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
1812 correct .eh_frame permissions.
1813 * configure: Regenerate.
1815 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
1817 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
1818 irrelevant options.
1820 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1822 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1823 Display warning message for -mno-speculate-indirect-jumps.
1825 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
1827 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
1828 Undocumented.
1829 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
1831 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
1833 PR tree-optimization/84225
1834 * tree-eh.c (find_trapping_overflow): Only call
1835 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
1837 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1839 PR target/84145
1840 * config/i386/i386.c: Reimplement the check of possible options
1841 -mibt/-mshstk conbination. Change error messages.
1842 * doc/invoke.texi: Fix a typo: remove extra '='.
1844 2018-02-06 Marek Polacek <polacek@redhat.com>
1846 PR tree-optimization/84228
1847 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
1849 2018-02-06 Tamar Christina <tamar.christina@arm.com>
1851 PR target/82641
1852 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
1853 emitted arch directives.
1854 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
1855 __ARM_FEATURE_COPROC before changing architectures.
1857 2018-02-06 Richard Biener <rguenther@suse.de>
1859 * config/i386/i386.c (print_reg): Fix typo.
1860 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
1862 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
1864 * configure: Regenerate.
1866 2018-02-05 Martin Sebor <msebor@redhat.com>
1868 PR tree-optimization/83369
1869 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
1870 inlining context.
1872 2018-02-05 Martin Liska <mliska@suse.cz>
1874 * doc/invoke.texi: Cherry-pick upstream r323995.
1876 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
1878 * ira.c (ira_init_register_move_cost): Adjust comment.
1880 2018-02-05 Martin Liska <mliska@suse.cz>
1882 PR gcov-profile/84137
1883 * doc/gcov.texi: Fix typo in documentation.
1885 2018-02-05 Martin Liska <mliska@suse.cz>
1887 PR gcov-profile/83879
1888 * doc/gcov.texi: Document necessity of --dynamic-list-data when
1889 using dlopen functionality.
1891 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
1893 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
1894 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
1895 _mm_maskz_range_ss, _mm_mask_range_round_ss,
1896 _mm_maskz_range_round_ss): New intrinsics.
1897 (__builtin_ia32_rangesd128_round)
1898 (__builtin_ia32_rangess128_round): Remove.
1899 (__builtin_ia32_rangesd128_mask_round,
1900 __builtin_ia32_rangess128_mask_round): New builtins.
1901 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
1902 __builtin_ia32_rangess128_round): Remove.
1903 (__builtin_ia32_rangesd128_mask_round,
1904 __builtin_ia32_rangess128_mask_round): New builtins.
1905 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
1906 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
1907 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
1908 "<round_saeonly_constraint>")): Changed to ...
1909 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
1910 "<round_saeonly_scalar_constraint>")): ... this.
1911 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
1912 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
1913 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
1914 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
1915 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
1917 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
1919 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
1920 options.
1921 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
1922 Remove all values except native, 8540 and 8548.
1924 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
1926 * config/i386/i386.c (ix86_output_function_return): Pass
1927 INVALID_REGNUM, instead of -1, as invalid register number to
1928 indirect_thunk_name and output_indirect_thunk.
1930 2018-02-02 Julia Koval <julia.koval@intel.com>
1932 * config.gcc: Add -march=icelake.
1933 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
1934 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
1935 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
1936 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
1937 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
1938 (processor_target_table): Add icelake.
1939 (ix86_option_override_internal): Handle new PTAs.
1940 (get_builtin_code_for_version): Handle icelake.
1941 (M_INTEL_COREI7_ICELAKE): New.
1942 (fold_builtin_cpu): Handle icelake.
1943 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
1944 * doc/invoke.texi: Add -march=icelake.
1946 2018-02-02 Julia Koval <julia.koval@intel.com>
1948 * config/i386/i386.c (ix86_option_override_internal): Change flags type
1949 to wide_int_bitmask.
1950 * wide-int-bitmask.h: New.
1952 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1954 PR target/84066
1955 * config/i386/i386.md: Replace Pmode with word_mode in
1956 builtin_setjmp_setup and builtin_longjmp to support x32.
1958 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
1960 PR target/56010
1961 PR target/83743
1962 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
1963 #include "opts.h".
1964 (rs6000_supported_cpu_names): New static variable.
1965 (linux_cpu_translation_table): Likewise.
1966 (elf_platform) <cpu>: Define new static variable and use it.
1967 Translate kernel AT_PLATFORM name to canonical name if needed.
1968 Error if platform name is unknown.
1970 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
1972 PR target/84089
1973 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
1975 2018-02-01 Jeff Law <law@redhat.com>
1977 PR target/84128
1978 * config/i386/i386.c (release_scratch_register_on_entry): Add new
1979 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
1980 the scratch if RELEASE_VIA_POP is false.
1981 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
1982 If we have to save a temporary register, decrement SIZE appropriately.
1983 Pass new arguments to release_scratch_register_on_entry.
1984 (ix86_adjust_stack_and_probe): Likewise.
1985 (ix86_emit_probe_stack_range): Pass new arguments to
1986 release_scratch_register_on_entry.
1988 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
1990 PR rtl-optimization/84157
1991 * combine.c (change_zero_ext): Use REG_P predicate in
1992 front of HARD_REGISTER_P predicate.
1994 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
1996 * config/avr/avr.c (avr_option_override): Move disabling of
1997 -fdelete-null-pointer-checks to...
1998 * common/config/avr/avr-common.c (avr_option_optimization_table):
1999 ...here.
2001 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2003 PR tree-optimization/81635
2004 * tree-data-ref.c (split_constant_offset_1): For types that
2005 wrap on overflow, try to use range info to prove that wrapping
2006 cannot occur.
2008 2018-02-01 Renlin Li <renlin.li@arm.com>
2010 PR target/83370
2011 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
2012 TAILCALL_ADDR_REGS.
2013 (aarch64_register_move_cost): Likewise.
2014 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
2015 TAILCALL_ADDR_REGS.
2016 (REG_CLASS_NAMES): Likewise.
2017 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
2018 TAILCALL_ADDR_REGS. Remove IP registers.
2019 * config/aarch64/aarch64.md (Ucs): Update register constraint.
2021 2018-02-01 Richard Biener <rguenther@suse.de>
2023 * domwalk.h (dom_walker::dom_walker): Add additional constructor
2024 for specifying RPO order and allow NULL for that.
2025 * domwalk.c (dom_walker::dom_walker): Likewise.
2026 (dom_walker::walk): Handle NULL RPO order.
2027 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
2028 in RPO order.
2029 (rewrite_update_dom_walker): Likewise.
2030 (mark_def_dom_walker): Likewise.
2032 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2034 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
2035 (aarch64_maybe_expand_sve_subreg_move): Declare.
2036 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
2037 * config/aarch64/predicates.md (aarch64_any_register_operand): New
2038 predicate.
2039 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
2040 that are semantically a reverse operation.
2041 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
2042 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
2043 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
2044 functions.
2045 (aarch64_can_change_mode_class): For big-endian, forbid changes
2046 between two SVE modes if they have different element sizes.
2048 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2050 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
2051 the TImode handling for big-endian targets.
2053 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2055 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
2056 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
2057 not just bytes.
2058 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
2059 Remove BSWAP handing for big-endian targets and use the form of
2060 LD1RQ appropariate for the mode.
2062 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2064 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
2065 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
2066 duplicated element.
2068 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2070 PR tearget/83845
2071 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
2072 check for operands that need to go through aarch64_sve_reload_be.
2074 2018-02-01 Jakub Jelinek <jakub@redhat.com>
2076 PR tree-optimization/81661
2077 PR tree-optimization/84117
2078 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
2079 * tree-eh.c: Include gimplify.h.
2080 (find_trapping_overflow, replace_trapping_overflow,
2081 rewrite_to_non_trapping_overflow): New functions.
2082 * tree-vect-loop.c: Include tree-eh.h.
2083 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
2084 * tree-data-ref.c: Include tree-eh.h.
2085 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
2087 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
2089 PR rtl-optimization/84123
2090 * combine.c (change_zero_ext): Check if hard register satisfies
2091 can_change_dest_mode before calling gen_lowpart_SUBREG.
2093 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2095 PR target/82444
2096 * ira.c (ira_init_register_move_cost): Remove assert.
2098 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2100 PR rtl-optimization/84071
2101 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2102 * doc/tm.texi: Regenerate.
2104 2018-01-31 Richard Biener <rguenther@suse.de>
2106 PR tree-optimization/84132
2107 * tree-data-ref.c (analyze_miv_subscript): Properly
2108 check whether evolution_function_is_affine_multivariate_p
2109 before calling gcd_of_steps_may_divide_p.
2111 2018-01-31 Julia Koval <julia.koval@intel.com>
2113 PR target/83618
2114 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2115 * config/i386/i386.md (rdpid_rex64) New.
2116 (rdpid): Make 32bit only.
2118 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2120 PR lto/84105
2121 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2122 an IDENTIFIER_NODE for FUNCTION_TYPE's.
2124 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2126 Revert
2127 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
2129 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
2131 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2133 PR rtl-optimization/84071
2134 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
2135 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
2137 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2139 * config/arc/arc.c (arc_handle_aux_attribute): New function.
2140 (arc_attribute_table): Add 'aux' attribute.
2141 (arc_in_small_data_p): Consider aux like variables.
2142 (arc_is_aux_reg_p): New function.
2143 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
2144 (arc_get_aux_arg): New function.
2145 (prepare_move_operands): Handle aux-register access.
2146 (arc_handle_aux_attribute): New function.
2147 * doc/extend.texi (ARC Variable attributes): Add subsection.
2149 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2151 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
2152 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
2153 (arc_attribute_table): Add 'uncached' attribute.
2154 (arc_print_operand): Print '.di' flag for uncached memory
2155 accesses.
2156 (arc_in_small_data_p): Do not consider for small data the uncached
2157 types.
2158 (arc_is_uncached_mem_p): New function.
2159 * config/arc/predicates.md (compact_store_memory_operand): Check
2160 for uncached memory accesses.
2161 (nonvol_nonimm_operand): Likewise.
2162 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
2164 2018-01-31 Jakub Jelinek <jakub@redhat.com>
2166 PR c/84100
2167 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
2168 falign-loops=): Add Optimization flag.
2170 2018-01-30 Jeff Law <law@redhat.com>
2172 PR target/84064
2173 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
2174 INT_REGISTERS_SAVED. Check it prior to calling
2175 get_scratch_register_on_entry.
2176 (ix86_adjust_stack_and_probe): Similarly.
2177 (ix86_emit_probe_stack_range): Similarly.
2178 (ix86_expand_prologue): Corresponding changes.
2180 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2182 PR target/40411
2183 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
2184 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
2186 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
2188 PR target/84112
2189 * lra-constraints.c (curr_insn_transform): Process AND in the
2190 address.
2192 2018-01-30 Jakub Jelinek <jakub@redhat.com>
2194 PR rtl-optimization/83986
2195 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
2196 dependence against last_pending_memory_flush in addition to
2197 pending_jump_insns.
2199 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
2201 PR tree-optimization/81611
2202 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
2203 copies.
2205 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2207 PR target/83758
2208 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
2209 a reg rtx.
2211 2018-01-30 Richard Biener <rguenther@suse.de>
2212 Jakub Jelinek <jakub@redhat.com>
2214 PR tree-optimization/84111
2215 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
2216 inner loops added during recursion, as they don't have up-to-date
2217 SSA form.
2219 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2221 PR ipa/81360
2222 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
2223 (can_inline_edge_by_limits_p): ... here.
2224 (can_early_inline_edge_p, check_callers,
2225 update_caller_keys, update_callee_keys, recursive_inlining,
2226 add_new_edges_to_heap, speculation_useful_p,
2227 inline_small_functions,
2228 inline_small_functions, flatten_function,
2229 inline_to_all_callers_1): Update.
2231 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2233 * profile-count.c (profile_count::combine_with_ipa_count): Handle
2234 zeros correctly.
2236 2018-01-30 Richard Biener <rguenther@suse.de>
2238 PR tree-optimization/83008
2239 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
2240 invariant and constant vector uses in stmts when they need
2241 more than one stmt.
2243 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2245 PR bootstrap/84017
2246 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
2247 * configure: Regenerate.
2249 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2251 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
2252 pattern.
2253 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
2254 Use gen_rtx_REG rather than gen_lowpart.
2256 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2258 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
2259 rather than 0 when creating partial subregs.
2261 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2263 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
2264 of usage.
2266 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2268 PR target/81550
2269 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
2270 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
2271 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
2272 flags. This restores the settings used before the 2017-07-24.
2273 Turning off pre increment/decrement/modify allows IVOPTS to
2274 optimize DF/SF loops where the index is an int.
2276 2018-01-29 Richard Biener <rguenther@suse.de>
2277 Kelvin Nilsen <kelvin@gcc.gnu.org>
2279 PR bootstrap/80867
2280 * tree-vect-stmts.c (vectorizable_call): Don't call
2281 targetm.vectorize_builtin_md_vectorized_function if callee is
2282 NULL.
2284 2018-01-22 Carl Love <cel@us.ibm.com>
2286 * doc/extend.tex: Fix typo in second arg in
2287 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
2289 2018-01-29 Richard Biener <rguenther@suse.de>
2291 PR tree-optimization/84086
2292 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
2293 (flush_ssaname_freelist): When SSA names were released reset
2294 the SCEV hash table.
2296 2018-01-29 Richard Biener <rguenther@suse.de>
2298 PR tree-optimization/84057
2299 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
2300 removed paths when removing edges.
2302 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
2304 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
2305 -mfunction-return=@var{choice}.
2307 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
2309 PR diagnostic/84034
2310 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
2311 Handle CR like TAB.
2312 (layout::print_source_line): Likewise.
2313 (test_get_line_width_without_trailing_whitespace): Add test cases.
2315 2018-01-27 Jakub Jelinek <jakub@redhat.com>
2317 PR middle-end/84040
2318 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
2319 debug insns.
2321 2018-01-26 Jim Wilson <jimw@sifive.com>
2323 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
2325 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
2326 specified.
2328 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2330 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
2331 and CMP + SUB-immediate -> SUBS.
2333 2018-01-26 Martin Sebor <msebor@redhat.com>
2335 PR tree-optimization/83896
2336 * tree-ssa-strlen.c (get_string_len): Rename...
2337 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
2338 Avoid assuming length is constant.
2339 (handle_char_store): Use HOST_WIDE_INT for string length.
2341 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
2343 PR target/81763
2344 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
2345 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
2347 2018-01-26 Richard Biener <rguenther@suse.de>
2349 PR rtl-optimization/84003
2350 * dse.c (record_store): Only record redundant stores when
2351 the earlier store aliases at least all accesses the later one does.
2353 2018-01-26 Jakub Jelinek <jakub@redhat.com>
2355 PR rtl-optimization/83985
2356 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
2357 REG_CFA_RESTORE insns.
2358 (delete_unmarked_insns): Don't ignore separate shrink wrapping
2359 REG_CFA_RESTORE insns here.
2361 PR c/83989
2362 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
2363 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
2365 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2367 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
2368 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
2369 (arc_init): Likewise.
2370 (arc_override_options): Likewise.
2371 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
2372 value.
2373 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
2374 support.
2375 * config/arc/arc.h (TARGET_DBNZ): Define.
2376 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
2377 properly set the tune attribute.
2378 (dbnz): Use TARGET_DBNZ guard.
2379 * config/arc/arc.opt (mtune): Add core3 option.
2381 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2383 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
2384 recognize new pic like addresses.
2385 (arc_delegitimize_address): Clean up.
2387 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2389 * config/arc/arc-arches.def: Option mrf16 valid for all
2390 architectures.
2391 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
2392 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
2393 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
2394 * config/arc/arc-tables.opt: Regenerate.
2395 * config/arc/arc.c (arc_conditional_register_usage): Handle
2396 reduced register file case.
2397 (arc_file_start): Set must have build attributes.
2398 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
2399 mrf16 option value.
2400 * config/arc/arc.opt (mrf16): Add new option.
2401 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
2402 * config/arc/genmultilib.awk: Handle new mrf16 option.
2403 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
2404 * config/arc/t-multilib: Regenerate.
2405 * doc/invoke.texi (ARC Options): Document mrf16 option.
2407 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2409 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
2410 * config/arc/arc.c (arc_handle_secure_attribute): New function.
2411 (arc_attribute_table): Add 'secure_call' attribute.
2412 (arc_print_operand): Print secure call operand.
2413 (arc_function_ok_for_sibcall): Don't optimize tail calls when
2414 secure.
2415 (arc_is_secure_call_p): New function. * config/arc/arc.md
2416 (call_i): Add support for sjli instruction.
2417 (call_value_i): Likewise.
2418 * config/arc/constraints.md (Csc): New constraint.
2420 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2421 John Eric Martin <John.Martin@emmicro-us.com>
2423 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
2424 * config/arc/arc.c (_arc_jli_section): New struct.
2425 (arc_jli_section): New type.
2426 (rc_jli_sections): New static variable.
2427 (arc_handle_jli_attribute): New function.
2428 (arc_attribute_table): Add jli_always and jli_fixed attribute.
2429 (arc_file_end): New function.
2430 (TARGET_ASM_FILE_END): Define.
2431 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
2432 (arc_add_jli_section): New function.
2433 (jli_call_scan): Likewise.
2434 (arc_reorg): Call jli_call_scan.
2435 (arc_output_addsi): Remove 'S' from printing asm operand.
2436 (arc_is_jli_call_p): New function.
2437 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
2438 operand.
2439 (movhi_insn): Likewise.
2440 (movsi_insn): Likewise.
2441 (movsi_set_cc_insn): Likewise.
2442 (loadqi_update): Likewise.
2443 (load_zeroextendqisi_update): Likewise.
2444 (load_signextendqisi_update): Likewise.
2445 (loadhi_update): Likewise.
2446 (load_zeroextendhisi_update): Likewise.
2447 (load_signextendhisi_update): Likewise.
2448 (loadsi_update): Likewise.
2449 (loadsf_update): Likewise.
2450 (movsicc_insn): Likewise.
2451 (bset_insn): Likewise.
2452 (bxor_insn): Likewise.
2453 (bclr_insn): Likewise.
2454 (bmsk_insn): Likewise.
2455 (bicsi3_insn): Likewise.
2456 (cmpsi_cc_c_insn): Likewise.
2457 (movsi_ne): Likewise.
2458 (movsi_cond_exec): Likewise.
2459 (clrsbsi2): Likewise.
2460 (norm_f): Likewise.
2461 (normw): Likewise.
2462 (swap): Likewise.
2463 (divaw): Likewise.
2464 (flag): Likewise.
2465 (sr): Likewise.
2466 (kflag): Likewise.
2467 (ffs): Likewise.
2468 (ffs_f): Likewise.
2469 (fls): Likewise.
2470 (call_i): Remove 'S' asm letter, add jli instruction.
2471 (call_value_i): Likewise.
2472 * config/arc/arc.op (mjli-always): New option.
2473 * config/arc/constraints.md (Cji): New constraint.
2474 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
2475 operand.
2476 (subsf3_fpx): Likewise.
2477 (mulsf3_fpx): Likewise.
2478 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
2479 asm operand.
2480 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
2481 function attrbutes.
2482 * doc/invoke.texi (ARC): Document mjli-always option.
2484 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
2486 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
2487 avoid addition with 0 and use incw and decw where possible.
2489 2018-01-26 Richard Biener <rguenther@suse.de>
2491 PR tree-optimization/81082
2492 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
2493 association if it requires casting to unsigned.
2494 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
2495 from fold_plusminus_mult_expr to catch important cases late when
2496 range info is available.
2498 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2500 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
2501 * configure.ac (hidden_linkonce): New test.
2502 * configure: Regenerate.
2503 * config.in: Regenerate.
2505 2018-01-26 Julia Koval <julia.koval@intel.com>
2507 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
2508 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
2509 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
2510 _mm_mask_bitshuffle_epi64_mask): Fix type.
2511 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
2512 USI_FTYPE_V4DI_V4DI_USI): Remove.
2513 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
2514 __builtin_ia32_vpshufbitqmb256_mask,
2515 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
2516 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
2517 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
2519 2018-01-26 Alan Modra <amodra@gmail.com>
2521 PR target/84033
2522 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
2523 UNSPEC_VBPERMQ. Sort other unspecs.
2525 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
2527 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
2529 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
2531 PR middle-end/83055
2532 * predict.c (drop_profile): Do not push/pop cfun; update also
2533 node->count.
2534 (handle_missing_profiles): Fix logic looking for zero profiles.
2536 2018-01-25 Jakub Jelinek <jakub@redhat.com>
2538 PR middle-end/83977
2539 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
2540 on functions with #pragma omp declare simd or functions with simd
2541 attribute.
2542 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
2543 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
2544 Remove trailing \n from warning_at calls.
2546 2018-01-25 Tom de Vries <tom@codesourcery.com>
2548 PR target/84028
2549 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2550 for neutered workers.
2552 2018-01-24 Joseph Myers <joseph@codesourcery.com>
2554 PR target/68467
2555 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
2556 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
2558 2018-01-24 Jeff Law <law@redhat.com>
2560 PR target/83994
2561 * i386.c (get_probe_interval): Move to earlier point.
2562 (ix86_compute_frame_layout): If -fstack-clash-protection and
2563 the frame is larger than the probe interval, then use pushes
2564 to save registers rather than reg->mem moves.
2565 (ix86_expand_prologue): Remove conditional for int_registers_saved
2566 assertion.
2568 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
2570 PR target/84014
2571 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
2572 min/max for never referenced object.
2574 2018-01-24 Jakub Jelinek <jakub@redhat.com>
2576 PR middle-end/83977
2577 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
2578 here.
2579 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
2580 attributes from DECL_ATTRIBUTES (decl) without affecting
2581 DECL_ATTRIBUTES (current_function_decl).
2582 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
2583 functions with non-NULL DECL_ABSTRACT_ORIGIN.
2585 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
2587 PR tree-optimization/83979
2588 * fold-const.c (fold_comparison): Use constant_boolean_node
2589 instead of boolean_{true,false}_node.
2591 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
2593 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
2594 with zero counts.
2596 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2598 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2599 Simplify the clause that sets the length attribute.
2600 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2601 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
2602 clause that sets the length attribute.
2603 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2605 2018-01-24 Tom de Vries <tom@codesourcery.com>
2607 PR target/83589
2608 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
2609 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
2610 Add strict parameter.
2611 (prevent_branch_around_nothing): Insert dummy insn between branch to
2612 label and label with no ptx insn inbetween.
2613 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
2615 2018-01-24 Tom de Vries <tom@codesourcery.com>
2617 PR target/81352
2618 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2619 for neutered threads in warp.
2620 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
2622 2018-01-24 Richard Biener <rguenther@suse.de>
2624 PR tree-optimization/83176
2625 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
2626 operands.
2628 2018-01-24 Richard Biener <rguenther@suse.de>
2630 PR tree-optimization/82819
2631 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
2632 code generating pluses that are no-ops in the target precision.
2634 2018-01-24 Richard Biener <rguenther@suse.de>
2636 PR middle-end/84000
2637 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
2639 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2641 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
2642 to merge probabilities.
2643 * predict.c (probably_never_executed): Also mark as cold functions
2644 with global 0 profile and guessed local profile.
2645 * profile-count.c (profile_probability::combine_with_count): New
2646 member function.
2647 * profile-count.h (profile_probability::operator*,
2648 profile_probability::operator*=, profile_probability::operator/,
2649 profile_probability::operator/=): Reduce precision to adjusted
2650 and set value to guessed on contradictory divisions.
2651 (profile_probability::combine_with_freq): Remove.
2652 (profile_probability::combine_wiht_count): Declare.
2653 (profile_count::force_nonzero):: Set to adjusted.
2654 (profile_count::probability_in):: Set quality to adjusted.
2655 * tree-ssa-tail-merge.c (replace_block_by): Use
2656 combine_with_count.
2658 2018-01-23 Andrew Waterman <andrew@sifive.com>
2659 Jim Wilson <jimw@sifive.com>
2661 * config/riscv/riscv.c (riscv_stack_boundary): New.
2662 (riscv_option_override): Set riscv_stack_boundary. Handle
2663 riscv_preferred_stack_boundary_arg.
2664 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
2665 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
2666 (STACK_BOUNDARY): Set to riscv_stack_boundary.
2667 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
2668 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
2669 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
2671 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
2673 PR target/83905
2674 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
2675 of struct ix86_frame.
2676 (ix86_expand_epilogue): Likewise. Add a local variable for
2677 the reg_save_offset field in struct ix86_frame.
2679 2018-01-23 Bin Cheng <bin.cheng@arm.com>
2681 PR tree-optimization/82604
2682 * tree-loop-distribution.c (enum partition_kind): New enum item
2683 PKIND_PARTIAL_MEMSET.
2684 (partition_builtin_p): Support above new enum item.
2685 (generate_code_for_partition): Ditto.
2686 (compute_access_range): Differentiate cases that equality can be
2687 proven at all loops, the innermost loops or no loops.
