1 2018-04-30 Richard Biener <rguenther@suse.de>
3 * tree-chrec.h (evolution_function_is_constant_p): Remove
5 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
7 2018-04-30 Richard Biener <rguenther@suse.de>
10 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
12 2018-04-30 Richard Biener <rguenther@suse.de>
14 PR tree-optimization/28364
15 PR tree-optimization/85275
16 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
17 copying first exit test.
19 2018-04-28 Mark Wielaard <mark@klomp.org>
21 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
23 (dwarf_AT): Handle DW_AT_addr_base.
24 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
26 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
29 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
30 (*ashl<dwi>3_doubleword_mask_1): Ditto.
31 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
32 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
34 2018-04-28 Richard Biener <rguenther@suse.de>
36 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
37 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
38 to reflect use. Only add interesting stmts.
40 2018-04-27 Martin Jambor <mjambor@suse.cz>
43 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
44 the jump function allows for passing through aggregate values.
46 2018-04-27 David Malcolm <dmalcolm@redhat.com>
48 * input.h (in_system_header_at): Convert from macro to inline
50 (from_macro_expansion_at): Likewise.
51 (from_macro_definition_at): Likewise.
53 2018-04-27 Jeff Law <law@redhat.com>
55 * config.gcc: Mark tile* targets as deprecated/obsolete.
57 2018-04-27 Richard Biener <rguenther@suse.de>
59 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
62 2018-04-27 Richard Biener <rguenther@suse.de>
64 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
66 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
68 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
69 with Yd constraint. Set "preferred_for_speed" attribute from
70 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
72 (*movdi_internal): Ditto.
73 (movti_interunit splitters): Remove
74 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
75 (movdi_interunit splitters): Ditto.
76 * config/i386/constraints.md (Ye): Remove.
77 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
79 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
82 * config/aarch64/constraints.md (Usg): Limit to 31.
85 2018-04-27 Jakub Jelinek <jakub@redhat.com>
87 PR tree-optimization/85529
88 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
89 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
90 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
91 zero extension or masking of the MSB bit.
92 (optimize_range_tests): Add FIRST_BB argument, pass it through
93 to optimize_range_tests_var_bound.
94 (maybe_optimize_range_tests, reassociate_bb): Adjust
95 optimize_range_tests callers.
97 2018-04-26 Richard Biener <rguenther@suse.de>
98 Jakub Jelinek <jakub@redhat.com>
100 * cgraph.h (symbol_table): Just declare debug method here.
101 * symtab.c (symbol_table::debug): Define.
103 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
105 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
107 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
109 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
110 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
111 (*movdi_internal): Substitute Yi and Yj constraint with x
112 and Ym and Yn constraint with y constraint. Update "isa"
113 attribute and set "preferred_for_speed" attribute from
114 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
115 (*movsi_internal): Ditto.
116 (*movdf_internal): Ditto.
117 (*movsf_internal): Ditto.
118 (*zero_extendsidi2): Ditto.
119 * config/i386/sse.md (vec_set<mode>_0): Ditto.
120 (sse2_loadld): Ditto.
121 (*vec_extract<ssevecmodelower>_0): Ditto.
122 (*vec_extractv4si_0_zext_sse4): Ditto.
123 (vec_concatv2di): Ditto.
124 (*vec_dup<mode>): Ditto.
125 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
126 * config/i386/constraints.md (Yi): Remove.
131 2018-04-26 Nathan Sidwell <nathan@acm.org>
133 * dumpfile.c (dump_open): New.
134 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
135 (dump_finish): Detect stdio/stderr by value not name.
137 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
139 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
141 2018-04-26 Tom de Vries <tom@codesourcery.com>
144 * config/nvptx/nvptx.c (verify_neutering_jumps)
145 (verify_neutering_labels): New function
146 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
148 2018-04-26 Tom de Vries <tom@codesourcery.com>
151 * config/nvptx/nvptx.c (needs_neutering_p): New function.
152 (nvptx_single): Use needs_neutering_p to skip over insns that do not
155 2018-04-26 Richard Biener <rguenther@suse.de>
156 Tom de Vries <tom@codesourcery.com>
159 * lto-streamer-out.c (output_function): Fixup loops if required to match
160 discovery done in the reader.
162 2018-04-26 Richard Biener <rguenther@suse.de>
164 PR tree-optimization/85116
165 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
166 have a loop exit from the single latch predecessor. Remove
167 case of header with just condition.
168 (ch_base::copy_headers): Exclude infinite loops from any
170 (pass_ch::execute): Record exits.
172 2018-04-26 Richard Biener <rguenther@suse.de>
174 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
175 prologue cost vector and pass it to vect_get_load_cost.
176 (vect_get_peeling_costs_all_drs): Likewise.
177 (vect_peeling_hash_get_lowest_cost): Likewise.
178 (vect_enhance_data_refs_alignment): Likewise.
180 2018-04-26 Richard Biener <rguenther@suse.de>
183 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
184 checking of integer<->pointer conversions.
185 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
186 sign-/zero-extending pointer types.
187 (expand_omp_for_static_chunk): Likewise.
189 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
190 Jean Lee <xiaoyur347@gmail.com>
192 * config/mips/mips.c (mips_asan_shadow_offset): New function.
193 (TARGET_ASAN_SHADOW_OFFSET): Define.
194 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
195 true for -fsanitize=address.
197 2018-04-25 Mark Wielaard <mark@klomp.org>
199 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
202 2018-04-25 Jakub Jelinek <jakub@redhat.com>
204 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
205 than "alu", remove explicit "memory" and "imm_disp" attributes.
206 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
209 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
210 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
213 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
216 * config/i386/i386.c (ix86_expand_builtin): Change memory
217 operand to XI, extend p0 to Pmode.
218 * config/i386/i386.md: Change unspec volatile and operand
219 1 mode to XI, change operand 0 mode to P.
221 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
223 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
224 GET_MODE_MASK before any checking.
225 (nds32_can_use_bset_p): Likewise.
226 (nds32_can_use_btgl_p): Likewise.
228 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
230 * config/nds32/nds32-doubleword.md: New define_split pattern for
231 illegal register number.
233 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
235 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
237 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
239 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
241 2018-04-25 Richard Biener <rguenther@suse.de>
243 * lto-streamer.h (LTO_major_version): Bump to 8.
245 2018-04-25 Jakub Jelinek <jakub@redhat.com>
247 * BASE-VER: Set to 9.0.0.
249 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
251 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
252 in __abskf2 and __powikf2.
254 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
257 * config/aarch64/constraints.md (Usg, Usj): New constraints.
258 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
259 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
260 Use the above on operand 2. Reindent.
261 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
263 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
266 * common/config/i386/i386-common.c (ix86_handle_option): Don't
268 * config/i386/i386.opt (mcet): Removed.
269 * doc/install.texi: Remove -mcet documentation.
270 * doc/invoke.texi: Likewise.
272 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
275 * doc/install.texi: Remove -mcet from bootstrap-cet.
277 2018-04-24 Jakub Jelinek <jakub@redhat.com>
280 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
281 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
285 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
286 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
287 containing a CONST_VECTOR.
289 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
291 * doc/install.texi: Update newlib dependency for nvptx.
293 2018-04-24 Jakub Jelinek <jakub@redhat.com>
296 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
297 instead of INTVAL when shifting x left.
299 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
301 PR tree-optimization/85478
302 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
303 vect_grouped_store_supported for single element vectors.
305 2018-04-24 Richard Biener <rguenther@suse.de>
308 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
309 load cost increase to the case of non-constant step.
311 2018-04-24 Jakub Jelinek <jakub@redhat.com>
314 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
315 destination if any_malformed_asm.
317 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
320 * expr.c (store_field): In the bitfield case, if the value comes from
321 a function call and is returned in registers by means of a PARALLEL,
322 do not change the mode of the temporary unless BLKmode and VOIDmode.
324 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
326 PR rtl-optimization/85423
327 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
328 dependencies to debug insns when the previous insn is non-debug.
330 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
332 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
333 enums into a single definition.
334 (fls): Fix predicates and printing.
337 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
339 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
340 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
341 and short u6 immediate.
342 (check_if_valid_sleep_operand): Remove.
343 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
345 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
347 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
348 flag_always_save_lp condition.
349 * config/nds32/nds32.opt (malways-save-lp): New option.
351 2018-04-22 Shiva Chen <shiva0217@gmail.com>
353 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
354 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
355 * config/nds32/nds32.h
356 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
357 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
359 2018-04-22 Shiva Chen <shiva0217@gmail.com>
361 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
362 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
364 2018-04-22 Shiva Chen <shiva0217@gmail.com>
365 Chung-Ju Wu <jasonwucj@gmail.com>
367 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
369 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
370 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
372 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
374 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
376 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
378 * config/nds32/nds32-protos.h (nds32_data_alignment,
379 nds32_local_alignment): Declare.
380 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
381 nds32_local_alignment): New functions.
382 (TARGET_CONSTANT_ALIGNMENT): Define.
383 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
385 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
387 * config/nds32/nds32.c
388 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
389 (TARGET_MODES_TIEABLE_P): Likewise.
391 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
393 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
396 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
397 Chung-Ju Wu <jasonwucj@gmail.com>
399 * config/nds32/constants.md (unspec_volatile_element): Add enum values
400 for unaligned access.
401 * config/nds32/nds32-intrinsic.c: Implementation of expanding
403 * config/nds32/nds32-intrinsic.md: Likewise.
404 * config/nds32/nds32_intrinsic.h: Likewise.
405 * config/nds32/nds32.h (nds32_builtins): Likewise.
406 * config/nds32/nds32.opt (munaligned-access): New option.
407 * config/nds32/nds32.c (nds32_asm_file_start): Display
408 flag_unaligned_access status.
410 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
412 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
413 -mno-relax is present.
414 * config/riscv/linux.h (LINK_SPEC): Ditto.
416 2018-04-20 Martin Sebor <msebor@redhat.com>
419 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
421 (gimple_fold_builtin_stxcpy_chk): Same.
422 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
424 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
427 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
428 __powikf2 when long double is IEEE 128-bit.
430 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
432 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
433 step to make sure stack always aligned.
435 2018-04-20 Carl Love <cel@us.ibm.com>
438 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
441 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
442 Tom de Vries <tom@codesourcery.com>
445 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
446 Emit insns for calls too.
447 (nvptx_find_par): Always look for worker-level predecessor insn.
448 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
450 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
451 (nvptx_process_pars): Propagate frames for calls.
453 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
456 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
458 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
459 (ix86_handle_option): Don't handle OPT_mibt.
460 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
462 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
464 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
465 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
466 (ix86_target_macros): Define __CET__ with flag_cf_protection
468 * config/i386/i386.c (isa2_opts): Remove -mibt.
469 * config/i386/i386.h (TARGET_IBT): Removed.
470 (TARGET_IBT_P): Likewise.
471 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
472 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
473 * config/i386/i386.opt (mcet): Update help message.
476 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
477 -mcet as an alias for -mshstk.
479 2018-04-20 Richard Biener <rguenther@suse.de>
482 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
483 complexity by forcing a single use of the multiply operand.
485 2018-04-20 Martin Jambor <mjambor@suse.cz>
488 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
489 recursion dependency to only apply to non-clones.
491 2018-04-20 Martin Jambor <mjambor@suse.cz>
494 * ipa-cp.c (create_specialized_node): Check that clones of
495 self-recursive edges exist during IPA-CP.
497 2018-04-19 Toon Moene <toon@moene.org>
499 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
502 2018-04-19 Jakub Jelinek <jakub@redhat.com>
504 PR tree-optimization/85467
505 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
506 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
507 VECTOR_CST element to type.
509 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
512 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
513 * config/i386/i386.md (builtin_setjmp_setup): Removed.
514 (builtin_longjmp): Likewise.
515 (save_stack_nonlocal): New pattern.
516 (restore_stack_nonlocal): Likewise.
518 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
521 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
522 Replace ASM_OUTPUT_LABEL with fprintf.
524 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
527 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
528 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
529 * config/i386/i386-c.c (ix86_target_macros_internal): Also
530 define __IBT__ and __SHSTK__ for -fcf-protection.
531 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
533 (ix86_trampoline_init): Likewise.
534 (x86_output_mi_thunk): Likewise.
535 (ix86_notrack_prefixed_insn_p): Likewise.
536 (ix86_option_override_internal): Don't disallow -fcf-protection.
537 * config/i386/i386.md (rdssp<mode>): Also enable for
539 (incssp<mode>): Likewise.
540 (nop_endbr): Likewise.
541 * config/i386/i386.opt (mcet): Change help message to built-in
545 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
546 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
547 enable CET built-in functions.
549 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
551 * common/config/i386/i386-common.c
552 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
553 OPTION_MASK_ISA_MOVDIRI_UNSET,
554 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
555 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
556 * config.gcc (movdirintrin.h): New header.
557 * config/i386/cpuid.h (bit_MOVDIRI,
558 bit_MOVDIR64B): New bits.
559 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
561 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
562 (VOID, PVOID, PCVOID)): New function types.
563 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
564 __builtin_ia32_directstoreu_u64,
565 __builtin_ia32_movdir64b): New builtins.
566 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
567 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
569 (ix86_valid_target_attribute_inner_p): Ditto.
570 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
571 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
572 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
573 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
574 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
575 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
576 (movdiri<mode>, movdir64b_<mode>): New.
577 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
578 * config/i386/immintrin.h: Include movdirintrin.h.
579 * config/i386/movdirintrin.h: New file.
580 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
582 2018-04-19 Richard Biener <rguenther@suse.de>
585 * cfg.c (clear_bb_flags): When loop state says we have
586 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
588 2018-04-19 Richard Biener <rguenther@suse.de>
590 PR tree-optimization/84737
591 * tree-vect-data-refs.c (vect_copy_ref_info): New function
592 copying restrict info.
593 (vect_setup_realignment): Use it.
594 * tree-vectorizer.h (vect_copy_ref_info): Declare.
595 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
596 the first DR to all generated stores.
597 (vectorizable_load): Likewise for loads.
599 2018-04-19 Jakub Jelinek <jakub@redhat.com>
601 PR tree-optimization/85446
602 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
603 the integral and pointer types to have the same precision.
605 * doc/install.texi: Document --disable-cet being the default and
608 2018-04-18 Martin Liska <mliska@suse.cz>
610 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
613 2018-04-18 Martin Liska <mliska@suse.cz>
616 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
619 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
620 arguments if they are comparable.
622 2018-04-18 Martin Liska <mliska@suse.cz>
625 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
628 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
631 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
634 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
635 ENDBR after calling __morestack.
637 2018-04-18 David Malcolm <dmalcolm@redhat.com>
640 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
641 by using gcc_base_ver to generate a gcc_driver_version, and use
642 it when generating GCC_DRIVER_NAME.
643 * configure: Regenerate.
645 2018-04-18 Jakub Jelinek <jakub@redhat.com>
648 * config.gcc: Obsolete powerpc*-*-*spe*.
650 2018-04-17 Jakub Jelinek <jakub@redhat.com>
653 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
654 (stabstr_D): Change type of unum from unsigned int to
655 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
658 2018-04-17 Jim Wilson <jimw@sifive.com>
661 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
662 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
663 Set arg_pointer_offset after using pretend_args_size.
665 2018-04-17 Jakub Jelinek <jakub@redhat.com>
667 PR rtl-optimization/85431
668 * dse.c (record_store): Ignore zero width stores.
671 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
672 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
673 __builtin_stack_restore rather than after it.
674 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
675 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
676 argument instead of virtual_dynamic_stack_rtx.
678 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
680 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
682 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
683 Add note to error message to explain internal mapping of overloaded
684 built-in function name to non-overloaded built-in function name.
685 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
688 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
691 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
692 where the inputs overlap with the output.
694 2018-04-17 Jakub Jelinek <jakub@redhat.com>
696 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
697 (=v, v) alternative and explicit "memory" attribute.
698 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
699 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
701 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
702 "sselog1" type instead of "sselog".
703 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
704 "sselog". Remove explicit "memory" attribute.
705 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
706 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
708 (vec_extract_hi_v32hi): Merge all alternatives into one, use
709 "sselog1" type instead of "sselog". Remove explicit "memory"
711 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
712 use "sselog1" type instead of "sselog". Remove explicit "memory"
714 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
715 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
717 (vec_extract_hi_v64qi): Merge all alternatives into one, use
718 "sselog1" type instead of "sselog". Remove explicit "memory"
720 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
721 use "sselog1" type instead of "sselog". Remove explicit "memory"
725 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
728 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
731 2018-04-17 Martin Jambor <mjambor@suse.cz>
734 * ipa-cp.c (create_specialized_node): Call
735 expand_all_artificial_thunks if necessary.
737 2018-04-17 Martin Liska <mliska@suse.cz>
740 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
741 in message, remote space in between '_G' and '('.
743 2018-04-17 Jakub Jelinek <jakub@redhat.com>
746 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
747 avx512f_vmcmp<mode>3<round_saeonly_name>,
748 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
749 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
750 avx512f_rndscale<mode><round_saeonly_name>,
751 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
752 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
753 Use %<iptr>2 instead of %2 for -masm=intel.
754 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
755 avx512f_vcvttss2usi<round_saeonly_name>,
756 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
758 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
759 avx512f_vcvttsd2usi<round_saeonly_name>,
760 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
761 Use %q1 instead of %1 for -masm=intel.
762 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
763 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
764 of %3 for -masm=intel.
765 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
767 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
769 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
771 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
772 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
774 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
776 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
777 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
778 %g1 and one with %0 and %1.
779 (avx512er_vmrcp28<mode><round_saeonly_name>,
780 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
782 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
783 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
784 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
785 of %0 and %{%4%} for -masm=intel.
786 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
787 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
788 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
789 order of %0 and %{%5%}%{z%} for -masm=intel.
791 2018-04-17 Jan Hubicka <jh@suse.cz>
794 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
796 2018-04-17 Martin Liska <mliska@suse.cz>
799 * multiple_target.c (create_dispatcher_calls): Set apostrophes
800 for target_clone error message. Make default implementation
801 clone to be a local declaration.
802 (separate_attrs): Add new argument and check for an empty
804 (expand_target_clones): Handle it.
805 (ipa_target_clone): Make redirection just for target_clones
808 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
809 Tom de Vries <tom@codesourcery.com>
812 * omp-expand.c (expand_oacc_for): Add dummy false branch for
813 tiled basic blocks without omp continue statements.
815 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
818 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
819 vec_extract expression as having side effects to make sure it gets
822 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
825 * config/i386/i386.c (get_builtin_code_for_version): Check
828 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
831 * gcc/config.gcc: Support "skylake".
832 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
834 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
835 (processor_target_table): Add "skylake".
836 (ix86_option_override_internal): Add "skylake".
837 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
838 PROCESSOR_CANNONLAKE.
839 (get_builtin_code_for_version): Fix priority for
840 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
841 PROCESSOR_SKYLAKE-AVX512.
842 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
843 (processor_type): Add PROCESSOR_SKYLAKE.
845 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
846 Jason Merrill <jason@redhat.com>
849 * convert.c (convert_to_integer_1): Use direct recursion for
850 enumeral types and types with a precision less than the number
851 of bits in their mode.
853 2018-04-16 Julia Koval <julia.koval@intel.com>
856 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
857 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
859 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
862 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
863 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
864 and -mno-direct-move.
866 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
869 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
870 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
871 Ensure negative shifts result in {0}.
873 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
875 PR rtl-optimization/79916
876 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
877 regs (if any) to define how to gnerate SD moves when LRA is in
880 2018-04-13 Jakub Jelinek <jakub@redhat.com>
882 PR rtl-optimization/85393
883 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
884 * except.c (expand_dw2_landing_pad_for_region): Make static.
885 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
886 a label and unconditional jump to old_bb, rather than
887 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
890 PR rtl-optimization/85376
891 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
892 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
893 instead of a specific value.
895 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
896 Bin Cheng <bin.cheng@arm.com>
898 PR tree-optimization/82965
899 PR tree-optimization/83991
900 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
901 by_profile_only parameter.
902 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
903 information if the loop was predicted to iterate too many times.
904 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
906 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
909 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
912 2018-04-13 Martin Liska <mliska@suse.cz>
913 Jakub Jelinek <jakub@redhat.com>
916 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
917 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
918 * builtins.c (expand_builtin_memory_copy_args): Use
919 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
920 handle dest_addr == pc_rtx.
922 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
925 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
926 asked to not generate direct moves.
927 (fix_trunc<mode>si2_stfiwx): Similar.
928 (fix_trunc<mode>si2_internal): Similar.
930 2018-04-12 Jakub Jelinek <jakub@redhat.com>
933 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
934 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
935 lookup if dest in some wider mode is known to be const0_rtx and
936 if so, record permanent equivalence for it to be ZERO_EXTEND of
937 the narrower mode destination.
939 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
941 * lto-streamer-out.c (output_function): Revert 259346.
942 * omp-expand.c (expand_oacc_for): Likewise.
944 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
946 PR rtl-optimization/85354
947 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
948 * sel-sched.c (sel_global_init): ... here.
950 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
953 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
954 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
955 mode for PE-COFF targets.
956 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
957 (i386_pe_asm_lto_end): Likewise.
958 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
959 (TARGET_ASM_LTO_END): Likewise.
960 * config/i386/winnt.c (saved_debug_info_level): New static variable.
961 (i386_pe_asm_lto_start): New function.
962 (i386_pe_asm_lto_end): Likewise.
964 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
965 Richard Biener <rguenther@suse.de>
968 * lto-streamer-out.c (output_function): Fix CFG loop state before
970 * omp-expand.c (expand_oacc_for): Handle calls to internal
971 functions like regular functions.
973 2018-04-12 Richard Biener <rguenther@suse.de>
976 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
977 for the early LTO debug to properly generate references to it
978 during DIE emission. Do not re-use that for the skeleton for
980 (dwarf2out_early_finish): Likewise.
982 2018-04-12 Jakub Jelinek <jakub@redhat.com>
986 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
987 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
988 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
989 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
990 and output is a reg, avoid creating invalid lowpart subreg, but
991 instead split into a 512-bit move. Don't split if not AVX512VL,
992 input is xmm16+ reg and output is a mem.
993 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
994 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
995 xmm16+ reg and output is a mem.
997 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
999 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
1000 also for flag_dwarf2_cfi_asm.
1002 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1004 PR rtl-optimization/85342
1005 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
1006 a bool scalar var inside of the loop instead. Don't try to update
1007 recog_data.operand after failed apply_change_group.
1009 2018-04-12 Tom de Vries <tom@codesourcery.com>
1012 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
1013 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
1014 array with flexible array member as array without given dimension.
1015 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
1016 argument for undefined param to true.
1018 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
1021 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1022 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
1023 from PowerPC section.
1024 * config/rs6000/sysv4.opt (mcall-): Improve help text.
1025 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
1026 help text that is too long.
1027 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
1028 help text that is too long.
1029 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
1030 help text that is too long.
1032 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
1034 * config/alpha/alpha.md (stack_probe_internal): Rename
1035 from "probe_stack". Update all callers.
1037 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1039 PR rtl-optimization/84566
1040 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
1041 sched_macro_fuse_insns.
1043 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1046 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
1047 (compute_block_dependences): ... from here.
1049 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1051 PR tree-optimization/85331
1052 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
1053 from int to HOST_WIDE_INT.
1055 2018-04-11 Martin Jambor <mjambor@suse.cz>
1058 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
1059 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
1060 not the same as the source val.
1061 (cgraph_edge_brings_value_p): New parameter.
1062 (gather_edges_for_value): Pass destination value to
1063 cgraph_edge_brings_value_p.
1064 (perhaps_add_new_callers): Likewise.
1065 (get_info_about_necessary_edges): Likewise and exclude values brought
1066 only by self-recursive edges.
1067 (create_specialized_node): Redirect only clones of self-calling edges.
1068 (+self_recursive_pass_through_p): New function.
1069 (find_more_scalar_values_for_callers_subset): Use it.
1070 (find_aggregate_values_for_callers_subset): Likewise.
1071 (known_aggs_to_agg_replacement_list): Removed.
1072 (decide_whether_version_node): Re-calculate known constants for all
1073 remaining context clones.
1075 2018-04-11 Richard Biener <rguenther@suse.de>
1078 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
1079 from early DWARF output.
1080 (dwarf2out_early_finish): Output line info unconditionally into
1081 early DWARF and add reference to it.
1083 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1086 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
1087 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
1088 other than V2DFmode using iptr mode attribute.
1089 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
1091 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1093 PR rtl-optimization/84659
1094 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
1096 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1099 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
1101 (output_loc_list): Pass address of a dummy size variable even in the
1102 locview handling loop.
1103 (index_location_lists): Add comment on why skip_loc_list_entry can't
1106 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
1109 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
1112 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
1115 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1116 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
1117 and -mstring-compare-inline-limit.
1119 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1122 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
1123 for stack clash protection in a register whenever we need it to be in
1126 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1128 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
1129 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
1131 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1134 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
1136 (mlong-double-): Ditto.
1137 * config/rs6000/sysv4.opt (msdata=): Ditto.
1138 (mtls-size=): Ditto.
1140 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
1142 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1143 erroneous entries for
1144 "vector int vec_ldl (int, long int *)", and
1145 "vector unsigned int vec_ldl (int, unsigned long int *)".
1146 Add comments and entries for
1147 "vector bool char vec_ldl (int, bool char *)",
1148 "vector bool short vec_ldl (int, bool short *)",
1149 "vector bool int vec_ldl (int, bool int *)",
1150 "vector bool long long vec_ldl (int, bool long long *)",
1151 "vector pixel vec_ldl (int, pixel *)",
1152 "vector long long vec_ldl (int, long long *)",
1153 "vector unsigned long long vec_ldl (int, unsigned long long *)".
1154 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
1155 type tree bool_long_long_type_node and correct definition of
1156 bool_V2DI_type_node to make reference to this new type tree.
1157 (rs6000_mangle_type): Replace erroneous reference to
1158 bool_long_type_node with bool_long_long_type_node.
1159 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
1160 comments to emphasize sign distinctions for char and int types and
1161 replace RS6000_BTI_bool_long constant with
1162 RS6000_BTI_bool_long_long constant. Also add comment to restrict
1163 use of RS6000_BTI_pixel.
1164 (bool_long_type_node): Remove this macro definition.
1165 (bool_long_long_type_node): New macro definition
1167 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1169 PR rtl-optimization/85300
1170 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
1171 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
1172 simplify_unary_operation fails.
1174 2018-04-10 Martin Liska <mliska@suse.cz>
1176 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
1177 cgraph_edge and ipa_ref.
1179 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1183 * config/i386/sse.md
1184 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
1185 computation of the VEC_MERGE selector from mask.
1186 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
1187 Fix decoding of the VEC_MERGE selector into mask.
1189 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1191 PR tree-optimization/85286
1192 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
1194 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1196 * final.c (final_1): Set insn_last_address as well as
1197 insn_current_address.
1199 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1202 * explow.c (emit_stack_probe): Call validize_mem on memory location
1203 before passing it to gen_probe_stack. Create address operand and
1204 legitimize it for the probe_stack_address case.
1206 2018-04-09 Jan Hubicka <jh@suse.cz>
1209 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
1210 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
1211 * tree.c (free_lang_data_in_type): Fix handling of binfos;
1213 (free_lang_data): Rebuild type inheritance graph.
1215 2018-04-09 Martin Sebor <msebor@redhat.com>
1217 * invoke.texi (-finline-small-functions): Mention other optimization
1219 (-findirect-inlining, -fpartial-inlining): Same.
1220 (-finline-functions-called-once): Same.
1221 (-freorder-blocks-and-partition): Same.
1223 2018-04-09 Jan Hubicka <jh@suse.cz>
1226 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
1227 jumps; choose last target that matches the criteria (i.e.
1228 no partition changes for non-crossing jumps).
1229 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
1230 support for redirecting crossing jumps to non-crossing.
1232 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
1234 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
1235 also for naked functions.
1237 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
1239 * config/arc/arc.md (add_shift): New pattern.
1240 (add_shift2): Likewise.
1241 (sub_shift): Likewise.
1242 (sub_shift_cmp0_noout): Likewise.
1243 (compare_si_ashiftsi): Likewise.
1244 (xbfu_cmp0_noout): New combine pattern.
1245 (xbfu_cmp0"): Likewise.
1246 (movsi_set_cc_insn): Place the predicable variant first.
1247 (commutative_binary_cmp0_noout): Remove clobber.
1248 (commutative_binary_cmp0): New pattern.
1249 (noncommutative_binary_cmp0): Likewise.
1250 (noncommutative_binary_cmp0_noout): Likewise.
1251 (noncommutative_binary_comparison_result_used): Removed.
1252 (rsub_cmp0): New pattern.
1253 (rsub_cmp0_noout): Likewise.
1254 (extzvsi): Changed, keep only meaningful variants.
1255 (SQH, SEZ): New iterators.
1256 (SQH_postfix): New mode attribute.
1257 (SEZ_prefix): New code attribute.
1258 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
1259 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
1260 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
1262 (noncommutative_operator): Check the availability of barrel
1265 2018-04-09 Richard Biener <rguenther@suse.de>
1267 PR tree-optimization/85284
1268 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
1269 Only use the niter constraining form of simple_iv when the exit
1272 2018-04-09 Tom de Vries <tom@codesourcery.com>
1275 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1276 (define_expand "*memory_barrier"): New define_expand.
1277 (define_insn "memory_barrier"): New insn.
1279 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1281 PR rtl-optimization/80463
1282 PR rtl-optimization/83972
1283 PR rtl-optimization/83480
1285 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
1286 correct producer for the insn.
1287 (tidy_control_flow): Fixup seqnos in case of debug insns.
1289 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1291 PR rtl-optimization/83913
1293 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
1294 different sched-times when merging exprs.
1296 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1298 PR rtl-optimization/83962
1300 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
1301 tidy_fallthru_edge and tidy_control_flow.
1303 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1305 PR rtl-optimization/83530
1307 * sel-sched.c (force_next_insn): New global variable.
1308 (remove_insn_for_debug): When force_next_insn is true, also leave only
1309 next insn in the ready list.
1310 (sel_sched_region): When the region wasn't scheduled, make another pass
1311 over it with force_next_insn set to 1.
1313 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1315 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
1317 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1318 for interrupt control.
1319 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1320 functions for interrupt control.
1321 * config/nds32/nds32-intrinsic.md: Likewise.
1322 * config/nds32/nds32_intrinsic.h: Likewise.
1323 * config/nds32/nds32.h (nds32_builtins): Likewise.
1325 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1327 * config/nds32/nds32.c (nds32_init_machine_status,
1328 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
1329 strict_aligned_p field.
1330 (nds32_expand_to_rtl_hook): New function.
1331 (TARGET_EXPAND_TO_RTL_HOOK): Define.
1332 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
1334 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1335 Chung-Ju Wu <jasonwucj@gmail.com>
1337 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
1338 * config/nds32/nds32-n7.md: New file.
1339 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
1340 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
1342 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
1343 * config/nds32/nds32.md (pipeline_model): Add n7.
1344 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
1345 * config/nds32/pipelines.md: Include n7 settings.
1347 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1348 Chung-Ju Wu <jasonwucj@gmail.com>
1350 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
1351 * config/nds32/nds32-e8.md: New file.
1352 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
1353 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
1355 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
1356 * config/nds32/nds32.md (pipeline_model): Add e8.
1357 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
1358 * config/nds32/pipelines.md: Include e8 settings.
1360 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1361 Chung-Ju Wu <jasonwucj@gmail.com>
1363 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
1364 * config/nds32/nds32-n8.md: New file.
1365 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
1366 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
1368 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
1369 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
1370 * config/nds32/nds32.md (pipeline_model): Add n8.
1371 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
1372 * config/nds32/pipelines.md: Include n8 settings.
1374 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1375 Chung-Ju Wu <jasonwucj@gmail.com>
1377 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
1378 * config/nds32/nds32-n9-2r1w.md: New file.
1379 * config/nds32/nds32-n9-3r2w.md: New file.
1380 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
1381 nds32_register_ports): New or modify for cpu n9.
1382 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
1384 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
1385 * config/nds32/nds32-utils.c: New file.
1386 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
1387 TARGET_MUL_SLOW): Define.
1388 * config/nds32/nds32.md (pipeline_model): New attribute.
1389 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
1390 New options that support cpu n9.
1391 * config/nds32/pipelines.md: Include n9 settings.
1392 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
1394 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1396 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
1397 information if necessary.
1398 (output_cond_branch_compare_zero): Likewise.
1399 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
1400 (nds32_target_alignment): Refine for alignment.
1401 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
1402 (FUNCTION_BOUNDARY): Modify.
1403 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
1405 * config/nds32/nds32.opt (malways-align, malign-functions): New.
1407 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1409 * config/nds32/constants.md (unspec_volatile_element): Add values for
1410 TLB operation and data prefetch.
1411 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1412 functions for TLB operation and data prefetch.
1413 * config/nds32/nds32-intrinsic.md: Likewise.
1414 * config/nds32/nds32_intrinsic.h: Likewise.
1415 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
1416 (nds32_print_operand): Likewise.
1417 * config/nds32/nds32.h (nds32_builtins): Likewise.
1419 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
1420 Andrew Pinski <pinsika@gcc.gnu.org>
1423 * match.pd: Use constant_boolean_node of correct type instead of
1424 boolean_true_node or boolean_false_node for simplifying
1425 pointer comparisons to zero.
1427 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1429 PR tree-optimization/80021
1430 * tree.c (verify_type_variant): Make error call in verify_variant_match
1431 translatable and remove final full stop.
1433 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1435 * config/nds32/constants.md (unspec_volatile_element): Add
1436 UNSPEC_VOLATILE_EH_RETURN.
1437 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
1438 nds32_output_stack_pop): Support dwarf exception handling process.
1439 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
1440 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
1441 exception handling process.
1442 (nds32_compute_stack_frame): Likewise.
1443 (nds32_return_addr_rtx): Likewise.
1444 (nds32_initial_elimination_offset): Likewise.
1445 (nds32_expand_prologue): Likewise.
1446 (nds32_expand_epilogue): Likewise.
1447 (nds32_dynamic_chain_address): New function.
1448 * config/nds32/nds32.h (machine_function): Add fields for dwarf
1450 (DYNAMIC_CHAIN_ADDRESS): Define.
1451 (EH_RETURN_DATA_REGNO): Define.
1452 (EH_RETURN_STACKADJ_RTX): Define.
1453 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
1454 patterns for dwarf exception handling.
1456 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1458 * config/nds32/nds32.h: Clean up obsolete macros.
1460 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1462 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1463 Add enum values for particular instructions.
1464 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1465 particular intrinsic functions.
1466 * config/nds32/nds32-intrinsic.md: Likewise.
1467 * config/nds32/nds32_intrinsic.h: Likewise.
1468 * config/nds32/nds32.h (nds32_builtins): Likewise.
1469 * config/nds32/nds32.md (type): Add pbsad and pbsada.
1470 (btst, ave): New patterns for particular instructions.
1472 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1474 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1475 Add enum values for atomic load/store and memory sync.
1476 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
1478 * config/nds32/nds32-intrinsic.md: Likewise.
1479 * config/nds32/nds32_intrinsic.h: Likewise.
1480 * config/nds32/nds32.h (nds32_builtins): Likewise.
1482 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1484 PR tree-optimization/85257
1485 * fold-const.c (native_encode_vector): If not all elts could fit
1486 and off is -1, return 0 rather than offset.
1487 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
1488 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
1489 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
1490 adjust buffer in native_interpret_expr call.
1492 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1494 * config/nds32/constants.md (unspec_volatile_element): Add cache
1495 control enum values.
1496 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
1497 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
1498 * config/nds32/nds32.c (nds32_cctl_names): New.
1499 (nds32_print_operand): Handle cache control register names.
1500 * config/nds32/nds32.h (nds32_builtins): New enum values.
1501 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
1503 * config/nds32/nds32.md (type): Add mmu.
1504 * config/nds32/pipelines.md (simple_insn): Add mmu.
1506 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1508 * config/nds32/nds32.md (type): Remove call.
1509 * config/nds32/pipelines.md (simple_insn): Likewise.
1511 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1513 * config/nds32/constants.md (unspec_volatile_element): Add
1514 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
1515 UNSPEC_VOLATILE_FMFCFG.
1516 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
1517 description for fmfcfg and fmfcsr.
1518 (bdesc_1arg): Add fmtcsr.
1519 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
1520 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
1521 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
1522 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
1523 unspec_fmfcfg): New patterns.
1524 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
1525 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
1526 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
1527 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
1528 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
1529 __nds32__fmfcfg): Define.
1531 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1533 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
1534 intrinsic register names.
1535 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
1536 intrinsic register enum values and macros.
1538 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1540 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
1541 for load/store addressing form.
1542 (nds32_print_operand_address): Likewise.
1544 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
1547 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
1548 based on LABEL_REF. Remove useless assertion.
1549 (pic_address_needs_scratch): Fix formatting.
1550 (sparc_legitimize_pic_address): Minor tweaks.
1551 (sparc_delegitimize_address): Adjust assertion accordingly.
1552 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
1553 into symbolic_operand.
1554 (movsi_high_pic_label_ref): Likewise.
1555 (movsi_lo_sum_pic_label_ref): Likewise.
1556 (movdi_pic_label_ref): Likewise.
1557 (movdi_high_pic_label_ref): Likewise.
1558 (movdi_lo_sum_pic_label_ref): Likewise.
1560 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
1562 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
1563 custom LIB_SPEC setup.
1565 2018-04-06 Ruslan Bukin <br@bsdpad.com>
1566 Kito Cheng <kito.cheng@gmail.com>
1568 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
1569 * config/riscv/freebsd.h: New.
1571 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1573 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
1574 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
1577 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1578 Kito Cheng <kito.cheng@gmail.com>
1580 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
1581 nds32_output_call, nds32_symbol_binds_local_p): New functions.
1582 * config/nds32/nds32-protos.h (nds32_output_call,
1583 nds32_output_return): Declare.
1584 * config/nds32/nds32.md: Refine all the call and return patterns.
1586 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1589 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
1590 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
1592 PR rtl-optimization/84872
1593 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
1594 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
1597 2018-04-06 Tamar Christina <tamar.christina@arm.com>
1599 * expr.c (copy_blkmode_to_reg): Revert 254862.
1600 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
1602 2018-04-06 Richard Biener <rguenther@suse.de>
1605 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
1606 after seeing a component reference with an adjacent field. Treat
1607 refs to arrays at struct end of external decls similar to
1608 refs to unconstrained commons.
1610 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1613 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
1614 look through SAVE_EXPRs with non-side-effects argument. Adjust
1616 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
1619 2018-04-06 Richard Biener <rguenther@suse.de>
1622 * alias.c (find_base_term): New wrapper around find_base_term
1623 unwinding CSELIB_VAL_PTR changes.
1624 (find_base_term): Do not restore CSELIB_VAL_PTR during the
1627 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1629 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
1631 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
1632 constant definitions.
1633 ("nop"): lr 0,0 -> nopr r0
1634 ("nop_lr0", "nop_lr1"): New insn definitions.
1636 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1638 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
1639 NDS32_V3PUSH_AVAILABLE_P macro.
1641 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
1642 Chung-Ju Wu <jasonwucj@gmail.com>
1644 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
1645 (nds32*-*-*): Add float and fpu_config into supported_defaults.
1646 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
1647 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
1648 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
1649 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
1650 * config/nds32/constraints.md: New constraints and checking for hard
1651 float configuration.
1652 * config/nds32/iterators.md: New mode iterator and attribute for hard
1653 float configuration.
1654 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
1656 * config/nds32/nds32-fpu.md: New file.
1657 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
1658 deal with hard float code generation.
1659 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
1661 (abi_type, float_reg_number): New enum type.
1662 * config/nds32/nds32-predicates.c: New predicates for hard float.
1663 * config/nds32/nds32-protos.h: Declare functions for hard float.
1664 * config/nds32/nds32.c: Implementation for hard float configuration.
1665 * config/nds32/nds32.h: Definitions for hard float configuration.
1666 * config/nds32/nds32.md: Include hard float machine description and
1667 modify patterns for hard float configuration.
1668 * config/nds32/nds32.opt: New options for hard float configuration.
1669 * config/nds32/predicates.md: New predicates for hard float
1672 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1674 * common/config/nds32/nds32-common.c
1675 (nds32_option_optimization_table): Enable -mreleax-hint by default.
1677 2018-04-05 Jakub Jelinek <jakub@redhat.com>
1680 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
1681 CONSTRUCTOR_ELT (ctor, ...)->value.
1683 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
1686 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
1688 2018-04-05 Tom de Vries <tom@codesourcery.com>
1691 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
1694 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1695 Kito Cheng <kito.cheng@gmail.com>
1697 * config/nds32/constraints.md (U33): Fine-tune checking condition.
1698 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
1699 * config/nds32/nds32.h (nds32_16bit_address_type): Add
1700 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
1702 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1703 Kito Cheng <kito.cheng@gmail.com>
1705 * config/nds32/constraints.md (Ufe): New memory constraint.
1706 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
1707 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
1708 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
1710 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
1711 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
1713 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1715 * config/nds32/nds32.md: Use optimize_size in the condition for
1716 alu-shift instructions.
1718 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1720 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
1722 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1724 * config/nds32/nds32.md (negsi2): Refine pattern.
1726 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1727 Chung-Ju Wu <jasonwucj@gmail.com>
1729 * config/nds32/iterators.md (shift_rotate): New code iterator.
1730 (shift): New code attribute.
1731 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
1732 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
1733 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
1734 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
1735 bit-wise operations.
1736 (andsi3, *andsi3): Ditto.
1737 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
1738 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
1739 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
1740 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
1741 nds32_ior_operand, nds32_xor_operand): New predicates.
1743 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1745 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
1746 (addsi3, subsi3): ... this.
1748 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1750 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
1752 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1754 * config/nds32/nds32.md: Adjust indention.
1756 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1758 * config/nds32/nds32.md (feature): New attribute.
1760 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1762 * config/nds32/nds32.md (subtype): New attribute.
1764 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
1767 * config/arm/arm-builtins.c (arm_expand_builtin): Change
1768 expansion to perform a bitwise AND of the argument followed by a
1769 boolean negation of the result.
1771 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
1773 PR rtl-optimization/84878
1774 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
1775 the basic block. Assert the use reference is not artificial and that
1776 it has an associated insn.
1778 2018-04-04 Michael Matz <matz@suse.de>
1780 * builtins.c (compute_objsize): Pass correct operand
1781 to array_at_struct_end_p.
1783 2018-04-04 Richard Biener <rguenther@suse.de>
1786 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
1787 from contexts for DINFO_LEVEL_TERSE and below.
1789 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1791 * config/nds32/nds32-doubleword.md (move_<mode>): Require
1792 resiter_operand condition.
1793 * config/nds32/nds32.md (*move<mode>): Ditto.
1795 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1796 Monk Chiang <sh.chiang04@gmail.com>
1798 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
1800 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1802 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
1804 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1805 Kito Cheng <kito.cheng@gmail.com>
1807 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
1808 nds32_cond_code_str, output_cond_branch,
1809 output_cond_branch_compare_zero, nds32_expand_cbranch,
1810 nds32_expand_cstore, nds32_expand_movcc,
1811 nds32_output_cbranchsi4_equality_zero,
1812 nds32_output_cbranchsi4_equality_reg,
1813 nds32_output_cbranchsi4_equality_reg_or_const_int,
1814 nds32_output_cbranchsi4_greater_less_zero: New functions.