2688 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
2689 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
2690 (finalize_partitions, distribute_loop): Don't fuse partition of
2691 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
2692 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
2693 parloop is enabled.
2695 2018-01-23 Martin Liska <mliska@suse.cz>
2697 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
2698 order to ignore the predictor.
2699 (PRED_POLYMORPHIC_CALL): Likewise.
2700 (PRED_RECURSIVE_CALL): Likewise.
2702 2018-01-23 Martin Liska <mliska@suse.cz>
2704 * tree-profile.c (tree_profiling): Print function header to
2705 aware reader which function we are working on.
2706 * value-prof.c (gimple_find_values_to_profile): Do not print
2707 not interesting value histograms.
2709 2018-01-23 Martin Liska <mliska@suse.cz>
2711 * profile-count.h (enum profile_quality): Add
2712 profile_uninitialized as the first value. Do not number values
2713 as they are zero based.
2714 (profile_count::verify): Update sanity check.
2715 (profile_probability::verify): Likewise.
2717 2018-01-23 Nathan Sidwell <nathan@acm.org>
2719 * doc/invoke.texi (ffor-scope): Deprecate.
2721 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2723 PR tree-optimization/83510
2724 * domwalk.c (set_all_edges_as_executable): New function.
2725 (dom_walker::dom_walker): Convert bool param
2726 "skip_unreachable_blocks" to enum reachability. Move setup of
2727 edge flags to set_all_edges_as_executable and only do it when
2728 reachability is REACHABLE_BLOCKS.
2729 * domwalk.h (enum dom_walker::reachability): New enum.
2730 (dom_walker::dom_walker): Convert bool param
2731 "skip_unreachable_blocks" to enum reachability.
2732 (set_all_edges_as_executable): New decl.
2733 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
2734 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
2735 "reachability".
2736 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
2737 but converting true to REACHABLE_BLOCKS.
2738 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
2739 * tree-vrp.c
2740 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
2741 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
2742 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
2743 REACHABLE_BLOCKS.
2744 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
2745 if check_all_array_refs will be called.
2747 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2749 * tree.c (selftest::test_location_wrappers): Add more test
2750 coverage.
2752 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2754 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
2755 (selftest::test_bit_in_range): Likewise.
2757 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
2759 PR testsuite/83888
2760 * doc/sourcebuild.texi (vect_float): Say that the selector
2761 only describes the situation when -funsafe-math-optimizations is on.
2762 (vect_float_strict): Document.
2764 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
2766 PR tree-optimization/83965
2767 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
2768 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
2769 instead of checking only for a reduction.
2770 (vect_recog_widen_sum_pattern): Likewise.
2772 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2774 * predict.c (probably_never_executed): Only use precise profile info.
2775 (compute_function_frequency): Skip after inlining hack since we now
2776 have quality checking.
2778 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2780 * profile-count.h (profile_probability::very_unlikely,
2781 profile_probability::unlikely, profile_probability::even): Set
2782 precision to guessed.
2784 2018-01-23 Richard Biener <rguenther@suse.de>
2786 PR tree-optimization/83963
2787 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
2788 Properly terminate dominator walk when crossing the exit edge not
2789 when visiting its source block.
2791 2018-01-23 Jakub Jelinek <jakub@redhat.com>
2793 PR c++/83918
2794 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
2795 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
2797 2018-01-22 Jakub Jelinek <jakub@redhat.com>
2799 PR tree-optimization/83957
2800 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
2801 semicolon after for body surrounded by braces.
2803 PR tree-optimization/83081
2804 * profile-count.h (profile_probability::split): New method.
2805 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
2806 Use profile_probability::split.
2807 (do_compare_rtx_and_jump): Fix adjustment of probabilities
2808 when splitting a single conditional jump into 2.
2810 2018-01-22 David Malcolm <dmalcolm@redhat.com>
2812 PR tree-optimization/69452
2813 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
2814 decl.
2816 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2818 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
2819 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
2820 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
2822 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2824 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
2825 * config/rl78/rl78.md (movdi): New define_expand.
2826 * config/rl78/rl78.c (rl78_split_movdi): New function.
2828 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
2830 PR target/83862
2831 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
2832 no longer used.
2833 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
2834 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
2835 128-bit to produce an UNSPEC move to get the double word with the
2836 signbit and then a shift directly to do signbit.
2837 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
2838 implementation with a new version that just does either a direct
2839 move or a regular move. Move memory interface to separate insns.
2840 Move insns so they are next to the expander.
2841 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
2842 with signbit move. Split big and little endian case.
2843 (signbit<mode>2_dm_mem_le): Likewise.
2844 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
2845 (signbit<mode>2_dm2): Likewise.
2847 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2849 * config/rl78/rl78.md (anddi3): New define_expand.
2851 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2853 * config/rl78/rl78.md (umindi3): New define_expand.
2855 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2857 * config/rl78/rl78.md (smindi3): New define_expand.
2859 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2861 * config/rl78/rl78.md (smaxdi3): New define_expand.
2863 2018-01-22 Carl Love <cel@us.ibm.com>
2865 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
2866 LVX_V1TI): Add macro expansion.
2867 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
2868 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
2869 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
2870 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
2871 Change check to determine if the instruction is a byte reversing
2872 entry. Fix typo in comment.
2873 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
2874 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
2875 Add def_builtin calls for new builtins.
2876 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
2877 Add define_insn expansion.
2879 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2881 * config/rl78/rl78.md (umaxdi3): New define_expand.
2883 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2885 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
2886 for non-QImode registers.
2888 2018-01-22 Richard Biener <rguenther@suse.de>
2890 PR tree-optimization/83963
2891 * graphite-scop-detection.c (scop_detection::get_sese): Delay
2892 including the loop exit block.
2893 (scop_detection::merge_sese): Likewise.
2894 (scop_detection::add_scop): Do it here instead.
2896 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2898 * doc/sourcebuild.texi (arm_softfloat): Document.
2900 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
2902 PR gcc/77734
2903 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
2904 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
2905 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
2907 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2908 David Edelsohn <dje.gcc@gmail.com>
2910 PR target/83946
2911 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2912 Change "crset eq" to "crset 2".
2913 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2914 (*call_indirect_aix<mode>_nospec): Likewise.
2915 (*call_value_indirect_aix<mode>_nospec): Likewise.
2916 (*call_indirect_elfv2<mode>_nospec): Likewise.
2917 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
2918 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
2919 change assembly output from . to $.
2920 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2921 (indirect_jump<mode>_nospec): Change assembly output from . to $.
2922 (*tablejump<mode>_internal1_nospec): Likewise.
2924 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
2926 PR target/80870
2927 * config/sh/sh_optimize_sett_clrt.cc:
2928 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
2930 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
2932 PR tree-optimization/83940
2933 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
2934 offset_dt to vect_constant_def rather than vect_unknown_def_type.
2935 (vect_check_load_store_mask): Add a mask_dt_out parameter and
2936 use it to pass back the definition type.
2937 (vect_check_store_rhs): Likewise rhs_dt_out.
2938 (vect_build_gather_load_calls): Add a mask_dt argument and use
2939 it instead of a call to vect_is_simple_use.
2940 (vectorizable_store): Update calls to vect_check_load_store_mask
2941 and vect_check_store_rhs. Use the dt returned by the latter instead
2942 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
2943 instead of calls to vect_is_simple_use. Pass the scalar rather
2944 than the vector operand to vect_is_simple_use when handling
2945 second and subsequent copies of an rhs value.
2946 (vectorizable_load): Update calls to vect_check_load_store_mask
2947 and vect_build_gather_load_calls. Use the cached mask_dt and
2948 gs_info.offset_dt instead of calls to vect_is_simple_use.
2950 2018-01-20 Jakub Jelinek <jakub@redhat.com>
2952 PR middle-end/83945
2953 * tree-emutls.c: Include gimplify.h.
2954 (lower_emutls_2): New function.
2955 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
2956 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
2957 it before further processing.
2959 PR target/83930
2960 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
2961 UINTVAL (trueop1) instead of INTVAL (op1).
2963 2018-01-19 Jakub Jelinek <jakub@redhat.com>
2965 PR debug/81570
2966 PR debug/83728
2967 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
2968 INCOMING_FRAME_SP_OFFSET if not defined.
2969 (scan_trace): Add ENTRY argument. If true and
2970 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
2971 emit a note to adjust the CFA offset.
2972 (create_cfi_notes): Adjust scan_trace callers.
2973 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
2974 INCOMING_FRAME_SP_OFFSET in the CIE.
2975 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
2976 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
2977 Likewise.
2978 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
2979 * doc/tm.texi: Regenerated.
2981 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2983 PR rtl-optimization/83147
2984 * lra-constraints.c (remove_inheritance_pseudos): Use
2985 lra_substitute_pseudo_within_insn.
2987 2018-01-19 Tom de Vries <tom@codesourcery.com>
2988 Cesar Philippidis <cesar@codesourcery.com>
2990 PR target/83920
2991 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
2993 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
2995 PR target/83790
2996 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
2997 spaces for function labels.
2999 2018-01-19 Martin Liska <mliska@suse.cz>
3001 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
3002 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
3003 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
3004 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
3005 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
3006 (PRED_CONST_RETURN): Change from 69 to 65.
3007 (PRED_NULL_RETURN): Change from 91 to 71.
3008 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
3009 (PRED_LOOP_GUARD): Change from 66 to 73.
3011 2018-01-19 Martin Liska <mliska@suse.cz>
3013 * predict.c (predict_insn_def): Add new assert.
3014 (struct branch_predictor): Change type to signed integer.
3015 (test_prediction_value_range): Amend test to cover
3016 PROB_UNINITIALIZED.
3017 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
3018 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
3019 (PRED_LOOP_ITERATIONS_MAX): Likewise.
3020 (PRED_LOOP_IV_COMPARE): Likewise.
3021 * predict.h (PROB_UNINITIALIZED): Define new constant.
3023 2018-01-19 Martin Liska <mliska@suse.cz>
3025 * predict.c (dump_prediction): Add new format for
3026 analyze_brprob.py script which is enabled with -details
3027 suboption.
3028 * profile-count.h (precise_p): New function.
3030 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3032 PR tree-optimization/83922
3033 * tree-vect-loop.c (vect_verify_full_masking): Return false if
3034 there are no statements that need masking.
3035 (vect_active_double_reduction_p): New function.
3036 (vect_analyze_loop_operations): Use it when handling phis that
3037 are not in the loop header.
3039 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3041 PR tree-optimization/83914
3042 * tree-vect-loop.c (vectorizable_induction): Don't convert
3043 init_expr or apply the peeling adjustment for inductions
3044 that are nested within the vectorized loop.
3046 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3048 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
3049 instead of NEG.
3051 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3053 PR sanitizer/81715
3054 PR testsuite/83882
3055 * function.h (gimplify_parameters): Add gimple_seq * argument.
3056 * function.c: Include gimple.h and options.h.
3057 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
3058 for the added local temporaries if needed.
3059 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
3060 if there are any parameter cleanups, wrap whole body into a
3061 try/finally with the cleanups.
3063 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
3065 PR target/82964
3066 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
3067 Use GET_MODE_CLASS for scalar floating point.
3069 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3071 PR ipa/82256
3072 patch by PaX Team
3073 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3074 Fix call of call_cgraph_insertion_hooks.
3076 2018-01-18 Martin Sebor <msebor@redhat.com>
3078 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
3080 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3082 PR ipa/83619
3083 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
3084 frequencies.
3086 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
3088 PR other/70268
3089 * common.opt: (-ffile-prefix-map): New option.
3090 * opts.c (common_handle_option): Defer it.
3091 * opts-global.c (handle_common_deferred_options): Handle it.
3092 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3093 * file-prefix-map.h: New file.
3094 (remap_debug_filename, add_debug_prefix_map): ...here.
3095 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3096 * final.c (debug_prefix_map, add_debug_prefix_map
3097 remap_debug_filename): Move to...
3098 * file-prefix-map.c: New file.
3099 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3100 generalize, get rid of alloca(), use strrchr() instead of strchr().
3101 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3102 Implement in terms of add_prefix_map().
3103 (remap_macro_filename, remap_debug_filename): Implement in term of
3104 remap_filename().
3105 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3106 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3107 * dbxout.c: Include file-prefix-map.h.
3108 * varasm.c: Likewise.
3109 * vmsdbgout.c: Likewise.
3110 * xcoffout.c: Likewise.
3111 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3112 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3113 * doc/invoke.texi (-ffile-prefix-map): Document.
3114 (-fdebug-prefix-map): Update description.
3116 2018-01-18 Martin Liska <mliska@suse.cz>
3118 * config/i386/i386.c (indirect_thunk_name): Document that also
3119 lfence is emitted.
3120 (output_indirect_thunk): Document why both instructions
3121 (pause and lfence) are generated.
3123 2018-01-18 Richard Biener <rguenther@suse.de>
3125 PR tree-optimization/83887
3126 * graphite-scop-detection.c
3127 (scop_detection::get_nearest_dom_with_single_entry): Remove.
3128 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
3129 (scop_detection::merge_sese): Re-implement with a flood-fill
3130 algorithm that properly finds a SESE region if it exists.
3132 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3134 PR c/61240
3135 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
3136 pointer_diff optimizations use view_convert instead of convert.
3138 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3140 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3141 Generate different code for -mno-speculate-indirect-jumps.
3142 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3143 (*call_indirect_aix<mode>): Disable for
3144 -mno-speculate-indirect-jumps.
3145 (*call_indirect_aix<mode>_nospec): New define_insn.
3146 (*call_value_indirect_aix<mode>): Disable for
3147 -mno-speculate-indirect-jumps.
3148 (*call_value_indirect_aix<mode>_nospec): New define_insn.
3149 (*sibcall_nonlocal_sysv<mode>): Generate different code for
3150 -mno-speculate-indirect-jumps.
3151 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3153 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3155 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
3156 long double type, set the flags for noting the default long double
3157 type, even if we don't pass or return a long double type.
3159 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
3161 PR ipa/83051
3162 * ipa-inline.c (flatten_function): Do not overwrite final inlining
3163 failure.
3165 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
3167 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
3168 support for merge[hl].
3169 (fold_mergehl_helper): New helper function.
3170 (tree-vector-builder.h): New #include for tree_vector_builder usage.
3171 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
3172 (altivec_vmrglw_direct): Add xxmrglw insn.
3174 2018-01-17 Andrew Waterman <andrew@sifive.com>
3176 * config/riscv/riscv.c (riscv_conditional_register_usage): If
3177 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
3179 2018-01-17 David Malcolm <dmalcolm@redhat.com>
3181 PR lto/83121
3182 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
3183 call the lto_location_cache before reading the
3184 DECL_SOURCE_LOCATION of the types.
3186 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3187 Richard Sandiford <richard.sandiford@linaro.org>
3189 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
3190 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
3191 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
3192 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
3193 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
3194 Add declaration.
3195 * config/aarch64/constraints.md (aarch64_movti_operand):
3196 Limit immediates.
3197 * config/aarch64/predicates.md (Uti): Add new constraint.
3199 2018-01-17 Carl Love <cel@us.ibm.com>
3201 * config/rs6000/vsx.md (define_expand xl_len_r,
3202 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
3203 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
3204 lxvll.
3205 (define_expand, define_insn): Move the shift left from the
3206 define_insn to the define_expand for lxvl and stxvl instructions.
3207 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
3208 and XL_LEN_R definitions to PURE.
3210 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
3212 * config/i386/i386.c (indirect_thunk_name): Declare regno
3213 as unsigned int. Compare regno with INVALID_REGNUM.
3214 (output_indirect_thunk): Ditto.
3215 (output_indirect_thunk_function): Ditto.
3216 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
3217 in the call to output_indirect_thunk_function.
3219 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
3221 PR middle-end/83884
3222 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
3223 rather than the size of inner_type to determine the stack slot size
3224 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
3226 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
3228 PR target/83546
3229 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
3230 to PTA_SILVERMONT.
3232 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3234 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
3235 endian Linux systems to optionally enable multilibs for selecting
3236 the long double type if the user configured an explicit type.
3237 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
3238 have no long double multilibs if not defined.
3239 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
3240 warn if the user used -mabi={ieee,ibm}longdouble and we built
3241 multilibs for long double.
3242 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
3243 appropriate multilib option.
3244 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
3245 multilib options.
3246 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
3247 for building long double multilibs.
3248 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
3250 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
3252 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
3253 copies.
3255 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
3256 64 bits.
3257 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
3258 128 bits.
3260 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
3261 variables.
3263 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
3264 return value.
3266 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3268 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
3269 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
3271 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3273 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
3274 different rtl trees depending on TARGET_64BIT.
3275 (rs6000_gen_lvx): Likewise.
3277 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3279 * config/visium/visium.md (nop): Tweak comment.
3280 (hazard_nop): Likewise.
3282 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3284 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
3285 -mspeculate-indirect-jumps.
3286 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
3287 for -mno-speculate-indirect-jumps.
3288 (*call_indirect_elfv2<mode>_nospec): New define_insn.
3289 (*call_value_indirect_elfv2<mode>): Disable for
3290 -mno-speculate-indirect-jumps.
3291 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
3292 (indirect_jump): Emit different RTL for
3293 -mno-speculate-indirect-jumps.
3294 (*indirect_jump<mode>): Disable for
3295 -mno-speculate-indirect-jumps.
3296 (*indirect_jump<mode>_nospec): New define_insn.
3297 (tablejump): Emit different RTL for
3298 -mno-speculate-indirect-jumps.
3299 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
3300 (tablejumpsi_nospec): New define_expand.
3301 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
3302 (tablejumpdi_nospec): New define_expand.
3303 (*tablejump<mode>_internal1): Disable for
3304 -mno-speculate-indirect-jumps.
3305 (*tablejump<mode>_internal1_nospec): New define_insn.
3306 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
3307 option.
3309 2018-01-16 Artyom Skrobov tyomitch@gmail.com
3311 * caller-save.c (insert_save): Drop unnecessary parameter. All
3312 callers updated.
3314 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3315 Richard Biener <rguenth@suse.de>
3317 PR libgomp/83590
3318 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
3319 return early, inline manually is_gimple_sizepos. Make sure if we
3320 call gimplify_expr we don't end up with a gimple constant.
3321 * tree.c (variably_modified_type_p): Don't return true for
3322 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
3323 * gimplify.h (is_gimple_sizepos): Remove.
3325 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3327 PR tree-optimization/83857
3328 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
3329 vectorizable_live_operation for pure SLP statements.
3330 (vectorizable_live_operation): Handle PHIs.
3332 2018-01-16 Richard Biener <rguenther@suse.de>
3334 PR tree-optimization/83867
3335 * tree-vect-stmts.c (vect_transform_stmt): Precompute
3336 nested_in_vect_loop_p since the scalar stmt may get invalidated.
3338 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3340 PR c/83844
3341 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
3342 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
3343 If off is not INTEGER_CST, issue a may not be aligned warning
3344 rather than isn't aligned. Use isn%'t rather than isn't.
3345 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
3346 into MULT_EXPR.
3347 <case MULT_EXPR>: Improve the case when bottom and one of the
3348 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
3349 operand, in that case check if the other operand is multiple of
3350 bottom divided by the INTEGER_CST operand.
3352 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3354 PR target/83858
3355 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
3356 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
3357 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
3358 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
3359 * config/pa/pa.c (pa_function_arg_advance): Likewise.
3360 (pa_function_arg, pa_arg_partial_bytes): Likewise.
3361 (pa_function_arg_size): New function.
3363 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3365 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
3366 in a separate statement.
3368 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3370 PR tree-optimization/83847
3371 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
3372 group gathers and scatters.
3374 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3376 PR rtl-optimization/86620
3377 * params.def (max-sched-ready-insns): Bump minimum value to 1.
3379 PR rtl-optimization/83213
3380 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
3381 to last if both are JUMP_INSNs.
3383 PR tree-optimization/83843
3384 * gimple-ssa-store-merging.c
3385 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
3386 store_immediate_info for bswap/nop orig_stores.
3388 2018-01-15 Andrew Waterman <andrew@sifive.com>
3390 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
3391 !TARGET_MUL.
3392 <UDIV>: Increase cost if !TARGET_DIV.
3394 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
3396 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
3397 (define_attr "cr_logical_3op"): New.
3398 (cceq_ior_compare): Adjust.
3399 (cceq_ior_compare_complement): Adjust.
3400 (*cceq_rev_compare): Adjust.
3401 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
3402 (is_cracked_insn): Adjust.
3403 (insn_must_be_first_in_group): Adjust.
3404 * config/rs6000/40x.md: Adjust.
3405 * config/rs6000/440.md: Adjust.
3406 * config/rs6000/476.md: Adjust.
3407 * config/rs6000/601.md: Adjust.
3408 * config/rs6000/603.md: Adjust.
3409 * config/rs6000/6xx.md: Adjust.
3410 * config/rs6000/7450.md: Adjust.
3411 * config/rs6000/7xx.md: Adjust.
3412 * config/rs6000/8540.md: Adjust.
3413 * config/rs6000/cell.md: Adjust.
3414 * config/rs6000/e300c2c3.md: Adjust.
3415 * config/rs6000/e500mc.md: Adjust.
3416 * config/rs6000/e500mc64.md: Adjust.
3417 * config/rs6000/e5500.md: Adjust.
3418 * config/rs6000/e6500.md: Adjust.
3419 * config/rs6000/mpc.md: Adjust.
3420 * config/rs6000/power4.md: Adjust.
3421 * config/rs6000/power5.md: Adjust.
3422 * config/rs6000/power6.md: Adjust.
3423 * config/rs6000/power7.md: Adjust.
3424 * config/rs6000/power8.md: Adjust.
3425 * config/rs6000/power9.md: Adjust.
3426 * config/rs6000/rs64.md: Adjust.
3427 * config/rs6000/titan.md: Adjust.
3429 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3431 * config/i386/predicates.md (indirect_branch_operand): Rewrite
3432 ix86_indirect_branch_register logic.
3434 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3436 * config/i386/constraints.md (Bs): Update
3437 ix86_indirect_branch_register check. Don't check
3438 ix86_indirect_branch_register with GOT_memory_operand.
3439 (Bw): Likewise.
3440 * config/i386/predicates.md (GOT_memory_operand): Don't check
3441 ix86_indirect_branch_register here.
3442 (GOT32_symbol_operand): Likewise.
3444 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3446 * config/i386/predicates.md (constant_call_address_operand):
3447 Rewrite ix86_indirect_branch_register logic.
3448 (sibcall_insn_operand): Likewise.
3450 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3452 * config/i386/constraints.md (Bs): Replace
3453 ix86_indirect_branch_thunk_register with
3454 ix86_indirect_branch_register.
3455 (Bw): Likewise.
3456 * config/i386/i386.md (indirect_jump): Likewise.
3457 (tablejump): Likewise.
3458 (*sibcall_memory): Likewise.
3459 (*sibcall_value_memory): Likewise.
3460 Peepholes of indirect call and jump via memory: Likewise.
3461 * config/i386/i386.opt: Likewise.
3462 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3463 (GOT_memory_operand): Likewise.
3464 (call_insn_operand): Likewise.
3465 (sibcall_insn_operand): Likewise.
3466 (GOT32_symbol_operand): Likewise.
3468 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3470 PR middle-end/83837
3471 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
3472 type rather than type addr's type points to.
3473 (expand_omp_atomic_mutex): Likewise.
3474 (expand_omp_atomic): Likewise.
3476 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3478 PR target/83839
3479 * config/i386/i386.c (output_indirect_thunk_function): Use
3480 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
3481 for __x86_return_thunk.
3483 2018-01-15 Richard Biener <rguenther@suse.de>
3485 PR middle-end/83850
3486 * expmed.c (extract_bit_field_1): Fix typo.
3488 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3490 PR target/83687
3491 * config/arm/iterators.md (VF): New mode iterator.
3492 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
3493 Remove integer-related logic from pattern.
3494 (neon_vabd<mode>_3): Likewise.
3496 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3498 PR middle-end/82694
3499 * common.opt (fstrict-overflow): No longer an alias.
3500 (fwrapv-pointer): New option.
3501 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
3502 also for pointer types based on flag_wrapv_pointer.
3503 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
3504 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
3505 opts->x_flag_wrapv got set.
3506 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
3507 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
3508 POINTER_TYPE_OVERFLOW_UNDEFINED.
3509 * match.pd: Likewise in address comparison pattern.
3510 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
3512 2018-01-15 Richard Biener <rguenther@suse.de>
3514 PR lto/83804
3515 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
3516 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
3517 Reset type names to their identifier if their TYPE_DECL doesn't
3518 have linkage (and thus is used for ODR and devirt).
3519 (save_debug_info_for_decl): Remove.
3520 (save_debug_info_for_type): Likewise.
3521 (add_tree_to_fld_list): Adjust.
3522 * tree-pretty-print.c (dump_generic_node): Make dumping of
3523 type names more robust.
3525 2018-01-15 Richard Biener <rguenther@suse.de>
3527 * BASE-VER: Bump to 8.0.1.