1815 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
1816 nds32_expand_cstore, nds32_expand_movcc,
1817 nds32_output_cbranchsi4_equality_zero,
1818 nds32_output_cbranchsi4_equality_reg,
1819 nds32_output_cbranchsi4_equality_reg_or_const_int,
1820 nds32_output_cbranchsi4_greater_less_zero): Declare.
1821 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
1822 nds32_rimm11s_operand): New predicates.
1823 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
1824 * config/nds32/nds32.md: Rewrite all the branch and conditional move
1827 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1829 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
1830 * config/nds32/nds32.md: Ditto.
1831 * config/nds32/pipelines.md: Ditto.
1833 2018-04-04 Richard Biener <rguenther@suse.de>
1835 PR tree-optimization/85168
1836 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
1837 propagating abnormals.
1839 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1841 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
1843 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1844 Kito Cheng <kito.cheng@gmail.com>
1846 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
1847 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
1848 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
1849 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
1850 * config/nds32/nds32.md (sibcall_internal): New.
1851 (sibcall_register): Remove.
1852 (sibcall_immediate): Remove.
1853 (sibcall_value_internal): New.
1854 (sibcall_value_register): Remove.
1855 (sibcall_value_immediate): Remove.
1856 * config/nds32/predicates.md (nds32_general_register_operand): New.
1857 (nds32_call_address_operand): New.
1859 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1861 PR rtl-optimization/85167
1862 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
1863 bb_defs if *split_p, instead preinitialize it to NULL.
1865 PR tree-optimization/85156
1866 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
1867 evaluating the argument multiple times.
1869 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
1871 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
1873 (_mm_cvtpd_ps): Likewise.
1874 (_mm_cvttpd_epi32): Likewise.
1875 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
1876 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
1877 vector, pixel, and bool following altivec.h include.
1879 2018-04-03 Martin Sebor <msebor@redhat.com>
1881 * doc/extend.texi (Common Function Attributes): Clarify.
1882 (const attribute): Likewise.
1883 (pure attribute): Likewise.
1885 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1888 * config/i386/i386.c (ix86_expand_vector_set): Use
1889 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
1891 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
1893 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
1894 instructions when changing rounding bits to preserve precision bits
1895 in the x87 control word.
1897 2018-04-03 Martin Liska <mliska@suse.cz>
1899 PR tree-optimization/82491
1900 * rtl.h (strip_offset_and_add): Replace += suboffset with
1901 poly_uint64 () + suboffset.
1903 2018-03-29 Martin Liska <mliska@suse.cz>
1904 Martin Jambor <mjambor@suse.cz>
1907 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
1908 param_type is not an integral or pointer type.
1910 2018-04-03 Richard Biener <rguenther@suse.de>
1912 * sese.h (recompute_all_dominators): Remove.
1914 2018-04-02 Martin Sebor <msebor@redhat.com>
1916 * doc/invoke.texi (-Wrestrict): Fix typos.
1918 2018-04-02 Jim Wilson <jimw@sifive.com>
1920 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
1921 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
1922 (<optab>di3, <optab>si3_extend): Likewise.
1923 (<optab>si3_mask, <optab>si3_mask_1): New.
1924 (<optab>di3_mask, <optab>di3_mask_1): New.
1925 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
1926 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
1927 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
1929 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
1931 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
1934 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
1936 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
1937 (nds32_canonicalize_comparison): New function.
1939 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1940 Kito Cheng <kito.cheng@gmail.com>
1941 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1943 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
1944 * config/nds32/constants.md (unspec_volatile_element): Add
1945 UNSPEC_VOLATILE_RELAX_GROUP.
1946 * config/nds32/nds32-relax-opt.c: New file.
1947 * config/nds32/nds32-predicates.c
1948 (nds32_symbol_load_store_p): New function.
1949 * config/nds32/nds32-protos.h
1950 (nds32_symbol_load_store_p): Declare function.
1951 (make_pass_nds32_relax_opt): Declare new rtl pass function.
1952 * config/nds32/nds32.c
1953 (nds32_register_pass): New function to register pass.
1954 (nds32_register_passes): New function to register passes.
1955 * config/nds32/nds32.md (relax_group): New pattern.
1956 * config/nds32/nds32.opt (mrelax-hint): New option.
1957 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
1959 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
1961 * config/nds32/t-nds32: Modify files dependency.
1963 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1965 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
1966 (PROFILE_HOOK): Define its implementation.
1968 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1970 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
1971 type and 32-bit size.
1973 2018-04-01 Jakub Jelinek <jakub@redhat.com>
1976 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
1977 (V_128_256): New mode iterator.
1978 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
1979 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
1980 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
1982 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
1983 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
1985 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
1988 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
1989 NaN inputs correctly.
1991 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
1994 * config/rs6000/vsx.md (??r): New mode attribute.
1995 (*vsx_mov<mode>_64bit): Use it.
1996 (*vsx_mov<mode>_32bit): Likewise.
1998 2018-03-30 Martin Sebor <msebor@redhat.com>
2000 PR tree-optimization/84818
2001 * builtins.c (check_access): Use warning_n.
2003 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2006 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
2008 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
2011 2018-03-30 Julia Koval <julia.koval@intel.com>
2014 * x86-tune.def (movx, partial_reg_dependency): Enable for
2017 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
2020 * lra-constraints.c (process_alt_operands): Move setting
2021 this_alternative_matches below.
2023 2018-03-29 Martin Liska <mliska@suse.cz>
2026 * doc/invoke.texi: Document how LTO works with debug info.
2027 Describe auto-load support of binutils. Mention 'x86-64'
2028 as valid option value of -march option.
2030 2018-03-29 Jakub Jelinek <jakub@redhat.com>
2032 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
2035 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
2036 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
2037 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
2040 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2043 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
2044 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
2045 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
2046 for RS6000_BTM_POWERPC64.
2047 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
2048 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
2049 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
2054 2018-03-28 Carl Love <cel@us.ibm.com>
2057 2017-09-27 Carl Love <cel@us.ibm.com>
2059 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
2060 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
2061 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
2064 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2066 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
2067 instead of __vector bool.
2068 (_mm_max_pu8): Likewise.
2069 (_mm_min_pi16): Likewise.
2071 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2074 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
2075 (DIVWEUO): Likewise.
2077 (DIVDEUO): Likewise.
2078 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
2079 DIVWEUO and DIVDEUO.
2080 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
2081 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
2082 (div_extend): Likewise.
2083 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
2085 (__builtin_divweuo): Likewise.
2086 (__builtin_divdeo): Likewise.
2087 (__builtin_divdeuo): Likewise.
2089 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2092 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
2093 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
2095 PR tree-optimization/82004
2096 * gimple-match-head.c (optimize_pow_to_exp): New function.
2097 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
2098 Don't fold to exp if optimize_pow_to_exp is false.
2100 2018-03-28 Martin Liska <mliska@suse.cz>
2103 * calls.c (initialize_argument_information): Fix trailing space.
2104 * common.opt: Fix typo and provide better explanation for
2105 -fsanitize-coverage option.
2106 * config/i386/i386.opt: Fix typo.
2108 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2109 Martin Liska <mliska@suse.cz>
2112 * gimplify.c (asan_poison_variable): Don't do the check for
2113 gimplify_omp_ctxp here.
2114 (gimplify_decl_expr): Do it here.
2115 (gimplify_target_expr): Likewise.
2117 2018-03-28 Martin Liska <mliska@suse.cz>
2120 * config/i386/i386.c (ix86_function_arg_advance): Do not call
2121 chkp_type_bounds_count if MPX is not enabled.
2123 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
2125 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
2127 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
2130 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
2131 function to create the function decl for complex long double
2132 multiply and divide for -mabi=ieeelongdouble.
2133 (init_float128_ieee): Call it.
2135 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
2138 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
2139 -fcf-protection=branch -mibt.
2140 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
2142 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2145 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
2147 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
2150 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
2151 extern array declarations.
2153 2018-03-27 Richard Biener <rguenther@suse.de>
2156 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
2157 explicit single_use checks.
2159 2018-03-27 Richard Biener <rguenther@suse.de>
2161 PR tree-optimization/85082
2162 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
2165 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2167 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
2168 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
2169 Turn on fasynchronous-unwind-tables and funwind-tables.
2171 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
2174 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
2175 (*bmi_blsr_<mode>_ccz): Ditto.
2177 2018-03-26 Tom de Vries <tom@codesourcery.com>
2179 PR tree-optimization/85063
2180 * omp-general.c (offloading_function_p): New function. Factor out
2182 * omp-offload.c (pass_omp_target_link::gate): ... here.
2183 * omp-general.h (offloading_function_p): Declare.
2184 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
2185 with attribute omp declare target for offloading functions.
2187 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
2189 PR tree-optimization/84005
2190 * tree-data-ref.h (get_base_for_alignment): Declare.
2191 * tree-data-ref.c (get_base_for_alignment_1): New function.
2192 (get_base_for_alignment): Likewise.
2193 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
2194 get_base_for_alignment to find a suitable base object, instead
2195 of always using drb->base_address.
2197 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2200 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
2201 known size by default.
2203 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
2206 * lra-constraints.c (process_alt_operands): Don't match BLKmode
2207 and non BLKmode operands.
2209 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2212 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
2213 Clean up attributes.
2215 2018-03-23 Richard Biener <rguenther@suse.de>
2218 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
2219 we are going to emit early debug for LTO.
2221 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2224 * function.c (match_asm_constraints_1): Don't optimize if input
2225 doesn't satisfy general_operand predicate for output's mode.
2228 * alias.c (write_dependence_p): Don't require for x_canonicalized
2229 non-VOIDmode if x has VOIDmode.
2232 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
2233 just don't try to optimize it rather than assert it never happens.
2235 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2237 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
2238 macro expansions for definition of ST_INTERNAL_<mode> and
2239 LD_INTERNAL_<mode> builtins.
2240 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
2242 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
2244 (altivec_expand_st_builtin): Likewise.
2245 (altivec_expand_builtin): Remove calls to deleted functions.
2246 (rs6000_address_for_altivec): Delete this function.
2247 * config/rs6000/vector.md: Remove expands for
2248 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
2250 2018-03-22 Sudakshina Das <sudi.das@arm.com>
2253 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
2254 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
2255 re-computing once computed.
2256 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
2257 (arm_init_machine_status): Initialize
2258 machine->static_chain_stack_bytes.
2260 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2263 * doc/extend.texi: Add four new prototypes for vec_ld.
2264 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
2265 definitions for more logical presentation.
2266 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
2267 entries for V1TI variants of __builtin_altivec_ld builtin.
2268 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
2269 handling of V1TI variant of LVX icode pattern.
2270 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
2271 (rs6000_gimple_fold_builtin): Likewise.
2272 (altivec_init_builtins): Add code to define
2273 __builtin_altivec_lvx_v1ti function.
2275 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2278 * function.c (match_asm_constraints_1): Don't do the optimization
2279 if input isn't a REG, SUBREG, MEM or constant.
2281 2018-03-22 Tom de Vries <tom@codesourcery.com>
2283 PR tree-optimization/84956
2284 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
2285 bb_has_abnormal_pred.
2287 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2290 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
2291 DECL_INITIAL (decl) to decl at the end.
2292 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
2295 2018-03-21 Joseph Myers <joseph@codesourcery.com>
2297 * doc/extend.texi (__builtin_tgmath): Document when complex
2298 integer types are treated as _Complex _Float64.
2300 2018-03-21 Tom de Vries <tom@codesourcery.com>
2302 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
2304 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2306 PR tree-optimization/84960
2307 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
2308 if it is ENTRY block, move them into single succ of ENTRY in that case.
2310 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
2312 PR tree-optimization/84811
2313 * poly-int.h (poly_span_traits): Remove the T3 parameter and
2314 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
2315 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
2316 (known_subrange_p): Update accordingly. Cast each value involved
2317 in the size comparison, rather than casting the result of the
2320 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2322 PR tree-optimization/84982
2323 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
2324 by flipping the least significant bit rather than all bits from
2325 bitpos to bitpos + bitsize - 1.
2327 2018-03-21 Nathan Sidwell <nathan@acm.org>
2329 * doc/extend.texi (Deprecated Features): Remove mention of
2330 long-deleted deprecations.
2332 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2335 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
2336 * configure: Regenerate.
2338 2018-03-21 Tom de Vries <tom@codesourcery.com>
2340 PR tree-optimization/83126
2341 * tree-parloops.c (num_phis): New function.
2342 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
2344 2018-03-21 Nathan Sidwell <nathan@acm.org>
2346 * doc/extend.texi (Deprecated Features): Update deprecated flags,
2347 mention anon-struct/union members and trailing attributes.
2349 2018-03-21 Bin Cheng <bin.cheng@arm.com>
2351 PR tree-optimization/84969
2352 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
2353 builtin memset partitions if they set different rhs values.
2355 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2357 PR rtl-optimization/84989
2358 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
2359 VEC_DUPLICATE with scalar result mode.
2361 2018-03-21 Martin Liska <mliska@suse.cz>
2364 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
2365 not intended return statement.
2367 2018-03-21 Martin Liska <mliska@suse.cz>
2370 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
2371 (chkp_find_bound_slots_1): Limit number of iterations.
2373 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
2376 * Minor grammar fixes for x86 options.
2378 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2381 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
2382 holding REG_CFA_RESTORE notes, instead turn them into a USE.
2384 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
2387 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
2388 (altivec_lvx_<mode>_1op): Likewise.
2389 (altivec_stvx_<mode>_2op): Likewise.
2390 (altivec_stvx_<mode>_1op): Likewise.
2391 (altivec_lvx_<VM2:mode>): New define_expand.
2392 (altivec_stvx_<VM2:mode>): Likewise.
2393 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
2394 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2395 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
2396 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2397 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
2398 (rs6000_gen_lvx): Likewise.
2399 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
2400 (altivec_expand_stv_builtin): Likewise.
2401 (altivec_expand_builtin): Likewise.
2402 * config/rs6000/vector.md: Likewise.
2404 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2407 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
2410 2018-03-20 Richard Biener <rguenther@suse.de>
2413 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
2414 sign-conversions as zero, fall back to standard scalar_stmt
2417 2018-03-20 Martin Liska <mliska@suse.cz>
2420 * predict.c (rebuild_frequencies): Handle case when we have
2421 PROFILE_ABSENT, but flag_guess_branch_prob is false.
2423 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2426 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
2427 flag_section_anchors.
2428 * varasm.c (use_blocks_for_decl_p): Remove hack for
2429 dw2_force_const_mem.
2432 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
2434 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
2435 be created, use lowpart_subreg of operands[0] rather than operands[0]
2437 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
2438 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
2439 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
2440 and n constraint instead of aarch64_shift_imm_di and Usd.
2441 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
2442 (*aarch64_<optab>_reg_minus<mode>3): ... this.
2444 2018-03-20 Sudakshina Das <sudi.das@arm.com>
2447 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
2448 to favor GPR over NEON registers.
2449 (<shift>di3_neon): Likewise.
2451 2018-03-20 Tom de Vries <tom@codesourcery.com>
2454 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
2455 (nvptx_process_pars): Emit bar.sync asap and alap.
2457 2018-03-20 Tom de Vries <tom@codesourcery.com>
2460 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
2461 seen_label if seen_label is already set.
2463 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2466 * config/i386/i386.c (fold_builtin_cpu): For features above 31
2467 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
2468 Use 1U instead of 1. Formatting fixes.
2471 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
2472 instead of TREE_TYPE (s1) for the return value.
2474 2018-03-19 Jakub Jelinek <jakub@redhat.com>
2476 PR tree-optimization/84946
2477 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
2478 bitsize + bitsize in poly_uint64 rather than poly_int64.
2481 * dwarf2asm.c: Include fold-const.c.
2482 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
2483 of decl rather than decl itself.
2485 PR rtl-optimization/84643
2486 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
2488 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
2491 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
2492 calling assemble_variable.
2494 2018-03-19 Sudakshina Das <sudi.das@arm.com>
2497 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
2498 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
2500 2018-03-19 Jim Wilson <jimw@sifive.com>
2503 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
2504 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
2505 (riscv_first_stack_step): Likewise.
2506 (riscv_option_override): Use STACK_BOUNDARY instead of
2508 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
2510 (BIGGEST_ALIGNMENT): Set to 128.
2511 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
2512 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
2515 2018-03-19 Richard Biener <rguenther@suse.de>
2517 PR tree-optimization/84933
2518 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
2519 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
2521 2018-03-19 Richard Biener <rguenther@suse.de>
2523 PR tree-optimization/84859
2524 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
2525 (cond_if_else_store_replacement): Perform sinking operation on
2526 single-store BBs regardless of MAX_STORES_TO_SINK setting.
2527 Generalize what a BB with a single eligible store is.
2529 2018-03-19 Richard Biener <rguenther@suse.de>
2531 PR tree-optimization/84929
2532 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
2533 chrec_is_positive against non-chrec arg.
2535 2018-03-19 Tamar Christina <tamar.christina@arm.com>
2538 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
2540 2018-03-18 Martin Liska <mliska@suse.cz>
2542 PR rtl-optimization/84635
2543 * regrename.c (build_def_use): Use matches_mode only when
2546 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
2548 PR tree-optimization/84913
2549 * tree-vect-loop.c (vectorizable_reduction): Don't try to
2550 vectorize chains of COND_EXPRs.
2552 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2554 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
2556 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2558 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
2560 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2562 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
2564 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2565 Kito Cheng <kito.cheng@gmail.com>
2567 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
2568 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
2569 (nds32_adjust_reg_alloc_order): New function.
2570 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
2572 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
2574 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
2575 nds32_print_operand, nds32_print_operand_address): Use
2576 HOST_WIDE_INT_PRINT_DEC instead.
2578 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2580 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
2582 2018-03-17 Jakub Jelinek <jakub@redhat.com>
2585 * config/i386/i386.c (initial_ix86_tune_features,
2586 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
2588 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
2589 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
2590 rather than 1u << ix86_tune. Formatting fix.
2591 (ix86_option_override_internal): Change ix86_arch_mask from
2592 unsigned int to unsigned HOST_WIDE_INT, initialize to
2593 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
2594 (ix86_function_specific_restore): Likewise.
2596 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2599 * postreload.c (reload_combine_recognize_pattern): Perform
2600 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
2601 truncate_int_for_mode the result for the destination's mode.
2604 * hsa-gen.c (mem_type_for_type): Fix comment typo.
2605 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
2607 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2610 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
2613 * lra-assigns.c (lra_split_hard_reg_for): Don't use
2614 regno_allocno_class_array and sorted_pseudos.
2615 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
2616 insns where regno is used.
2618 2018-03-16 Martin Liska <mliska@suse.cz>
2621 * multiple_target.c (create_dispatcher_calls): Redirect
2622 reference in the symbol table.
2624 2018-03-16 Martin Liska <mliska@suse.cz>
2627 * multiple_target.c (create_dispatcher_calls): Redirect also
2630 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2634 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
2635 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
2636 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
2638 2018-03-16 Julia Koval <julia.koval@intel.com>
2640 * doc/invoke.texi (Skylake Server): Add CLWB.
2641 Cannonlake): Remove CLWB.
2643 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2645 PR tree-optimization/84841
2646 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
2648 (FLOAT_ONE_CONST_TYPE): Define.
2649 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
2650 (sort_by_operand_rank): Put entries with higher constant_type last
2651 rather than first to match comments.
2653 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
2655 * config/nios2/nios2.md (movsi_internal): Fix thinko in
2658 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2661 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
2662 check and use address of target always.
2664 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
2667 * config/i386/i386.c (indirect_thunk_needed): Update comments.
2668 (indirect_thunk_bnd_needed): Likewise.
2669 (indirect_thunks_used): Likewise.
2670 (indirect_thunks_bnd_used): Likewise.
2671 (indirect_return_needed): New.
2672 (indirect_return_bnd_needed): Likewise.
2673 (output_indirect_thunk_function): Add a bool argument for
2675 (output_indirect_thunk_function): Don't generate alias for
2676 function return thunk.
2677 (ix86_code_end): Call output_indirect_thunk_function to generate
2678 function return thunks.
2679 (ix86_output_function_return): Set indirect_return_bnd_needed
2680 and indirect_return_needed instead of indirect_thunk_bnd_needed
2681 and indirect_thunk_needed.
2683 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
2685 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
2686 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
2687 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
2689 2018-03-15 David Malcolm <dmalcolm@redhat.com>
2690 Paul Hua <paul.hua.gm@gmail.com>
2693 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
2695 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
2697 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
2698 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
2701 2018-03-15 Tamar Christina <tamar.christina@arm.com>
2704 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
2705 instead of GET_MODE_SIZE when comparing Units.
2707 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2710 * varasm.c (hash_section): Return an unchangeble hash value
2711 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
2712 Return !aarch64_can_use_per_function_literal_pools_p ().
2714 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2717 * optabs.c (emit_conditional_move): Pass address of cmode's copy
2718 rather than address of cmode as last argument to prepare_cmp_insn.
2720 2018-03-15 Julia Koval <julia.koval@intel.com>
2722 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
2723 F_AVX512VNNI, F_AVX512BITALG): New.
2725 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
2728 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
2729 insn for floating-point loads and stores.
2731 2018-03-14 Carl Love <cel@us.ibm.com>
2733 * config/rs6000/rs6000-c.c: Add macro definitions for
2734 ALTIVEC_BUILTIN_VEC_PERMXOR.
2735 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
2736 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
2737 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
2738 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
2740 * config/doc/extend.texi: Add prototypes for vec_permxor.
2742 2018-03-14 David Malcolm <dmalcolm@redhat.com>
2745 * diagnostic-show-locus.c (class layout_point): Convert m_line
2746 from int to linenum_type.
2747 (line_span::comparator): Use linenum "compare" function when
2748 comparing line numbers.
2749 (test_line_span): New function.
2750 (layout_range::contains_point): Convert param "row" from int to
2752 (layout_range::intersects_line_p): Likewise.
2753 (layout::will_show_line_p): Likewise.
2754 (layout::print_source_line): Likewise.
2755 (layout::should_print_annotation_line_p): Likewise.
2756 (layout::print_annotation_line): Likewise.
2757 (layout::print_leading_fixits): Likewise.
2758 (layout::annotation_line_showed_range_p): Likewise.
2759 (struct line_corrections): Likewise for field m_row.
2760 (line_corrections::line_corrections): Likewise for param "row".
2761 (layout::print_trailing_fixits): Likewise.
2762 (layout::get_state_at_point): Likewise.
2763 (layout::get_x_bound_for_row): Likewise.
2764 (layout::print_line): Likewise.
2765 (diagnostic_show_locus): Likewise for locals "last_line" and
2767 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
2768 * input.c (selftest::test_linenum_comparisons): New function.
2769 (selftest::input_c_tests): Call it.
2770 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
2771 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
2772 * selftest.h (ASSERT_GT): New macro.
2773 (ASSERT_GT_AT): New macro.
2774 (ASSERT_LT): New macro.
2775 (ASSERT_LT_AT): New macro.
2777 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
2779 PR rtl-optimization/84780
2780 * combine.c (distribute_links): Don't make a link based on pc_rtx.
2782 2018-03-14 Martin Liska <mliska@suse.cz>
2784 * tree.c (record_node_allocation_statistics): Use
2785 get_stats_node_kind.
2786 (get_stats_node_kind): New function extracted from
2787 record_node_allocation_statistics.
2788 (free_node): Use get_stats_node_kind.
2790 2018-03-14 Richard Biener <rguenther@suse.de>
2792 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
2793 that the value-set of ANTIC_IN doesn't grow.
2796 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2798 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2799 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2801 2018-03-14 Julia Koval <julia.koval@intel.com>
2803 * config.gcc (icelake-client, icelake-server): New.
2805 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
2806 (initial_ix86_arch_features): Ditto.
2807 (PTA_SKYLAKE): Add SGX.
2808 (PTA_ICELAKE): Remove.
2809 (PTA_ICELAKE_CLIENT): New.
2810 (PTA_ICELAKE_SERVER): New.
2811 (ix86_option_override_internal): Split up icelake on icelake client and
2813 (get_builtin_code_for_version): Ditto.
2814 (fold_builtin_cpu): Ditto.
2815 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
2816 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
2817 * config/i386/i386.h (processor_type): Ditto.
2818 * doc/invoke.texi: Ditto.
2820 2018-03-14 Jakub Jelinek <jakub@redhat.com>
2823 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
2824 INTEGER_CST offset, add it together with bitpos / 8 and
2825 sign extend based on POINTER_SIZE.
2829 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
2832 * config/i386/constraints.md (Yc): New register constraint.
2833 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
2834 Use Yc constraint for alternative 2 of operand 0. Remove
2835 preferred_for_speed attribute.
2837 2018-03-14 Richard Biener <rguenther@suse.de>
2839 PR tree-optimization/84830
2840 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
2841 with the old one to avoid oscillations.
2843 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
2846 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
2848 (assign_by_spills): Return a flag of reload assignment failure.
2849 Do not process the reload assignment failures. Do not spill other
2850 reload pseudos if they has the same reg class. Update n if
2852 (lra_assign): Add a return arg. Set up from the result of
2853 assign_by_spills call.
2854 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2855 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2856 usage_insns if it is not NULL.
2857 (spill_hard_reg_in_range): New function.
2858 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2859 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2860 function prototypes.
2861 (lra_assign): Change prototype.
2862 * lra.c (lra): Add code to deal with fails by splitting hard reg
2865 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
2867 * config/riscv/riscv.opt (mrelax): New option.
2868 * config/riscv/riscv.c (riscv_file_start): Emit ".option
2869 "norelax" when riscv_mrelax is disabled.
2870 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
2872 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2875 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
2876 reassociation for int modes.
2878 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2880 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
2881 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
2883 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
2884 * config/aarch64/aarch64-sve.md
2885 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
2886 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
2887 (*extend<mode><Vwide>2): Rename to...
2888 (aarch64_sve_extend<mode><Vwide>2): ...this.
2889 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
2890 renaming the old pattern to...
2891 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
2893 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
2894 define_expand, renaming the old pattern to...
2895 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
2896 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
2897 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
2898 account when deciding which SVE instruction the optab should use.
2899 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
2901 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2903 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
2904 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
2905 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
2906 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
2907 (tlsdesc_small_<mode>): Turn a define_expand and use
2908 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
2909 (tlsdesc_small_advsimd_<mode>): ...this.
2910 (tlsdesc_small_sve_<mode>): New pattern.
2912 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2914 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
2915 (UNSPEC_UMUL_HIGHPART): New constants.
2916 (MUL_HIGHPART): New int iteraor.
2917 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
2918 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
2920 (*<su>mul<mode>3_highpart): New define_insn.
2922 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2925 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2928 2018-03-13 Martin Liska <mliska@suse.cz>
2931 * (sem_item_optimizer::sem_item_optimizer): Initialize new
2933 (sem_item_optimizer::~sem_item_optimizer): Release it.
2934 (sem_item_optimizer::merge_classes): Register variable aliases.
2935 (sem_item_optimizer::fixup_pt_set): New function.
2936 (sem_item_optimizer::fixup_points_to_sets): Likewise.
2937 * ipa-icf.h: Declare new variables and functions.
2939 2018-03-13 Jakub Jelinek <jakub@redhat.com>
2942 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
2943 integer_pow2p@2 and test integer_pow2p in condition.
2944 (A < 0 ? C : 0): Similarly for @1.
2947 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
2948 characters starting at p contain '\0' character, don't look beyond
2952 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
2953 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
2956 * reg-stack.c (change_stack): Change update_end var from int to
2957 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
2958 also call set_block_for_insn on the newly added insns and rescan.
2961 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
2962 on the last operand.
2965 * tree.c (stabilize_reference_1): Return save_expr (e) for
2966 STATEMENT_LIST even if it doesn't have side-effects.
2968 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
2970 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
2972 2018-03-12 Renlin Li <renlin.li@arm.com>
2974 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
2975 aarch64_output_scalar_simd_mov_immediate.
2977 2018-03-12 Martin Sebor <msebor@redhat.com>
2979 PR tree-optimization/83456
2980 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
2981 for perfectly overlapping calls to memcpy.
2982 (gimple_fold_builtin_memory_chk): Same.
2983 (gimple_fold_builtin_strcpy): Handle no-warning.
2984 (gimple_fold_builtin_stxcpy_chk): Same.
2985 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
2987 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2989 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
2990 parameter. Use it for SFmode.
2991 (rs6000_function_arg_advance_1): Adjust.
2992 (rs6000_function_arg): Adjust.
2993 (rs6000_gimplify_va_arg): Pass false for that new parameter.
2995 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2997 PR rtl-optimization/84169
2998 PR rtl-optimization/84780
2999 * combine.c (can_combine_p): Check for a 2-insn combination whether
3000 the destination register is used between the two insns, too.
3002 2018-03-12 Richard Biener <rguenther@suse.de>
3004 PR tree-optimization/84803
3005 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
3006 for refs DR analysis didn't process.
3008 2018-03-12 Richard Biener <rguenther@suse.de>
3010 PR tree-optimization/84777
3011 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
3012 force-vectorize loops ignore whether we are optimizing for size.
3014 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
3016 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
3017 (TARGET_MD_ASM_ADJUST): Define.
3019 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
3020 Kito Cheng <kito.cheng@gmail.com>
3021 Chung-Ju Wu <jasonwucj@gmail.com>
3023 * config/nds32/nds32.c (nds32_compute_stack_frame,
3024 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
3025 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
3026 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
3027 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
3028 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
3029 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
3030 * config/nds32/nds32.md (prologue, epilogue): Use macro
3031 NDS32_V3PUSH_AVAILABLE_P to do checking.
3033 2018-03-11 Jakub Jelinek <jakub@redhat.com>
3036 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
3037 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
3038 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
3039 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
3040 addition of most attributes on !orig_type_die or the attribute not
3041 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
3043 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3044 Chung-Ju Wu <jasonwucj@gmail.com>
3046 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
3048 * config/nds32/nds32.opt (mvh): New option.
3050 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3051 Chung-Ju Wu <jasonwucj@gmail.com>
3053 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
3055 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
3056 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
3059 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3060 Chung-Ju Wu <jasonwucj@gmail.com>
3062 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
3064 * config/nds32/nds32-multiple.md (strlensi): New pattern.
3065 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
3067 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
3068 Kito Cheng <kito.cheng@gmail.com>
3069 Chung-Ju Wu <jasonwucj@gmail.com>
3071 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
3072 UNSPEC_FFMISM and UNSPEC_FLMISM.
3073 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
3074 for ffb, ffmism and flmism.
3075 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
3076 (unspec_ffmism): Ditto.
3077 (unspec_flmism): Ditto.
3078 (nds32_expand_builtin_impl): Check if string extension is available.
3079 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
3080 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
3082 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
3085 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3088 * lra-assigns.c (assign_by_spills): Return a flag of reload
3089 assignment failure. Do not process the reload assignment
3090 failures. Do not spill other reload pseudos if they has the same
3092 (lra_assign): Add a return arg. Set up from the result of
3093 assign_by_spills call.
3094 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3095 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3096 usage_insns if it is not NULL.
3097 (spill_hard_reg_in_range): New function.
3098 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3099 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3100 function prototypes.
3101 (lra_assign): Change prototype.
3102 * lra.c (lra): Add code to deal with fails by splitting hard reg
3105 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
3108 * config/i386/i386.opt: Replace Enforcment with Enforcement.
3110 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
3113 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
3114 (dw_val_node): Add val_symbolic_view.
3115 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
3116 (symview_upper_bound): New.
3117 (new_line_info_table): Initialize symviews_since_reset.
3118 (dwarf2out_source_line): Count symviews_since_reset and set
3119 symview_upper_bound.
3120 (dw_val_equal_p): Handle symview.
3121 (add_AT_symview): New.
3122 (print_dw_val): Handle symview.
3123 (attr_checksum, attr_checksum_ordered): Likewise.
3124 (same_dw_val_p, size_of_die): Likewise.
3125 (value_format, output_die): Likewise.
3126 (add_high_low_attributes): Use add_AT_symview for entry_view.
3127 (dwarf2out_finish): Reset symview_upper_bound, clear
3130 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
3133 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
3134 Add strict argument and use it.
3135 (rs6000_split_multireg_move): Update for new strict argument.
3136 (mem_operand_gpr): Disallow all non-offsettable addresses.
3137 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
3139 2018-03-09 Jakub Jelinek <jakub@redhat.com>
3142 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
3143 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
3144 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
3147 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
3148 decl, use remap_type if we want to use the type.
3150 2018-03-09 Martin Sebor <msebor@redhat.com>
3152 PR tree-optimization/84526
3153 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3155 (builtin_access::generic_overlap): Be prepared to handle non-array
3158 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
3160 PR rtl-optimization/84682
3161 * lra-constraints.c (process_address_1): Check is_address flag
3162 for address constraints.
3163 (process_alt_operands): Likewise.
3164 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
3165 preprocess_constraints.
3166 * recog.h (preprocess_constraints): Add oploc parameter.
3168 * recog.c (preprocess_constraints): Test address_operand for
3169 CT_ADDRESS constraints.
3171 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3174 * lra-assigns.c (assign_by_spills): Return a flag of reload
3175 assignment failure. Do not process the reload assignment
3176 failures. Do not spill other reload pseudos if they has the same
3178 (lra_assign): Add a return arg. Set up from the result of
3179 assign_by_spills call.
3180 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3181 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3182 usage_insns if it is not NULL.
3183 (spill_hard_reg_in_range): New function.
3184 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3185 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3186 function prototypes.
3187 (lra_assign): Change prototype.
3188 * lra.c (lra): Add code to deal with fails by splitting hard reg
3191 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3194 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
3195 Accept complain bool parameter. Only emit errors if it is true.
3196 (arm_parse_cpu_option_name): Likewise.
3197 (arm_target_thumb_only): Adjust callers of the above.
3198 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
3199 prototype to take a default true bool parameter.
3200 (arm_parse_arch_option_name): Likewise.
3202 2018-03-09 David Malcolm <dmalcolm@redhat.com>
3203 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
3207 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
3208 * configure: Regenerate.
3209 * configure.ac ("linker --version-script option"): New.
3210 ("linker soname option"): New.
3212 2018-03-09 Richard Biener <rguenther@suse.de>
3214 PR tree-optimization/84775
3215 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
3216 immediate uses of predicate stmts and mark them modified.
3219 PR tree-optimization/84178
3220 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3222 (version_loop_for_if_conversion): Delay update_ssa call.
3223 (tree_if_conversion): Delay update_ssa until after predicate
3226 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
3229 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
3230 when the function accesses prior frames.
3232 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3235 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
3236 gen_llsym, otherwise call maybe_gen_llsym.
3239 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
3240 has ',' character inside of it.
3242 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3245 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
3246 as clobbering CC_REGNUM.
3248 2018-03-08 Richard Biener <rguenther@suse.de>
3251 * tree-scalar-evolution.c: Include tree-into-ssa.h.
3252 (follow_copies_to_constant): Do not follow SSA names registered
3255 2018-03-08 Richard Biener <rguenther@suse.de>
3257 PR tree-optimization/84178
3258 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3260 (version_loop_for_if_conversion): Delay update_ssa call.
3261 (tree_if_conversion): Delay update_ssa until after predicate
3264 2018-03-08 David Malcolm <dmalcolm@redhat.com>
3266 PR tree-optimization/84178
3267 * tree-if-conv.c (release_bb_predicate): Remove the
3268 the assertion that the stmts have NULL use_ops.
3269 Discard the statements, asserting that they haven't
3270 yet been added to a BB.
3272 2018-03-08 Richard Biener <rguenther@suse.de>
3274 PR tree-optimization/84746
3275 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
3276 (phi_translate): Pass in destination ANTIC_OUT set.
3277 (phi_translate_1): Likewise. For a simplified result lookup
3278 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
3279 (phi_translate_set): Adjust.
3280 (do_pre_regular_insertion): Likewise.
3281 (do_pre_partial_partial_insertion): Likewise.
3283 2018-03-08 Martin Liska <mliska@suse.cz>
3285 PR gcov-profile/84735
3286 * doc/gcov.texi: Document usage of profile files.
3287 * gcov-io.h: Document changes in the format.
3289 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
3293 * dwarf2out.c (struct dw_line_info_table): Update comments for
3295 (FORCE_RESET_NEXT_VIEW): New.
3296 (FORCE_RESETTING_VIEW_P): New.
3297 (RESETTING_VIEW_P): Check for -1 too.
3298 (ZERO_VIEW_P): Likewise.
3299 (new_line_info_table): Force-reset next view.
3300 (dwarf2out_begin_function): Likewise.
3301 (dwarf2out_source_line): Simplify zero_view_p initialization.
3302 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
3303 view directly. Omit view when omitting .loc at line 0.
3305 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3307 PR tree-optimization/84740
3308 * tree-switch-conversion.c (process_switch): Call build_constructors
3309 only if info.phi_count is non-zero.
3311 PR tree-optimization/84739
3312 * tree-tailcall.c (find_tail_calls): Check call arguments against
3313 DECL_ARGUMENTS (current_function_decl) rather than
3314 DECL_ARGUMENTS (func) when checking for tail recursion.
3316 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3318 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
3319 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
3320 Volker Reichelt's entry and add entries for people that perform
3321 GCC fuzzy testing and report numerous bugs.
3323 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
3326 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
3327 readonly data in sdata, if that is disabled.
3328 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
3329 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3330 -mreadonly-in-sdata option.
3332 2018-03-07 Martin Sebor <msebor@redhat.com>
3334 PR tree-optimization/84468
3335 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
3336 basic block when looking for nul assignment.
3338 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
3341 * except.h (output_function_exception_table): Adjust prototype.
3342 * except.c (output_function_exception_table): Remove FNNAME parameter
3343 and add SECTION parameter. Ouput one part of the table at a time.
3344 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
3345 the first part of the exception table and emit unwind directives.
3346 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
3347 (i386_pe_seh_cold_init): Likewise.
3348 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
3349 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
3350 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
3351 (ix86_output_call_insn): Emit a nop in one more case for SEH.
3352 * config/i386/winnt.c: Include except.h.
3353 (struct seh_frame_state): Add reg_offset, after_prologue and
3354 in_cold_section fields.
3355 (i386_pe_seh_end_prologue): Set seh->after_prologue.
3356 (i386_pe_seh_cold_init): New function.
3357 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
3358 to seh->in_cold_section.
3359 (seh_emit_push): Record the offset of the push.
3360 (seh_emit_save): Record the offet of the save.
3361 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
3362 Test seh->after_prologue to disregard the epilogue.
3363 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
3364 (i386_pe_end_cold_function): New function.
3366 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3369 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
3370 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
3373 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
3375 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
3376 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
3379 * multiple_target.c: Include tree-inline.h and intl.h.
3380 (expand_target_clones): Diagnose and fail if node->definition and
3381 !tree_versionable_function_p (node->decl).
3383 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
3385 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
3387 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
3388 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
3389 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
3391 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3394 * combine.c (try_combine): Use reg_or_subregno instead of handling
3395 just paradoxical SUBREGs and REGs.
3397 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
3399 * config/arc/arc.c (arc_finalize_pic): Remove function.
3400 (arc_must_save_register): We use single base PIC register, remove
3401 checks to save/restore the PIC register.
3402 (arc_expand_prologue): Likewise.
3403 * config/arc/arc-protos.h (arc_set_default_type_attributes):
3405 (arc_verify_short): Likewise.
3406 (arc_attr_type): Likewise.
3407 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
3408 (walk_stores): Likewise.
3409 (arc_address_cost): Make it static.
3410 (arc_verify_short): Likewise.
3411 (branch_dest): Likewise.
3412 (arc_attr_type): Likewise.
3413 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
3414 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
3415 (arc_final_prescan_insn): Remove inserting the nops due to
3416 hardware hazards. It is done in reorg step.
3417 (insn_length_variant_t): Remove.
3418 (insn_length_parameters_t): Likewise.
3419 (arc_insn_length_parameters): Likewise.
3420 (arc_get_insn_variants): Likewise.
3421 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
3423 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3426 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
3429 PR tree-optimization/84687
3430 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
3432 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
3434 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3436 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
3437 Rename to ppc_speculation_barrier.
3438 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
3439 __builtin_ppc_speculation_barrier.
3441 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3444 * combine.c (combine_simplify_rtx): Don't try to simplify if
3445 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
3448 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
3450 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
3451 to 32 bytes when compiling for POWER9.
3453 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3456 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
3457 regparm >= 3 with no arg reg available also for calls with
3458 flag_force_indirect_call. Pass decl to ix86_function_regparm.
3461 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
3463 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
3465 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
3468 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
3470 2018-03-05 Richard Biener <rguenther@suse.de>
3472 PR tree-optimization/84486
3473 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
3474 When inserting a __builtin_assume_aligned call set the LHS
3475 SSA name alignment info accordingly.
3477 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
3479 PR tree-optimization/84114
3480 * config/aarch64/aarch64.c (aarch64_reassociation_width)
3481 Avoid reassociation of FLOAT_MODE addition.
3483 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
3485 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
3486 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
3487 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
3488 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
3489 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
3490 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
3491 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
3493 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
3494 __builtin_ia32_wbinvd): New builtins.
3495 (SPECIAL_ARGS2): New.
3496 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
3497 (SPECIAL_ARGS2): New.
3498 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
3499 (ix86_valid_target_attribute_inner_p): Ditto.
3500 (ix86_init_mmx_sse_builtins): Add special_args2.
3501 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
3502 TARGET_WBNOINVD_P): New.
3503 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
3504 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
3505 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
3506 * config/i386/immintrin.h (_wbinvd): New intrinsic.
3507 * config/i386/pconfigintrin.h: New file.
3508 * config/i386/wbnoinvdintrin.h: Ditto.
3509 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
3510 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
3512 2018-03-05 Richard Biener <rguenther@suse.de>
3514 PR tree-optimization/84670
3515 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3517 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3518 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3519 (compute_antic_aux): Only assert the number of values in ANTIC_IN
3520 doesn't grow if all successors (recursively) were visited at least
3523 2018-03-05 Richard Biener <rguenther@suse.de>
3525 PR tree-optimization/84650
3526 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
3527 if executed in the loop pipeline.
3529 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
3531 * doc/configfiles.texi (Configuration Files): Move info about
3532 conditionalizing $target-protos.h to...
3533 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
3534 differs from $target-protos.h.
3536 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
3537 Chung-Ju Wu <jasonwucj@gmail.com>
3539 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
3540 * config/nds32/nds32-multiple.md (setmemsi): Define.
3541 * config/nds32/nds32-memory-manipulation.c
3542 (nds32_gen_dup_4_byte_to_word_value): New.
3543 (emit_setmem_word_loop): New.
3544 (emit_setmem_byte_loop): New.
3545 (nds32_expand_setmem_loop): New.
3546 (nds32_expand_setmem_loop_v3m): New.
3547 (nds32_expand_setmem_unroll): New.
3548 (nds32_expand_setmem): New.
3550 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3551 Chung-Ju Wu <jasonwucj@gmail.com>
3553 * config/nds32/nds32-memory-manipulation.c
3554 (nds32_emit_load_store): New.
3555 (nds32_emit_post_inc_load_store): New.
3556 (nds32_emit_mem_move): New.
3557 (nds32_emit_mem_move_block): New.
3558 (nds32_expand_movmemsi_loop_unknown_size): New.
3559 (nds32_expand_movmemsi_loop_known_size): New.
3560 (nds32_expand_movmemsi_loop): New.
3561 (nds32_expand_movmemsi_unroll): New.
3562 (nds32_expand_movmemqi): Rename ...
3563 (nds32_expand_movmemsi): ... to this.
3564 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
3565 (movmemsi): ... to this.