3529 2018-01-14 Martin Sebor <msebor@redhat.com>
3531 PR other/83508
3532 * builtins.c (check_access): Avoid warning when the no-warning bit
3533 is set.
3535 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
3537 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
3538 * ira-color (allocno_hard_regs_compare): Likewise.
3540 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
3542 PR target/83013
3543 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
3544 Use .pushsection/.popsection.
3546 2018-01-14 Martin Sebor <msebor@redhat.com>
3548 PR c++/81327
3549 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
3551 2018-01-14 Jakub Jelinek <jakub@redhat.com>
3553 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
3554 entry from extra_headers.
3555 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
3556 extra_headers, make the list bitwise identical to the i?86-*-* one.
3558 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3560 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
3561 -mcmodel=large with -mindirect-branch=thunk,
3562 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
3563 -mfunction-return=thunk-extern.
3564 * doc/invoke.texi: Document -mcmodel=large is incompatible with
3565 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
3566 -mfunction-return=thunk and -mfunction-return=thunk-extern.
3568 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3570 * config/i386/i386.c (print_reg): Print the name of the full
3571 integer register without '%'.
3572 (ix86_print_operand): Handle 'V'.
3573 * doc/extend.texi: Document 'V' modifier.
3575 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3577 * config/i386/constraints.md (Bs): Disallow memory operand for
3578 -mindirect-branch-register.
3579 (Bw): Likewise.
3580 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3581 (GOT_memory_operand): Likewise.
3582 (call_insn_operand): Likewise.
3583 (sibcall_insn_operand): Likewise.
3584 (GOT32_symbol_operand): Likewise.
3585 * config/i386/i386.md (indirect_jump): Call convert_memory_address
3586 for -mindirect-branch-register.
3587 (tablejump): Likewise.
3588 (*sibcall_memory): Likewise.
3589 (*sibcall_value_memory): Likewise.
3590 Disallow peepholes of indirect call and jump via memory for
3591 -mindirect-branch-register.
3592 (*call_pop): Replace m with Bw.
3593 (*call_value_pop): Likewise.
3594 (*sibcall_pop_memory): Replace m with Bs.
3595 * config/i386/i386.opt (mindirect-branch-register): New option.
3596 * doc/invoke.texi: Document -mindirect-branch-register option.
3598 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3600 * config/i386/i386-protos.h (ix86_output_function_return): New.
3601 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
3602 set function_return_type.
3603 (indirect_thunk_name): Add ret_p to indicate thunk for function
3604 return.
3605 (output_indirect_thunk_function): Pass false to
3606 indirect_thunk_name.
3607 (ix86_output_indirect_branch_via_reg): Likewise.
3608 (ix86_output_indirect_branch_via_push): Likewise.
3609 (output_indirect_thunk_function): Create alias for function
3610 return thunk if regno < 0.
3611 (ix86_output_function_return): New function.
3612 (ix86_handle_fndecl_attribute): Handle function_return.
3613 (ix86_attribute_table): Add function_return.
3614 * config/i386/i386.h (machine_function): Add
3615 function_return_type.
3616 * config/i386/i386.md (simple_return_internal): Use
3617 ix86_output_function_return.
3618 (simple_return_internal_long): Likewise.
3619 * config/i386/i386.opt (mfunction-return=): New option.
3620 (indirect_branch): Mention -mfunction-return=.
3621 * doc/extend.texi: Document function_return function attribute.
3622 * doc/invoke.texi: Document -mfunction-return= option.
3624 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3626 * config/i386/i386-opts.h (indirect_branch): New.
3627 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
3628 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
3629 with local indirect jump when converting indirect call and jump.
3630 (ix86_set_indirect_branch_type): New.
3631 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
3632 (indirectlabelno): New.
3633 (indirect_thunk_needed): Likewise.
3634 (indirect_thunk_bnd_needed): Likewise.
3635 (indirect_thunks_used): Likewise.
3636 (indirect_thunks_bnd_used): Likewise.
3637 (INDIRECT_LABEL): Likewise.
3638 (indirect_thunk_name): Likewise.
3639 (output_indirect_thunk): Likewise.
3640 (output_indirect_thunk_function): Likewise.
3641 (ix86_output_indirect_branch_via_reg): Likewise.
3642 (ix86_output_indirect_branch_via_push): Likewise.
3643 (ix86_output_indirect_branch): Likewise.
3644 (ix86_output_indirect_jmp): Likewise.
3645 (ix86_code_end): Call output_indirect_thunk_function if needed.
3646 (ix86_output_call_insn): Call ix86_output_indirect_branch if
3647 needed.
3648 (ix86_handle_fndecl_attribute): Handle indirect_branch.
3649 (ix86_attribute_table): Add indirect_branch.
3650 * config/i386/i386.h (machine_function): Add indirect_branch_type
3651 and has_local_indirect_jump.
3652 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
3653 to true.
3654 (tablejump): Likewise.
3655 (*indirect_jump): Use ix86_output_indirect_jmp.
3656 (*tablejump_1): Likewise.
3657 (simple_return_indirect_internal): Likewise.
3658 * config/i386/i386.opt (mindirect-branch=): New option.
3659 (indirect_branch): New.
3660 (keep): Likewise.
3661 (thunk): Likewise.
3662 (thunk-inline): Likewise.
3663 (thunk-extern): Likewise.
3664 * doc/extend.texi: Document indirect_branch function attribute.
3665 * doc/invoke.texi: Document -mindirect-branch= option.
3667 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
3669 PR ipa/83051
3670 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
3672 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
3674 * ipa-inline.c (want_inline_small_function_p): Return false if
3675 inlining has already failed with CIF_FINAL_ERROR.
3676 (update_caller_keys): Call want_inline_small_function_p before
3677 can_inline_edge_p.
3678 (update_callee_keys): Likewise.
3680 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3682 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
3683 New function.
3684 (rs6000_quadword_masked_address_p): Likewise.
3685 (quad_aligned_load_p): Likewise.
3686 (quad_aligned_store_p): Likewise.
3687 (const_load_sequence_p): Add comment to describe the outer-most loop.
3688 (mimic_memory_attributes_and_flags): New function.
3689 (rs6000_gen_stvx): Likewise.
3690 (replace_swapped_aligned_store): Likewise.
3691 (rs6000_gen_lvx): Likewise.
3692 (replace_swapped_aligned_load): Likewise.
3693 (replace_swapped_load_constant): Capitalize argument name in
3694 comment describing this function.
3695 (rs6000_analyze_swaps): Add a third pass to search for vector loads
3696 and stores that access quad-word aligned addresses and replace
3697 with stvx or lvx instructions when appropriate.
3698 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
3699 New function prototype.
3700 (rs6000_quadword_masked_address_p): Likewise.
3701 (rs6000_gen_lvx): Likewise.
3702 (rs6000_gen_stvx): Likewise.
3703 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
3704 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
3705 when memory address is aligned.
3706 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
3707 this split to select lvx instruction when memory address is aligned.
3708 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
3709 instruction when memory address is aligned.
3710 (*vsx_le_perm_load_v16qi): Likewise.
3711 (four unnamed splitters): Modify to select the stvx instruction
3712 when memory is aligned.
3714 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
3716 * predict.c (determine_unlikely_bbs): Handle correctly BBs
3717 which appears in the queue multiple times.
3719 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3720 Alan Hayward <alan.hayward@arm.com>
3721 David Sherwood <david.sherwood@arm.com>
3723 * tree-vectorizer.h (vec_lower_bound): New structure.
3724 (_loop_vec_info): Add check_nonzero and lower_bounds.
3725 (LOOP_VINFO_CHECK_NONZERO): New macro.
3726 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
3727 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
3728 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
3729 fields. Make seg_len the distance travelled, not including the
3730 access size.
3731 (dr_direction_indicator): Declare.
3732 (dr_zero_step_indicator): Likewise.
3733 (dr_known_forward_stride_p): Likewise.
3734 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
3735 tree-ssanames.h.
3736 (runtime_alias_check_p): Allow runtime alias checks with
3737 variable strides.
3738 (operator ==): Compare access_size and align.
3739 (prune_runtime_alias_test_list): Rework for new distinction between
3740 the access_size and seg_len.
3741 (create_intersect_range_checks_index): Likewise. Cope with polynomial
3742 segment lengths.
3743 (get_segment_min_max): New function.
3744 (create_intersect_range_checks): Use it.
3745 (dr_step_indicator): New function.
3746 (dr_direction_indicator): Likewise.
3747 (dr_zero_step_indicator): Likewise.
3748 (dr_known_forward_stride_p): Likewise.
3749 * tree-loop-distribution.c (data_ref_segment_size): Return
3750 DR_STEP * (niters - 1).
3751 (compute_alias_check_pairs): Update call to the dr_with_seg_len
3752 constructor.
3753 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
3754 (vect_preserves_scalar_order_p): New function, split out from...
3755 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
3756 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
3757 (vect_vfa_access_size): New function.
3758 (vect_vfa_align): Likewise.
3759 (vect_compile_time_alias): Take access_size_a and access_b arguments.
3760 (dump_lower_bound): New function.
3761 (vect_check_lower_bound): Likewise.
3762 (vect_small_gap_p): Likewise.
3763 (vectorizable_with_step_bound_p): Likewise.
3764 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
3765 depencies if the vectorization factor is 1. Convert the checks
3766 for nonzero steps into checks on the bounds of DR_STEP. Try using
3767 a bunds check for variable steps if the minimum required step is
3768 relatively small. Update calls to the dr_with_seg_len
3769 constructor and to vect_compile_time_alias.
3770 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
3771 function.
3772 (vect_loop_versioning): Call it.
3773 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
3774 when retrying.
3775 (vect_estimate_min_profitable_iters): Account for any bounds checks.
3777 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3778 Alan Hayward <alan.hayward@arm.com>
3779 David Sherwood <david.sherwood@arm.com>
3781 * doc/sourcebuild.texi (vect_scatter_store): Document.
3782 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
3783 optabs.
3784 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
3785 Document.
3786 * genopinit.c (main): Add supports_vec_scatter_store and
3787 supports_vec_scatter_store_cached to target_optabs.
3788 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
3789 IFN_MASK_SCATTER_STORE.
3790 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
3791 functions.
3792 * internal-fn.h (internal_store_fn_p): Declare.
3793 (internal_fn_stored_value_index): Likewise.
3794 * internal-fn.c (scatter_store_direct): New macro.
3795 (expand_scatter_store_optab_fn): New function.
3796 (direct_scatter_store_optab_supported_p): New macro.
3797 (internal_store_fn_p): New function.
3798 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
3799 IFN_MASK_SCATTER_STORE.
3800 (internal_fn_mask_index): Likewise.
3801 (internal_fn_stored_value_index): New function.
3802 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
3803 for scatter stores.
3804 * optabs-query.h (supports_vec_scatter_store_p): Declare.
3805 * optabs-query.c (supports_vec_scatter_store_p): New function.
3806 * tree-vectorizer.h (vect_get_store_rhs): Declare.
3807 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
3808 true for scatter stores.
3809 (vect_gather_scatter_fn_p): Handle scatter stores too.
3810 (vect_check_gather_scatter): Consider using scatter stores if
3811 supports_vec_scatter_store_p.
3812 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
3813 scatter stores too.
3814 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
3815 internal_fn_stored_value_index.
3816 (check_load_store_masking): Handle scatter stores too.
3817 (vect_get_store_rhs): Make public.
3818 (vectorizable_call): Use internal_store_fn_p.
3819 (vectorizable_store): Handle scatter store internal functions.
3820 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
3821 when deciding whether the end of the group has been reached.
3822 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
3823 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
3824 (mask_scatter_store<mode>): New insns.
3826 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3827 Alan Hayward <alan.hayward@arm.com>
3828 David Sherwood <david.sherwood@arm.com>
3830 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
3831 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
3832 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
3833 function.
3834 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
3835 Use vect_truncate_gather_scatter_offset if we can't treat the
3836 operation as a normal gather load or scatter store.
3837 (get_group_load_store_type): Take the gather_scatter_info
3838 as argument. Try using a gather load or scatter store for
3839 single-element groups.
3840 (get_load_store_type): Update calls to get_group_load_store_type
3841 and vect_use_strided_gather_scatters_p.
3843 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3844 Alan Hayward <alan.hayward@arm.com>
3845 David Sherwood <david.sherwood@arm.com>
3847 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
3848 optional tree argument.
3849 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
3850 null target hooks.
3851 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
3852 but continue to use the current value as a fallback.
3853 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
3854 to compare the updates.
3855 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
3856 (get_load_store_type): Use it when handling a strided access.
3857 (vect_get_strided_load_store_ops): New function.
3858 (vect_get_data_ptr_increment): Likewise.
3859 (vectorizable_load): Handle strided gather loads. Always pass
3860 a step to vect_create_data_ref_ptr and bump_vector_ptr.
3862 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3863 Alan Hayward <alan.hayward@arm.com>
3864 David Sherwood <david.sherwood@arm.com>
3866 * doc/md.texi (gather_load@var{m}): Document.
3867 (mask_gather_load@var{m}): Likewise.
3868 * genopinit.c (main): Add supports_vec_gather_load and
3869 supports_vec_gather_load_cached to target_optabs.
3870 * optabs-tree.c (init_tree_optimization_optabs): Use
3871 ggc_cleared_alloc to allocate target_optabs.
3872 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
3873 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
3874 functions.
3875 * internal-fn.h (internal_load_fn_p): Declare.
3876 (internal_gather_scatter_fn_p): Likewise.
3877 (internal_fn_mask_index): Likewise.
3878 (internal_gather_scatter_fn_supported_p): Likewise.
3879 * internal-fn.c (gather_load_direct): New macro.
3880 (expand_gather_load_optab_fn): New function.
3881 (direct_gather_load_optab_supported_p): New macro.
3882 (direct_internal_fn_optab): New function.
3883 (internal_load_fn_p): Likewise.
3884 (internal_gather_scatter_fn_p): Likewise.
3885 (internal_fn_mask_index): Likewise.
3886 (internal_gather_scatter_fn_supported_p): Likewise.
3887 * optabs-query.c (supports_at_least_one_mode_p): New function.
3888 (supports_vec_gather_load_p): Likewise.
3889 * optabs-query.h (supports_vec_gather_load_p): Declare.
3890 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
3891 and memory_type field.
3892 (NUM_PATTERNS): Bump to 15.
3893 * tree-vect-data-refs.c: Include internal-fn.h.
3894 (vect_gather_scatter_fn_p): New function.
3895 (vect_describe_gather_scatter_call): Likewise.
3896 (vect_check_gather_scatter): Try using internal functions for
3897 gather loads. Recognize existing calls to a gather load function.
3898 (vect_analyze_data_refs): Consider using gather loads if
3899 supports_vec_gather_load_p.
3900 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
3901 (vect_get_gather_scatter_offset_type): Likewise.
3902 (vect_convert_mask_for_vectype): Likewise.
3903 (vect_add_conversion_to_patterm): Likewise.
3904 (vect_try_gather_scatter_pattern): Likewise.
3905 (vect_recog_gather_scatter_pattern): New pattern recognizer.
3906 (vect_vect_recog_func_ptrs): Add it.
3907 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
3908 internal_fn_mask_index and internal_gather_scatter_fn_p.
3909 (check_load_store_masking): Take the gather_scatter_info as an
3910 argument and handle gather loads.
3911 (vect_get_gather_scatter_ops): New function.
3912 (vectorizable_call): Check internal_load_fn_p.
3913 (vectorizable_load): Likewise. Handle gather load internal
3914 functions.
3915 (vectorizable_store): Update call to check_load_store_masking.
3916 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
3917 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
3918 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
3919 (aarch64_gather_scale_operand_d): New predicates.
3920 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
3921 (mask_gather_load<mode>): New insns.
3923 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3924 Alan Hayward <alan.hayward@arm.com>
3925 David Sherwood <david.sherwood@arm.com>
3927 * optabs.def (fold_left_plus_optab): New optab.
3928 * doc/md.texi (fold_left_plus_@var{m}): Document.
3929 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
3930 * internal-fn.c (fold_left_direct): Define.
3931 (expand_fold_left_optab_fn): Likewise.
3932 (direct_fold_left_optab_supported_p): Likewise.
3933 * fold-const-call.c (fold_const_fold_left): New function.
3934 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
3935 * tree-parloops.c (valid_reduction_p): New function.
3936 (gather_scalar_reductions): Use it.
3937 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
3938 (vect_finish_replace_stmt): Declare.
3939 * tree-vect-loop.c (fold_left_reduction_fn): New function.
3940 (needs_fold_left_reduction_p): New function, split out from...
3941 (vect_is_simple_reduction): ...here. Accept reductions that
3942 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
3943 (vect_force_simple_reduction): Also store the reduction type in
3944 the assignment's STMT_VINFO_REDUC_TYPE.
3945 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
3946 (merge_with_identity): New function.
3947 (vect_expand_fold_left): Likewise.
3948 (vectorize_fold_left_reduction): Likewise.
3949 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
3950 scalar phi in place for it. Check for target support and reject
3951 cases that would reassociate the operation. Defer the transform
3952 phase to vectorize_fold_left_reduction.
3953 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
3954 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
3955 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
3957 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3959 * tree-if-conv.c (predicate_mem_writes): Remove redundant
3960 call to ifc_temp_var.
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 * target.def (legitimize_address_displacement): Take the original
3967 offset as a poly_int.
3968 * targhooks.h (default_legitimize_address_displacement): Update
3969 accordingly.
3970 * targhooks.c (default_legitimize_address_displacement): Likewise.
3971 * doc/tm.texi: Regenerate.
3972 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
3973 as an argument, moving assert of ad->disp == ad->disp_term to...
3974 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
3975 Try calling targetm.legitimize_address_displacement before expanding
3976 the address rather than afterwards, and adjust for the new interface.
3977 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
3978 Match the new hook interface. Handle SVE addresses.
3979 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
3980 new hook interface.
3982 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3984 * Makefile.in (OBJS): Add early-remat.o.
3985 * target.def (select_early_remat_modes): New hook.
3986 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
3987 * doc/tm.texi: Regenerate.
3988 * targhooks.h (default_select_early_remat_modes): Declare.
3989 * targhooks.c (default_select_early_remat_modes): New function.
3990 * timevar.def (TV_EARLY_REMAT): New timevar.
3991 * passes.def (pass_early_remat): New pass.
3992 * tree-pass.h (make_pass_early_remat): Declare.
3993 * early-remat.c: New file.
3994 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
3995 function.
3996 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
3998 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3999 Alan Hayward <alan.hayward@arm.com>
4000 David Sherwood <david.sherwood@arm.com>
4002 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
4003 vfm1 with a bound_epilog parameter.
4004 (vect_do_peeling): Update calls accordingly, and move the prologue
4005 call earlier in the function. Treat the base bound_epilog as 0 for
4006 fully-masked loops and retain vf - 1 for other loops. Add 1 to
4007 this base when peeling for gaps.
4008 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
4009 with fully-masked loops.
4010 (vect_estimate_min_profitable_iters): Handle the single peeled
4011 iteration in that case.
4013 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4014 Alan Hayward <alan.hayward@arm.com>
4015 David Sherwood <david.sherwood@arm.com>
4017 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
4018 single-element interleaving even if the size is not a power of 2.
4019 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
4020 accesses for single-element interleaving if the group size is
4021 not a power of 2.
4023 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4024 Alan Hayward <alan.hayward@arm.com>
4025 David Sherwood <david.sherwood@arm.com>
4027 * doc/md.texi (fold_extract_last_@var{m}): Document.
4028 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
4029 * optabs.def (fold_extract_last_optab): New optab.
4030 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
4031 * internal-fn.c (fold_extract_direct): New macro.
4032 (expand_fold_extract_optab_fn): Likewise.
4033 (direct_fold_extract_optab_supported_p): Likewise.
4034 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
4035 * tree-vect-loop.c (vect_model_reduction_cost): Handle
4036 EXTRACT_LAST_REDUCTION.
4037 (get_initial_def_for_reduction): Do not create an initial vector
4038 for EXTRACT_LAST_REDUCTION reductions.
4039 (vectorizable_reduction): Leave the scalar phi in place for
4040 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
4041 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
4042 epilogue code for EXTRACT_LAST_REDUCTION and defer the
4043 transform phase to vectorizable_condition.
4044 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
4045 split out from...
4046 (vect_finish_stmt_generation): ...here.
4047 (vect_finish_replace_stmt): New function.
4048 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
4049 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
4050 pattern.
4051 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
4053 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4054 Alan Hayward <alan.hayward@arm.com>
4055 David Sherwood <david.sherwood@arm.com>
4057 * doc/md.texi (extract_last_@var{m}): Document.
4058 * optabs.def (extract_last_optab): New optab.
4059 * internal-fn.def (EXTRACT_LAST): New internal function.
4060 * internal-fn.c (cond_unary_direct): New macro.
4061 (expand_cond_unary_optab_fn): Likewise.
4062 (direct_cond_unary_optab_supported_p): Likewise.
4063 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
4064 loops using EXTRACT_LAST.
4065 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
4066 (extract_last_<mode>): ...this optab.
4067 (vec_extract<mode><Vel>): Update accordingly.
4069 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4070 Alan Hayward <alan.hayward@arm.com>
4071 David Sherwood <david.sherwood@arm.com>
4073 * target.def (empty_mask_is_expensive): New hook.
4074 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
4075 * doc/tm.texi: Regenerate.
4076 * targhooks.h (default_empty_mask_is_expensive): Declare.
4077 * targhooks.c (default_empty_mask_is_expensive): New function.
4078 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
4079 if the target says that empty masks are expensive.
4080 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
4081 New function.
4082 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
4084 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4085 Alan Hayward <alan.hayward@arm.com>
4086 David Sherwood <david.sherwood@arm.com>
4088 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
4089 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
4090 (vect_use_loop_mask_for_alignment_p): New function.
4091 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
4092 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4093 niters_skip argument. Make sure that the first niters_skip elements
4094 of the first iteration are inactive.
4095 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4096 Update call to vect_set_loop_masks_directly.
4097 (get_misalign_in_elems): New function, split out from...
4098 (vect_gen_prolog_loop_niters): ...here.
4099 (vect_update_init_of_dr): Take a code argument that specifies whether
4100 the adjustment should be added or subtracted.
4101 (vect_update_init_of_drs): Likewise.
4102 (vect_prepare_for_masked_peels): New function.
4103 (vect_do_peeling): Skip prologue peeling if we're using a mask
4104 instead. Update call to vect_update_inits_of_drs.
4105 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4106 mask_skip_niters.
4107 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4108 alignment. Do not include the number of peeled iterations in
4109 the minimum threshold in that case.
4110 (vectorizable_induction): Adjust the start value down by
4111 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4112 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4113 Take the number of skipped iterations into account when calculating
4114 the loop bounds.
4115 * tree-vect-stmts.c (vect_gen_while_not): New function.
4117 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4118 Alan Hayward <alan.hayward@arm.com>
4119 David Sherwood <david.sherwood@arm.com>
4121 * doc/sourcebuild.texi (vect_fully_masked): Document.
4122 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
4123 default value to 0.
4124 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
4125 split out from...
4126 (vect_analyze_loop_2): ...here. Don't check the vectorization
4127 factor against the number of loop iterations if the loop is
4128 fully-masked.
4130 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4131 Alan Hayward <alan.hayward@arm.com>
4132 David Sherwood <david.sherwood@arm.com>
4134 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
4135 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
4136 (dump_groups): Update accordingly.
4137 (iv_use::mem_type): New member variable.
4138 (address_p): New function.
4139 (record_use): Add a mem_type argument and initialize the new
4140 mem_type field.
4141 (record_group_use): Add a mem_type argument. Use address_p.
4142 Remove obsolete null checks of base_object. Update call to record_use.
4143 (find_interesting_uses_op): Update call to record_group_use.
4144 (find_interesting_uses_cond): Likewise.
4145 (find_interesting_uses_address): Likewise.
4146 (get_mem_type_for_internal_fn): New function.
4147 (find_address_like_use): Likewise.
4148 (find_interesting_uses_stmt): Try find_address_like_use before
4149 calling find_interesting_uses_op.
4150 (addr_offset_valid_p): Use the iv mem_type field as the type
4151 of the addressed memory.
4152 (add_autoinc_candidates): Likewise.
4153 (get_address_cost): Likewise.
4154 (split_small_address_groups_p): Use address_p.
4155 (split_address_groups): Likewise.
4156 (add_iv_candidate_for_use): Likewise.
4157 (autoinc_possible_for_pair): Likewise.
4158 (rewrite_groups): Likewise.
4159 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
4160 (determine_group_iv_cost): Update after split of USE_ADDRESS.
4161 (get_alias_ptr_type_for_ptr_address): New function.
4162 (rewrite_use_address): Rewrite address uses in calls that were
4163 identified by find_address_like_use.
4165 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4166 Alan Hayward <alan.hayward@arm.com>
4167 David Sherwood <david.sherwood@arm.com>
4169 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
4170 TARGET_MEM_REFs.
4171 * gimple-expr.h (is_gimple_addressable: Likewise.
4172 * gimple-expr.c (is_gimple_address): Likewise.
4173 * internal-fn.c (expand_call_mem_ref): New function.
4174 (expand_mask_load_optab_fn): Use it.
4175 (expand_mask_store_optab_fn): Likewise.
4177 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4178 Alan Hayward <alan.hayward@arm.com>
4179 David Sherwood <david.sherwood@arm.com>
4181 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
4182 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
4183 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
4184 (cond_umax@var{mode}): Document.
4185 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
4186 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
4187 (cond_umin_optab, cond_umax_optab): New optabs.
4188 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
4189 (COND_IOR, COND_XOR): New internal functions.
4190 * internal-fn.h (get_conditional_internal_fn): Declare.
4191 * internal-fn.c (cond_binary_direct): New macro.
4192 (expand_cond_binary_optab_fn): Likewise.
4193 (direct_cond_binary_optab_supported_p): Likewise.
4194 (get_conditional_internal_fn): New function.
4195 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
4196 Cope with reduction statements that are vectorized as calls rather
4197 than assignments.
4198 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
4199 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
4200 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
4201 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4202 (UNSPEC_COND_EOR): New unspecs.
4203 (optab): Add mappings for them.
4204 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
4205 (sve_int_op, sve_fp_op): New int attributes.
4207 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4208 Alan Hayward <alan.hayward@arm.com>
4209 David Sherwood <david.sherwood@arm.com>
4211 * optabs.def (while_ult_optab): New optab.
4212 * doc/md.texi (while_ult@var{m}@var{n}): Document.
4213 * internal-fn.def (WHILE_ULT): New internal function.
4214 * internal-fn.h (direct_internal_fn_supported_p): New override
4215 that takes two types as argument.
4216 * internal-fn.c (while_direct): New macro.
4217 (expand_while_optab_fn): New function.
4218 (convert_optab_supported_p): Likewise.
4219 (direct_while_optab_supported_p): New macro.
4220 * wide-int.h (wi::udiv_ceil): New function.
4221 * tree-vectorizer.h (rgroup_masks): New structure.
4222 (vec_loop_masks): New typedef.
4223 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
4224 and fully_masked_p.
4225 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
4226 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
4227 (vect_max_vf): New function.
4228 (slpeel_make_loop_iterate_ntimes): Delete.
4229 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
4230 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
4231 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
4232 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
4233 internal-fn.h, stor-layout.h and optabs-query.h.
4234 (vect_set_loop_mask): New function.
4235 (add_preheader_seq): Likewise.
4236 (add_header_seq): Likewise.
4237 (interleave_supported_p): Likewise.
4238 (vect_maybe_permute_loop_masks): Likewise.
4239 (vect_set_loop_masks_directly): Likewise.
4240 (vect_set_loop_condition_masked): Likewise.
4241 (vect_set_loop_condition_unmasked): New function, split out from
4242 slpeel_make_loop_iterate_ntimes.
4243 (slpeel_make_loop_iterate_ntimes): Rename to..
4244 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
4245 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
4246 (vect_do_peeling): Update call accordingly.
4247 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
4248 loops.
4249 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4250 mask_compare_type, can_fully_mask_p and fully_masked_p.
4251 (release_vec_loop_masks): New function.
4252 (_loop_vec_info): Use it to free the loop masks.
4253 (can_produce_all_loop_masks_p): New function.
4254 (vect_get_max_nscalars_per_iter): Likewise.
4255 (vect_verify_full_masking): Likewise.
4256 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
4257 retries, and free the mask rgroups before retrying. Check loop-wide
4258 reasons for disallowing fully-masked loops. Make the final decision
4259 about whether use a fully-masked loop or not.
4260 (vect_estimate_min_profitable_iters): Do not assume that peeling
4261 for the number of iterations will be needed for fully-masked loops.
4262 (vectorizable_reduction): Disable fully-masked loops.
4263 (vectorizable_live_operation): Likewise.
4264 (vect_halve_mask_nunits): New function.
4265 (vect_double_mask_nunits): Likewise.
4266 (vect_record_loop_mask): Likewise.
4267 (vect_get_loop_mask): Likewise.
4268 (vect_transform_loop): Handle the case in which the final loop
4269 iteration might handle a partial vector. Call vect_set_loop_condition
4270 instead of slpeel_make_loop_iterate_ntimes.
4271 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
4272 (check_load_store_masking): New function.
4273 (prepare_load_store_mask): Likewise.
4274 (vectorizable_store): Handle fully-masked loops.
4275 (vectorizable_load): Likewise.
4276 (supportable_widening_operation): Use vect_halve_mask_nunits for
4277 booleans.
4278 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
4279 (vect_gen_while): New function.
4280 * config/aarch64/aarch64.md (umax<mode>3): New expander.
4281 (aarch64_uqdec<mode>): New insn.
4283 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4284 Alan Hayward <alan.hayward@arm.com>
4285 David Sherwood <david.sherwood@arm.com>
4287 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
4288 (reduc_xor_scal_optab): New optabs.
4289 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
4290 (reduc_xor_scal_@var{m}): Document.
4291 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
4292 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
4293 internal functions.
4294 * fold-const-call.c (fold_const_call): Handle them.
4295 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
4296 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
4297 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
4298 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
4299 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
4300 (UNSPEC_XORV): New unspecs.
4301 (optab): Add entries for them.
4302 (BITWISEV): New int iterator.
4303 (bit_reduc_op): New int attributes.
4305 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4306 Alan Hayward <alan.hayward@arm.com>
4307 David Sherwood <david.sherwood@arm.com>
4309 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
4310 * internal-fn.def (VEC_SHL_INSERT): New internal function.
4311 * optabs.def (vec_shl_insert_optab): New optab.
4312 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
4313 (duplicate_and_interleave): Likewise.
4314 * tree-vect-loop.c: Include internal-fn.h.
4315 (neutral_op_for_slp_reduction): New function, split out from
4316 get_initial_defs_for_reduction.
4317 (get_initial_def_for_reduction): Handle option 2 for variable-length
4318 vectors by loading the neutral value into a vector and then shifting
4319 the initial value into element 0.
4320 (get_initial_defs_for_reduction): Replace the code argument with
4321 the neutral value calculated by neutral_op_for_slp_reduction.
4322 Use gimple_build_vector for constant-length vectors.
4323 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
4324 but the first group_size elements have a neutral value.
4325 Use duplicate_and_interleave otherwise.
4326 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
4327 Update call to get_initial_defs_for_reduction. Handle SLP
4328 reductions for variable-length vectors by creating one vector
4329 result for each scalar result, with the elements associated
4330 with other scalar results stubbed out with the neutral value.
4331 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
4332 Require IFN_VEC_SHL_INSERT for double reductions on
4333 variable-length vectors, or SLP reductions that have
4334 a neutral value. Require can_duplicate_and_interleave_p
4335 support for variable-length unchained SLP reductions if there
4336 is no neutral value, such as for MIN/MAX reductions. Also require
4337 the number of vector elements to be a multiple of the number of
4338 SLP statements when doing variable-length unchained SLP reductions.
4339 Update call to vect_create_epilog_for_reduction.
4340 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
4341 and remove initial values.
4342 (duplicate_and_interleave): Make public.
4343 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
4344 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
4346 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4347 Alan Hayward <alan.hayward@arm.com>
4348 David Sherwood <david.sherwood@arm.com>
4350 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
4351 (can_duplicate_and_interleave_p): New function.
4352 (vect_get_and_check_slp_defs): Take the vector of statements
4353 rather than just the current one. Remove excess parentheses.
4354 Restriction rejectinon of vect_constant_def and vect_external_def
4355 for variable-length vectors to boolean types, or types for which
4356 can_duplicate_and_interleave_p is false.
4357 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
4358 (duplicate_and_interleave): New function.
4359 (vect_get_constant_vectors): Use gimple_build_vector for
4360 constant-length vectors and suitable variable-length constant
4361 vectors. Use duplicate_and_interleave for other variable-length
4362 vectors. Don't defer the update when inserting new statements.
4364 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4365 Alan Hayward <alan.hayward@arm.com>
4366 David Sherwood <david.sherwood@arm.com>
4368 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
4369 min_profitable_iters doesn't go negative.
4371 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4372 Alan Hayward <alan.hayward@arm.com>
4373 David Sherwood <david.sherwood@arm.com>
4375 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
4376 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
4377 * optabs.def (vec_mask_load_lanes_optab): New optab.
4378 (vec_mask_store_lanes_optab): Likewise.
4379 * internal-fn.def (MASK_LOAD_LANES): New internal function.
4380 (MASK_STORE_LANES): Likewise.
4381 * internal-fn.c (mask_load_lanes_direct): New macro.
4382 (mask_store_lanes_direct): Likewise.
4383 (expand_mask_load_optab_fn): Handle masked operations.
4384 (expand_mask_load_lanes_optab_fn): New macro.
4385 (expand_mask_store_optab_fn): Handle masked operations.
4386 (expand_mask_store_lanes_optab_fn): New macro.
4387 (direct_mask_load_lanes_optab_supported_p): Likewise.
4388 (direct_mask_store_lanes_optab_supported_p): Likewise.
4389 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
4390 parameter.
4391 (vect_load_lanes_supported): Likewise.
4392 * tree-vect-data-refs.c (strip_conversion): New function.
4393 (can_group_stmts_p): Likewise.
4394 (vect_analyze_data_ref_accesses): Use it instead of checking
4395 for a pair of assignments.
4396 (vect_store_lanes_supported): Take a masked_p parameter.
4397 (vect_load_lanes_supported): Likewise.
4398 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
4399 vect_store_lanes_supported and vect_load_lanes_supported.
4400 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4401 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
4402 parameter. Don't allow gaps for masked accesses.
4403 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
4404 and vect_load_lanes_supported.
4405 (get_load_store_type): Take a masked_p parameter and update
4406 call to get_group_load_store_type.
4407 (vectorizable_store): Update call to get_load_store_type.
4408 Handle IFN_MASK_STORE_LANES.
4409 (vectorizable_load): Update call to get_load_store_type.
4410 Handle IFN_MASK_LOAD_LANES.
4412 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4413 Alan Hayward <alan.hayward@arm.com>
4414 David Sherwood <david.sherwood@arm.com>
4416 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
4417 modes for SVE.
4418 * config/aarch64/aarch64-protos.h
4419 (aarch64_sve_struct_memory_operand_p): Declare.
4420 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
4421 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
4422 (VPRED, vpred): Handle SVE structure modes.
4423 * config/aarch64/constraints.md (Utx): New constraint.
4424 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
4425 (aarch64_sve_struct_nonimmediate_operand): New predicates.
4426 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
4427 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
4428 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
4429 structure modes. Split into pieces after RA.
4430 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
4431 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
4432 New patterns.
4433 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
4434 SVE structure modes.
4435 (aarch64_classify_address): Likewise.
4436 (sizetochar): Move earlier in file.
4437 (aarch64_print_operand): Handle SVE register lists.
4438 (aarch64_array_mode): New function.
4439 (aarch64_sve_struct_memory_operand_p): Likewise.
4440 (TARGET_ARRAY_MODE): Redefine.
4442 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4443 Alan Hayward <alan.hayward@arm.com>
4444 David Sherwood <david.sherwood@arm.com>
4446 * target.def (array_mode): New target hook.
4447 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
4448 * doc/tm.texi: Regenerate.
4449 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
4450 * hooks.c (hook_optmode_mode_uhwi_none): New function.
4451 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
4452 targetm.array_mode.
4453 * stor-layout.c (mode_for_array): Likewise. Support polynomial
4454 type sizes.
4456 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4457 Alan Hayward <alan.hayward@arm.com>
4458 David Sherwood <david.sherwood@arm.com>
4460 * fold-const.c (fold_binary_loc): Check the argument types
4461 rather than the result type when testing for a vector operation.
4463 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4465 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
4466 * doc/tm.texi: Regenerate.
4468 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4469 Alan Hayward <alan.hayward@arm.com>
4470 David Sherwood <david.sherwood@arm.com>
4472 * doc/invoke.texi (-msve-vector-bits=): Document new option.
4473 (sve): Document new AArch64 extension.
4474 * doc/md.texi (w): Extend the description of the AArch64
4475 constraint to include SVE vectors.
4476 (Upl, Upa): Document new AArch64 predicate constraints.
4477 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
4478 enum.
4479 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
4480 (msve-vector-bits=): New option.
4481 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
4482 SVE when these are disabled.
4483 (sve): New extension.
4484 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
4485 modes. Adjust their number of units based on aarch64_sve_vg.
4486 (MAX_BITSIZE_MODE_ANY_MODE): Define.
4487 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
4488 aarch64_addr_query_type.
4489 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
4490 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
4491 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
4492 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
4493 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
4494 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
4495 (aarch64_simd_imm_zero_p): Delete.
4496 (aarch64_check_zero_based_sve_index_immediate): Declare.
4497 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4498 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4499 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4500 (aarch64_sve_float_mul_immediate_p): Likewise.
4501 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4502 rather than an rtx.
4503 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
4504 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
4505 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
4506 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
4507 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
4508 (aarch64_regmode_natural_size): Likewise.
4509 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
4510 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
4511 left one place.
4512 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
4513 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
4514 for VG and the SVE predicate registers.
4515 (V_ALIASES): Add a "z"-prefixed alias.
4516 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
4517 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
4518 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
4519 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
4520 (REG_CLASS_NAMES): Add entries for them.
4521 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
4522 and the predicate registers.
4523 (aarch64_sve_vg): Declare.
4524 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
4525 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
4526 (REGMODE_NATURAL_SIZE): Define.
4527 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4528 SVE macros.
4529 * config/aarch64/aarch64.c: Include cfgrtl.h.
4530 (simd_immediate_info): Add a constructor for series vectors,
4531 and an associated step field.
4532 (aarch64_sve_vg): New variable.
4533 (aarch64_dbx_register_number): Handle VG and the predicate registers.
4534 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
4535 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
4536 (VEC_ANY_DATA, VEC_STRUCT): New constants.
4537 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
4538 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
4539 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
4540 (aarch64_get_mask_mode): New functions.
4541 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
4542 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
4543 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
4544 predicate modes and predicate registers. Explicitly restrict
4545 GPRs to modes of 16 bytes or smaller. Only allow FP registers
4546 to store a vector mode if it is recognized by
4547 aarch64_classify_vector_mode.
4548 (aarch64_regmode_natural_size): New function.
4549 (aarch64_hard_regno_caller_save_mode): Return the original mode
4550 for predicates.
4551 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
4552 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
4553 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
4554 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
4555 functions.
4556 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
4557 does not overlap dest if the function is frame-related. Handle
4558 SVE constants.
4559 (aarch64_split_add_offset): New function.
4560 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
4561 them aarch64_add_offset.
4562 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
4563 and update call to aarch64_sub_sp.
4564 (aarch64_add_cfa_expression): New function.
4565 (aarch64_expand_prologue): Pass extra temporary registers to the
4566 functions above. Handle the case in which we need to emit new
4567 DW_CFA_expressions for registers that were originally saved
4568 relative to the stack pointer, but now have to be expressed
4569 relative to the frame pointer.
4570 (aarch64_output_mi_thunk): Pass extra temporary registers to the
4571 functions above.
4572 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
4573 IP0 and IP1 values for SVE frames.
4574 (aarch64_expand_vec_series): New function.
4575 (aarch64_expand_sve_widened_duplicate): Likewise.
4576 (aarch64_expand_sve_const_vector): Likewise.
4577 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
4578 Handle SVE constants. Use emit_move_insn to move a force_const_mem
4579 into the register, rather than emitting a SET directly.
4580 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
4581 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
4582 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
4583 (offset_9bit_signed_scaled_p): New functions.
4584 (aarch64_replicate_bitmask_imm): New function.
4585 (aarch64_bitmask_imm): Use it.
4586 (aarch64_cannot_force_const_mem): Reject expressions involving
4587 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
4588 (aarch64_classify_index): Handle SVE indices, by requiring
4589 a plain register index with a scale that matches the element size.
4590 (aarch64_classify_address): Handle SVE addresses. Assert that
4591 the mode of the address is VOIDmode or an integer mode.
4592 Update call to aarch64_classify_symbol.
4593 (aarch64_classify_symbolic_expression): Update call to
4594 aarch64_classify_symbol.
4595 (aarch64_const_vec_all_in_range_p): New function.
4596 (aarch64_print_vector_float_operand): Likewise.
4597 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
4598 "vN" for FP registers with SVE modes. Handle (const ...) vectors
4599 and the FP immediates 1.0 and 0.5.
4600 (aarch64_print_address_internal): Handle SVE addresses.
4601 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
4602 (aarch64_regno_regclass): Handle predicate registers.
4603 (aarch64_secondary_reload): Handle big-endian reloads of SVE
4604 data modes.
4605 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
4606 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
4607 (aarch64_convert_sve_vector_bits): New function.
4608 (aarch64_override_options): Use it to handle -msve-vector-bits=.
4609 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4610 rather than an rtx.
4611 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
4612 Handle SVE vector and predicate modes. Accept VL-based constants
4613 that need only one temporary register, and VL offsets that require
4614 no temporary registers.
4615 (aarch64_conditional_register_usage): Mark the predicate registers
4616 as fixed if SVE isn't available.
4617 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
4618 Return true for SVE vector and predicate modes.
4619 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
4620 rather than an unsigned int. Handle SVE modes.
4621 (aarch64_preferred_simd_mode): Update call accordingly. Handle
4622 SVE modes.
4623 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
4624 if SVE is enabled.
4625 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4626 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4627 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4628 (aarch64_sve_float_mul_immediate_p): New functions.
4629 (aarch64_sve_valid_immediate): New function.
4630 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
4631 Explicitly reject structure modes. Check for INDEX constants.
4632 Handle PTRUE and PFALSE constants.
4633 (aarch64_check_zero_based_sve_index_immediate): New function.
4634 (aarch64_simd_imm_zero_p): Delete.
4635 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
4636 vector modes. Accept constants in the range of CNT[BHWD].
4637 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
4638 ask for an Advanced SIMD mode.
4639 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
4640 (aarch64_simd_vector_alignment): Handle SVE predicates.
4641 (aarch64_vectorize_preferred_vector_alignment): New function.
4642 (aarch64_simd_vector_alignment_reachable): Use it instead of
4643 the vector size.
4644 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
4645 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
4646 functions.
4647 (MAX_VECT_LEN): Delete.
4648 (expand_vec_perm_d): Add a vec_flags field.
4649 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
4650 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
4651 (aarch64_evpc_ext): Don't apply a big-endian lane correction
4652 for SVE modes.
4653 (aarch64_evpc_rev): Rename to...
4654 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
4655 (aarch64_evpc_rev_global): New function.
4656 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
4657 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
4658 MAX_VECT_LEN.
4659 (aarch64_evpc_sve_tbl): New function.
4660 (aarch64_expand_vec_perm_const_1): Update after rename of
4661 aarch64_evpc_rev. Handle SVE permutes too, trying
4662 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
4663 than aarch64_evpc_tbl.
4664 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
4665 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
4666 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
4667 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
4668 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
4669 (aarch64_expand_sve_vcond): New functions.
4670 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
4671 of aarch64_vector_mode_p.
4672 (aarch64_dwarf_poly_indeterminate_value): New function.
4673 (aarch64_compute_pressure_classes): Likewise.
4674 (aarch64_can_change_mode_class): Likewise.
4675 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
4676 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
4677 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
4678 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
4679 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
4680 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
4681 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
4682 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
4683 constraints.
4684 (Dn, Dl, Dr): Accept const as well as const_vector.
4685 (Dz): Likewise. Compare against CONST0_RTX.
4686 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
4687 of "vector" where appropriate.
4688 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
4689 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
4690 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
4691 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
4692 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
4693 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
4694 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
4695 (v_int_equiv): Extend to SVE modes.
4696 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
4697 mode attributes.
4698 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
4699 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
4700 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
4701 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
4702 (SVE_COND_FP_CMP): New int iterators.
4703 (perm_hilo): Handle the new unpack unspecs.
4704 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
4705 attributes.
4706 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
4707 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
4708 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
4709 (aarch64_equality_operator, aarch64_constant_vector_operand)
4710 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
4711 (aarch64_sve_nonimmediate_operand): Likewise.
4712 (aarch64_sve_general_operand): Likewise.
4713 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
4714 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
4715 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
4716 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
4717 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
4718 (aarch64_sve_float_arith_immediate): Likewise.
4719 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
4720 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
4721 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
4722 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
4723 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
4724 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
4725 (aarch64_sve_float_arith_operand): Likewise.
4726 (aarch64_sve_float_arith_with_sub_operand): Likewise.
4727 (aarch64_sve_float_mul_operand): Likewise.
4728 (aarch64_sve_vec_perm_operand): Likewise.
4729 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
4730 (aarch64_mov_operand): Accept const_poly_int and const_vector.
4731 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
4732 as well as const_vector.
4733 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
4734 in file. Use CONST0_RTX and CONSTM1_RTX.
4735 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
4736 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
4737 Use aarch64_simd_imm_zero.
4738 * config/aarch64/aarch64-sve.md: New file.
4739 * config/aarch64/aarch64.md: Include it.
4740 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
4741 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
4742 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
4743 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
4744 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
4745 (sve): New attribute.
4746 (enabled): Disable instructions with the sve attribute unless
4747 TARGET_SVE.
4748 (movqi, movhi): Pass CONST_POLY_INT operaneds through
4749 aarch64_expand_mov_immediate.
4750 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
4751 CNT[BHSD] immediates.
4752 (movti): Split CONST_POLY_INT moves into two halves.
4753 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
4754 Split additions that need a temporary here if the destination
4755 is the stack pointer.
4756 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
4757 (*add<mode>3_poly_1): New instruction.
4758 (set_clobber_cc): New expander.
4760 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4762 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
4763 parameter and use it instead of GET_MODE_SIZE (innermode). Use
4764 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
4765 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
4766 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
4767 Change innermode from fixed_mode_size to machine_mode.
4768 (simplify_subreg): Update call accordingly. Handle a constant-sized
4769 subreg of a variable-length CONST_VECTOR.
4771 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4772 Alan Hayward <alan.hayward@arm.com>
4773 David Sherwood <david.sherwood@arm.com>
4775 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
4776 (add_offset_to_base): New function, split out from...
4777 (create_mem_ref): ...here. When handling a scale other than 1,
4778 check first whether the address is valid without the offset.
4779 Add it into the base if so, leaving the index and scale as-is.
4781 2018-01-12 Jakub Jelinek <jakub@redhat.com>
4783 PR c++/83778
4784 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
4785 fold_for_warn before checking if arg2 is INTEGER_CST.
4787 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
4789 * config/rs6000/predicates.md (load_multiple_operation): Delete.
4790 (store_multiple_operation): Delete.
4791 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
4792 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
4793 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
4794 guarded by TARGET_STRING.
4795 (rs6000_output_load_multiple): Delete.
4796 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
4797 OPTION_MASK_STRING / TARGET_STRING handling.
4798 (print_operand) <'N', 'O'>: Add comment that these are unused now.
4799 (const rs6000_opt_masks) <"string">: Change mask to 0.
4800 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
4801 (MASK_STRING): Delete.
4802 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
4803 parts. Simplify.
4804 (load_multiple): Delete.
4805 (*ldmsi8): Delete.
4806 (*ldmsi7): Delete.
4807 (*ldmsi6): Delete.
4808 (*ldmsi5): Delete.
4809 (*ldmsi4): Delete.
4810 (*ldmsi3): Delete.
4811 (store_multiple): Delete.
4812 (*stmsi8): Delete.
4813 (*stmsi7): Delete.
4814 (*stmsi6): Delete.
4815 (*stmsi5): Delete.
4816 (*stmsi4): Delete.
4817 (*stmsi3): Delete.
4818 (movmemsi_8reg): Delete.
4819 (corresponding unnamed define_insn): Delete.
4820 (movmemsi_6reg): Delete.
4821 (corresponding unnamed define_insn): Delete.
4822 (movmemsi_4reg): Delete.
4823 (corresponding unnamed define_insn): Delete.
4824 (movmemsi_2reg): Delete.
4825 (corresponding unnamed define_insn): Delete.
4826 (movmemsi_1reg): Delete.
4827 (corresponding unnamed define_insn): Delete.
4828 * config/rs6000/rs6000.opt (mno-string): New.
4829 (mstring): Replace by deprecation warning stub.
4830 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
4832 2018-01-12 Jakub Jelinek <jakub@redhat.com>
4834 * regrename.c (regrename_do_replace): If replacing the same
4835 reg multiple times, try to reuse last created gen_raw_REG.
4837 PR debug/81155
4838 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
4839 main to workaround a bug in GDB.
4841 2018-01-12 Tom de Vries <tom@codesourcery.com>
4843 PR target/83737
4844 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
4846 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
4848 PR rtl-optimization/80481
4849 * ira-color.c (get_cap_member): New function.