3566 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
3567 (nds32_expand_movmemsi): ... to this.
3569 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3570 Monk Chiang <sh.chiang04@gmail.com>
3571 Chung-Ju Wu <jasonwucj@gmail.com>
3573 * config/nds32/nds32-protos.h
3574 (nds32_expand_load_multiple): New arguments.
3575 (nds32_expand_store_multiple): Ditto.
3576 (nds32_valid_multiple_load_store): Rename ...
3577 (nds32_valid_multiple_load_store_p): ... to this.
3578 * config/nds32/nds32-memory-manipulation.c
3579 (nds32_expand_load_multiple): Refine implementation.
3580 (nds32_expand_store_multiple): Ditto.
3581 * config/nds32/nds32-multiple.md
3582 (load_multiple): Update nds32_expand_load_multiple interface.
3583 (store_multiple): Update nds32_expand_store_multiple interface.
3584 * config/nds32/nds32-predicates.c
3585 (nds32_valid_multiple_load_store): Rename ...
3586 (nds32_valid_multiple_load_store_p): ... to this and refine
3588 * config/nds32/predicates.md
3589 (nds32_load_multiple_and_update_address_operation): New predicate.
3590 (nds32_store_multiple_and_update_address_operation): New predicate.
3592 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3593 Chung-Ju Wu <jasonwucj@gmail.com>
3595 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
3596 (combo): New attribute.
3597 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
3599 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
3601 * config/nds32/nds32.opt: Change -mcmodel= default value.
3603 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
3604 Monk Chiang <sh.chiang04@gmail.com>
3605 Chung-Ju Wu <jasonwucj@gmail.com>
3607 * config/nds32/constants.md (unspec_element): New enum.
3608 * config/nds32/constraints.md (Umw): New constraint.
3609 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
3610 * config/nds32/nds32-intrinsic.md: Likewise.
3611 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
3612 (nds32_valid_smw_lwm_base_p): New.
3613 (nds32_output_smw_single_word): New.
3614 (nds32_output_lmw_single_word): New.
3615 (nds32_expand_unaligned_load): New.
3616 (nds32_expand_unaligned_store): New.
3617 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
3618 (nds32_output_smw_single_word): Declare.
3619 (nds32_output_lmw_single_word): Declare.
3620 (nds32_expand_unaligned_load): Declare.
3621 (nds32_expand_unaligned_store): Declare.
3622 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
3623 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
3624 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
3625 NDS32_BUILTIN_UASTORE_DW.
3626 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
3629 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
3630 Kito Cheng <kito.cheng@gmail.com>
3631 Chung-Ju Wu <jasonwucj@gmail.com>
3633 * config/nds32/nds32-intrinsic.c
3634 (nds32_expand_builtin_null_ftype_reg): Delete.
3635 (nds32_expand_builtin_reg_ftype_imm): Ditto.
3636 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
3637 (nds32_read_argument): New.
3638 (nds32_legitimize_target): Ditto.
3639 (nds32_legitimize_argument): Ditto.
3640 (nds32_check_constant_argument): Ditto.
3641 (nds32_expand_unop_builtin): Ditto.
3642 (nds32_expand_unopimm_builtin): Ditto.
3643 (nds32_expand_binop_builtin): Ditto.
3644 (nds32_builtin_decl_impl): Ditto.
3645 (builtin_description): Ditto.
3646 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
3647 (nds32_init_builtins_impl): Ditto.
3648 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
3649 (nds32_builtin_decl): New.
3650 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
3651 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
3653 2018-03-02 Jeff Law <law@redhat.com>
3655 * reorg.c (stop_search_p): Handle DEBUG_INSN.
3656 (redundant_insn, fill_simple_delay_slots): Likewise.
3657 (fill_slots_from_thread): Likewise.
3658 * resource.c (mark_referenced_resources): Likewise.
3659 (mark_set_resources, find_dead_or_set_registers): Likewise.
3661 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3663 * substring-locations.h (format_warning_va): Formatting fix for
3665 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
3667 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
3668 * substring-locations.c: Include intl.h.
3669 (format_warning_va): Turned into small wrapper around
3670 format_warning_n_va, renamed to ...
3671 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
3672 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
3674 (format_warning_at_substring_n): New function.
3675 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
3676 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
3677 format_warning_at_substring with just a shorter name instead of
3678 const function pointer.
3679 (fmtwarn_n): New function.
3680 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
3681 appropriate, get rid of all the fmtstr temporaries, move conditionals
3682 with G_() wrapped string literals directly into fmtwarn arguments,
3683 cast dir.len to (int), formatting fixes.
3685 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
3687 * doc/invoke.texi: Remove "Cilk Plus" references.
3689 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3690 Richard Biener <rguenther@suse.de>
3693 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
3694 for error or warning attributes if CALL_FROM_THUNK_P is set.
3697 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3700 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
3701 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
3704 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
3705 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
3707 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
3708 instead of -1U in last predictors element's probability member.
3710 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3713 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3714 arguments if they are comparable.
3716 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3718 PR tree-optimization/84634
3719 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
3720 masks and masked_loop_p with a single loop_masks, making sure it's
3721 null for bb vectorization.
3723 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3725 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3726 (vect_analyze_data_ref_access): Use loop->safe_len rather than
3727 loop->force_vectorize to check whether there is no alias.
3729 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3732 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
3734 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
3736 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
3737 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
3738 instead of a loop around prev_real_insn.
3739 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
3743 * config/i386/i386.c (ix86_print_operand): Use conditional
3744 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
3747 2018-03-02 Richard Biener <rguenther@suse.de>
3749 PR tree-optimization/84427
3750 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
3751 (bitmap_set_subtract_values): Rewrite to handle multiple
3754 (prune_clobbered_mems): Likewise.
3755 (phi_translate): Take edge instead of pred/phiblock.
3756 (phi_translate_1): Likewise.
3757 (phi_translate_set): Likewise. Insert all translated
3758 exprs for a value into the set, keeping possibly multiple
3759 expressions per value.
3760 (compute_antic_aux): Adjust for phi_translate changes.
3761 When intersecting union the expressions and prune those
3762 not in the final value set, keeping possibly multiple
3763 expressions per value. Do not use value-insertion
3764 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
3765 all expressions. Add verification that the value-sets
3766 only shrink during iteration.
3767 (compute_partial_antic_aux): Adjust for the phi_translate changes.
3768 (do_pre_regular_insertion): Likewise.
3769 (do_pre_partial_partial_insertion): Likewise.
3771 2018-03-02 Richard Biener <rguenther@suse.de>
3774 * config/darwin.c (saved_debug_info_level): New static global.
3775 (darwin_asm_lto_start): Disable debug info generation for LTO out.
3776 (darwin_asm_lto_end): Restore debug info generation settings.
3778 2018-03-01 Martin Liska <mliska@suse.cz>
3781 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
3784 2018-03-01 Richard Biener <rguenther@suse.de>
3787 * dwarf2out.c (gen_variable_die): Properly handle late VLA
3788 type annotation with LTO when debug was disabled at compile-time.
3790 2018-03-01 Matthew Fortune <mfortune@gmail.com>
3792 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
3794 (mips_final_postscan_insn): Likewise.
3796 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
3798 PR rtl-optimization/84528
3799 * alias.c (init_alias_target): Add commentary.
3800 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
3801 a unique base value if the frame pointer is not eliminated
3802 to the stack pointer.
3804 2018-03-01 Tom de Vries <tom@codesourcery.com>
3806 PR rtl-optimization/83327
3807 * lra-int.h (hard_regs_spilled_into): Declare.
3808 * lra.c (hard_regs_spilled_into): Define.
3809 (init_reg_info): Init hard_regs_spilled_into.
3810 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
3811 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
3812 (process_bb_lives): Handle hard_regs_spilled_into.
3813 (lra_create_live_ranges_1): Before doing liveness propagation, clear
3814 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
3816 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
3818 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
3819 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
3820 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
3821 * config/rs6000/aix72.h: New file.
3823 2018-02-28 Jakub Jelinek <jakub@redhat.com>
3825 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
3826 instead of warning_at with conditional singular and plural messages
3830 * stor-layout.c (update_alignment_for_field): For
3831 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
3832 && !DECL_PACKED (field), do the alignment update, just use
3833 only desired_align instead of MAX (type_align, desired_align)
3835 (place_field): Don't do known_align < desired_align handling
3836 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
3837 is non-NULL, instead do it after rli->prev_field handling and
3838 only if not within a bitfield word. For DECL_PACKED (field)
3839 use type_align of BITS_PER_UNIT.
3841 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
3843 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
3844 superfluous parentheses and trailing spaces.
3846 2018-02-28 Richard Biener <rguenther@suse.de>
3848 PR tree-optimization/84584
3849 * graphite-scop-detection.c (scop_detection::add_scop): Discard
3850 SCoPs with fake exit edge.
3852 2018-02-28 Martin Liska <mliska@suse.cz>
3855 * timevar.c (timer::print): Fix format to properly print 100%
3858 2018-02-28 Richard Biener <rguenther@suse.de>
3861 * genmatch.c (capture_info::walk_match): Do not mark
3862 captured expressions without operands as expr_p given
3863 they act more like predicates and should be subject to
3864 "lost tail" side-effect preserving.
3866 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
3868 PR rtl-optimization/81611
3869 * auto-inc-dec.c (attempt_change): Move dead note from
3870 mem_insn if it's the next use of regno
3871 (find_address): Take address use of reg holding
3872 non-incremented value. Add parm to limit search to the named
3874 (merge_in_block): Attempt to use a mem insn that is the next
3875 use of the original regno.
3877 2018-02-27 Martin Sebor <msebor@redhat.com>
3880 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
3881 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
3883 2018-02-27 Martin Sebor <msebor@redhat.com>
3885 PR translation/84207
3886 * diagnostic-core.h (warning_n, error_n, inform_n): Change
3887 n argument to unsigned HOST_WIDE_INT.
3888 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
3889 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
3890 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
3891 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
3893 2018-02-27 Richard Biener <rguenther@suse.de>
3895 PR tree-optimization/84512
3896 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
3897 Do not use the estimate returned from record_stmt_cost for
3898 the scalar iteration cost but sum properly using add_stmt_cost.
3900 2018-02-27 Richard Biener <rguenther@suse.de>
3902 PR tree-optimization/84466
3903 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3904 Adjust last change to less strictly validate use operands.
3906 2018-02-27 Martin Liska <mliska@suse.cz>
3908 PR gcov-profile/84548
3909 * gcov.c (process_file): Allow partial overlap and consider it
3910 also as group functions.
3911 (output_lines): Properly calculate range of lines for a group.
3913 2018-02-27 Martin Liska <mliska@suse.cz>
3915 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
3916 'ggc' suffixes. Change first column width.
3917 (timer::print): Fix formatting of the column.
3919 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
3921 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
3922 preserve inline entry blocks for the sake of debug inline
3923 entry point markers alone.
3924 (remove_unused_locals): Suggest in comments a better place to
3925 force the preservation of inline entry blocks that are
3926 otherwise unused, but do not preserve them.
3928 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3930 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
3932 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3935 * config/i386/constraints.md (Bs): Replace
3936 ix86_indirect_branch_register with
3937 TARGET_INDIRECT_BRANCH_REGISTER.
3939 * config/i386/i386.md (indirect_jump): Likewise.
3940 (tablejump): Likewise.
3941 (*sibcall_memory): Likewise.
3942 (*sibcall_value_memory): Likewise.
3943 Peepholes of indirect call and jump via memory: Likewise.
3944 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
3945 (*sibcall_value_GOT_32): Likewise.
3946 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3947 (GOT_memory_operand): Likewise.
3948 (call_insn_operand): Likewise.
3949 (sibcall_insn_operand): Likewise.
3950 (GOT32_symbol_operand): Likewise.
3951 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
3953 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3955 PR rtl-optimization/83496
3956 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
3957 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
3958 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
3959 redundant insn, if any.
3960 (relax_delay_slots): Likewise.
3961 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
3963 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
3965 PR tree-optimization/83965
3966 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
3967 that grouped statements are part of a reduction chain. Return
3968 true if the statement is not marked as a reduction itself but
3970 (vect_recog_dot_prod_pattern): Don't check whether the statement
3971 is part of a group here.
3972 (vect_recog_sad_pattern): Likewise.
3973 (vect_recog_widen_sum_pattern): Likewise.
3975 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3978 * final.c (rest_of_clean_state): Also look for calls inside sequences.
3980 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3983 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
3985 (ix86_output_indirect_function_return): New prototype.
3986 (ix86_split_simple_return_pop_internal): Likewise.
3987 * config/i386/i386.c (indirect_return_via_cx): New.
3988 (indirect_return_via_cx_bnd): Likewise.
3989 (indirect_thunk_name): Handle return va CX_REG.
3990 (output_indirect_thunk_function): Create alias for
3991 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
3992 (ix86_output_indirect_jmp): Remove the bool argument.
3993 (ix86_output_indirect_function_return): New function.
3994 (ix86_split_simple_return_pop_internal): Likewise.
3995 * config/i386/i386.md (*indirect_jump): Don't pass false
3996 to ix86_output_indirect_jmp.
3997 (*tablejump_1): Likewise.
3998 (simple_return_pop_internal): Change it to define_insn_and_split.
3999 Call ix86_split_simple_return_pop_internal to split it for
4001 (simple_return_indirect_internal): Call
4002 ix86_output_indirect_function_return instead of
4003 ix86_output_indirect_jmp.
4005 2018-02-26 Jakub Jelinek <jakub@redhat.com>
4008 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
4009 memset and value initialization afterwards.
4011 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
4013 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
4015 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4018 * common/config/aarch64/aarch64-common.c
4019 (aarch_option_optimization_table[]): Switch
4020 off fomit-frame-pointer
4022 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4023 Chung-Ju Wu <jasonwucj@gmail.com>
4025 * config/nds32/nds32-multiple.md (load_multiple): Disallow
4027 (store_multiple): Ditto.
4029 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4031 * config.gcc: Add --with-cpu support for nds32 target.
4032 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
4033 * config/nds32/nds32.opt: Add -mcpu= option.
4035 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
4037 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
4038 isel=yes): Warn for these deprecated options.
4040 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
4042 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
4043 ISA_2_5_MASKS_EMBEDDED.
4045 2018-02-23 Jakub Jelinek <jakub@redhat.com>
4047 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
4048 p->max as pointers rather than using iterative_hash_expr.
4050 2018-02-23 Carl Love <cel@us.ibm.com>
4052 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
4053 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
4055 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
4056 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
4057 P8V_BUILTIN_VEC_VUNSIGNED2.
4059 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
4062 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
4063 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
4065 * lra-constraints.c (curr_insn_transform): Set up
4066 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
4068 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
4069 * lra-eliminations.c (spill_pseudos): Ditto.
4070 (process_insn_for_elimination): Ditto.
4071 * lra-lives.c (reg_early_clobber_p): Use the new macros.
4072 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
4073 LRA_NON_CLOBBERED_ALT.
4075 2018-02-22 Martin Sebor <msebor@redhat.com>
4077 PR tree-optimization/84480
4078 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
4079 to maybe_diag_stxncpy_trunc. Call it.
4080 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
4081 from gimple_fold_builtin_strcpy. Print inlining stack.
4082 (handle_builtin_stxncpy): Print inlining stack.
4083 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
4085 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
4088 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
4089 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
4090 and -fcheck-pointer-bounds are used together.
4091 (indirect_thunk_prefix): New enum.
4092 (indirect_thunk_need_prefix): New function.
4093 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
4094 "_nt" instead of "_bnd" for NOTRACK prefix.
4095 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
4096 (output_indirect_thunk_function): Likewise.
4098 (ix86_code_end): Update output_indirect_thunk_function calls.
4099 (ix86_output_indirect_branch_via_reg): Replace
4100 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
4101 (ix86_output_indirect_branch_via_push): Likewise.
4102 (ix86_output_function_return): Likewise.
4103 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
4104 incompatible with -fcf-protection=branch and
4105 -fcheck-pointer-bounds.
4107 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4110 * config/aarch64/aarch64.c (aarch64_print_address_internal):
4111 Change gcc_assert call to output_operand_lossage.
4113 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4115 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
4117 2018-02-22 DJ Delorie <dj@redhat.com>
4118 Sebastian Perta <sebastian.perta@renesas.com>
4119 Oleg Endo <olegendo@gcc.gnu.org>
4121 * config/rx/rx.c (rx_rtx_costs): New function.
4122 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
4124 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
4126 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
4128 2018-02-22 Martin Liska <mliska@suse.cz>
4131 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
4132 Add "native" as a possible value.
4134 2018-02-22 Martin Liska <mliska@suse.cz>
4137 * config/i386/i386.c (ix86_option_override_internal):
4138 Add "native" as a possible value for -march and -mtune.
4140 2018-02-22 Jakub Jelinek <jakub@redhat.com>
4143 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
4144 to all type variants.
4146 PR tree-optimization/84503
4147 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
4148 width as info->bitpos + info->bitsize - start.
4149 (merged_store_group::merge_overlapping): Simplify width computation.
4150 (check_no_overlap): New function.
4151 (imm_store_chain_info::try_coalesce_bswap): Compute expected
4152 start + width and last_order of the group, fail if check_no_overlap
4154 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
4155 to group if check_no_overlap fails.
4157 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4159 * config/rs6000/altivec.md: Delete contraint arguments to
4160 define_expand, define_split, and define_peephole2, and in
4161 define_insn_and_split if always unused.
4162 * config/rs6000/darwin.md: Ditto.
4163 * config/rs6000/dfp.md: Ditto.
4164 * config/rs6000/rs6000.md: Ditto.
4165 * config/rs6000/sync.md: Ditto.
4166 * config/rs6000/vector.md: Ditto.
4167 * config/rs6000/vsx.md: Ditto.
4169 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4171 * config/rs6000/altivec.md: Write output control strings as braced
4172 blocks instead of double-quoted strings.
4173 * config/rs6000/darwin.md: Ditto.
4174 * config/rs6000/rs6000.md: Ditto.
4175 * config/rs6000/vector.md: Ditto.
4176 * config/rs6000/vsx.md: Ditto.
4178 2018-02-21 Jason Merrill <jason@redhat.com>
4180 PR c++/84314 - ICE with templates and fastcall attribute.
4181 * attribs.c (build_type_attribute_qual_variant): Remove assert.
4183 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4185 * ipa-cp.c (determine_versionability): Fix comment typos.
4187 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4190 * ipa-cp.c (determine_versionability): Do not version functions caling
4193 2018-02-21 Martin Liska <mliska@suse.cz>
4196 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
4197 Add "native" as a possible value.
4198 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
4199 the macro when native cpu detection is available.
4201 2018-02-21 Martin Liska <mliska@suse.cz>
4204 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
4205 Add "native" as a possible value.
4206 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
4207 when native cpu detection is available.
4209 2018-02-21 Jakub Jelinek <jakub@redhat.com>
4210 Martin Sebor <msebor@redhat.com>
4212 PR tree-optimization/84478
4213 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
4215 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
4216 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
4217 support which is conservatively correct, for 2 only stay conservative
4218 for maxlen. Formatting and comment capitalization fixes. Add STRICT
4219 argument to the 2 argument get_range_strlen, adjust 6 arg
4220 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
4222 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
4223 (gimple_fold_builtin_strlen): Pass true as last argument to
4226 2018-02-20 Martin Sebor <msebor@redhat.com>
4229 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
4230 (builtin_memref::set_base_and_offset): Same. Handle inner references.
4231 (builtin_memref::builtin_memref): Factor out parts into
4232 set_base_and_offset and call it.
4234 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4237 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
4238 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
4239 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
4240 search at the associated MODE_INT.
4242 2018-02-20 Jeff Law <law@redhat.com>
4245 PR tree-optimization/81592
4247 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
4248 for range data rather than using global data.
4249 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
4250 range data rather than using global data.
4251 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
4252 pass it to children as needed.
4253 (struct directive::fmtresult): Similarly.
4254 (struct directive::set_width): Similarly.
4255 (struct directive::set_precision): Similarly.
4256 (format_integer, format_directive, parse_directive): Similarly.
4257 (format_none): Accept unnamed vr_values parameter.
4258 (format_percent, format_floating, format_character): Similarly.
4259 (format_string, format_plain): Similarly.
4260 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
4261 the EVRP range analyzer for range data rather than using global data.
4262 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
4263 gimple-ssa-evrp-analyze.h
4264 (class sprintf_dom_walker): Add after_dom_children member function.
4265 Add evrp_range_analyzer member.
4266 (sprintf_dom_walker::before_dom_children): Call into the EVRP
4267 range analyzer as needed.
4268 (sprintf_dom_walker::after_dom_children): New member function.
4269 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
4271 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
4272 (evrp_range_analyzer::pop_to_marker): Likewise.
4274 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4276 PR tree-optimization/84419
4277 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
4278 with the required type if its current type is compatible but
4281 2018-02-20 Jakub Jelinek <jakub@redhat.com>
4284 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
4285 after vectorization.
4287 2018-02-20 Martin Liska <mliska@suse.cz>
4290 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
4291 possible values if we don't have a hint.
4293 2018-02-20 Martin Liska <mliska@suse.cz>
4297 * final.c (shorten_branches): Build align_tab array with one
4299 * opts.c (finish_options): Add alignment option limit check.
4300 (MAX_CODE_ALIGN): Likewise.
4301 (MAX_CODE_ALIGN_VALUE): Likewise.
4302 * doc/invoke.texi: Document maximum allowed option value for
4303 all -falign-* options.
4305 2018-02-19 Jakub Jelinek <jakub@redhat.com>
4308 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
4309 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
4310 * var-tracking.c (emit_note_insn_var_location): Remove all references
4311 to NOTE_INSN_CALL_ARG_LOCATION.
4312 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
4313 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
4314 Use copy_rtx_if_shared.
4315 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
4316 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
4317 (dwarf2out_var_location): Remove handling of
4318 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
4320 * final.c (final_scan_insn): Remove all references to
4321 NOTE_INSN_CALL_ARG_LOCATION.
4322 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
4323 before dumping final insns.