4850 (allocnos_conflict_by_live_ranges_p): Use it.
4851 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
4852 (setup_slot_coalesced_allocno_live_ranges): Ditto.
4854 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
4856 PR target/83628
4857 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
4858 (*saddl_se_1): Ditto.
4859 (*ssubsi_1): Ditto.
4860 (*ssubl_se_1): Ditto.
4862 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4864 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
4865 rather than wi::to_widest for DR_INITs.
4866 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
4867 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
4868 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
4869 INTEGER_CSTs.
4870 (vect_analyze_group_access_1): Note that here.
4872 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4874 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
4875 polynomial type sizes.
4877 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4879 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
4880 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
4881 (gimple_add_tmp_var): Likewise.
4883 2018-01-12 Martin Liska <mliska@suse.cz>
4885 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
4886 (gimple_alloc_sizes): Likewise.
4887 (dump_gimple_statistics): Use PRIu64 in printf format.
4888 * gimple.h: Change uint64_t to int.
4890 2018-01-12 Martin Liska <mliska@suse.cz>
4892 * tree-core.h: Use uint64_t instead of int.
4893 * tree.c (tree_node_counts): Likewise.
4894 (tree_node_sizes): Likewise.
4895 (dump_tree_statistics): Use PRIu64 in printf format.
4897 2018-01-12 Martin Liska <mliska@suse.cz>
4899 * Makefile.in: As qsort_chk is implemented in vec.c, add
4900 vec.o to linkage of gencfn-macros.
4901 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
4902 passing the info to record_node_allocation_statistics.
4903 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
4904 and pass the info.
4905 * ggc-common.c (struct ggc_usage): Add operator== and use
4906 it in operator< and compare function.
4907 * mem-stats.h (struct mem_usage): Likewise.
4908 * vec.c (struct vec_usage): Remove operator< and compare
4909 function. Can be simply inherited.
4911 2018-01-12 Martin Jambor <mjambor@suse.cz>
4913 PR target/81616
4914 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
4915 * tree-ssa-math-opts.c: Include domwalk.h.
4916 (convert_mult_to_fma_1): New function.
4917 (fma_transformation_info): New type.
4918 (fma_deferring_state): Likewise.
4919 (cancel_fma_deferring): New function.
4920 (result_of_phi): Likewise.
4921 (last_fma_candidate_feeds_initial_phi): Likewise.
4922 (convert_mult_to_fma): Added deferring logic, split actual
4923 transformation to convert_mult_to_fma_1.
4924 (math_opts_dom_walker): New type.
4925 (math_opts_dom_walker::after_dom_children): New method, body moved
4926 here from pass_optimize_widening_mul::execute, added deferring logic
4927 bits.
4928 (pass_optimize_widening_mul::execute): Moved most of code to
4929 math_opts_dom_walker::after_dom_children.
4930 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
4931 * config/i386/i386.c (ix86_option_override_internal): Added
4932 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
4934 2018-01-12 Richard Biener <rguenther@suse.de>
4936 PR debug/83157
4937 * dwarf2out.c (gen_variable_die): Do not reset old_die for
4938 inline instance vars.
4940 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
4942 PR target/81819
4943 * config/rx/rx.c (rx_is_restricted_memory_address):
4944 Handle SUBREG case.
4946 2018-01-12 Richard Biener <rguenther@suse.de>
4948 PR tree-optimization/80846
4949 * target.def (split_reduction): New target hook.
4950 * targhooks.c (default_split_reduction): New function.
4951 * targhooks.h (default_split_reduction): Declare.
4952 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
4953 target requests first reduce vectors by combining low and high
4954 parts.
4955 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
4956 (get_vectype_for_scalar_type_and_size): Export.
4957 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
4958 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
4959 * doc/tm.texi: Regenerate.
4960 * config/i386/i386.c (ix86_split_reduction): Implement
4961 TARGET_VECTORIZE_SPLIT_REDUCTION.
4963 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4965 PR target/83368
4966 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
4967 in PIC mode except for TARGET_VXWORKS_RTP.
4968 * config/sparc/sparc.c: Include cfgrtl.h.
4969 (TARGET_INIT_PIC_REG): Define.
4970 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
4971 (sparc_pic_register_p): New predicate.
4972 (sparc_legitimate_address_p): Use it.
4973 (sparc_legitimize_pic_address): Likewise.
4974 (sparc_delegitimize_address): Likewise.
4975 (sparc_mode_dependent_address_p): Likewise.
4976 (gen_load_pcrel_sym): Remove 4th parameter.
4977 (load_got_register): Adjust call to above. Remove obsolete stuff.
4978 (sparc_expand_prologue): Do not call load_got_register here.
4979 (sparc_flat_expand_prologue): Likewise.
4980 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
4981 (sparc_use_pseudo_pic_reg): New function.
4982 (sparc_init_pic_reg): Likewise.
4983 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
4984 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
4986 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
4988 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
4989 Add item for branch_cost.
4991 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4993 PR rtl-optimization/83565
4994 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
4995 not extend the result to a larger mode for rotate operations.
4996 (num_sign_bit_copies1): Likewise.
4998 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5000 PR target/40411
5001 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
5002 -symbolic.
5003 Use values-Xc.o for -pedantic.
5004 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
5006 2018-01-12 Martin Liska <mliska@suse.cz>
5008 PR ipa/83054
5009 * ipa-devirt.c (final_warning_record::grow_type_warnings):
5010 New function.
5011 (possible_polymorphic_call_targets): Use it.
5012 (ipa_devirt): Likewise.
5014 2018-01-12 Martin Liska <mliska@suse.cz>
5016 * profile-count.h (enum profile_quality): Use 0 as invalid
5017 enum value of profile_quality.
5019 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
5021 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
5022 -mext-string options.
5024 2018-01-12 Richard Biener <rguenther@suse.de>
5026 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
5027 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
5028 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
5029 Likewise.
5030 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
5032 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5034 * configure.ac (--with-long-double-format): Add support for the
5035 configuration option to change the default long double format on
5036 PowerPC systems.
5037 * config.gcc (powerpc*-linux*-*): Likewise.
5038 * configure: Regenerate.
5039 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
5040 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
5041 used without modification.
5043 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5045 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
5046 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
5047 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
5048 MISC_BUILTIN_SPEC_BARRIER.
5049 (rs6000_init_builtins): Likewise.
5050 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
5051 enum value.
5052 (speculation_barrier): New define_insn.
5053 * doc/extend.texi: Document __builtin_speculation_barrier.
5055 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5057 PR target/83203
5058 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
5059 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
5060 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
5061 iterators.
5062 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
5063 integral modes instead of "ss" and "sd".
5064 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
5065 vectors with 32-bit and 64-bit elements.
5066 (vecdupssescalarmodesuffix): New mode attribute.
5067 (vec_dup<mode>): Use it.
5069 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
5071 PR target/83330
5072 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
5073 frame if argument is passed on stack.
5075 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5077 PR target/82682
5078 * ree.c (combine_reaching_defs): Optimize also
5079 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
5080 reg2=any_extend(exp); reg1=reg2;, formatting fix.
5082 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5084 PR middle-end/83189
5085 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
5087 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5089 PR middle-end/83718
5090 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
5091 after they are computed.
5093 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5095 PR tree-optimization/83695
5096 * gimple-loop-linterchange.cc
5097 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5098 reset cached scev information after interchange.
5099 (pass_linterchange::execute): Remove call to scev_reset_htab.
5101 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5103 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5104 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5105 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5106 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5107 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5108 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5109 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5110 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5111 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5112 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5113 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5114 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5115 (V_lane_reg): Likewise.
5116 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5117 New define_expand.
5118 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5119 (vfmal_lane_low<mode>_intrinsic,
5120 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5121 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5122 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
5123 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
5124 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
5125 vfmsl_lane_high<mode>_intrinsic): New define_insns.
5127 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5129 * config/arm/arm-cpus.in (fp16fml): New feature.
5130 (ALL_SIMD): Add fp16fml.
5131 (armv8.2-a): Add fp16fml as an option.
5132 (armv8.3-a): Likewise.
5133 (armv8.4-a): Add fp16fml as part of fp16.
5134 * config/arm/arm.h (TARGET_FP16FML): Define.
5135 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
5136 when appropriate.
5137 * config/arm/arm-modes.def (V2HF): Define.
5138 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
5139 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
5140 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
5141 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
5142 vfmsl_low, vfmsl_high): New set of builtins.
5143 * config/arm/iterators.md (PLUSMINUS): New code iterator.
5144 (vfml_op): New code attribute.
5145 (VFMLHALVES): New int iterator.
5146 (VFML, VFMLSEL): New mode attributes.
5147 (V_reg): Define mapping for V2HF.
5148 (V_hi, V_lo): New mode attributes.
5149 (VF_constraint): Likewise.
5150 (vfml_half, vfml_half_selector): New int attributes.
5151 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
5152 define_expand.
5153 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
5154 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
5155 New define_insn.
5156 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
5157 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
5158 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
5159 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
5160 documentation.
5161 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
5162 Document new effective target and option set.
5164 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5166 * config/arm/arm-cpus.in (armv8_4): New feature.
5167 (ARMv8_4a): New fgroup.
5168 (armv8.4-a): New arch.
5169 * config/arm/arm-tables.opt: Regenerate.
5170 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
5171 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
5172 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
5173 Add matching rules for -march=armv8.4-a and extensions.
5174 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
5176 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
5178 PR target/81821
5179 * config/rx/rx.md (BW): New mode attribute.
5180 (sync_lock_test_and_setsi): Add mode suffix to insn output.
5182 2018-01-11 Richard Biener <rguenther@suse.de>
5184 PR tree-optimization/83435
5185 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
5186 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
5187 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
5189 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5190 Alan Hayward <alan.hayward@arm.com>
5191 David Sherwood <david.sherwood@arm.com>
5193 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
5194 field.
5195 (aarch64_classify_address): Initialize it. Track polynomial offsets.
5196 (aarch64_print_address_internal): Use it to check for a zero offset.
5198 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5199 Alan Hayward <alan.hayward@arm.com>
5200 David Sherwood <david.sherwood@arm.com>
5202 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
5203 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
5204 Return a poly_int64 rather than a HOST_WIDE_INT.
5205 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
5206 rather than a HOST_WIDE_INT.
5207 * config/aarch64/aarch64.h (aarch64_frame): Protect with
5208 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
5209 hard_fp_offset, frame_size, initial_adjust, callee_offset and
5210 final_offset from HOST_WIDE_INT to poly_int64.
5211 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
5212 to_constant when getting the number of units in an Advanced SIMD
5213 mode.
5214 (aarch64_builtin_vectorized_function): Check for a constant number
5215 of units.
5216 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
5217 GET_MODE_SIZE.
5218 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
5219 attribute instead of GET_MODE_NUNITS.
5220 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
5221 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
5222 GET_MODE_SIZE for fixed-size registers.
5223 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
5224 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
5225 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
5226 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
5227 (aarch64_print_operand, aarch64_print_address_internal)
5228 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
5229 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
5230 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
5231 Handle polynomial GET_MODE_SIZE.
5232 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
5233 wider than SImode without modification.
5234 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
5235 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
5236 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
5237 passing and returning SVE modes.
5238 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
5239 rather than GEN_INT.
5240 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
5241 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
5242 (aarch64_allocate_and_probe_stack_space): Likewise.
5243 (aarch64_layout_frame): Cope with polynomial offsets.
5244 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
5245 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
5246 polynomial offsets.
5247 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
5248 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
5249 poly_int64 rather than a HOST_WIDE_INT.
5250 (aarch64_get_separate_components, aarch64_process_components)
5251 (aarch64_expand_prologue, aarch64_expand_epilogue)
5252 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
5253 (aarch64_anchor_offset): New function, split out from...
5254 (aarch64_legitimize_address): ...here.
5255 (aarch64_builtin_vectorization_cost): Handle polynomial
5256 TYPE_VECTOR_SUBPARTS.
5257 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
5258 GET_MODE_NUNITS.
5259 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
5260 number of elements from the PARALLEL rather than the mode.
5261 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
5262 rather than GET_MODE_BITSIZE.
5263 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
5264 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
5265 (aarch64_expand_vec_perm_const_1): Handle polynomial
5266 d->perm.length () and d->perm elements.
5267 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
5268 Apply to_constant to d->perm elements.
5269 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
5270 polynomial CONST_VECTOR_NUNITS.
5271 (aarch64_move_pointer): Take amount as a poly_int64 rather
5272 than an int.
5273 (aarch64_progress_pointer): Avoid temporary variable.
5274 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
5275 the mode attribute instead of GET_MODE.
5277 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5278 Alan Hayward <alan.hayward@arm.com>
5279 David Sherwood <david.sherwood@arm.com>
5281 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
5282 x exists before using it.
5283 (aarch64_add_constant_internal): Rename to...
5284 (aarch64_add_offset_1): ...this. Replace regnum with separate
5285 src and dest rtxes. Handle the case in which they're different,
5286 including when the offset is zero. Replace scratchreg with an rtx.
5287 Use 2 additions if there is no spare register into which we can
5288 move a 16-bit constant.
5289 (aarch64_add_constant): Delete.
5290 (aarch64_add_offset): Replace reg with separate src and dest
5291 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
5292 Use aarch64_add_offset_1.
5293 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
5294 an rtx rather than an int. Take the delta as a poly_int64
5295 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
5296 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
5297 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
5298 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
5299 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
5300 and aarch64_add_sp.
5301 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
5302 aarch64_add_constant.
5304 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5306 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
5307 Use scalar_float_mode.
5309 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5311 * config/aarch64/aarch64-simd.md
5312 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
5313 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
5314 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
5315 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
5316 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
5317 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
5318 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
5319 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
5320 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
5321 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
5323 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5325 PR target/83514
5326 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
5327 targ_options->x_arm_arch_string is non NULL.
5329 2018-01-11 Tamar Christina <tamar.christina@arm.com>
5331 * config/aarch64/aarch64.h
5332 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
5334 2018-01-11 Sudakshina Das <sudi.das@arm.com>
5336 PR target/82096
5337 * expmed.c (emit_store_flag_force): Swap if const op0
5338 and change VOIDmode to mode of op0.
5340 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5342 PR rtl-optimization/83761
5343 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
5344 than bytes to mode_for_size.
5346 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5348 PR middle-end/83189
5349 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
5350 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
5351 profile.
5353 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5355 PR middle-end/83575
5356 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
5357 when in layout mode.
5358 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
5359 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
5360 partition fixup.
5362 2018-01-10 Michael Collison <michael.collison@arm.com>
5364 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
5365 * config/aarch64/aarch64-option-extension.def: Add
5366 AARCH64_OPT_EXTENSION of 'fp16fml'.
5367 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5368 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
5369 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
5370 * config/aarch64/constraints.md (Ui7): New constraint.
5371 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
5372 (VFMLA_SEL_W): Ditto.
5373 (f16quad): Ditto.
5374 (f16mac1): Ditto.
5375 (VFMLA16_LOW): New int iterator.
5376 (VFMLA16_HIGH): Ditto.
5377 (UNSPEC_FMLAL): New unspec.
5378 (UNSPEC_FMLSL): Ditto.
5379 (UNSPEC_FMLAL2): Ditto.
5380 (UNSPEC_FMLSL2): Ditto.
5381 (f16mac): New code attribute.
5382 * config/aarch64/aarch64-simd-builtins.def
5383 (aarch64_fmlal_lowv2sf): Ditto.
5384 (aarch64_fmlsl_lowv2sf): Ditto.
5385 (aarch64_fmlalq_lowv4sf): Ditto.
5386 (aarch64_fmlslq_lowv4sf): Ditto.
5387 (aarch64_fmlal_highv2sf): Ditto.
5388 (aarch64_fmlsl_highv2sf): Ditto.
5389 (aarch64_fmlalq_highv4sf): Ditto.
5390 (aarch64_fmlslq_highv4sf): Ditto.
5391 (aarch64_fmlal_lane_lowv2sf): Ditto.
5392 (aarch64_fmlsl_lane_lowv2sf): Ditto.
5393 (aarch64_fmlal_laneq_lowv2sf): Ditto.
5394 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
5395 (aarch64_fmlalq_lane_lowv4sf): Ditto.
5396 (aarch64_fmlsl_lane_lowv4sf): Ditto.
5397 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
5398 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
5399 (aarch64_fmlal_lane_highv2sf): Ditto.
5400 (aarch64_fmlsl_lane_highv2sf): Ditto.
5401 (aarch64_fmlal_laneq_highv2sf): Ditto.
5402 (aarch64_fmlsl_laneq_highv2sf): Ditto.
5403 (aarch64_fmlalq_lane_highv4sf): Ditto.
5404 (aarch64_fmlsl_lane_highv4sf): Ditto.
5405 (aarch64_fmlalq_laneq_highv4sf): Ditto.
5406 (aarch64_fmlsl_laneq_highv4sf): Ditto.
5407 * config/aarch64/aarch64-simd.md:
5408 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
5409 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5410 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
5411 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5412 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
5413 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
5414 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
5415 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
5416 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
5417 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
5418 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
5419 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
5420 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
5421 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
5422 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
5423 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
5424 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
5425 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
5426 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
5427 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
5428 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
5429 (vfmlsl_low_u32): Ditto.
5430 (vfmlalq_low_u32): Ditto.
5431 (vfmlslq_low_u32): Ditto.
5432 (vfmlal_high_u32): Ditto.
5433 (vfmlsl_high_u32): Ditto.
5434 (vfmlalq_high_u32): Ditto.
5435 (vfmlslq_high_u32): Ditto.
5436 (vfmlal_lane_low_u32): Ditto.
5437 (vfmlsl_lane_low_u32): Ditto.
5438 (vfmlal_laneq_low_u32): Ditto.
5439 (vfmlsl_laneq_low_u32): Ditto.
5440 (vfmlalq_lane_low_u32): Ditto.
5441 (vfmlslq_lane_low_u32): Ditto.
5442 (vfmlalq_laneq_low_u32): Ditto.
5443 (vfmlslq_laneq_low_u32): Ditto.
5444 (vfmlal_lane_high_u32): Ditto.
5445 (vfmlsl_lane_high_u32): Ditto.
5446 (vfmlal_laneq_high_u32): Ditto.
5447 (vfmlsl_laneq_high_u32): Ditto.
5448 (vfmlalq_lane_high_u32): Ditto.
5449 (vfmlslq_lane_high_u32): Ditto.
5450 (vfmlalq_laneq_high_u32): Ditto.
5451 (vfmlslq_laneq_high_u32): Ditto.
5452 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
5453 (AARCH64_FL_FOR_ARCH8_4): New.
5454 (AARCH64_ISA_F16FML): New ISA flag.
5455 (TARGET_F16FML): New feature flag for fp16fml.
5456 (doc/invoke.texi): Document new fp16fml option.
5458 2018-01-10 Michael Collison <michael.collison@arm.com>
5460 * config/aarch64/aarch64-builtins.c:
5461 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
5462 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5463 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
5464 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
5465 (AARCH64_ISA_SHA3): New ISA flag.
5466 (TARGET_SHA3): New feature flag for sha3.
5467 * config/aarch64/iterators.md (sha512_op): New int attribute.
5468 (CRYPTO_SHA512): New int iterator.
5469 (UNSPEC_SHA512H): New unspec.
5470 (UNSPEC_SHA512H2): Ditto.
5471 (UNSPEC_SHA512SU0): Ditto.
5472 (UNSPEC_SHA512SU1): Ditto.
5473 * config/aarch64/aarch64-simd-builtins.def
5474 (aarch64_crypto_sha512hqv2di): New builtin.
5475 (aarch64_crypto_sha512h2qv2di): Ditto.
5476 (aarch64_crypto_sha512su0qv2di): Ditto.
5477 (aarch64_crypto_sha512su1qv2di): Ditto.
5478 (aarch64_eor3qv8hi): Ditto.
5479 (aarch64_rax1qv2di): Ditto.
5480 (aarch64_xarqv2di): Ditto.
5481 (aarch64_bcaxqv8hi): Ditto.
5482 * config/aarch64/aarch64-simd.md:
5483 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
5484 (aarch64_crypto_sha512su0qv2di): Ditto.
5485 (aarch64_crypto_sha512su1qv2di): Ditto.
5486 (aarch64_eor3qv8hi): Ditto.
5487 (aarch64_rax1qv2di): Ditto.
5488 (aarch64_xarqv2di): Ditto.
5489 (aarch64_bcaxqv8hi): Ditto.
5490 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
5491 (vsha512h2q_u64): Ditto.
5492 (vsha512su0q_u64): Ditto.
5493 (vsha512su1q_u64): Ditto.
5494 (veor3q_u16): Ditto.
5495 (vrax1q_u64): Ditto.
5496 (vxarq_u64): Ditto.
5497 (vbcaxq_u16): Ditto.
5498 * config/arm/types.md (crypto_sha512): New type attribute.
5499 (crypto_sha3): Ditto.
5500 (doc/invoke.texi): Document new sha3 option.
5502 2018-01-10 Michael Collison <michael.collison@arm.com>
5504 * config/aarch64/aarch64-builtins.c:
5505 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
5506 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5507 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
5508 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
5509 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
5510 (AARCH64_ISA_SM4): New ISA flag.
5511 (TARGET_SM4): New feature flag for sm4.
5512 * config/aarch64/aarch64-simd-builtins.def
5513 (aarch64_sm3ss1qv4si): Ditto.
5514 (aarch64_sm3tt1aq4si): Ditto.
5515 (aarch64_sm3tt1bq4si): Ditto.
5516 (aarch64_sm3tt2aq4si): Ditto.
5517 (aarch64_sm3tt2bq4si): Ditto.
5518 (aarch64_sm3partw1qv4si): Ditto.
5519 (aarch64_sm3partw2qv4si): Ditto.
5520 (aarch64_sm4eqv4si): Ditto.
5521 (aarch64_sm4ekeyqv4si): Ditto.
5522 * config/aarch64/aarch64-simd.md:
5523 (aarch64_sm3ss1qv4si): Ditto.
5524 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
5525 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
5526 (aarch64_sm4eqv4si): Ditto.
5527 (aarch64_sm4ekeyqv4si): Ditto.
5528 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
5529 (sm3part_op): Ditto.
5530 (CRYPTO_SM3TT): Ditto.
5531 (CRYPTO_SM3PART): Ditto.
5532 (UNSPEC_SM3SS1): New unspec.
5533 (UNSPEC_SM3TT1A): Ditto.
5534 (UNSPEC_SM3TT1B): Ditto.
5535 (UNSPEC_SM3TT2A): Ditto.
5536 (UNSPEC_SM3TT2B): Ditto.
5537 (UNSPEC_SM3PARTW1): Ditto.
5538 (UNSPEC_SM3PARTW2): Ditto.
5539 (UNSPEC_SM4E): Ditto.
5540 (UNSPEC_SM4EKEY): Ditto.
5541 * config/aarch64/constraints.md (Ui2): New constraint.
5542 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
5543 * config/arm/types.md (crypto_sm3): New type attribute.
5544 (crypto_sm4): Ditto.
5545 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
5546 (vsm3tt1aq_u32): Ditto.
5547 (vsm3tt1bq_u32): Ditto.
5548 (vsm3tt2aq_u32): Ditto.
5549 (vsm3tt2bq_u32): Ditto.
5550 (vsm3partw1q_u32): Ditto.
5551 (vsm3partw2q_u32): Ditto.
5552 (vsm4eq_u32): Ditto.
5553 (vsm4ekeyq_u32): Ditto.
5554 (doc/invoke.texi): Document new sm4 option.
5556 2018-01-10 Michael Collison <michael.collison@arm.com>
5558 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
5559 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
5560 (AARCH64_FL_FOR_ARCH8_4): New.
5561 (AARCH64_FL_V8_4): New flag.
5562 (doc/invoke.texi): Document new armv8.4-a option.
5564 2018-01-10 Michael Collison <michael.collison@arm.com>
5566 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5567 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
5568 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
5569 * config/aarch64/aarch64-option-extension.def: Add
5570 AARCH64_OPT_EXTENSION of 'sha2'.
5571 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
5572 (crypto): Disable sha2 and aes if crypto disabled.
5573 (crypto): Enable aes and sha2 if enabled.
5574 (simd): Disable sha2 and aes if simd disabled.
5575 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
5576 New flags.
5577 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
5578 (TARGET_SHA2): New feature flag for sha2.
5579 (TARGET_AES): New feature flag for aes.
5580 * config/aarch64/aarch64-simd.md:
5581 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
5582 conditional on TARGET_AES.
5583 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
5584 (aarch64_crypto_sha1hsi): Make pattern conditional
5585 on TARGET_SHA2.
5586 (aarch64_crypto_sha1hv4si): Ditto.
5587 (aarch64_be_crypto_sha1hv4si): Ditto.
5588 (aarch64_crypto_sha1su1v4si): Ditto.
5589 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
5590 (aarch64_crypto_sha1su0v4si): Ditto.
5591 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
5592 (aarch64_crypto_sha256su0v4si): Ditto.