4324 * except.c (emit_note_eh_region_end): Remove all references to
4325 NOTE_INSN_CALL_ARG_LOCATION.
4326 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
4327 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
4328 * config/arc/arc.c (hwloop_optimize): Likewise.
4329 * config/arm/arm.c (create_fix_barrier): Likewise.
4330 * config/s390/s390.c (s390_chunkify_start): Likewise.
4331 * config/sh/sh.c (find_barrier): Likewise.
4332 * config/i386/i386.c (rest_of_insert_endbranch,
4333 ix86_seh_fixup_eh_fallthru): Likewise.
4334 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
4335 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
4336 * config/frv/frv.c (frv_function_prologue): Likewise.
4337 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
4339 (note_outside_basic_block_p): Remove all references to
4340 NOTE_INSN_CALL_ARG_LOCATION.
4341 * gengtype.c (adjust_field_rtx_def): Likewise.
4342 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
4344 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
4345 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
4348 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
4351 PR tree-optimization/84452
4352 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
4353 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
4356 2018-02-19 Martin Liska <mliska@suse.cz>
4359 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
4361 2018-02-19 Martin Liska <mliska@suse.cz>
4362 Richard Sandiford <richard.sandiford@linaro.org>
4364 PR tree-optimization/82491
4365 * gimple-fold.c (get_base_constructor): Make earlier bail out
4368 2018-02-19 Carl Love <cel@us.ibm.com>
4370 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
4371 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
4373 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
4374 P8V_BUILTIN_VEC_NEG.
4376 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
4378 * config/rl78/rl78.md (movdf): New define expand.
4380 2018-02-19 Martin Liska <mliska@suse.cz>
4383 * doc/invoke.texi: Fix typo.
4384 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
4386 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
4388 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
4389 handle rs6000_single_float and rs6000_double_float specially for
4392 2018-02-16 Jeff Law <law@redhat.com>
4394 * config/rx/rx.c (add_pop_cfi_notes): New function.;
4397 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4400 * ipa-inline.c (inline_small_functions): Fix a typo.
4402 2018-02-16 Nathan Sidwell <nathan@acm.org>
4404 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
4406 2018-02-16 Carl Love <cel@us.ibm.com>
4408 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
4409 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
4410 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
4411 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
4412 expansion to P8V_BUILTIN_VEC_FLOAT2.
4414 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
4416 PR rtl-optimization/70023
4417 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
4418 src_regno into account.
4420 2018-02-16 Carl Love <cel@us.ibm.com>
4422 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
4423 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
4424 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
4425 * config/rs6000/rs6000.c: Remove case statements for
4426 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
4427 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
4428 and P9V_BUILTIN_VEC_VINSERT4B.
4429 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
4430 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
4431 * config/rs6000/vsx.md:
4432 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
4435 2018-02-16 Carl Love <cel@us.ibm.com>
4437 * config/rs6000/altivec.h: Add builtin names vec_extract4b
4439 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
4441 * config/rs6000/rs6000-c.c: Add the definitions for
4442 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
4443 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
4444 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
4445 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
4446 definition for insert4b and define insn *insert3b_internal.
4447 * doc/extend.texi: Add documentation for vec_extract4b.
4449 2018-02-16 Nathan Sidwell <nathan@acm.org>
4451 * doc/extend.texi (Backwards Compatibility): Mention friend
4452 injection. Note for-scope is deprecated.
4453 * doc/invoke.texi (-ffriend-injection): Deprecate.
4455 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
4457 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
4458 that moved to I2, also allow destinations that are a paradoxical
4459 subreg (instead of a normal reg).
4461 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
4464 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
4467 2018-02-16 Richard Biener <rguenther@suse.de>
4469 PR tree-optimization/84037
4470 PR tree-optimization/84016
4472 * config/i386/i386.c (ix86_builtin_vectorization_cost):
4473 Adjust vec_construct for the fact we need additional higher latency
4474 128bit inserts for AVX256 and AVX512 vector builds.
4475 (ix86_add_stmt_cost): Scale vector construction cost for
4478 2018-02-16 Richard Biener <rguenther@suse.de>
4480 PR tree-optimization/84417
4481 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
4482 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
4483 (non_rewritable_lvalue_p): Likewise, use poly-ints.
4485 2018-02-16 Martin Liska <mliska@suse.cz>
4488 * internal-fn.def (ASAN_CHECK): Set proper flags.
4489 (ASAN_MARK): Likewise.
4491 2018-02-16 Julia Koval <julia.koval@intel.com>
4493 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
4494 from PTA_CANNONLAKE.
4496 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4499 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
4500 Use ++iter rather than iter++ for std::list iterators.
4501 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
4502 defer deleting them until all nodes in the forest are processed. Do
4503 free even leaf nodes. Change to_process into auto_vec.
4506 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
4507 * vec.h (vec_default_construct): Use memset instead of placement new
4508 if BROKEN_VALUE_INITIALIZATION is defined.
4509 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
4510 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
4513 PR rtl-optimization/83723
4514 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
4515 * lra.c (lra_substitute_pseudo): Likewise. If true, use
4516 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
4518 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
4520 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
4522 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
4524 PR rtl-optimization/81443
4525 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
4526 from inner REGs to paradoxical SUBREGs.
4528 2018-02-16 Richard Biener <rguenther@suse.de>
4530 PR tree-optimization/84399
4531 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4532 For operands we can analyze at their definition make sure we can
4533 analyze them at each use as well.
4535 2018-02-16 Richard Biener <rguenther@suse.de>
4537 PR tree-optimization/84190
4538 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
4539 volatile accesses if the decl isn't volatile.
4541 2018-02-15 Jason Merrill <jason@redhat.com>
4543 PR c++/84314 - ICE with templates and fastcall attribute.
4544 * attribs.c (build_type_attribute_qual_variant): Don't clobber
4545 TYPE_CANONICAL on an existing type.
4547 2018-02-15 Jakub Jelinek <jakub@redhat.com>
4549 PR tree-optimization/84383
4550 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
4551 dstoff nor call operand_equal_p if dstbase is NULL.
4553 PR tree-optimization/84334
4554 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
4555 also a CONSTANT_CLASS_P, punt.
4557 2018-02-14 Jim Wilson <jimw@sifive.com>
4559 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
4560 first SMALL_OPERAND check. New local min_second_step. Move assert
4561 to where locals are set. Add TARGET_RVC support.
4562 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
4564 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
4566 * doc/invoke.texi: Correct -Wformat-overflow code sample.
4568 2018-02-14 Martin Sebor <msebor@redhat.com>
4570 PR tree-optimization/83698
4571 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
4572 arrays constrain the offset range to their bounds.
4573 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
4574 (builtin_access::overlap): Avoid setting the size of overlap if it's
4576 (maybe_diag_overlap): Also consider arrays when deciding what values
4577 of offsets to include in diagnostics.
4579 2018-02-14 Martin Sebor <msebor@redhat.com>
4582 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
4583 that correspond to the kind of a declaration.
4585 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
4588 * config/pa/pa.md: Load address of PIC label using the linkage table
4589 if the label is nonlocal.
4591 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
4593 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
4594 warning message if user requests -maltivec=be.
4595 * doc/invoke.texi: Document deprecation of -maltivec=be.
4597 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
4600 * config/rs6000/rs6000-c.c: Update definitions for
4601 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
4602 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
4604 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4607 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
4608 add _get_ssp intrinsics. Remove argument from
4609 __builtin_ia32_rdssp[d|q].
4610 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
4611 * config/i386/i386-builtin.def: Remove argument from
4612 __builtin_ia32_rdssp[d|q].
4613 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
4614 ix86_expand_special_args_builtin for _rdssp[d|q].
4615 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
4616 Clear register before usage.
4617 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
4618 Add documentation for new _get_ssp and _inc_ssp intrinsics.
4620 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
4622 PR tree-optimization/84357
4623 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
4624 operand 1 of an ARRAY_REF too.
4626 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
4629 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
4630 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4632 (set_of_reg): New struct.
4633 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
4634 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
4635 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4637 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
4638 Split into bitclr, bitset, bitinvert patterns if appropriate.
4639 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
4640 use rx_fuse_in_memory_bitop.
4641 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
4642 to named insn, correct maximum insn length.
4644 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
4647 * machmode.def: Define a complex mode for PARTIAL_INT.
4648 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
4650 * doc/rtl.texi: Document CSPImode.
4651 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
4653 (msp430_hard_regno_nregs_with_padding): Likewise.
4655 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
4658 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
4660 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
4662 PR rtl-optimization/84169
4663 * combine.c (try_combine): New variable split_i2i3. Set it to true if
4664 we generated a parallel as new i3 and we split that to new i2 and i3
4665 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
4666 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
4667 those to i2, not i1. Partially rewrite this scan code.
4669 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4672 * stor-layout.c (place_field): For variable length fields, adjust
4673 offset_align afterwards not just based on the field's alignment,
4674 but also on the size.
4677 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
4678 of exps and logs in the use_exp2 case.
4680 2018-02-13 Jeff Law <law@redhat.com>
4682 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
4685 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
4688 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
4692 * common.opt (gas-loc-support, gas-locview-support): New.
4693 (ginline-points, ginternal-reset-location-views): New.
4694 * doc/invoke.texi: Document them. Use @itemx where intended.
4695 (gvariable-location-views): Adjust.
4696 * target.def (reset_location_view): New.
4697 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
4698 (TARGET_RESET_LOCATION_VIEW): New.
4699 * doc/tm.texi: Rebuilt.
4700 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
4701 (dwarf2out_default_as_locview_support): New.
4702 (output_asm_line_debug_info): Use option variables.
4703 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
4704 (output_loc_list): Likewise.
4705 (add_high_low_attributes): Check option variables.
4706 Don't output entry view attribute in strict mode.
4707 (gen_inlined_subroutine_die): Check option variables.
4708 (dwarf2out_inline_entry): Likewise.
4709 (init_sections_and_labels): Likewise.
4710 (dwarf2out_early_finish): Likewise.
4711 (maybe_reset_location_view): New, from...
4712 (dwarf2out_var_location): ... here. Call it.
4713 * debug.h (dwarf2out_default_as_loc_support): Declare.
4714 (dwarf2out_default_as_locview_support): Declare.
4715 * hooks.c (hook_int_rtx_insn_0): New.
4716 * hooks.h (hook_int_rtx_insn_0): Declare.
4717 * toplev.c (process_options): Take -gas-loc-support and
4718 -gas-locview-support from dwarf2out. Enable
4719 -gvariable-location-views by default only with locview
4720 assembler support. Enable -ginternal-reset-location-views by
4721 default only if the target defines the corresponding hook.
4722 Enable -ginline-points by default if location views are
4723 enabled; force it disabled if statement frontiers are
4725 * tree-inline.c (expand_call_inline): Check option variables.
4726 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
4728 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
4730 PR tree-optimization/84321
4731 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
4732 handling. Also check whether the anti-range contains any values
4733 that satisfy the mask; switch to a VR_RANGE if not.
4735 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
4738 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
4740 2018-02-13 Martin Jambor <mjambor@suse.cz>
4743 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
4744 of call statements, also set location of a load to a temporary.
4746 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
4748 * config/rl78/rl78.c (add_vector_labels): New function.
4749 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
4750 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
4751 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
4752 which checks that no arguments are passed.
4753 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
4754 * doc/extend.texi: Documentation for the new attribute.
4756 2018-02-13 Andreas Schwab <schwab@suse.de>
4758 * config/riscv/linux.h (CPP_SPEC): Define.
4760 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4763 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
4764 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
4765 OPTION_MASK_ISA_AES as first argument to def_builtin_const
4766 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
4767 instead of OPTION_MASK_ISA_PCLMUL as first argument to
4768 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
4769 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
4770 temporarily for AES and PCLMUL builtins.
4772 PR tree-optimization/84339
4773 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
4774 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
4778 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
4779 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
4780 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
4782 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
4784 * omp-simd-clone.h: New file.
4785 * omp-simd-clone.c: Include omp-simd-clone.h.
4786 (expand_simd_clones): No longer static.
4787 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
4788 cgraph.h and omp-simd-clone.h.
4789 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
4790 (vect_recog_widen_shift_pattern): Formatting fix.
4791 (vect_pattern_recog_1): Don't check optab for calls.
4794 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
4795 operands[2] into a REG before using gen_lowpart on it.
4797 2018-02-12 Jeff Law <law@redhat.com>
4800 * config/sh/sh.c (find_barrier): Consider a sibling call
4803 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
4804 successfully back substituting a reg.
4806 2018-02-12 Richard Biener <rguenther@suse.de>
4808 PR tree-optimization/84037
4809 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
4810 parameter, move visited init to caller.
4811 (vect_slp_analyze_operations): Separate cost from validity
4812 check, initialize visited once for all instances.
4813 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
4815 * tree-vect-stmts.c (vect_model_simple_cost): Make early
4817 (vect_model_promotion_demotion_cost): Likewise.
4818 (vectorizable_bswap): Guard cost modeling with !slp_node
4819 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
4821 (vectorizable_call): Likewise.
4822 (vectorizable_conversion): Likewise.
4823 (vectorizable_assignment): Likewise.
4824 (vectorizable_shift): Likewise.
4825 (vectorizable_operation): Likewise.
4826 (vectorizable_store): Likewise.
4827 (vectorizable_load): Likewise.
4828 (vectorizable_condition): Likewise.
4829 (vectorizable_comparison): Likewise.
4831 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
4834 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
4835 (ASAN_MARK): Fix fnspec to account for return value, change pointer
4836 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
4838 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4841 * params.def (inline-min-speedup): Increase from 8 to 15.
4842 (max-inline-insns-auto): Decrease from 40 to 30.
4843 * ipa-split.c (consider_split): Add some buffer for function to
4844 be considered inlining candidate.
4845 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
4848 2018-02-12 Richard Biener <rguenther@suse.de>
4850 PR tree-optimization/84037
4851 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
4852 matched stmts if we cannot swap the non-matched ones.
4854 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4856 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
4857 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
4858 _mm_maskz_scalef_round_ss): New intrinsics.
4859 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
4860 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
4861 __builtin_ia32_scalefss_round): Remove.
4862 (__builtin_ia32_scalefsd_mask_round,
4863 __builtin_ia32_scalefss_mask_round): New intrinsics.
4864 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
4865 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
4866 ((match_operand:VF_128 2 "<round_nimm_predicate>"
4867 "<round_constraint>")): Changed to ...
4868 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
4869 "<round_scalar_constraint>")): ... this.
4870 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
4871 %0, %1, %2<round_op3>}"): Changed to ...
4872 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
4873 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
4874 %2<round_scalar_mask_op3>}"): ... this.
4875 * config/i386/subst.md (round_scalar_nimm_predicate): New.
4877 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4879 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
4880 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
4881 (_mm_maskz_sqrt_round_ss): New intrinsics.
4882 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
4883 (__builtin_ia32_sqrtsd_mask_round)
4884 (__builtin_ia32_sqrtss_mask_round): New builtins.
4885 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
4886 (__builtin_ia32_sqrtss_round): Remove.
4887 (__builtin_ia32_sqrtsd_mask_round)
4888 (__builtin_ia32_sqrtss_mask_round): New builtins.
4889 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
4890 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
4891 ((match_operand:VF_128 1 "vector_operand"
4892 "xBm,<round_constraint>")): Changed to ...
4893 ((match_operand:VF_128 1 "vector_operand"
4894 "xBm,<round_scalar_constraint>")): ... this.
4895 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
4896 %0, %2, %<iptr>1<round_op3>}): Changed to ...
4897 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
4898 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
4899 %<iptr>1<round_scalar_mask_op3>}): ... this.
4900 ((set_attr "prefix" "<round_prefix>")): Changed to ...
4901 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
4903 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
4906 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
4907 Cast vec_cmpeq result to correct type.
4908 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
4909 Cast vec_cmpgt result to correct type.
4911 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
4913 * final.c (final_scan_insn_1): Renamed from...
4914 (final_scan_insn): ... this. New wrapper, to recover
4915 seen from the outermost call in recursive ones.
4916 * config/sparc/sparc.c (output_return): Drop seen from call.
4917 (output_sibcall): Likewise.
4918 * config/visium/visium.c (output_branch): Likewise.
4920 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
4922 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
4925 2018-02-10 Alan Modra <amodra@gmail.com>
4928 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
4929 Specify LR as an input.
4931 2018-02-10 Jakub Jelinek <jakub@redhat.com>
4934 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
4935 remove_member_access_dummy_vars): New functions.
4936 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
4937 lower_omp_1, execute_lower_omp): Use them.
4939 PR rtl-optimization/84308
4940 * shrink-wrap.c (spread_components): Release todo vector.
4942 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
4944 PR rtl-optimization/57193
4945 * ira-color.c (struct allocno_color_data): Add member
4946 conflict_allocno_hard_prefs.
4947 (update_conflict_allocno_hard_prefs): New.
4948 (bucket_allocno_compare_func): Add a preference based on
4949 conflict_allocno_hard_prefs.
4950 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
4951 (color_allocnos): Remove a dead code. Initiate
4952 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
4954 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4957 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
4958 constraint from =wa to wa. Avoid a subreg on the output operand,
4959 instead use a pseudo and subreg it in a move.
4960 (p9_xxbrd_<mode>): Changed to ...
4961 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
4962 (p9_xxbrd_v2df): New expander.
4963 (p9_xxbrw_<mode>): Changed to ...
4964 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
4965 (p9_xxbrw_v4sf): New expander.
4967 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4969 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
4971 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
4974 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
4975 multiply in 32-bit mode.
4976 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
4977 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
4980 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4982 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
4983 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
4984 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
4985 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
4987 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
4990 * dwarf2out.c (is_trivial_indirect_ref): New function.
4991 (dwarf2out_late_global_decl): Do not generate a location
4992 attribute for variables that have a non-trivial DECL_VALUE_EXPR
4993 and that are not defined in the current unit.
4995 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4997 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
4998 instead of a libcall for UNORDERED.
5000 2018-02-09 Tamar Christina <tamar.christina@arm.com>
5003 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
5004 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
5006 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5009 * config/s390/s390.c (s390_set_current_function): Invoke
5010 s390_indirect_branch_settings also if fndecl didn't change.
5012 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5014 * config/rs6000/rs6000.md (blockage): Set length to zero.
5016 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5018 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
5020 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5023 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
5024 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
5028 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
5029 PARALLEL incoming that failed vt_get_decl_and_offset check.
5032 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
5033 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
5035 (get_variable_section): For decls in named .bss* sections pass true as
5036 second argument to bss_initializer_p.
5038 2018-02-09 Marek Polacek <polacek@redhat.com>
5039 Jakub Jelinek <jakub@redhat.com>
5042 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
5043 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
5044 Sync some changes from cxx_fold_indirect_ref.
5046 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5048 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
5050 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
5051 (BLOCK_INLINE_ENTRY_LABEL): New.
5052 (dwarf2out_var_location): Disregard inline entry markers.
5053 (inline_entry_data): New struct.
5054 (inline_entry_data_hasher): New hashtable type.
5055 (inline_entry_data_hasher::hash): New.
5056 (inline_entry_data_hasher::equal): New.
5057 (inline_entry_data_table): New variable.
5058 (add_high_low_attributes): Add DW_AT_entry_pc and
5059 DW_AT_GNU_entry_view attributes if a pending entry is found
5060 in inline_entry_data_table. Add old entry_pc attribute only
5061 if debug nonbinding markers are disabled.
5062 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
5063 markers are enabled.
5064 (block_within_block_p, dwarf2out_inline_entry): New.
5065 (dwarf2out_finish): Check that no entries remained in
5066 inline_entry_data_table.
5067 * final.c (reemit_insn_block_notes): Handle inline entry notes.
5068 (final_scan_insn, notice_source_line): Likewise.
5069 (rest_of_clean_state): Skip inline entry markers.
5070 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
5072 * gimple.c (gimple_build_debug_inline_entry): New.
5073 * gimple.h (enum gimple_debug_subcode): Add
5074 GIMPLE_DEBUG_INLINE_ENTRY.
5075 (gimple_build_debug_inline_entry): Declare.
5076 (gimple_debug_inline_entry_p): New.
5077 (gimple_debug_nonbind_marker_p): Adjust.
5078 * insn-notes.def (INLINE_ENTRY): New.
5079 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
5080 inline entry marker notes.
5081 (print_insn): Likewise.
5082 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
5083 (INSN_DEBUG_MARKER_KIND): Likewise.
5084 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
5085 * tree-inline.c (expand_call_inline): Build and insert
5086 debug_inline_entry stmt.
5087 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
5088 inline entry blocks early, if nonbind markers are enabled.
5089 (dump_scope_block): Dump fragment info.
5090 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
5091 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
5092 (gimple_build_debug_inline_entry): New.
5093 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
5094 Enable/disable inline entry points too.
5095 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
5096 (DEBUG_INSN): Describe inline entry markers.
5098 * common.opt (gvariable-location-views): New.
5099 (gvariable-location-views=incompat5): New.
5100 * config.in: Rebuilt.
5101 * configure: Rebuilt.
5102 * configure.ac: Test assembler for view support.
5103 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
5104 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
5105 * dwarf2out.c (var_loc_view): New typedef.
5106 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
5107 (dwarf2out_locviews_in_attribute): New.
5108 (dwarf2out_locviews_in_loclist): New.
5109 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
5110 (enum dw_line_info_opcode): Add LI_adv_address.
5111 (struct dw_line_info_table): Add view.
5112 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
5113 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
5114 (zero_view_p): New variable.
5115 (ZERO_VIEW_P): New macro.
5116 (output_asm_line_debug_info): New.
5117 (struct var_loc_node): Add view.
5118 (add_AT_view_list, AT_loc_list): New.
5119 (add_var_loc_to_decl): Add view param. Test it against last.