5593 (aarch64_crypto_sha256su1v4si): Ditto.
5594 (doc/invoke.texi): Document new aes and sha2 options.
5596 2018-01-10 Martin Sebor <msebor@redhat.com>
5598 PR tree-optimization/83781
5599 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
5600 as string arrays.
5602 2018-01-11 Martin Sebor <msebor@gmail.com>
5603 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5605 PR tree-optimization/83501
5606 PR tree-optimization/81703
5608 * tree-ssa-strlen.c (get_string_cst): Rename...
5609 (get_string_len): ...to this. Handle global constants.
5610 (handle_char_store): Adjust.
5612 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
5613 Jim Wilson <jimw@sifive.com>
5615 * config/riscv/riscv-protos.h (riscv_output_return): New.
5616 * config/riscv/riscv.c (struct machine_function): New naked_p field.
5617 (riscv_attribute_table, riscv_output_return),
5618 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
5619 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
5620 (riscv_compute_frame_info): Only compute frame->mask if not a naked
5621 function.
5622 (riscv_expand_prologue): Add early return for naked function.
5623 (riscv_expand_epilogue): Likewise.
5624 (riscv_function_ok_for_sibcall): Return false for naked function.
5625 (riscv_set_current_function): New.
5626 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
5627 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
5628 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
5629 * doc/extend.texi (RISC-V Function Attributes): New.
5631 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
5633 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
5634 check for 128-bit long double before checking TCmode.
5635 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
5636 128-bit long doubles before checking TFmode or TCmode.
5637 (FLOAT128_IBM_P): Likewise.
5639 2018-01-10 Martin Sebor <msebor@redhat.com>
5641 PR tree-optimization/83671
5642 * builtins.c (c_strlen): Unconditionally return zero for the empty
5643 string.
5644 Use -Warray-bounds for warnings.
5645 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
5646 for non-constant array indices with COMPONENT_REF, arrays of
5647 arrays, and pointers to arrays.
5648 (gimple_fold_builtin_strlen): Determine and set length range for
5649 non-constant character arrays.
5651 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
5653 PR middle-end/81897
5654 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
5655 empty blocks.
5657 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
5659 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
5661 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5663 PR target/83399
5664 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
5665 VECTOR_MEM_ALTIVEC_OR_VSX_P.
5666 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
5667 indexed_or_indirect_operand predicate.
5668 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
5669 (*vsx_le_perm_load_v8hi): Likewise.
5670 (*vsx_le_perm_load_v16qi): Likewise.
5671 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
5672 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
5673 (*vsx_le_perm_store_v8hi): Likewise.
5674 (*vsx_le_perm_store_v16qi): Likewise.
5675 (eight unnamed splitters): Likewise.
5677 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5679 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
5680 * config/rs6000/emmintrin.h: Likewise.
5681 * config/rs6000/mmintrin.h: Likewise.
5682 * config/rs6000/xmmintrin.h: Likewise.
5684 2018-01-10 David Malcolm <dmalcolm@redhat.com>
5686 PR c++/43486
5687 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
5688 "public_flag".
5689 * tree.c (tree_nop_conversion): Return true for location wrapper
5690 nodes.
5691 (maybe_wrap_with_location): New function.
5692 (selftest::check_strip_nops): New function.
5693 (selftest::test_location_wrappers): New function.
5694 (selftest::tree_c_tests): Call it.
5695 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
5696 (maybe_wrap_with_location): New decl.
5697 (EXPR_LOCATION_WRAPPER_P): New macro.
5698 (location_wrapper_p): New inline function.
5699 (tree_strip_any_location_wrapper): New inline function.
5701 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
5703 PR target/83735
5704 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
5705 stack_realign_offset for the largest alignment of stack slot
5706 actually used.
5707 (ix86_find_max_used_stack_alignment): New function.
5708 (ix86_finalize_stack_frame_flags): Use it. Set
5709 max_used_stack_alignment if we don't realign stack.
5710 * config/i386/i386.h (machine_function): Add
5711 max_used_stack_alignment.
5713 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
5715 * config/arm/arm.opt (-mbranch-cost): New option.
5716 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
5717 account.
5719 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
5721 PR target/83629
5722 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
5723 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
5725 2018-01-10 Richard Biener <rguenther@suse.de>
5727 PR debug/83765
5728 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
5729 early out so it also covers the case where we have a non-NULL
5730 origin.
5732 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
5734 PR tree-optimization/83753
5735 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
5736 for non-strided grouped accesses if the number of elements is 1.
5738 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5740 PR target/81616
5741 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
5742 * i386.h (TARGET_USE_GATHER): Define.
5743 * x86-tune.def (X86_TUNE_USE_GATHER): New.
5745 2018-01-10 Martin Liska <mliska@suse.cz>
5747 PR bootstrap/82831
5748 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
5749 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
5750 partitioning.
5751 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
5752 CLEANUP_NO_PARTITIONING is not set.
5754 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
5756 * doc/rtl.texi: Remove documentation of (const ...) wrappers
5757 for vectors, as a partial revert of r254296.
5758 * rtl.h (const_vec_p): Delete.
5759 (const_vec_duplicate_p): Don't test for vector CONSTs.
5760 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
5761 * expmed.c (make_tree): Likewise.
5763 Revert:
5764 * common.md (E, F): Use CONSTANT_P instead of checking for
5765 CONST_VECTOR.
5766 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
5767 checking for CONST_VECTOR.
5769 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5771 PR middle-end/83575
5772 * predict.c (force_edge_cold): Handle in more sane way edges
5773 with no prediction.
5775 2018-01-09 Carl Love <cel@us.ibm.com>
5777 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
5778 V4SI, V4SF types.
5779 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
5780 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
5781 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
5782 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
5783 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
5784 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
5785 * config/rs6000/rs6000-protos.h: Add extern defition for
5786 rs6000_generate_float2_double_code.
5787 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
5788 function.
5789 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
5790 (float2_v2df): Add define_expand.
5792 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
5794 PR target/83628
5795 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
5796 op_mode in the force_to_mode call.
5798 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5800 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
5801 instead of checking each element individually.
5802 (aarch64_evpc_uzp): Likewise.
5803 (aarch64_evpc_zip): Likewise.
5804 (aarch64_evpc_ext): Likewise.
5805 (aarch64_evpc_rev): Likewise.
5806 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
5807 instead of checking each element individually. Return true without
5808 generating rtl if
5809 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
5810 whether all selected elements come from the same input, instead of
5811 checking each element individually. Remove calls to gen_rtx_REG,
5812 start_sequence and end_sequence and instead assert that no rtl is
5813 generated.
5815 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5817 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
5818 order of HIGH and CONST checks.
5820 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5822 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
5823 if the destination isn't an SSA_NAME.
5825 2018-01-09 Richard Biener <rguenther@suse.de>
5827 PR tree-optimization/83668
5828 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
5829 move prologue...
5830 (canonicalize_loop_form): ... here, renamed from ...
5831 (canonicalize_loop_closed_ssa_form): ... this and amended to
5832 swap successor edges for loop exit blocks to make us use
5833 the RPO order we need for initial schedule generation.
5835 2018-01-09 Joseph Myers <joseph@codesourcery.com>
5837 PR tree-optimization/64811
5838 * match.pd: When optimizing comparisons with Inf, avoid
5839 introducing or losing exceptions from comparisons with NaN.
5841 2018-01-09 Martin Liska <mliska@suse.cz>
5843 PR sanitizer/82517
5844 * asan.c (shadow_mem_size): Add gcc_assert.
5846 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
5848 Don't save registers in main().
5850 PR target/83738
5851 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
5852 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
5853 * config/avr/avr.c (avr_set_current_function): Don't error if
5854 naked, OS_task or OS_main are specified at the same time.
5855 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
5856 OS_main.
5857 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
5858 attribute.
5859 * common/config/avr/avr-common.c (avr_option_optimization_table):
5860 Switch on -mmain-is-OS_task for optimizing compilations.
5862 2018-01-09 Richard Biener <rguenther@suse.de>
5864 PR tree-optimization/83572
5865 * graphite.c: Include cfganal.h.
5866 (graphite_transform_loops): Connect infinite loops to exit
5867 and remove fake edges at the end.
5869 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5871 * ipa-inline.c (edge_badness): Revert accidental checkin.
5873 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5875 PR ipa/80763
5876 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
5877 symbols; not inline clones.
5879 2018-01-09 Jakub Jelinek <jakub@redhat.com>
5881 PR target/83507
5882 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
5883 hard registers. Formatting fixes.
5885 PR preprocessor/83722
5886 * gcc.c (try_generate_repro): Pass
5887 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
5888 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
5889 do_report_bug.
5891 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
5892 Kito Cheng <kito.cheng@gmail.com>
5894 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
5895 (riscv_leaf_function_p): Delete.
5896 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
5898 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5900 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
5901 function.
5902 (do_ifelse): New function.
5903 (do_isel): New function.
5904 (do_sub3): New function.
5905 (do_add3): New function.
5906 (do_load_mask_compare): New function.
5907 (do_overlap_load_compare): New function.
5908 (expand_compare_loop): New function.
5909 (expand_block_compare): Call expand_compare_loop() when appropriate.
5910 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
5911 option description.
5912 (-mblock-compare-inline-loop-limit): New option.
5914 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5916 PR target/83677
5917 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
5918 Reverse order of second and third operands in first alternative.
5919 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
5920 of first and second elements in UNSPEC_VPERMR vector.
5921 (altivec_expand_vec_perm_le): Likewise.
5923 2018-01-08 Jeff Law <law@redhat.com>
5925 PR rtl-optimizatin/81308
5926 * tree-switch-conversion.c (cfg_altered): New file scoped static.
5927 (process_switch): If group_case_labels makes a change, then set
5928 cfg_altered.
5929 (pass_convert_switch::execute): If a switch is converted, then
5930 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
5932 PR rtl-optimization/81308
5933 * recog.c (split_all_insns): Conditionally cleanup the CFG after
5934 splitting insns.
5936 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
5938 PR target/83663 - Revert r255946
5939 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
5940 generation for cases where splatting a value is not useful.
5941 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
5942 across a vec_duplicate and a paradoxical subreg forming a vector
5943 mode to a vec_concat.
5945 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5947 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
5948 -march=armv8.3-a variants.
5949 * config/arm/t-multilib: Likewise.
5950 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
5952 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5954 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
5955 to generate rtl.
5956 (cceq_ior_compare_complement): Give it a name so I can use it, and
5957 change boolean_or_operator predicate to boolean_operator so it can
5958 be used to generate a crand.
5959 (eqne): New code iterator.
5960 (bd/bd_neg): New code_attrs.
5961 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
5962 a single define_insn.
5963 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
5964 decrement (bdnzt/bdnzf/bdzt/bdzf).
5965 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
5966 with the new names of the branch decrement patterns, and added the
5967 names of the branch decrement conditional patterns.
5969 2018-01-08 Richard Biener <rguenther@suse.de>
5971 PR tree-optimization/83563
5972 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
5973 cache.
5975 2018-01-08 Richard Biener <rguenther@suse.de>
5977 PR middle-end/83713
5978 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
5980 2018-01-08 Richard Biener <rguenther@suse.de>
5982 PR tree-optimization/83685
5983 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
5984 references to abnormals.
5986 2018-01-08 Richard Biener <rguenther@suse.de>
5988 PR lto/83719
5989 * dwarf2out.c (output_indirect_strings): Handle empty
5990 skeleton_debug_str_hash.
5991 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
5993 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
5995 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
5996 (emit_store_direct): Likewise.
5997 (arc_trampoline_adjust_address): Likewise.
5998 (arc_asm_trampoline_template): New function.
5999 (arc_initialize_trampoline): Use asm_trampoline_template.
6000 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
6001 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
6002 * config/arc/arc.md (flush_icache): Delete pattern.
6004 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6006 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
6007 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
6008 munaligned-access.
6010 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6012 PR target/83681
6013 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
6014 by not USED_FOR_TARGET.
6015 (make_pass_resolve_sw_modes): Likewise.
6017 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6019 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
6020 USED_FOR_TARGET.
6022 2018-01-08 Richard Biener <rguenther@suse.de>
6024 PR middle-end/83580
6025 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
6027 2018-01-08 Richard Biener <rguenther@suse.de>
6029 PR middle-end/83517
6030 * match.pd ((t * 2) / 2) -> t): Add missing :c.
6032 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
6034 PR middle-end/81897
6035 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
6036 basic blocks with a small number of successors.
6037 (convert_control_dep_chain_into_preds): Improve handling of
6038 forwarder blocks.
6039 (dump_predicates): Split apart into...
6040 (dump_pred_chain): ...here...
6041 (dump_pred_info): ...and here.
6042 (can_one_predicate_be_invalidated_p): Add debugging printfs.
6043 (can_chain_union_be_invalidated_p): Improve check for invalidation
6044 of paths.
6045 (uninit_uses_cannot_happen): Avoid unnecessary if
6046 convert_control_dep_chain_into_preds yielded nothing.
6048 2018-01-06 Martin Sebor <msebor@redhat.com>
6050 PR tree-optimization/83640
6051 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
6052 subtracting negative offset from size.
6053 (builtin_access::overlap): Adjust offset bounds of the access to fall
6054 within the size of the object if possible.
6056 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
6058 PR rtl-optimization/83699
6059 * expmed.c (extract_bit_field_1): Restrict the vector usage of
6060 extract_bit_field_as_subreg to cases in which the extracted
6061 value is also a vector.
6063 * lra-constraints.c (process_alt_operands): Test for the equivalence
6064 substitutions when detecting a possible reload cycle.
6066 2018-01-06 Jakub Jelinek <jakub@redhat.com>
6068 PR debug/83480
6069 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
6070 by default if flag_selective_schedling{,2}. Formatting fixes.
6072 PR rtl-optimization/83682
6073 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
6074 if it has non-VECTOR_MODE element mode.
6075 (vec_duplicate_p): Likewise.
6077 PR middle-end/83694
6078 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
6079 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
6081 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6083 PR target/83604
6084 * config/i386/i386-builtin.def
6085 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
6086 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
6087 Require also OPTION_MASK_ISA_AVX512F in addition to
6088 OPTION_MASK_ISA_GFNI.
6089 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
6090 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
6091 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
6092 to OPTION_MASK_ISA_GFNI.
6093 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6094 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6095 OPTION_MASK_ISA_AVX512BW.
6096 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6097 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6098 addition to OPTION_MASK_ISA_GFNI.
6099 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6100 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6101 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6102 to OPTION_MASK_ISA_GFNI.
6103 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6104 a requirement for all ISAs rather than any of them with a few
6105 exceptions.
6106 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6107 processing.
6108 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6109 bitmasks to be enabled with 3 exceptions, instead of requiring any
6110 enabled ISA with lots of exceptions.
6111 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6112 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6113 Change avx512bw in isa attribute to avx512f.
6114 * config/i386/sgxintrin.h: Add license boilerplate.
6115 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6116 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6117 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6118 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6119 defined.
6120 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6121 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6122 temporarily sse2 rather than sse if not enabled already.
6124 PR target/83604
6125 * config/i386/sse.md (VI248_VLBW): Rename to ...
6126 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
6127 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
6128 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
6129 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
6130 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
6131 mode iterator instead of VI248_VLBW.
6133 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
6135 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
6136 (record_modified): Skip clobbers; add debug output.
6137 (param_change_prob): Use sreal frequencies.
6139 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6141 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
6142 punt for user-aligned variables.
6144 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6146 * tree-chrec.c (chrec_contains_symbols): Return true for
6147 POLY_INT_CST.
6149 2018-01-05 Sudakshina Das <sudi.das@arm.com>
6151 PR target/82439
6152 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
6153 of (x|y) == x for BICS pattern.
6155 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6157 PR tree-optimization/83605
6158 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
6159 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
6160 can throw.
6162 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
6164 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
6165 * config/epiphany/rtems.h: New file.
6167 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6168 Uros Bizjak <ubizjak@gmail.com>
6170 PR target/83554
6171 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
6172 QIreg_operand instead of register_operand predicate.
6173 * config/i386/i386.c (ix86_rop_should_change_byte_p,
6174 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
6175 comments instead of -fmitigate[-_]rop.
6177 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6179 PR bootstrap/81926
6180 * cgraphunit.c (symbol_table::compile): Switch to text_section
6181 before calling assembly_start debug hook.
6182 * run-rtl-passes.c (run_rtl_passes): Likewise.
6183 Include output.h.
6185 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6187 * tree-vrp.c (extract_range_from_binary_expr_1): Check
6188 range_int_cst_p rather than !symbolic_range_p before calling
6189 extract_range_from_multiplicative_op_1.
6191 2018-01-04 Jeff Law <law@redhat.com>
6193 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
6194 redundant test in assertion.
6196 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6198 * doc/rtl.texi: Document machine_mode wrapper classes.
6200 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6202 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
6203 using tree_to_uhwi.
6205 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6207 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
6208 the VEC_PERM_EXPR fold to fail.
6210 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6212 PR debug/83585
6213 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
6214 to switched_sections.
6216 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6218 PR target/83680
6219 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
6220 test for d.testing.
6222 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
6224 PR target/83387
6225 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
6226 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
6228 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6230 PR debug/83666
6231 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
6232 is BLKmode and bitpos not zero or mode change is needed.
6234 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6236 PR target/83675
6237 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
6238 TARGET_VIS2.
6240 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
6242 PR target/83628
6243 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
6244 instead of MULT rtx. Update all corresponding splitters.
6245 (*saddl_se): Ditto.
6246 (*ssub<modesuffix>): Ditto.
6247 (*ssubl_se): Ditto.
6248 (*cmp_sadd_di): Update split patterns.
6249 (*cmp_sadd_si): Ditto.
6250 (*cmp_sadd_sidi): Ditto.
6251 (*cmp_ssub_di): Ditto.
6252 (*cmp_ssub_si): Ditto.
6253 (*cmp_ssub_sidi): Ditto.
6254 * config/alpha/predicates.md (const23_operand): New predicate.
6255 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
6256 Look for ASHIFT, not MULT inner operand.
6257 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
6259 2018-01-04 Martin Liska <mliska@suse.cz>
6261 PR gcov-profile/83669
6262 * gcov.c (output_intermediate_file): Add version to intermediate
6263 gcov file.
6264 * doc/gcov.texi: Document new field 'version' in intermediate
6265 file format. Fix location of '-k' option of gcov command.
6267 2018-01-04 Martin Liska <mliska@suse.cz>
6269 PR ipa/82352
6270 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
6272 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6274 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
6276 2018-01-03 Martin Sebor <msebor@redhat.com>
6278 PR tree-optimization/83655
6279 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
6280 checking calls with invalid arguments.
6282 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6284 * tree-vect-stmts.c (vect_get_store_rhs): New function.
6285 (vectorizable_mask_load_store): Delete.
6286 (vectorizable_call): Return false for masked loads and stores.
6287 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
6288 instead of gimple_assign_rhs1.
6289 (vectorizable_load): Handle IFN_MASK_LOAD.
6290 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
6292 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6294 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
6295 split out from..,
6296 (vectorizable_mask_load_store): ...here.
6297 (vectorizable_load): ...and here.
6299 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6301 * tree-vect-stmts.c (vect_build_all_ones_mask)
6302 (vect_build_zero_merge_argument): New functions, split out from...
6303 (vectorizable_load): ...here.
6305 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6307 * tree-vect-stmts.c (vect_check_store_rhs): New function,
6308 split out from...
6309 (vectorizable_mask_load_store): ...here.
6310 (vectorizable_store): ...and here.
6312 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6314 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
6315 split out from...
6316 (vectorizable_mask_load_store): ...here.
6318 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6320 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
6321 (vect_model_store_cost): Take a vec_load_store_type instead of a
6322 vect_def_type.
6323 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
6324 (vect_model_store_cost): Take a vec_load_store_type instead of a
6325 vect_def_type.
6326 (vectorizable_mask_load_store): Update accordingly.
6327 (vectorizable_store): Likewise.
6328 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
6330 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6332 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
6333 IFN_MASK_LOAD calls here rather than...
6334 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
6336 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6337 Alan Hayward <alan.hayward@arm.com>
6338 David Sherwood <david.sherwood@arm.com>
6340 * expmed.c (extract_bit_field_1): For vector extracts,
6341 fall back to extract_bit_field_as_subreg if vec_extract
6342 isn't available.
6344 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6345 Alan Hayward <alan.hayward@arm.com>
6346 David Sherwood <david.sherwood@arm.com>
6348 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
6349 they are variable or constant sized.
6350 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
6351 slots for constant-sized data.
6353 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6354 Alan Hayward <alan.hayward@arm.com>
6355 David Sherwood <david.sherwood@arm.com>
6357 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
6358 handling COND_EXPRs with boolean comparisons, try to find a better
6359 basis for the mask type than the boolean itself.
6361 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6363 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
6364 is calculated and how it can be overridden.
6365 * genmodes.c (max_bitsize_mode_any_mode): New variable.
6366 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
6367 if defined.
6368 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
6369 if nonzero.
6371 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6372 Alan Hayward <alan.hayward@arm.com>
6373 David Sherwood <david.sherwood@arm.com>
6375 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
6376 Remove the mode argument.
6377 (aarch64_simd_valid_immediate): Remove the mode and inverse
6378 arguments.
6379 * config/aarch64/iterators.md (bitsize): New iterator.
6380 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
6381 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
6382 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
6383 aarch64_simd_valid_immediate.
6384 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
6385 (aarch64_reg_or_bic_imm): Likewise.
6386 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
6387 with an insn_type enum and msl with a modifier_type enum.
6388 Replace element_width with a scalar_mode. Change the shift
6389 to unsigned int. Add constructors for scalar_float_mode and
6390 scalar_int_mode elements.
6391 (aarch64_vect_float_const_representable_p): Delete.
6392 (aarch64_can_const_movi_rtx_p)
6393 (aarch64_simd_scalar_immediate_valid_for_move)
6394 (aarch64_simd_make_constant): Update call to
6395 aarch64_simd_valid_immediate.
6396 (aarch64_advsimd_valid_immediate_hs): New function.
6397 (aarch64_advsimd_valid_immediate): Likewise.
6398 (aarch64_simd_valid_immediate): Remove mode and inverse
6399 arguments. Rewrite to use the above. Use const_vec_duplicate_p
6400 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
6401 and aarch64_float_const_representable_p on the result.
6402 (aarch64_output_simd_mov_immediate): Remove mode argument.
6403 Update call to aarch64_simd_valid_immediate and use of
6404 simd_immediate_info.
6405 (aarch64_output_scalar_simd_mov_immediate): Update call
6406 accordingly.
6408 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6409 Alan Hayward <alan.hayward@arm.com>
6410 David Sherwood <david.sherwood@arm.com>
6412 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
6413 (mode_nunits): Likewise CONST_MODE_NUNITS.
6414 * machmode.def (ADJUST_NUNITS): Document.
6415 * genmodes.c (mode_data::need_nunits_adj): New field.
6416 (blank_mode): Update accordingly.
6417 (adj_nunits): New variable.
6418 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
6419 parameter.
6420 (emit_mode_size_inline): Set need_bytesize_adj for all modes
6421 listed in adj_nunits.
6422 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
6423 listed in adj_nunits. Don't emit case statements for such modes.
6424 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
6425 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
6426 nothing if adj_nunits is nonnull.
6427 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
6428 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
6429 (emit_mode_fbit): Update use of print_maybe_const_decl.
6430 (emit_move_size): Likewise. Treat the array as non-const
6431 if adj_nunits.
6432 (emit_mode_adjustments): Handle adj_nunits.
6434 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6436 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
6437 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
6438 (VECTOR_MODES): Use it.
6439 (make_vector_modes): Take the prefix as an argument.
6441 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6442 Alan Hayward <alan.hayward@arm.com>
6443 David Sherwood <david.sherwood@arm.com>
6445 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
6446 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
6447 for MODE_VECTOR_BOOL.
6448 * machmode.def (VECTOR_BOOL_MODE): Document.
6449 * genmodes.c (VECTOR_BOOL_MODE): New macro.
6450 (make_vector_bool_mode): New function.
6451 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
6452 MODE_VECTOR_BOOL.
6453 * lto-streamer-in.c (lto_input_mode_table): Likewise.
6454 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
6455 Likewise.
6456 * stor-layout.c (int_mode_for_mode): Likewise.
6457 * tree.c (build_vector_type_for_mode): Likewise.
6458 * varasm.c (output_constant_pool_2): Likewise.
6459 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
6460 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
6461 for MODE_VECTOR_BOOL.
6462 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
6463 of mode class checks.
6464 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
6465 instead of a list of mode class checks.
6466 (expand_vector_scalar_condition): Likewise.
6467 (type_for_widest_vector_mode): Handle BImode as an inner mode.
6469 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6470 Alan Hayward <alan.hayward@arm.com>
6471 David Sherwood <david.sherwood@arm.com>
6473 * machmode.h (mode_size): Change from unsigned short to
6474 poly_uint16_pod.
6475 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
6476 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6477 or if measurement_type is not polynomial.