5120 (new_loc_list): Add view params. Record them.
5121 (AT_loc_list_ptr): Handle loc and view lists.
5122 (view_list_to_loc_list_val_node): New.
5123 (print_dw_val): Handle dw_val_class_view_list.
5124 (size_of_die): Likewise.
5125 (value_format): Likewise.
5126 (loc_list_has_views): New.
5127 (gen_llsym): Set vl_symbol too.
5128 (maybe_gen_llsym, skip_loc_list_entry): New.
5129 (dwarf2out_maybe_output_loclist_view_pair): New.
5130 (output_loc_list): Output view list or entries too.
5131 (output_view_list_offset): New.
5132 (output_die): Handle dw_val_class_view_list.
5133 (output_dwarf_version): New.
5134 (output_compilation_unit_header): Use it.
5135 (output_skeleton_debug_sections): Likewise.
5136 (output_rnglists, output_line_info): Likewise.
5137 (output_pubnames, output_aranges): Update version comments.
5138 (output_one_line_info_table): Output view numbers in asm comments.
5139 (dw_loc_list): Determine current endview, pass it to new_loc_list.
5140 Call maybe_gen_llsym.
5141 (loc_list_from_tree_1): Adjust.
5142 (add_AT_location_description): Create view list attribute if
5143 needed, check it's absent otherwise.
5144 (convert_cfa_to_fb_loc_list): Adjust.
5145 (maybe_emit_file): Call output_asm_line_debug_info for test.
5146 (dwarf2out_var_location): Reset views as needed. Precompute
5147 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
5148 attribute. Set view.
5149 (new_line_info_table): Reset next view.
5150 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
5151 (dwarf2out_source_line): Likewise. Output view resets and labels to
5152 the assembler, or select appropriate line info opcodes.
5153 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
5154 (optimize_string_length): Catch it. Adjust.
5155 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
5156 dw_val_class_view_list, and remove it if no longer needed.
5157 (hash_loc_list): Hash view numbers.
5158 (loc_list_hasher::equal): Compare them.
5159 (optimize_location_lists): Check whether a view list symbol is
5160 needed, and whether the locview attribute is present, and
5161 whether they match. Remove the locview attribute if no longer
5163 (index_location_lists): Call skip_loc_list_entry for test.
5164 (dwarf2out_finish): Call output_asm_line_debug_info for test.
5165 Use output_dwarf_version.
5166 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
5167 (struct dw_val_node): Add val_view_list.
5168 * final.c (SEEN_NEXT_VIEW): New.
5169 (set_next_view_needed): New.
5170 (clear_next_view_needed): New.
5171 (maybe_output_next_view): New.
5172 (final_start_function): Rename to...
5173 (final_start_function_1): ... this. Take pointer to FIRST,
5174 add SEEN parameter. Emit param bindings in the initial view.
5175 (final_start_function): Reintroduce SEEN-less interface.
5176 (final): Rename to...
5177 (final_1): ... this. Take SEEN parameter. Output final pending
5178 next view at the end.
5179 (final): Reintroduce seen-less interface.
5180 (final_scan_insn): Output pending next view before switching
5181 sections or ending a block. Mark the next view as needed when
5182 outputting variable locations. Notify debug backend of section
5183 changes, and of location view changes.
5184 (rest_of_handle_final): Adjust.
5185 * toplev.c (process_options): Autodetect value for debug variable
5186 location views option. Warn on incompat5 without -gdwarf-5.
5187 * doc/invoke.texi (gvariable-location-views): New.
5188 (gvariable-location-views=incompat5): New.
5189 (gno-variable-location-views): New.
5191 2018-02-08 David Malcolm <dmalcolm@redhat.com>
5193 PR tree-optimization/84136
5194 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
5195 that the result of find_edge is non-NULL.
5197 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5200 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
5201 storing integer register in SImode. Fix cost of 256 and 512
5202 byte aligned SSE register store.
5204 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5206 * config/i386/i386.c (ix86_multiplication_cost): Fix
5207 multiplication cost for TARGET_AVX512DQ.
5209 2018-02-08 Marek Polacek <polacek@redhat.com>
5211 PR tree-optimization/84238
5212 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
5215 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5217 PR tree-optimization/84265
5218 * tree-vect-stmts.c (vectorizable_store): Don't treat
5219 VMAT_CONTIGUOUS accesses as grouped.
5220 (vectorizable_load): Likewise.
5222 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5224 PR tree-optimization/81635
5225 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
5226 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
5227 (test_round_for_mask): New functions.
5228 (wide_int_cc_tests): Call test_round_for_mask.
5229 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
5230 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
5231 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
5232 range returned by get_range_info.
5234 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5237 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
5238 * symtab.c: Include builtins.h
5239 (symtab_node::output_to_lto_symbol_table_p): Move here
5240 from lto-streamer-out.c:output_symbol_p.
5241 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
5242 (output_symbol_p): Move all logic to symtab.c
5243 (produce_symtab): Update.
5245 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5247 * config/s390/s390-opts.h (enum indirect_branch): Define.
5248 * config/s390/s390-protos.h (s390_return_addr_from_memory)
5249 (s390_indirect_branch_via_thunk)
5250 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
5251 (enum s390_indirect_branch_type): Define.
5252 * config/s390/s390.c (struct s390_frame_layout, struct
5253 machine_function): Remove.
5254 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
5255 (indirect_branch_table_label_no, indirect_branch_table_name):
5257 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
5258 (enum s390_indirect_branch_option): Define.
5259 (s390_return_addr_from_memory): New function.
5260 (s390_handle_string_attribute): New function.
5261 (s390_attribute_table): Add new attribute handler.
5262 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
5263 (s390_indirect_branch_via_thunk): New function.
5264 (s390_indirect_branch_via_inline_thunk): New function.
5265 (s390_function_ok_for_sibcall): When jumping via thunk disallow
5266 sibling call optimization for non z10 compiles.
5267 (s390_emit_call): Force indirect branch target to be a single
5268 register. Add r1 clobber for non-z10 compiles.
5269 (s390_emit_epilogue): Emit return jump via return_use expander.
5270 (s390_reorg): Handle JUMP_INSNs as execute targets.
5271 (s390_option_override_internal): Perform validity checks for the
5272 new command line options.
5273 (s390_indirect_branch_attrvalue): New function.
5274 (s390_indirect_branch_settings): New function.
5275 (s390_set_current_function): Invoke s390_indirect_branch_settings.
5276 (s390_output_indirect_thunk_function): New function.
5277 (s390_code_end): Implement target hook.
5278 (s390_case_values_threshold): Implement target hook.
5279 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
5281 * config/s390/s390.h (struct s390_frame_layout)
5282 (struct machine_function): Move here from s390.c.
5283 (TARGET_INDIRECT_BRANCH_NOBP_RET)
5284 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
5285 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
5286 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
5287 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
5288 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
5289 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
5290 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
5291 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
5292 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
5293 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
5294 (mnemonic attribute): Add values which aren't recognized
5296 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
5297 pattern for branch conversion. Fix mnemonic attribute.
5298 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
5299 indirect branch via thunk if requested.
5300 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
5301 ("*indirect_jump"): Disable for branch conversion using out of
5303 ("indirect_jump_via_thunk<mode>_z10")
5304 ("indirect_jump_via_thunk<mode>")
5305 ("indirect_jump_via_inlinethunk<mode>_z10")
5306 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
5307 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
5308 ("casesi_jump_via_inlinethunk<mode>_z10")
5309 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
5310 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
5311 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
5312 ("*indirect2_jump"): Disable for branch conversion.
5313 ("casesi_jump"): Turn into expander and expand patterns for branch
5315 ("return_use"): New expander.
5316 ("*return"): Emit return via thunk and rename it to ...
5317 ("*return<mode>"): ... this one.
5318 * config/s390/s390.opt: Add new options and and enum for the
5321 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5323 * lra-constraints.c (match_reload): Unconditionally use
5324 gen_lowpart_SUBREG, rather than selecting between that
5325 and equivalent gen_rtx_SUBREG code.
5327 2018-02-08 Richard Biener <rguenther@suse.de>
5329 PR tree-optimization/84233
5330 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
5331 changed flag instead of boguously re-using phi_inserted.
5333 2018-02-08 Martin Jambor <mjambor@suse.cz>
5335 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
5336 static local variables.
5338 2018-02-08 Richard Biener <rguenther@suse.de>
5340 PR tree-optimization/84278
5341 * tree-vect-stmts.c (vectorizable_store): When looking for
5342 smaller vector types to perform grouped strided loads/stores
5343 make sure the mode is supported by the target.
5344 (vectorizable_load): Likewise.
5346 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5348 * config/aarch64/aarch64.c (aarch64_components_for_bb):
5349 Increase LDP/STP opportunities by adding adjacent callee-saves.
5351 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5353 PR rtl-optimization/84068
5354 PR rtl-optimization/83459
5355 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
5357 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
5359 PR tree-optimization/84224
5360 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
5361 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
5364 2018-02-07 Iain Sandoe <iain@codesourcery.com>
5367 * config/rs6000/altivec.md (*restore_world): Remove LR use.
5368 * config/rs6000/predicates.md (restore_world_operation): Adjust op
5369 count, remove one USE.
5371 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
5373 * doc/install.texi (Configuration): Document the
5374 --with-long-double-format={ibm,ieee} PowerPC configuration
5378 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
5379 Convert from define_expand to be define_insn_and_split. Rework
5380 float/double/_Float128 conversions to QI/HI/SImode to work with
5381 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
5382 conversions to QI/HImode types did a store and then a load to
5383 truncate the value. For conversions to VSX registers, don't split
5384 the insn, instead emit the code directly. Use the code iterator
5385 any_fix to combine signed and unsigned conversions.
5386 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
5387 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
5388 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5389 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
5390 (fix_<mode>di2_hw): Likewise.
5391 (fixuns_<mode>di2_hw): Likewise.
5392 (fix_<mode>si2_hw): Likewise.
5393 (fixuns_<mode>si2_hw): Likewise.
5394 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
5395 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5396 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
5397 fix<uns>_trunc<SFDF:mode>si2_p8.
5398 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
5400 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
5401 (fix<uns>_<mode>_mem): Likewise.
5402 (fctiw<u>z_<mode>_mem): Likewise.
5403 (fix<uns>_<mode>_mem): Likewise.
5404 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
5405 the register allocator from doing a direct move to the GPRs to do
5406 a store, and instead use the ISA 3.0 store byte/half-word from
5407 vector register instruction. For IEEE 128-bit floating point,
5408 also optimize stores of 32-bit ints.
5409 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
5411 2018-02-07 Alan Hayward <alan.hayward@arm.com>
5413 * genextract.c (push_pathstr_operand): New function to support
5415 (walk_rtx): Call push_pathstr_operand.
5416 (print_path): Support [a-zA-Z].
5418 2018-02-07 Richard Biener <rguenther@suse.de>
5420 PR tree-optimization/84037
5421 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
5422 (cse_and_gimplify_to_preheader): Declare.
5423 (vect_get_place_in_interleaving_chain): Likewise.
5424 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5426 (_loop_vec_info::~_loop_vec_info): Delete it.
5427 (cse_and_gimplify_to_preheader): New function.
5428 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
5429 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
5430 (vectorizable_load): Likewise. For grouped stores always base
5431 the IV on the first element.
5432 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
5433 condition before gimplifying.
5435 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5437 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
5438 *DIV_EXPR and *MOD_EXPR.
5440 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
5443 * config/i386/i386.c (ix86_option_override_internal): Mask out
5444 the CF_SET bit when checking -fcf-protection.
5446 2018-02-07 Tom de Vries <tom@codesourcery.com>
5449 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
5452 2018-02-07 Richard Biener <rguenther@suse.de>
5454 PR tree-optimization/84204
5455 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
5458 PR tree-optimization/84205
5459 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
5460 special-case isl_ast_op_zdiv_r.
5462 PR tree-optimization/84223
5463 * graphite-scop-detection.c (gather_bbs::before_dom_children):
5464 Only add conditions from within the region.
5465 (gather_bbs::after_dom_children): Adjust.
5467 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
5470 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
5471 * config/avr/avr.md: Only post-reload split REG-REG moves if
5472 either register is GENERAL_REG_P.
5474 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5476 PR tree-optimization/84235
5477 * tree-ssa-scopedtables.c
5478 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
5479 if the subtraction is performed in floating point type where NaNs are
5480 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
5481 build 1. Formatting fix.
5483 2018-02-06 Jakub Jelinek <jakub@redhat.com>
5486 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
5487 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
5488 and skip it regardless of bb boundaries. Use CALL_P macro,
5489 don't test INSN_P (insn) together with CALL_P or JUMP_P check
5490 unnecessarily, formatting fix.
5492 2018-02-06 Michael Collison <michael.collison@arm.com>
5494 * config/arm/thumb2.md:
5495 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
5496 (*thumb_mov_notscc): Ditto.
5498 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
5501 * config/rs6000/rs6000.md (su code attribute): Use "u" for
5502 unsigned_fix, not "s".
5504 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5506 * configure.ac (gcc_fn_eh_frame_ro): New function.
5507 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
5508 correct .eh_frame permissions.
5509 * configure: Regenerate.
5511 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
5513 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
5516 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5518 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5519 Display warning message for -mno-speculate-indirect-jumps.
5521 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
5523 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
5525 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
5527 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
5529 PR tree-optimization/84225
5530 * tree-eh.c (find_trapping_overflow): Only call
5531 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
5533 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5536 * config/i386/i386.c: Reimplement the check of possible options
5537 -mibt/-mshstk conbination. Change error messages.
5538 * doc/invoke.texi: Fix a typo: remove extra '='.
5540 2018-02-06 Marek Polacek <polacek@redhat.com>
5542 PR tree-optimization/84228
5543 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
5545 2018-02-06 Tamar Christina <tamar.christina@arm.com>
5548 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
5549 emitted arch directives.
5550 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
5551 __ARM_FEATURE_COPROC before changing architectures.
5553 2018-02-06 Richard Biener <rguenther@suse.de>
5555 * config/i386/i386.c (print_reg): Fix typo.
5556 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
5558 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
5560 * configure: Regenerate.
5562 2018-02-05 Martin Sebor <msebor@redhat.com>
5564 PR tree-optimization/83369
5565 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
5568 2018-02-05 Martin Liska <mliska@suse.cz>
5570 * doc/invoke.texi: Cherry-pick upstream r323995.
5572 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
5574 * ira.c (ira_init_register_move_cost): Adjust comment.
5576 2018-02-05 Martin Liska <mliska@suse.cz>
5578 PR gcov-profile/84137
5579 * doc/gcov.texi: Fix typo in documentation.
5581 2018-02-05 Martin Liska <mliska@suse.cz>
5583 PR gcov-profile/83879
5584 * doc/gcov.texi: Document necessity of --dynamic-list-data when
5585 using dlopen functionality.
5587 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
5589 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
5590 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
5591 _mm_maskz_range_ss, _mm_mask_range_round_ss,
5592 _mm_maskz_range_round_ss): New intrinsics.
5593 (__builtin_ia32_rangesd128_round)
5594 (__builtin_ia32_rangess128_round): Remove.
5595 (__builtin_ia32_rangesd128_mask_round,
5596 __builtin_ia32_rangess128_mask_round): New builtins.
5597 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
5598 __builtin_ia32_rangess128_round): Remove.
5599 (__builtin_ia32_rangesd128_mask_round,
5600 __builtin_ia32_rangess128_mask_round): New builtins.
5601 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
5602 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
5603 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
5604 "<round_saeonly_constraint>")): Changed to ...
5605 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
5606 "<round_saeonly_scalar_constraint>")): ... this.
5607 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
5608 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
5609 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
5610 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
5611 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
5613 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
5615 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
5617 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
5618 Remove all values except native, 8540 and 8548.
5620 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
5622 * config/i386/i386.c (ix86_output_function_return): Pass
5623 INVALID_REGNUM, instead of -1, as invalid register number to
5624 indirect_thunk_name and output_indirect_thunk.
5626 2018-02-02 Julia Koval <julia.koval@intel.com>
5628 * config.gcc: Add -march=icelake.
5629 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
5630 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
5631 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
5632 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
5633 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
5634 (processor_target_table): Add icelake.
5635 (ix86_option_override_internal): Handle new PTAs.
5636 (get_builtin_code_for_version): Handle icelake.
5637 (M_INTEL_COREI7_ICELAKE): New.
5638 (fold_builtin_cpu): Handle icelake.
5639 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
5640 * doc/invoke.texi: Add -march=icelake.
5642 2018-02-02 Julia Koval <julia.koval@intel.com>
5644 * config/i386/i386.c (ix86_option_override_internal): Change flags type
5645 to wide_int_bitmask.
5646 * wide-int-bitmask.h: New.
5648 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5651 * config/i386/i386.md: Replace Pmode with word_mode in
5652 builtin_setjmp_setup and builtin_longjmp to support x32.
5654 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
5658 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
5660 (rs6000_supported_cpu_names): New static variable.
5661 (linux_cpu_translation_table): Likewise.
5662 (elf_platform) <cpu>: Define new static variable and use it.
5663 Translate kernel AT_PLATFORM name to canonical name if needed.
5664 Error if platform name is unknown.
5666 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
5669 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
5671 2018-02-01 Jeff Law <law@redhat.com>
5674 * config/i386/i386.c (release_scratch_register_on_entry): Add new
5675 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
5676 the scratch if RELEASE_VIA_POP is false.
5677 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
5678 If we have to save a temporary register, decrement SIZE appropriately.
5679 Pass new arguments to release_scratch_register_on_entry.
5680 (ix86_adjust_stack_and_probe): Likewise.
5681 (ix86_emit_probe_stack_range): Pass new arguments to
5682 release_scratch_register_on_entry.
5684 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
5686 PR rtl-optimization/84157
5687 * combine.c (change_zero_ext): Use REG_P predicate in
5688 front of HARD_REGISTER_P predicate.
5690 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
5692 * config/avr/avr.c (avr_option_override): Move disabling of
5693 -fdelete-null-pointer-checks to...
5694 * common/config/avr/avr-common.c (avr_option_optimization_table):
5697 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5699 PR tree-optimization/81635
5700 * tree-data-ref.c (split_constant_offset_1): For types that
5701 wrap on overflow, try to use range info to prove that wrapping
5704 2018-02-01 Renlin Li <renlin.li@arm.com>
5707 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
5709 (aarch64_register_move_cost): Likewise.
5710 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
5712 (REG_CLASS_NAMES): Likewise.
5713 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
5714 TAILCALL_ADDR_REGS. Remove IP registers.
5715 * config/aarch64/aarch64.md (Ucs): Update register constraint.
5717 2018-02-01 Richard Biener <rguenther@suse.de>
5719 * domwalk.h (dom_walker::dom_walker): Add additional constructor
5720 for specifying RPO order and allow NULL for that.
5721 * domwalk.c (dom_walker::dom_walker): Likewise.
5722 (dom_walker::walk): Handle NULL RPO order.
5723 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
5725 (rewrite_update_dom_walker): Likewise.
5726 (mark_def_dom_walker): Likewise.
5728 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5730 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
5731 (aarch64_maybe_expand_sve_subreg_move): Declare.
5732 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
5733 * config/aarch64/predicates.md (aarch64_any_register_operand): New
5735 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
5736 that are semantically a reverse operation.
5737 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
5738 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
5739 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
5741 (aarch64_can_change_mode_class): For big-endian, forbid changes
5742 between two SVE modes if they have different element sizes.
5744 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5746 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
5747 the TImode handling for big-endian targets.
5749 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5751 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
5752 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
5754 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
5755 Remove BSWAP handing for big-endian targets and use the form of
5756 LD1RQ appropariate for the mode.
5758 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5760 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
5761 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
5764 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5767 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
5768 check for operands that need to go through aarch64_sve_reload_be.
5770 2018-02-01 Jakub Jelinek <jakub@redhat.com>
5772 PR tree-optimization/81661
5773 PR tree-optimization/84117
5774 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
5775 * tree-eh.c: Include gimplify.h.
5776 (find_trapping_overflow, replace_trapping_overflow,
5777 rewrite_to_non_trapping_overflow): New functions.
5778 * tree-vect-loop.c: Include tree-eh.h.
5779 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
5780 * tree-data-ref.c: Include tree-eh.h.
5781 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
5783 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
5785 PR rtl-optimization/84123
5786 * combine.c (change_zero_ext): Check if hard register satisfies
5787 can_change_dest_mode before calling gen_lowpart_SUBREG.
5789 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
5792 * ira.c (ira_init_register_move_cost): Remove assert.
5794 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5796 PR rtl-optimization/84071
5797 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
5798 * doc/tm.texi: Regenerate.
5800 2018-01-31 Richard Biener <rguenther@suse.de>
5802 PR tree-optimization/84132
5803 * tree-data-ref.c (analyze_miv_subscript): Properly
5804 check whether evolution_function_is_affine_multivariate_p
5805 before calling gcd_of_steps_may_divide_p.
5807 2018-01-31 Julia Koval <julia.koval@intel.com>
5810 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
5811 * config/i386/i386.md (rdpid_rex64) New.
5812 (rdpid): Make 32bit only.
5814 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
5817 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
5818 an IDENTIFIER_NODE for FUNCTION_TYPE's.
5820 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5823 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5825 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5827 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5829 PR rtl-optimization/84071
5830 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
5831 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
5833 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5835 * config/arc/arc.c (arc_handle_aux_attribute): New function.
5836 (arc_attribute_table): Add 'aux' attribute.
5837 (arc_in_small_data_p): Consider aux like variables.
5838 (arc_is_aux_reg_p): New function.
5839 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
5840 (arc_get_aux_arg): New function.
5841 (prepare_move_operands): Handle aux-register access.
5842 (arc_handle_aux_attribute): New function.
5843 * doc/extend.texi (ARC Variable attributes): Add subsection.
5845 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5847 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
5848 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
5849 (arc_attribute_table): Add 'uncached' attribute.