6478 (fixed_size_mode::includes_p): Check for constant-sized modes.
6479 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
6480 return a poly_uint16 rather than an unsigned short.
6481 (emit_mode_size): Change the type of mode_size from unsigned short
6482 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
6483 (emit_mode_adjustments): Cope with polynomial vector sizes.
6484 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6485 for GET_MODE_SIZE.
6486 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6487 for GET_MODE_SIZE.
6488 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
6489 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
6490 * caller-save.c (setup_save_areas): Likewise.
6491 (replace_reg_with_saved_mem): Likewise.
6492 * calls.c (emit_library_call_value_1): Likewise.
6493 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
6494 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
6495 (gen_lowpart_for_combine): Likewise.
6496 * convert.c (convert_to_integer_1): Likewise.
6497 * cse.c (equiv_constant, cse_insn): Likewise.
6498 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
6499 (cselib_subst_to_values): Likewise.
6500 * dce.c (word_dce_process_block): Likewise.
6501 * df-problems.c (df_word_lr_mark_ref): Likewise.
6502 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
6503 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
6504 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
6505 (rtl_for_decl_location): Likewise.
6506 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
6507 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
6508 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
6509 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
6510 (expand_expr_real_1): Likewise.
6511 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
6512 (pad_below): Likewise.
6513 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6514 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
6515 * ira.c (get_subreg_tracking_sizes): Likewise.
6516 * ira-build.c (ira_create_allocno_objects): Likewise.
6517 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
6518 (ira_sort_regnos_for_alter_reg): Likewise.
6519 * ira-costs.c (record_operand_costs): Likewise.
6520 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
6521 (resolve_simple_move): Likewise.
6522 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
6523 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
6524 (lra_constraints): Likewise.
6525 (CONST_POOL_OK_P): Reject variable-sized modes.
6526 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
6527 (add_pseudo_to_slot, lra_spill): Likewise.
6528 * omp-low.c (omp_clause_aligned_alignment): Likewise.
6529 * optabs-query.c (get_best_extraction_insn): Likewise.
6530 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6531 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
6532 (expand_mult_highpart, valid_multiword_target_p): Likewise.
6533 * recog.c (offsettable_address_addr_space_p): Likewise.
6534 * regcprop.c (maybe_mode_change): Likewise.
6535 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
6536 * regrename.c (build_def_use): Likewise.
6537 * regstat.c (dump_reg_info): Likewise.
6538 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
6539 (find_reloads, find_reloads_subreg_address): Likewise.
6540 * reload1.c (eliminate_regs_1): Likewise.
6541 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
6542 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
6543 (simplify_binary_operation_1, simplify_subreg): Likewise.
6544 * targhooks.c (default_function_arg_padding): Likewise.
6545 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
6546 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
6547 (verify_gimple_assign_ternary): Likewise.
6548 * tree-inline.c (estimate_move_cost): Likewise.
6549 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6550 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
6551 (get_address_cost_ainc): Likewise.
6552 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
6553 (vect_supportable_dr_alignment): Likewise.
6554 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6555 (vectorizable_reduction): Likewise.
6556 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
6557 (vectorizable_operation, vectorizable_load): Likewise.
6558 * tree.c (build_same_sized_truth_vector_type): Likewise.
6559 * valtrack.c (cleanup_auto_inc_dec): Likewise.
6560 * var-tracking.c (emit_note_insn_var_location): Likewise.
6561 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
6562 (ADDR_VEC_ALIGN): Likewise.
6564 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6565 Alan Hayward <alan.hayward@arm.com>
6566 David Sherwood <david.sherwood@arm.com>
6568 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
6569 unsigned short.
6570 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6571 or if measurement_type is polynomial.
6572 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
6573 * combine.c (make_extraction): Likewise.
6574 * dse.c (find_shift_sequence): Likewise.
6575 * dwarf2out.c (mem_loc_descriptor): Likewise.
6576 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
6577 (extract_bit_field, extract_low_bits): Likewise.
6578 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
6579 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
6580 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
6581 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
6582 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6583 * reload.c (find_reloads): Likewise.
6584 * reload1.c (alter_reg): Likewise.
6585 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
6586 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
6587 * tree-if-conv.c (predicate_mem_writes): Likewise.
6588 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
6589 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
6590 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
6591 * valtrack.c (dead_debug_insert_temp): Likewise.
6592 * varasm.c (mergeable_constant_section): Likewise.
6593 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
6595 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6596 Alan Hayward <alan.hayward@arm.com>
6597 David Sherwood <david.sherwood@arm.com>
6599 * expr.c (expand_assignment): Cope with polynomial mode sizes
6600 when assigning to a CONCAT.
6602 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6603 Alan Hayward <alan.hayward@arm.com>
6604 David Sherwood <david.sherwood@arm.com>
6606 * machmode.h (mode_precision): Change from unsigned short to
6607 poly_uint16_pod.
6608 (mode_to_precision): Return a poly_uint16 rather than an unsigned
6609 short.
6610 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
6611 or if measurement_type is not polynomial.
6612 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
6613 in which the mode is already known to be a scalar_int_mode.
6614 * genmodes.c (emit_mode_precision): Change the type of mode_precision
6615 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
6616 initializer.
6617 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6618 for GET_MODE_PRECISION.
6619 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6620 for GET_MODE_PRECISION.
6621 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
6622 as polynomial.
6623 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
6624 (expand_field_assignment, make_extraction): Likewise.
6625 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
6626 (get_last_value): Likewise.
6627 * convert.c (convert_to_integer_1): Likewise.
6628 * cse.c (cse_insn): Likewise.
6629 * expr.c (expand_expr_real_1): Likewise.
6630 * lra-constraints.c (simplify_operand_subreg): Likewise.
6631 * optabs-query.c (can_atomic_load_p): Likewise.
6632 * optabs.c (expand_atomic_load): Likewise.
6633 (expand_atomic_store): Likewise.
6634 * ree.c (combine_reaching_defs): Likewise.
6635 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
6636 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
6637 * tree.h (type_has_mode_precision_p): Likewise.
6638 * ubsan.c (instrument_si_overflow): Likewise.
6640 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6641 Alan Hayward <alan.hayward@arm.com>
6642 David Sherwood <david.sherwood@arm.com>
6644 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
6645 polynomial numbers of units.
6646 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
6647 (valid_vector_subparts_p): New function.
6648 (build_vector_type): Remove temporary shim and take the number
6649 of units as a poly_uint64 rather than an int.
6650 (build_opaque_vector_type): Take the number of units as a
6651 poly_uint64 rather than an int.
6652 * tree.c (build_vector_from_ctor): Handle polynomial
6653 TYPE_VECTOR_SUBPARTS.
6654 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
6655 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
6656 (build_vector_from_val): If the number of units is variable,
6657 use build_vec_duplicate_cst for constant operands and
6658 VEC_DUPLICATE_EXPR otherwise.
6659 (make_vector_type): Remove temporary is_constant ().
6660 (build_vector_type, build_opaque_vector_type): Take the number of
6661 units as a poly_uint64 rather than an int.
6662 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
6663 VECTOR_CST_NELTS.
6664 * cfgexpand.c (expand_debug_expr): Likewise.
6665 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
6666 (store_constructor, expand_expr_real_1): Likewise.
6667 (const_scalar_mask_from_tree): Likewise.
6668 * fold-const-call.c (fold_const_reduction): Likewise.
6669 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
6670 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
6671 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
6672 (fold_relational_const): Likewise.
6673 (native_interpret_vector): Likewise. Change the size from an
6674 int to an unsigned int.
6675 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
6676 TYPE_VECTOR_SUBPARTS.
6677 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
6678 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
6679 duplicating a non-constant operand into a variable-length vector.
6680 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
6681 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
6682 * ipa-icf.c (sem_variable::equals): Likewise.
6683 * match.pd: Likewise.
6684 * omp-simd-clone.c (simd_clone_subparts): Likewise.
6685 * print-tree.c (print_node): Likewise.
6686 * stor-layout.c (layout_type): Likewise.
6687 * targhooks.c (default_builtin_vectorization_cost): Likewise.
6688 * tree-cfg.c (verify_gimple_comparison): Likewise.
6689 (verify_gimple_assign_binary): Likewise.
6690 (verify_gimple_assign_ternary): Likewise.
6691 (verify_gimple_assign_single): Likewise.
6692 * tree-pretty-print.c (dump_generic_node): Likewise.
6693 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6694 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
6695 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
6696 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
6697 (vect_shift_permute_load_chain): Likewise.
6698 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
6699 (expand_vector_condition, optimize_vector_constructor): Likewise.
6700 (lower_vec_perm, get_compute_type): Likewise.
6701 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6702 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
6703 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
6704 (vect_recog_mask_conversion_pattern): Likewise.
6705 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
6706 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
6707 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
6708 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
6709 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
6710 (vectorizable_shift, vectorizable_operation, vectorizable_store)
6711 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
6712 (supportable_widening_operation): Likewise.
6713 (supportable_narrowing_operation): Likewise.
6714 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
6715 Likewise.
6716 * varasm.c (output_constant): Likewise.
6718 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6719 Alan Hayward <alan.hayward@arm.com>
6720 David Sherwood <david.sherwood@arm.com>
6722 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
6723 so that both the length == 3 and length != 3 cases set up their
6724 own permute vectors. Add comments explaining why we know the
6725 number of elements is constant.
6726 (vect_permute_load_chain): Likewise.
6728 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6729 Alan Hayward <alan.hayward@arm.com>
6730 David Sherwood <david.sherwood@arm.com>
6732 * machmode.h (mode_nunits): Change from unsigned char to
6733 poly_uint16_pod.
6734 (ONLY_FIXED_SIZE_MODES): New macro.
6735 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
6736 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
6737 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
6738 New typedefs.
6739 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
6740 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
6741 or if measurement_type is not polynomial.
6742 * genmodes.c (ZERO_COEFFS): New macro.
6743 (emit_mode_nunits_inline): Make mode_nunits_inline return a
6744 poly_uint16.
6745 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
6746 Use ZERO_COEFFS when emitting initializers.
6747 * data-streamer.h (bp_pack_poly_value): New function.
6748 (bp_unpack_poly_value): Likewise.
6749 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6750 for GET_MODE_NUNITS.
6751 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6752 for GET_MODE_NUNITS.
6753 * tree.c (make_vector_type): Remove temporary shim and make
6754 the real function take the number of units as a poly_uint64
6755 rather than an int.
6756 (build_vector_type_for_mode): Handle polynomial nunits.
6757 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
6758 * emit-rtl.c (const_vec_series_p_1): Likewise.
6759 (gen_rtx_CONST_VECTOR): Likewise.
6760 * fold-const.c (test_vec_duplicate_folding): Likewise.
6761 * genrecog.c (validate_pattern): Likewise.
6762 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
6763 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6764 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
6765 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
6766 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
6767 * rtlanal.c (subreg_get_info): Likewise.
6768 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
6769 (vect_grouped_load_supported): Likewise.
6770 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
6771 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
6772 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
6773 (simplify_const_unary_operation, simplify_binary_operation_1)
6774 (simplify_const_binary_operation, simplify_ternary_operation)
6775 (test_vector_ops_duplicate, test_vector_ops): Likewise.
6776 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
6777 instead of CONST_VECTOR_NUNITS.
6778 * varasm.c (output_constant_pool_2): Likewise.
6779 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
6780 explicit-encoded elements in the XVEC for variable-length vectors.
6782 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6784 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
6786 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6787 Alan Hayward <alan.hayward@arm.com>
6788 David Sherwood <david.sherwood@arm.com>
6790 * coretypes.h (fixed_size_mode): Declare.
6791 (fixed_size_mode_pod): New typedef.
6792 * builtins.h (target_builtins::x_apply_args_mode)
6793 (target_builtins::x_apply_result_mode): Change type to
6794 fixed_size_mode_pod.
6795 * builtins.c (apply_args_size, apply_result_size, result_vector)
6796 (expand_builtin_apply_args_1, expand_builtin_apply)
6797 (expand_builtin_return): Update accordingly.
6799 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6801 * cse.c (hash_rtx_cb): Hash only the encoded elements.
6802 * cselib.c (cselib_hash_rtx): Likewise.
6803 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
6804 CONST_VECTOR encoding.
6806 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6807 Jeff Law <law@redhat.com>
6809 PR target/83641
6810 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
6811 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
6812 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
6813 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
6815 PR target/83641
6816 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
6817 explicitly probe *sp in a noreturn function if there were any callee
6818 register saves or frame pointer is needed.
6820 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6822 PR debug/83621
6823 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
6824 BLKmode for ternary, binary or unary expressions.
6826 PR debug/83645
6827 * var-tracking.c (delete_vta_debug_insn): New inline function.
6828 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
6829 insns from get_insns () to NULL instead of each bb separately.
6830 Use delete_vta_debug_insn. No longer static.
6831 (vt_debug_insns_local, variable_tracking_main_1): Adjust
6832 delete_vta_debug_insns callers.
6833 * rtl.h (delete_vta_debug_insns): Declare.
6834 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
6835 instead of variable_tracking_main.
6837 2018-01-03 Martin Sebor <msebor@redhat.com>
6839 PR tree-optimization/83603
6840 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
6841 arguments past the endof the argument list in functions declared
6842 without a prototype.
6843 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
6844 Avoid checking when arguments are null.
6846 2018-01-03 Martin Sebor <msebor@redhat.com>
6848 PR c/83559
6849 * doc/extend.texi (attribute const): Fix a typo.
6850 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
6851 issuing -Wsuggest-attribute for void functions.
6853 2018-01-03 Martin Sebor <msebor@redhat.com>
6855 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
6856 offset_int::from instead of wide_int::to_shwi.
6857 (maybe_diag_overlap): Remove assertion.
6858 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
6859 * gimple-ssa-sprintf.c (format_directive): Same.
6860 (parse_directive): Same.
6861 (sprintf_dom_walker::compute_format_length): Same.
6862 (try_substitute_return_value): Same.
6864 2018-01-03 Jeff Law <law@redhat.com>
6866 PR middle-end/83654
6867 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
6868 non-constant residual for zero at runtime and avoid probing in
6869 that case. Reorganize code for trailing problem to mirror handling
6870 of the residual.
6872 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6874 PR tree-optimization/83501
6875 * tree-ssa-strlen.c (get_string_cst): New.
6876 (handle_char_store): Call get_string_cst.
6878 2018-01-03 Martin Liska <mliska@suse.cz>
6880 PR tree-optimization/83593
6881 * tree-ssa-strlen.c: Include tree-cfg.h.
6882 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
6883 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
6884 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
6885 to false.
6886 (strlen_dom_walker::before_dom_children): Call
6887 gimple_purge_dead_eh_edges. Dump tranformation with details
6888 dump flags.
6889 (strlen_dom_walker::before_dom_children): Update call by adding
6890 new argument cleanup_eh.
6891 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
6893 2018-01-03 Martin Liska <mliska@suse.cz>
6895 PR ipa/83549
6896 * cif-code.def (VARIADIC_THUNK): New enum value.
6897 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
6898 thunks.
6900 2018-01-03 Jan Beulich <jbeulich@suse.com>
6902 * sse.md (mov<mode>_internal): Tighten condition for when to use
6903 vmovdqu<ssescalarsize> for TI and OI modes.
6905 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6907 Update copyright years.
6909 2018-01-03 Martin Liska <mliska@suse.cz>
6911 PR ipa/83594
6912 * ipa-visibility.c (function_and_variable_visibility): Skip
6913 functions with noipa attribure.
6915 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6917 * gcc.c (process_command): Update copyright notice dates.
6918 * gcov-dump.c (print_version): Ditto.
6919 * gcov.c (print_version): Ditto.
6920 * gcov-tool.c (print_version): Ditto.
6921 * gengtype.c (create_file): Ditto.
6922 * doc/cpp.texi: Bump @copying's copyright year.
6923 * doc/cppinternals.texi: Ditto.
6924 * doc/gcc.texi: Ditto.
6925 * doc/gccint.texi: Ditto.
6926 * doc/gcov.texi: Ditto.
6927 * doc/install.texi: Ditto.
6928 * doc/invoke.texi: Ditto.
6930 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6932 * vector-builder.h (vector_builder::m_full_nelts): Change from
6933 unsigned int to poly_uint64.
6934 (vector_builder::full_nelts): Update prototype accordingly.
6935 (vector_builder::new_vector): Likewise.
6936 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
6937 (vector_builder::operator ==): Likewise.
6938 (vector_builder::finalize): Likewise.
6939 * int-vector-builder.h (int_vector_builder::int_vector_builder):
6940 Take the number of elements as a poly_uint64 rather than an
6941 unsigned int.
6942 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
6943 from unsigned int to poly_uint64.
6944 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
6945 (vec_perm_indices::new_vector): Likewise.
6946 (vec_perm_indices::length): Likewise.
6947 (vec_perm_indices::nelts_per_input): Likewise.
6948 (vec_perm_indices::input_nelts): Likewise.
6949 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
6950 number of elements per input as a poly_uint64 rather than an
6951 unsigned int. Use the original encoding for variable-length
6952 vectors, rather than clamping each individual element.
6953 For the second and subsequent elements in each pattern,
6954 clamp the step and base before clamping their sum.
6955 (vec_perm_indices::series_p): Handle polynomial element counts.
6956 (vec_perm_indices::all_in_range_p): Likewise.
6957 (vec_perm_indices_to_tree): Likewise.
6958 (vec_perm_indices_to_rtx): Likewise.
6959 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
6960 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
6961 (tree_vector_builder::new_binary_operation): Handle polynomial
6962 element counts. Return false if we need to know the number
6963 of elements at compile time.
6964 * fold-const.c (fold_vec_perm): Punt if the number of elements
6965 isn't known at compile time.
6967 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6969 * vec-perm-indices.h (vec_perm_builder): Change element type
6970 from HOST_WIDE_INT to poly_int64.
6971 (vec_perm_indices::element_type): Update accordingly.
6972 (vec_perm_indices::clamp): Handle polynomial element_types.
6973 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
6974 (vec_perm_indices::all_in_range_p): Likewise.
6975 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
6976 than shwi trees.
6977 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
6978 polynomial vec_perm_indices element types.
6979 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
6980 * fold-const.c (fold_vec_perm): Likewise.
6981 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
6982 * tree-vect-generic.c (lower_vec_perm): Likewise.
6983 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
6984 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
6985 element type to HOST_WIDE_INT.
6987 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6988 Alan Hayward <alan.hayward@arm.com>
6989 David Sherwood <david.sherwood@arm.com>
6991 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
6992 rather than an int. Use plus_constant.
6993 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
6994 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
6996 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6997 Alan Hayward <alan.hayward@arm.com>
6998 David Sherwood <david.sherwood@arm.com>
7000 * calls.c (emit_call_1, expand_call): Change struct_value_size from
7001 a HOST_WIDE_INT to a poly_int64.
7003 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7004 Alan Hayward <alan.hayward@arm.com>
7005 David Sherwood <david.sherwood@arm.com>
7007 * calls.c (load_register_parameters): Cope with polynomial
7008 mode sizes. Require a constant size for BLKmode parameters
7009 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
7010 forces a parameter to be padded at the lsb end in order to
7011 fill a complete number of words, require the parameter size
7012 to be ordered wrt UNITS_PER_WORD.
7014 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7015 Alan Hayward <alan.hayward@arm.com>
7016 David Sherwood <david.sherwood@arm.com>
7018 * reload1.c (spill_stack_slot_width): Change element type
7019 from unsigned int to poly_uint64_pod.
7020 (alter_reg): Treat mode sizes as polynomial.
7022 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7023 Alan Hayward <alan.hayward@arm.com>
7024 David Sherwood <david.sherwood@arm.com>
7026 * reload.c (complex_word_subreg_p): New function.
7027 (reload_inner_reg_of_subreg, push_reload): Use it.
7029 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7030 Alan Hayward <alan.hayward@arm.com>
7031 David Sherwood <david.sherwood@arm.com>
7033 * lra-constraints.c (process_alt_operands): Reject matched
7034 operands whose sizes aren't ordered.
7035 (match_reload): Refer to this check here.
7037 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7038 Alan Hayward <alan.hayward@arm.com>
7039 David Sherwood <david.sherwood@arm.com>
7041 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
7042 that the mode size is in the set {1, 2, 4, 8, 16}.
7044 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7045 Alan Hayward <alan.hayward@arm.com>
7046 David Sherwood <david.sherwood@arm.com>
7048 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
7049 Use plus_constant instead of gen_rtx_PLUS.
7051 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7052 Alan Hayward <alan.hayward@arm.com>
7053 David Sherwood <david.sherwood@arm.com>
7055 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
7056 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7057 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
7058 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
7059 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
7060 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
7061 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
7062 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
7063 * config/i386/i386.c (ix86_push_rounding): ...this new function.
7064 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
7065 a poly_int64.
7066 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
7067 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
7068 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
7069 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
7070 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
7071 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
7072 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
7073 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
7074 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
7075 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
7076 function.
7077 * expr.c (emit_move_resolve_push): Treat the input and result
7078 of PUSH_ROUNDING as a poly_int64.
7079 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
7080 (emit_push_insn): Likewise.
7081 * lra-eliminations.c (mark_not_eliminable): Likewise.
7082 * recog.c (push_operand): Likewise.
7083 * reload1.c (elimination_effects): Likewise.
7084 * rtlanal.c (nonzero_bits1): Likewise.
7085 * calls.c (store_one_arg): Likewise. Require the padding to be
7086 known at compile time.
7088 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7089 Alan Hayward <alan.hayward@arm.com>
7090 David Sherwood <david.sherwood@arm.com>
7092 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
7093 Use plus_constant instead of gen_rtx_PLUS.
7095 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7096 Alan Hayward <alan.hayward@arm.com>
7097 David Sherwood <david.sherwood@arm.com>
7099 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
7100 rather than an int.
7102 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7103 Alan Hayward <alan.hayward@arm.com>
7104 David Sherwood <david.sherwood@arm.com>
7106 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
7107 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7108 via stack temporaries. Treat the mode size as polynomial too.
7110 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7111 Alan Hayward <alan.hayward@arm.com>
7112 David Sherwood <david.sherwood@arm.com>
7114 * expr.c (expand_expr_real_2): When handling conversions involving
7115 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7116 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7117 as a poly_uint64 too.
7119 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7120 Alan Hayward <alan.hayward@arm.com>
7121 David Sherwood <david.sherwood@arm.com>
7123 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
7125 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7126 Alan Hayward <alan.hayward@arm.com>
7127 David Sherwood <david.sherwood@arm.com>
7129 * combine.c (can_change_dest_mode): Handle polynomial
7130 REGMODE_NATURAL_SIZE.
7131 * expmed.c (store_bit_field_1): Likewise.
7132 * expr.c (store_constructor): Likewise.
7133 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
7134 and polynomial REGMODE_NATURAL_SIZE.
7135 (gen_lowpart_common): Likewise.
7136 * reginfo.c (record_subregs_of_mode): Likewise.
7137 * rtlanal.c (read_modify_subreg_p): Likewise.
7139 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7140 Alan Hayward <alan.hayward@arm.com>
7141 David Sherwood <david.sherwood@arm.com>
7143 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
7144 numbers of elements.
7146 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7147 Alan Hayward <alan.hayward@arm.com>
7148 David Sherwood <david.sherwood@arm.com>
7150 * match.pd: Cope with polynomial numbers of vector elements.
7152 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7153 Alan Hayward <alan.hayward@arm.com>
7154 David Sherwood <david.sherwood@arm.com>
7156 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
7157 in a POINTER_PLUS_EXPR.
7159 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7160 Alan Hayward <alan.hayward@arm.com>
7161 David Sherwood <david.sherwood@arm.com>
7163 * omp-simd-clone.c (simd_clone_subparts): New function.
7164 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
7165 (ipa_simd_modify_function_body): Likewise.
7167 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7168 Alan Hayward <alan.hayward@arm.com>
7169 David Sherwood <david.sherwood@arm.com>
7171 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
7172 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
7173 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
7174 (expand_vector_condition, vector_element): Likewise.
7175 (subparts_gt): New function.
7176 (get_compute_type): Use subparts_gt.
7177 (count_type_subparts): Delete.
7178 (expand_vector_operations_1): Use subparts_gt instead of
7179 count_type_subparts.
7181 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7182 Alan Hayward <alan.hayward@arm.com>
7183 David Sherwood <david.sherwood@arm.com>
7185 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
7186 (vect_compile_time_alias): ...this new function. Do the calculation
7187 on poly_ints rather than trees.
7188 (vect_prune_runtime_alias_test_list): Update call accordingly.
7190 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7191 Alan Hayward <alan.hayward@arm.com>
7192 David Sherwood <david.sherwood@arm.com>
7194 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
7195 numbers of units.
7196 (vect_schedule_slp_instance): Likewise.
7198 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7199 Alan Hayward <alan.hayward@arm.com>
7200 David Sherwood <david.sherwood@arm.com>
7202 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
7203 constant and extern definitions for variable-length vectors.
7204 (vect_get_constant_vectors): Note that the number of units
7205 is known to be constant.
7207 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7208 Alan Hayward <alan.hayward@arm.com>
7209 David Sherwood <david.sherwood@arm.com>
7211 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
7212 of units as polynomial. Choose between WIDE and NARROW based
7213 on multiple_p.
7215 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7216 Alan Hayward <alan.hayward@arm.com>
7217 David Sherwood <david.sherwood@arm.com>
7219 * tree-vect-stmts.c (simd_clone_subparts): New function.
7220 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
7222 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7223 Alan Hayward <alan.hayward@arm.com>
7224 David Sherwood <david.sherwood@arm.com>
7226 * tree-vect-stmts.c (vectorizable_call): Treat the number of
7227 vectors as polynomial. Use build_index_vector for
7228 IFN_GOMP_SIMD_LANE.
7230 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7231 Alan Hayward <alan.hayward@arm.com>
7232 David Sherwood <david.sherwood@arm.com>
7234 * tree-vect-stmts.c (get_load_store_type): Treat the number of
7235 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
7236 for variable-length vectors.
7237 (vectorizable_mask_load_store): Treat the number of units as
7238 polynomial, asserting that it is constant if the condition has
7239 already been enforced.
7240 (vectorizable_store, vectorizable_load): Likewise.
7242 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7243 Alan Hayward <alan.hayward@arm.com>
7244 David Sherwood <david.sherwood@arm.com>
7246 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
7247 of units as polynomial. Punt if we can't tell at compile time
7248 which vector contains the final result.
7250 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7251 Alan Hayward <alan.hayward@arm.com>
7252 David Sherwood <david.sherwood@arm.com>
7254 * tree-vect-loop.c (vectorizable_induction): Treat the number
7255 of units as polynomial. Punt on SLP inductions. Use an integer
7256 VEC_SERIES_EXPR for variable-length integer reductions. Use a
7257 cast of such a series for variable-length floating-point
7258 reductions.
7260 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7261 Alan Hayward <alan.hayward@arm.com>
7262 David Sherwood <david.sherwood@arm.com>
7264 * tree.h (build_index_vector): Declare.
7265 * tree.c (build_index_vector): New function.
7266 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
7267 of units as polynomial, forcibly converting it to a constant if
7268 vectorizable_reduction has already enforced the condition.
7269 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
7270 to create a {1,2,3,...} vector.
7271 (vectorizable_reduction): Treat the number of units as polynomial.
7272 Choose vectype_in based on the largest scalar element size rather
7273 than the smallest number of units. Enforce the restrictions
7274 relied on above.
7276 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7277 Alan Hayward <alan.hayward@arm.com>
7278 David Sherwood <david.sherwood@arm.com>
7280 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
7281 number of units as polynomial.
7283 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7284 Alan Hayward <alan.hayward@arm.com>
7285 David Sherwood <david.sherwood@arm.com>
7287 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
7288 * target.def (autovectorize_vector_sizes): Return the vector sizes
7289 by pointer, using vector_sizes rather than a bitmask.
7290 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
7291 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
7292 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
7293 Likewise.
7294 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
7295 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
7296 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
7297 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
7298 * omp-general.c (omp_max_vf): Likewise.
7299 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7300 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
7301 * tree-vect-loop.c (vect_analyze_loop): Likewise.
7302 * tree-vect-slp.c (vect_slp_bb): Likewise.
7303 * doc/tm.texi: Regenerate.
7304 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
7305 to a poly_uint64.
7306 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
7307 the vector size as a poly_uint64 rather than an unsigned int.
7308 (current_vector_size): Change from an unsigned int to a poly_uint64.
7309 (get_vectype_for_scalar_type): Update accordingly.
7310 * tree.h (build_truth_vector_type): Take the size and number of
7311 units as a poly_uint64 rather than an unsigned int.
7312 (build_vector_type): Add a temporary overload that takes
7313 the number of units as a poly_uint64 rather than an unsigned int.
7314 * tree.c (make_vector_type): Likewise.
7315 (build_truth_vector_type): Take the number of units as a poly_uint64
7316 rather than an unsigned int.
7318 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7319 Alan Hayward <alan.hayward@arm.com>
7320 David Sherwood <david.sherwood@arm.com>
7322 * target.def (get_mask_mode): Take the number of units and length
7323 as poly_uint64s rather than unsigned ints.
7324 * targhooks.h (default_get_mask_mode): Update accordingly.
7325 * targhooks.c (default_get_mask_mode): Likewise.
7326 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
7327 * doc/tm.texi: Regenerate.
7329 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7330 Alan Hayward <alan.hayward@arm.com>
7331 David Sherwood <david.sherwood@arm.com>
7333 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
7334 * omp-general.c (omp_max_vf): Likewise.
7335 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
7336 (expand_omp_simd): Handle polynomial safelen.
7337 * omp-low.c (omplow_simd_context): Add a default constructor.
7338 (omplow_simd_context::max_vf): Change from int to poly_uint64.
7339 (lower_rec_simd_input_clauses): Update accordingly.
7340 (lower_rec_input_clauses): Likewise.
7342 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7343 Alan Hayward <alan.hayward@arm.com>
7344 David Sherwood <david.sherwood@arm.com>
7346 * tree-vectorizer.h (vect_nunits_for_cost): New function.
7347 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
7348 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
7349 (vect_analyze_slp_cost): Likewise.
7350 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
7351 (vect_model_load_cost): Likewise.
7353 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7354 Alan Hayward <alan.hayward@arm.com>
7355 David Sherwood <david.sherwood@arm.com>
7357 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
7358 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
7359 from an unsigned int * to a poly_uint64_pod *.
7360 (calculate_unrolling_factor): New function.
7361 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
7363 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7364 Alan Hayward <alan.hayward@arm.com>
7365 David Sherwood <david.sherwood@arm.com>
7367 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
7368 from an unsigned int to a poly_uint64.
7369 (_loop_vec_info::slp_unrolling_factor): Likewise.
7370 (_loop_vec_info::vectorization_factor): Change from an int
7371 to a poly_uint64.
7372 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
7373 (vect_get_num_vectors): New function.
7374 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
7375 (vect_get_num_copies): Use vect_get_num_vectors.
7376 (vect_analyze_data_ref_dependences): Change max_vf from an int *
7377 to an unsigned int *.
7378 (vect_analyze_data_refs): Change min_vf from an int * to a
7379 poly_uint64 *.
7380 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7381 than an unsigned HOST_WIDE_INT.
7382 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7383 (vect_analyze_data_ref_dependence): Change max_vf from an int *
7384 to an unsigned int *.
7385 (vect_analyze_data_ref_dependences): Likewise.
7386 (vect_compute_data_ref_alignment): Handle polynomial vf.
7387 (vect_enhance_data_refs_alignment): Likewise.
7388 (vect_prune_runtime_alias_test_list): Likewise.
7389 (vect_shift_permute_load_chain): Likewise.
7390 (vect_supportable_dr_alignment): Likewise.
7391 (dependence_distance_ge_vf): Take the vectorization factor as a
7392 poly_uint64 rather than an unsigned HOST_WIDE_INT.
7393 (vect_analyze_data_refs): Change min_vf from an int * to a
7394 poly_uint64 *.
7395 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
7396 vfm1 as a poly_uint64 rather than an int. Make the same change
7397 for the returned bound_scalar.
7398 (vect_gen_vector_loop_niters): Handle polynomial vf.
7399 (vect_do_peeling): Likewise. Update call to
7400 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
7401 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
7402 be constant.
7403 * tree-vect-loop.c (vect_determine_vectorization_factor)
7404 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
7405 (vect_get_known_peeling_cost): Likewise.
7406 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
7407 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
7408 (vect_transform_loop): Likewise. Use the lowest possible VF when
7409 updating the upper bounds of the loop.
7410 (vect_min_worthwhile_factor): Make static. Return an unsigned int
7411 rather than an int.
7412 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
7413 polynomial unroll factors.
7414 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
7415 (vect_make_slp_decision): Likewise.
7416 (vect_supported_load_permutation_p): Likewise, and polynomial
7417 vf too.
7418 (vect_analyze_slp_cost): Handle polynomial vf.
7419 (vect_slp_analyze_node_operations): Likewise.
7420 (vect_slp_analyze_bb_1): Likewise.
7421 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7422 than an unsigned HOST_WIDE_INT.
7423 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
7424 (vectorizable_load): Handle polynomial vf.
7425 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
7426 a poly_uint64.
7427 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
7429 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7430 Alan Hayward <alan.hayward@arm.com>
7431 David Sherwood <david.sherwood@arm.com>
7433 * match.pd: Handle bit operations involving three constants
7434 and try to fold one pair.
7436 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7438 * tree-vect-loop-manip.c: Include gimple-fold.h.
7439 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
7440 niters_maybe_zero parameters. Handle other cases besides a step of 1.
7441 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
7442 Add a path that uses a step of VF instead of 1, but disable it
7443 for now.
7444 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
7445 and niters_no_overflow parameters. Update calls to
7446 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
7447 Create a new SSA name if the latter choses to use a ste other
7448 than zero, and return it via niters_vector_mult_vf_var.
7449 * tree-vect-loop.c (vect_transform_loop): Update calls to
7450 vect_do_peeling, vect_gen_vector_loop_niters and
7451 slpeel_make_loop_iterate_ntimes.
7452 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
7453 (vect_gen_vector_loop_niters): Update declarations after above changes.
7455 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
7457 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
7458 128-bit round to integer instructions.
7459 (ceil<mode>2): Likewise.
7460 (btrunc<mode>2): Likewise.
7461 (round<mode>2): Likewise.
7463 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7465 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
7466 unaligned VSX load/store on P8/P9.
7467 (expand_block_clear): Allow the use of unaligned VSX
7468 load/store on P8/P9.
7470 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7472 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
7473 New function.
7474 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
7475 swap associated with both a load and a store.
7477 2018-01-02 Andrew Waterman <andrew@sifive.com>
7479 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
7480 * config/riscv/riscv.md (clear_cache): Use it.
7482 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
7484 * web.c: Remove out-of-date comment.
7486 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7488 * expr.c (fixup_args_size_notes): Check that any existing
7489 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
7490 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
7491 (emit_single_push_insn): ...here.
7493 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7495 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
7496 (const_vector_encoded_nelts): New function.
7497 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
7498 (const_vector_int_elt, const_vector_elt): Declare.
7499 * emit-rtl.c (const_vector_int_elt_1): New function.
7500 (const_vector_elt): Likewise.
7501 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
7502 of CONST_VECTOR_ELT.
7504 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7506 * expr.c: Include rtx-vector-builder.h.
7507 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
7508 directly on the tree encoding.
7509 (const_vector_from_tree): Likewise.
7510 * optabs.c: Include rtx-vector-builder.h.
7511 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
7512 sequence of "u" values.
7513 * vec-perm-indices.c: Include rtx-vector-builder.h.
7514 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
7515 directly on the vec_perm_indices encoding.
7517 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7519 * doc/rtl.texi (const_vector): Describe new encoding scheme.
7520 * Makefile.in (OBJS): Add rtx-vector-builder.o.
7521 * rtx-vector-builder.h: New file.
7522 * rtx-vector-builder.c: Likewise.
7523 * rtl.h (rtx_def::u2): Add a const_vector field.
7524 (CONST_VECTOR_NPATTERNS): New macro.
7525 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
7526 (CONST_VECTOR_DUPLICATE_P): Likewise.
7527 (CONST_VECTOR_STEPPED_P): Likewise.
7528 (CONST_VECTOR_ENCODED_ELT): Likewise.
7529 (const_vec_duplicate_p): Check for a duplicated vector encoding.
7530 (unwrap_const_vec_duplicate): Likewise.
7531 (const_vec_series_p): Check for a non-duplicated vector encoding.
7532 Say that the function only returns true for integer vectors.
7533 * emit-rtl.c: Include rtx-vector-builder.h.
7534 (gen_const_vec_duplicate_1): Delete.
7535 (gen_const_vector): Call gen_const_vec_duplicate instead of
7536 gen_const_vec_duplicate_1.
7537 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
7538 (gen_const_vec_duplicate): Use rtx_vector_builder.
7539 (gen_const_vec_series): Likewise.
7540 (gen_rtx_CONST_VECTOR): Likewise.
7541 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
7542 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7543 Build a new vector rather than modifying a CONST_VECTOR in-place.
7544 (handle_special_swappables): Update call accordingly.
7545 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
7546 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7547 Build a new vector rather than modifying a CONST_VECTOR in-place.
7548 (handle_special_swappables): Update call accordingly.
7550 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7552 * simplify-rtx.c (simplify_const_binary_operation): Use
7553 CONST_VECTOR_ELT instead of XVECEXP.
7555 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7557 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
7558 the selector elements to be different from the data elements
7559 if the selector is a VECTOR_CST.
7560 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
7561 ssizetype for the selector.
7563 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7565 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
7566 before testing each element individually.
7567 * tree-vect-generic.c (lower_vec_perm): Likewise.
7569 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7571 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
7572 * selftest-run-tests.c (selftest::run_tests): Call it.
7573 * vector-builder.h (vector_builder::operator ==): New function.
7574 (vector_builder::operator !=): Likewise.
7575 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
7576 (vec_perm_indices::all_from_input_p): New function.
7577 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7578 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
7579 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
7580 instead of reading the VECTOR_CST directly. Detect whether both
7581 vector inputs are the same before constructing the vec_perm_indices,
7582 and update the number of inputs argument accordingly. Use the
7583 utility functions added above. Only construct sel2 if we need to.
7585 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7587 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
7588 the broadcast of the low byte.
7589 (expand_mult_highpart): Use an explicit encoding for the permutes.
7590 * optabs-query.c (can_mult_highpart_p): Likewise.
7591 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
7592 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7593 (vectorizable_bswap): Likewise.
7594 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
7595 explicit encoding for the power-of-2 permutes.
7596 (vect_permute_store_chain): Likewise.
7597 (vect_grouped_load_supported): Likewise.
7598 (vect_permute_load_chain): Likewise.
7600 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7602 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
7603 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
7604 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
7605 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7606 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
7607 (vect_gen_perm_mask_any): Likewise.
7609 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7611 * int-vector-builder.h: New file.
7612 * vec-perm-indices.h: Include int-vector-builder.h.
7613 (vec_perm_indices): Redefine as an int_vector_builder.
7614 (auto_vec_perm_indices): Delete.
7615 (vec_perm_builder): Redefine as a stand-alone class.
7616 (vec_perm_indices::vec_perm_indices): New function.
7617 (vec_perm_indices::clamp): Likewise.
7618 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
7619 (vec_perm_indices::new_vector): New function.
7620 (vec_perm_indices::new_expanded_vector): Update for new
7621 vec_perm_indices class.
7622 (vec_perm_indices::rotate_inputs): New function.
7623 (vec_perm_indices::all_in_range_p): Operate directly on the
7624 encoded form, without computing elided elements.
7625 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
7626 encoding. Update for new vec_perm_indices class.
7627 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
7628 the given vec_perm_builder.
7629 (expand_vec_perm_var): Update vec_perm_builder constructor.
7630 (expand_mult_highpart): Use vec_perm_builder instead of
7631 auto_vec_perm_indices.
7632 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
7633 vec_perm_indices instead of auto_vec_perm_indices. Use a single
7634 or double series encoding as appropriate.
7635 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
7636 vec_perm_indices instead of auto_vec_perm_indices.
7637 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7638 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7639 (vect_permute_store_chain): Likewise.
7640 (vect_grouped_load_supported): Likewise.
7641 (vect_permute_load_chain): Likewise.
7642 (vect_shift_permute_load_chain): Likewise.
7643 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
7644 (vect_transform_slp_perm_load): Likewise.
7645 (vect_schedule_slp_instance): Likewise.
7646 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7647 (vectorizable_mask_load_store): Likewise.
7648 (vectorizable_bswap): Likewise.
7649 (vectorizable_store): Likewise.
7650 (vectorizable_load): Likewise.
7651 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
7652 vec_perm_indices instead of auto_vec_perm_indices. Use
7653 tree_to_vec_perm_builder to read the vector from a tree.
7654 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
7655 vec_perm_builder instead of a vec_perm_indices.
7656 (have_whole_vector_shift): Use vec_perm_builder and
7657 vec_perm_indices instead of auto_vec_perm_indices. Leave the
7658 truncation to calc_vec_perm_mask_for_shift.
7659 (vect_create_epilog_for_reduction): Likewise.
7660 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
7661 from auto_vec_perm_indices to vec_perm_indices.
7662 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7663 instead of changing individual elements.
7664 (aarch64_vectorize_vec_perm_const): Use new_vector to install
7665 the vector in d.perm.
7666 * config/arm/arm.c (expand_vec_perm_d::perm): Change
7667 from auto_vec_perm_indices to vec_perm_indices.
7668 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7669 instead of changing individual elements.
7670 (arm_vectorize_vec_perm_const): Use new_vector to install
7671 the vector in d.perm.
7672 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
7673 Update vec_perm_builder constructor.
7674 (rs6000_expand_interleave): Likewise.
7675 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
7676 (rs6000_expand_interleave): Likewise.
7678 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7680 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
7681 to qimode could truncate the indices.
7682 * optabs.c (expand_vec_perm_var): Likewise.
7684 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7686 * Makefile.in (OBJS): Add vec-perm-indices.o.
7687 * vec-perm-indices.h: New file.
7688 * vec-perm-indices.c: Likewise.
7689 * target.h (vec_perm_indices): Replace with a forward class
7690 declaration.
7691 (auto_vec_perm_indices): Move to vec-perm-indices.h.
7692 * optabs.h: Include vec-perm-indices.h.
7693 (expand_vec_perm): Delete.
7694 (selector_fits_mode_p, expand_vec_perm_var): Declare.
7695 (expand_vec_perm_const): Declare.
7696 * target.def (vec_perm_const_ok): Replace with...
7697 (vec_perm_const): ...this new hook.
7698 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
7699 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
7700 * doc/tm.texi: Regenerate.
7701 * optabs.def (vec_perm_const): Delete.
7702 * doc/md.texi (vec_perm_const): Likewise.
7703 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
7704 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
7705 expand_vec_perm for constant permutation vectors. Assert that
7706 the mode of variable permutation vectors is the integer equivalent
7707 of the mode that is being permuted.
7708 * optabs-query.h (selector_fits_mode_p): Declare.
7709 * optabs-query.c: Include vec-perm-indices.h.
7710 (selector_fits_mode_p): New function.
7711 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
7712 is defined, instead of checking whether the vec_perm_const_optab
7713 exists. Use targetm.vectorize.vec_perm_const instead of
7714 targetm.vectorize.vec_perm_const_ok. Check whether the indices
7715 fit in the vector mode before using a variable permute.
7716 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
7717 vec_perm_indices instead of an rtx.
7718 (expand_vec_perm): Replace with...
7719 (expand_vec_perm_const): ...this new function. Take the selector
7720 as a vec_perm_indices rather than an rtx. Also take the mode of
7721 the selector. Update call to shift_amt_for_vec_perm_mask.
7722 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
7723 Use vec_perm_indices::new_expanded_vector to expand the original
7724 selector into bytes. Check whether the indices fit in the vector
7725 mode before using a variable permute.
7726 (expand_vec_perm_var): Make global.
7727 (expand_mult_highpart): Use expand_vec_perm_const.
7728 * fold-const.c: Includes vec-perm-indices.h.
7729 * tree-ssa-forwprop.c: Likewise.
7730 * tree-vect-data-refs.c: Likewise.
7731 * tree-vect-generic.c: Likewise.
7732 * tree-vect-loop.c: Likewise.
7733 * tree-vect-slp.c: Likewise.
7734 * tree-vect-stmts.c: Likewise.
7735 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
7736 Delete.
7737 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
7738 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
7739 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
7740 (aarch64_vectorize_vec_perm_const): ...this new function.
7741 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7742 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7743 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
7744 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
7745 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7746 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7747 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
7748 into...
7749 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
7750 check for NEON modes.
7751 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
7752 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
7753 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
7754 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
7755 into...
7756 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
7757 the old VEC_PERM_CONST conditions.
7758 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
7759 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
7760 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
7761 (ia64_vectorize_vec_perm_const_ok): Merge into...
7762 (ia64_vectorize_vec_perm_const): ...this new function.
7763 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
7764 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
7765 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
7766 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
7767 * config/mips/mips.c (mips_expand_vec_perm_const)
7768 (mips_vectorize_vec_perm_const_ok): Merge into...
7769 (mips_vectorize_vec_perm_const): ...this new function.
7770 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
7771 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
7772 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
7773 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
7774 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
7775 (rs6000_expand_vec_perm_const): Delete.
7776 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
7777 Delete.
7778 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7779 (altivec_expand_vec_perm_const_le): Take each operand individually.
7780 Operate on constant selectors rather than rtxes.
7781 (altivec_expand_vec_perm_const): Likewise. Update call to
7782 altivec_expand_vec_perm_const_le.
7783 (rs6000_expand_vec_perm_const): Delete.
7784 (rs6000_vectorize_vec_perm_const_ok): Delete.
7785 (rs6000_vectorize_vec_perm_const): New function.
7786 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
7787 an element count and rtx array.
7788 (rs6000_expand_extract_even): Update call accordingly.
7789 (rs6000_expand_interleave): Likewise.
7790 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
7791 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
7792 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
7793 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
7794 (rs6000_expand_vec_perm_const): Delete.
7795 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7796 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7797 (altivec_expand_vec_perm_const_le): Take each operand individually.
7798 Operate on constant selectors rather than rtxes.
7799 (altivec_expand_vec_perm_const): Likewise. Update call to
7800 altivec_expand_vec_perm_const_le.
7801 (rs6000_expand_vec_perm_const): Delete.
7802 (rs6000_vectorize_vec_perm_const_ok): Delete.
7803 (rs6000_vectorize_vec_perm_const): New function. Remove stray
7804 reference to the SPE evmerge intructions.
7805 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
7806 an element count and rtx array.
7807 (rs6000_expand_extract_even): Update call accordingly.
7808 (rs6000_expand_interleave): Likewise.
7809 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
7810 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
7811 new function.
7812 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7814 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7816 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
7817 vector mode and that that mode matches the mode of the data
7818 being permuted.
7819 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
7820 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
7821 directly using expand_vec_perm_1 when forcing selectors into
7822 registers.
7823 (expand_vec_perm_var): New function, split out from expand_vec_perm.
7825 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7827 * optabs-query.h (can_vec_perm_p): Delete.
7828 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
7829 * optabs-query.c (can_vec_perm_p): Split into...
7830 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
7831 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
7832 particular selector is valid.
7833 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7834 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7835 (vect_grouped_load_supported): Likewise.
7836 (vect_shift_permute_load_chain): Likewise.
7837 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
7838 (vect_transform_slp_perm_load): Likewise.
7839 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7840 (vectorizable_bswap): Likewise.
7841 (vect_gen_perm_mask_checked): Likewise.
7842 * fold-const.c (fold_ternary_loc): Likewise. Don't take
7843 implementations of variable permutation vectors into account
7844 when deciding which selector to use.
7845 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
7846 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
7847 with a false third argument.
7848 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
7849 to test whether the constant selector is valid and can_vec_perm_var_p
7850 to test whether a variable selector is valid.
7852 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7854 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
7855 * optabs-query.c (can_vec_perm_p): Likewise.
7856 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
7857 instead of vec_perm_indices.
7858 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
7859 (vect_gen_perm_mask_checked): Likewise,
7860 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
7861 (vect_gen_perm_mask_checked): Likewise,
7863 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7865 * optabs-query.h (qimode_for_vec_perm): Declare.
7866 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
7867 (qimode_for_vec_perm): ...this new function.
7868 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
7870 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7872 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
7873 does not have a conditional at the top.
7875 2018-01-02 Richard Biener <rguenther@suse.de>
7877 * ipa-inline.c (big_speedup_p): Fix expression.
7879 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
7881 PR target/81616
7882 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
7883 for generic 4->6.
7885 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
7887 PR target/81616
7888 Generic tuning.
7889 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
7890 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
7891 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
7892 cond_taken_branch_cost 3->4.
7894 2018-01-01 Jakub Jelinek <jakub@redhat.com>
7896 PR tree-optimization/83581
7897 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
7898 TODO_cleanup_cfg if any changes have been made.
7900 PR middle-end/83608
7901 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
7902 convert_modes if target mode has the right side, but different mode
7903 class.
7905 PR middle-end/83609
7906 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
7907 last argument when extracting from CONCAT. If either from_real or
7908 from_imag is NULL, use expansion through memory. If result is not
7909 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
7910 the parts directly to inner mode, if even that fails, use expansion
7911 through memory.
7913 PR middle-end/83623
7914 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
7915 check for bswap in mode rather than HImode and use that in expand_unop
7916 too.
7917 \f
7918 Copyright (C) 2018 Free Software Foundation, Inc.
7920 Copying and distribution of this file, with or without modification,
7921 are permitted in any medium without royalty provided the copyright
7922 notice and this notice are preserved.