5850 (arc_print_operand): Print '.di' flag for uncached memory
5852 (arc_in_small_data_p): Do not consider for small data the uncached
5854 (arc_is_uncached_mem_p): New function.
5855 * config/arc/predicates.md (compact_store_memory_operand): Check
5856 for uncached memory accesses.
5857 (nonvol_nonimm_operand): Likewise.
5858 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
5860 2018-01-31 Jakub Jelinek <jakub@redhat.com>
5863 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
5864 falign-loops=): Add Optimization flag.
5866 2018-01-30 Jeff Law <law@redhat.com>
5869 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
5870 INT_REGISTERS_SAVED. Check it prior to calling
5871 get_scratch_register_on_entry.
5872 (ix86_adjust_stack_and_probe): Similarly.
5873 (ix86_emit_probe_stack_range): Similarly.
5874 (ix86_expand_prologue): Corresponding changes.
5876 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5879 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
5880 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
5882 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
5885 * lra-constraints.c (curr_insn_transform): Process AND in the
5888 2018-01-30 Jakub Jelinek <jakub@redhat.com>
5890 PR rtl-optimization/83986
5891 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
5892 dependence against last_pending_memory_flush in addition to
5895 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
5897 PR tree-optimization/81611
5898 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
5901 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5904 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
5907 2018-01-30 Richard Biener <rguenther@suse.de>
5908 Jakub Jelinek <jakub@redhat.com>
5910 PR tree-optimization/84111
5911 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
5912 inner loops added during recursion, as they don't have up-to-date
5915 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5918 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
5919 (can_inline_edge_by_limits_p): ... here.
5920 (can_early_inline_edge_p, check_callers,
5921 update_caller_keys, update_callee_keys, recursive_inlining,
5922 add_new_edges_to_heap, speculation_useful_p,
5923 inline_small_functions,
5924 inline_small_functions, flatten_function,
5925 inline_to_all_callers_1): Update.
5927 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5929 * profile-count.c (profile_count::combine_with_ipa_count): Handle
5932 2018-01-30 Richard Biener <rguenther@suse.de>
5934 PR tree-optimization/83008
5935 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
5936 invariant and constant vector uses in stmts when they need
5939 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5942 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
5943 * configure: Regenerate.
5945 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5947 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
5949 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
5950 Use gen_rtx_REG rather than gen_lowpart.
5952 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5954 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
5955 rather than 0 when creating partial subregs.
5957 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5959 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
5962 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
5965 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
5966 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
5967 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
5968 flags. This restores the settings used before the 2017-07-24.
5969 Turning off pre increment/decrement/modify allows IVOPTS to
5970 optimize DF/SF loops where the index is an int.
5972 2018-01-29 Richard Biener <rguenther@suse.de>
5973 Kelvin Nilsen <kelvin@gcc.gnu.org>
5976 * tree-vect-stmts.c (vectorizable_call): Don't call
5977 targetm.vectorize_builtin_md_vectorized_function if callee is
5980 2018-01-22 Carl Love <cel@us.ibm.com>
5982 * doc/extend.tex: Fix typo in second arg in
5983 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
5985 2018-01-29 Richard Biener <rguenther@suse.de>
5987 PR tree-optimization/84086
5988 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
5989 (flush_ssaname_freelist): When SSA names were released reset
5990 the SCEV hash table.
5992 2018-01-29 Richard Biener <rguenther@suse.de>
5994 PR tree-optimization/84057
5995 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
5996 removed paths when removing edges.
5998 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
6000 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
6001 -mfunction-return=@var{choice}.
6003 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
6006 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
6008 (layout::print_source_line): Likewise.
6009 (test_get_line_width_without_trailing_whitespace): Add test cases.
6011 2018-01-27 Jakub Jelinek <jakub@redhat.com>
6014 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
6017 2018-01-26 Jim Wilson <jimw@sifive.com>
6019 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
6021 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
6024 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6026 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
6027 and CMP + SUB-immediate -> SUBS.
6029 2018-01-26 Martin Sebor <msebor@redhat.com>
6031 PR tree-optimization/83896
6032 * tree-ssa-strlen.c (get_string_len): Rename...
6033 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
6034 Avoid assuming length is constant.
6035 (handle_char_store): Use HOST_WIDE_INT for string length.
6037 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
6040 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
6041 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
6043 2018-01-26 Richard Biener <rguenther@suse.de>
6045 PR rtl-optimization/84003
6046 * dse.c (record_store): Only record redundant stores when
6047 the earlier store aliases at least all accesses the later one does.
6049 2018-01-26 Jakub Jelinek <jakub@redhat.com>
6051 PR rtl-optimization/83985
6052 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
6053 REG_CFA_RESTORE insns.
6054 (delete_unmarked_insns): Don't ignore separate shrink wrapping
6055 REG_CFA_RESTORE insns here.
6058 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
6059 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
6061 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6063 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
6064 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
6065 (arc_init): Likewise.
6066 (arc_override_options): Likewise.
6067 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
6069 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
6071 * config/arc/arc.h (TARGET_DBNZ): Define.
6072 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
6073 properly set the tune attribute.
6074 (dbnz): Use TARGET_DBNZ guard.
6075 * config/arc/arc.opt (mtune): Add core3 option.
6077 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6079 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
6080 recognize new pic like addresses.
6081 (arc_delegitimize_address): Clean up.
6083 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6085 * config/arc/arc-arches.def: Option mrf16 valid for all
6087 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
6088 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
6089 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
6090 * config/arc/arc-tables.opt: Regenerate.
6091 * config/arc/arc.c (arc_conditional_register_usage): Handle
6092 reduced register file case.
6093 (arc_file_start): Set must have build attributes.
6094 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
6096 * config/arc/arc.opt (mrf16): Add new option.
6097 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
6098 * config/arc/genmultilib.awk: Handle new mrf16 option.
6099 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
6100 * config/arc/t-multilib: Regenerate.
6101 * doc/invoke.texi (ARC Options): Document mrf16 option.
6103 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6105 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
6106 * config/arc/arc.c (arc_handle_secure_attribute): New function.
6107 (arc_attribute_table): Add 'secure_call' attribute.
6108 (arc_print_operand): Print secure call operand.
6109 (arc_function_ok_for_sibcall): Don't optimize tail calls when
6111 (arc_is_secure_call_p): New function. * config/arc/arc.md
6112 (call_i): Add support for sjli instruction.
6113 (call_value_i): Likewise.
6114 * config/arc/constraints.md (Csc): New constraint.
6116 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6117 John Eric Martin <John.Martin@emmicro-us.com>
6119 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
6120 * config/arc/arc.c (_arc_jli_section): New struct.
6121 (arc_jli_section): New type.
6122 (rc_jli_sections): New static variable.
6123 (arc_handle_jli_attribute): New function.
6124 (arc_attribute_table): Add jli_always and jli_fixed attribute.
6125 (arc_file_end): New function.
6126 (TARGET_ASM_FILE_END): Define.
6127 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
6128 (arc_add_jli_section): New function.
6129 (jli_call_scan): Likewise.
6130 (arc_reorg): Call jli_call_scan.
6131 (arc_output_addsi): Remove 'S' from printing asm operand.
6132 (arc_is_jli_call_p): New function.
6133 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
6135 (movhi_insn): Likewise.
6136 (movsi_insn): Likewise.
6137 (movsi_set_cc_insn): Likewise.
6138 (loadqi_update): Likewise.
6139 (load_zeroextendqisi_update): Likewise.
6140 (load_signextendqisi_update): Likewise.
6141 (loadhi_update): Likewise.
6142 (load_zeroextendhisi_update): Likewise.
6143 (load_signextendhisi_update): Likewise.
6144 (loadsi_update): Likewise.
6145 (loadsf_update): Likewise.
6146 (movsicc_insn): Likewise.
6147 (bset_insn): Likewise.
6148 (bxor_insn): Likewise.
6149 (bclr_insn): Likewise.
6150 (bmsk_insn): Likewise.
6151 (bicsi3_insn): Likewise.
6152 (cmpsi_cc_c_insn): Likewise.
6153 (movsi_ne): Likewise.
6154 (movsi_cond_exec): Likewise.
6155 (clrsbsi2): Likewise.
6166 (call_i): Remove 'S' asm letter, add jli instruction.
6167 (call_value_i): Likewise.
6168 * config/arc/arc.op (mjli-always): New option.
6169 * config/arc/constraints.md (Cji): New constraint.
6170 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
6172 (subsf3_fpx): Likewise.
6173 (mulsf3_fpx): Likewise.
6174 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
6176 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
6178 * doc/invoke.texi (ARC): Document mjli-always option.
6180 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
6182 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
6183 avoid addition with 0 and use incw and decw where possible.
6185 2018-01-26 Richard Biener <rguenther@suse.de>
6187 PR tree-optimization/81082
6188 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
6189 association if it requires casting to unsigned.
6190 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
6191 from fold_plusminus_mult_expr to catch important cases late when
6192 range info is available.
6194 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6196 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
6197 * configure.ac (hidden_linkonce): New test.
6198 * configure: Regenerate.
6199 * config.in: Regenerate.
6201 2018-01-26 Julia Koval <julia.koval@intel.com>
6203 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
6204 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
6205 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
6206 _mm_mask_bitshuffle_epi64_mask): Fix type.
6207 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
6208 USI_FTYPE_V4DI_V4DI_USI): Remove.
6209 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
6210 __builtin_ia32_vpshufbitqmb256_mask,
6211 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
6212 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
6213 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
6215 2018-01-26 Alan Modra <amodra@gmail.com>
6218 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
6219 UNSPEC_VBPERMQ. Sort other unspecs.
6221 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
6223 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
6225 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
6228 * predict.c (drop_profile): Do not push/pop cfun; update also
6230 (handle_missing_profiles): Fix logic looking for zero profiles.
6232 2018-01-25 Jakub Jelinek <jakub@redhat.com>
6235 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
6236 on functions with #pragma omp declare simd or functions with simd
6238 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
6239 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
6240 Remove trailing \n from warning_at calls.
6242 2018-01-25 Tom de Vries <tom@codesourcery.com>
6245 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6246 for neutered workers.
6248 2018-01-24 Joseph Myers <joseph@codesourcery.com>
6251 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
6252 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
6254 2018-01-24 Jeff Law <law@redhat.com>
6257 * i386.c (get_probe_interval): Move to earlier point.
6258 (ix86_compute_frame_layout): If -fstack-clash-protection and
6259 the frame is larger than the probe interval, then use pushes
6260 to save registers rather than reg->mem moves.
6261 (ix86_expand_prologue): Remove conditional for int_registers_saved
6264 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
6267 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
6268 min/max for never referenced object.
6270 2018-01-24 Jakub Jelinek <jakub@redhat.com>
6273 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
6275 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
6276 attributes from DECL_ATTRIBUTES (decl) without affecting
6277 DECL_ATTRIBUTES (current_function_decl).
6278 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
6279 functions with non-NULL DECL_ABSTRACT_ORIGIN.
6281 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
6283 PR tree-optimization/83979
6284 * fold-const.c (fold_comparison): Use constant_boolean_node
6285 instead of boolean_{true,false}_node.
6287 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
6289 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
6292 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6294 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6295 Simplify the clause that sets the length attribute.
6296 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6297 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
6298 clause that sets the length attribute.
6299 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6301 2018-01-24 Tom de Vries <tom@codesourcery.com>
6304 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
6305 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
6306 Add strict parameter.
6307 (prevent_branch_around_nothing): Insert dummy insn between branch to
6308 label and label with no ptx insn inbetween.
6309 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
6311 2018-01-24 Tom de Vries <tom@codesourcery.com>
6314 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6315 for neutered threads in warp.
6316 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
6318 2018-01-24 Richard Biener <rguenther@suse.de>
6320 PR tree-optimization/83176
6321 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
6324 2018-01-24 Richard Biener <rguenther@suse.de>
6326 PR tree-optimization/82819
6327 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
6328 code generating pluses that are no-ops in the target precision.
6330 2018-01-24 Richard Biener <rguenther@suse.de>
6333 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
6335 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6337 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
6338 to merge probabilities.
6339 * predict.c (probably_never_executed): Also mark as cold functions
6340 with global 0 profile and guessed local profile.
6341 * profile-count.c (profile_probability::combine_with_count): New
6343 * profile-count.h (profile_probability::operator*,
6344 profile_probability::operator*=, profile_probability::operator/,
6345 profile_probability::operator/=): Reduce precision to adjusted
6346 and set value to guessed on contradictory divisions.
6347 (profile_probability::combine_with_freq): Remove.
6348 (profile_probability::combine_wiht_count): Declare.
6349 (profile_count::force_nonzero):: Set to adjusted.
6350 (profile_count::probability_in):: Set quality to adjusted.
6351 * tree-ssa-tail-merge.c (replace_block_by): Use
6354 2018-01-23 Andrew Waterman <andrew@sifive.com>
6355 Jim Wilson <jimw@sifive.com>
6357 * config/riscv/riscv.c (riscv_stack_boundary): New.
6358 (riscv_option_override): Set riscv_stack_boundary. Handle
6359 riscv_preferred_stack_boundary_arg.
6360 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
6361 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
6362 (STACK_BOUNDARY): Set to riscv_stack_boundary.
6363 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
6364 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
6365 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
6367 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
6370 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
6371 of struct ix86_frame.
6372 (ix86_expand_epilogue): Likewise. Add a local variable for
6373 the reg_save_offset field in struct ix86_frame.
6375 2018-01-23 Bin Cheng <bin.cheng@arm.com>
6377 PR tree-optimization/82604
6378 * tree-loop-distribution.c (enum partition_kind): New enum item
6379 PKIND_PARTIAL_MEMSET.
6380 (partition_builtin_p): Support above new enum item.
6381 (generate_code_for_partition): Ditto.
6382 (compute_access_range): Differentiate cases that equality can be
6383 proven at all loops, the innermost loops or no loops.
6384 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
6385 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
6386 (finalize_partitions, distribute_loop): Don't fuse partition of
6387 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
6388 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
6391 2018-01-23 Martin Liska <mliska@suse.cz>
6393 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
6394 order to ignore the predictor.
6395 (PRED_POLYMORPHIC_CALL): Likewise.
6396 (PRED_RECURSIVE_CALL): Likewise.
6398 2018-01-23 Martin Liska <mliska@suse.cz>
6400 * tree-profile.c (tree_profiling): Print function header to
6401 aware reader which function we are working on.
6402 * value-prof.c (gimple_find_values_to_profile): Do not print
6403 not interesting value histograms.
6405 2018-01-23 Martin Liska <mliska@suse.cz>
6407 * profile-count.h (enum profile_quality): Add
6408 profile_uninitialized as the first value. Do not number values
6409 as they are zero based.
6410 (profile_count::verify): Update sanity check.
6411 (profile_probability::verify): Likewise.
6413 2018-01-23 Nathan Sidwell <nathan@acm.org>
6415 * doc/invoke.texi (ffor-scope): Deprecate.
6417 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6419 PR tree-optimization/83510
6420 * domwalk.c (set_all_edges_as_executable): New function.
6421 (dom_walker::dom_walker): Convert bool param
6422 "skip_unreachable_blocks" to enum reachability. Move setup of
6423 edge flags to set_all_edges_as_executable and only do it when
6424 reachability is REACHABLE_BLOCKS.
6425 * domwalk.h (enum dom_walker::reachability): New enum.
6426 (dom_walker::dom_walker): Convert bool param
6427 "skip_unreachable_blocks" to enum reachability.
6428 (set_all_edges_as_executable): New decl.
6429 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
6430 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
6432 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
6433 but converting true to REACHABLE_BLOCKS.
6434 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
6436 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
6437 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
6438 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
6440 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
6441 if check_all_array_refs will be called.
6443 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6445 * tree.c (selftest::test_location_wrappers): Add more test
6448 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6450 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
6451 (selftest::test_bit_in_range): Likewise.
6453 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6456 * doc/sourcebuild.texi (vect_float): Say that the selector
6457 only describes the situation when -funsafe-math-optimizations is on.
6458 (vect_float_strict): Document.
6460 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6462 PR tree-optimization/83965
6463 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
6464 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
6465 instead of checking only for a reduction.
6466 (vect_recog_widen_sum_pattern): Likewise.
6468 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6470 * predict.c (probably_never_executed): Only use precise profile info.
6471 (compute_function_frequency): Skip after inlining hack since we now
6472 have quality checking.
6474 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6476 * profile-count.h (profile_probability::very_unlikely,
6477 profile_probability::unlikely, profile_probability::even): Set
6478 precision to guessed.
6480 2018-01-23 Richard Biener <rguenther@suse.de>
6482 PR tree-optimization/83963
6483 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
6484 Properly terminate dominator walk when crossing the exit edge not
6485 when visiting its source block.
6487 2018-01-23 Jakub Jelinek <jakub@redhat.com>
6490 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
6491 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
6493 2018-01-22 Jakub Jelinek <jakub@redhat.com>
6495 PR tree-optimization/83957
6496 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
6497 semicolon after for body surrounded by braces.
6499 PR tree-optimization/83081
6500 * profile-count.h (profile_probability::split): New method.
6501 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
6502 Use profile_probability::split.
6503 (do_compare_rtx_and_jump): Fix adjustment of probabilities
6504 when splitting a single conditional jump into 2.
6506 2018-01-22 David Malcolm <dmalcolm@redhat.com>
6508 PR tree-optimization/69452
6509 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
6512 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6514 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
6515 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
6516 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
6518 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6520 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
6521 * config/rl78/rl78.md (movdi): New define_expand.
6522 * config/rl78/rl78.c (rl78_split_movdi): New function.
6524 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
6527 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
6529 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
6530 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
6531 128-bit to produce an UNSPEC move to get the double word with the
6532 signbit and then a shift directly to do signbit.
6533 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
6534 implementation with a new version that just does either a direct
6535 move or a regular move. Move memory interface to separate insns.
6536 Move insns so they are next to the expander.
6537 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
6538 with signbit move. Split big and little endian case.
6539 (signbit<mode>2_dm_mem_le): Likewise.
6540 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
6541 (signbit<mode>2_dm2): Likewise.
6543 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6545 * config/rl78/rl78.md (anddi3): New define_expand.
6547 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6549 * config/rl78/rl78.md (umindi3): New define_expand.
6551 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6553 * config/rl78/rl78.md (smindi3): New define_expand.
6555 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6557 * config/rl78/rl78.md (smaxdi3): New define_expand.
6559 2018-01-22 Carl Love <cel@us.ibm.com>
6561 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
6562 LVX_V1TI): Add macro expansion.
6563 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
6564 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
6565 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
6566 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
6567 Change check to determine if the instruction is a byte reversing
6568 entry. Fix typo in comment.
6569 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
6570 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
6571 Add def_builtin calls for new builtins.
6572 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
6573 Add define_insn expansion.
6575 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6577 * config/rl78/rl78.md (umaxdi3): New define_expand.
6579 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6581 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
6582 for non-QImode registers.
6584 2018-01-22 Richard Biener <rguenther@suse.de>
6586 PR tree-optimization/83963
6587 * graphite-scop-detection.c (scop_detection::get_sese): Delay
6588 including the loop exit block.
6589 (scop_detection::merge_sese): Likewise.
6590 (scop_detection::add_scop): Do it here instead.
6592 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6594 * doc/sourcebuild.texi (arm_softfloat): Document.
6596 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
6599 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
6600 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
6601 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
6603 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6604 David Edelsohn <dje.gcc@gmail.com>
6607 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6608 Change "crset eq" to "crset 2".
6609 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6610 (*call_indirect_aix<mode>_nospec): Likewise.
6611 (*call_value_indirect_aix<mode>_nospec): Likewise.
6612 (*call_indirect_elfv2<mode>_nospec): Likewise.
6613 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
6614 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
6615 change assembly output from . to $.
6616 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6617 (indirect_jump<mode>_nospec): Change assembly output from . to $.
6618 (*tablejump<mode>_internal1_nospec): Likewise.
6620 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
6623 * config/sh/sh_optimize_sett_clrt.cc:
6624 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
6626 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
6628 PR tree-optimization/83940
6629 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
6630 offset_dt to vect_constant_def rather than vect_unknown_def_type.
6631 (vect_check_load_store_mask): Add a mask_dt_out parameter and
6632 use it to pass back the definition type.
6633 (vect_check_store_rhs): Likewise rhs_dt_out.
6634 (vect_build_gather_load_calls): Add a mask_dt argument and use
6635 it instead of a call to vect_is_simple_use.
6636 (vectorizable_store): Update calls to vect_check_load_store_mask
6637 and vect_check_store_rhs. Use the dt returned by the latter instead
6638 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
6639 instead of calls to vect_is_simple_use. Pass the scalar rather
6640 than the vector operand to vect_is_simple_use when handling
6641 second and subsequent copies of an rhs value.
6642 (vectorizable_load): Update calls to vect_check_load_store_mask
6643 and vect_build_gather_load_calls. Use the cached mask_dt and
6644 gs_info.offset_dt instead of calls to vect_is_simple_use.
6646 2018-01-20 Jakub Jelinek <jakub@redhat.com>
6649 * tree-emutls.c: Include gimplify.h.
6650 (lower_emutls_2): New function.
6651 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
6652 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
6653 it before further processing.
6656 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
6657 UINTVAL (trueop1) instead of INTVAL (op1).
6659 2018-01-19 Jakub Jelinek <jakub@redhat.com>
6663 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
6664 INCOMING_FRAME_SP_OFFSET if not defined.
6665 (scan_trace): Add ENTRY argument. If true and
6666 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
6667 emit a note to adjust the CFA offset.
6668 (create_cfi_notes): Adjust scan_trace callers.
6669 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
6670 INCOMING_FRAME_SP_OFFSET in the CIE.
6671 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
6672 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
6674 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
6675 * doc/tm.texi: Regenerated.
6677 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6679 PR rtl-optimization/83147
6680 * lra-constraints.c (remove_inheritance_pseudos): Use
6681 lra_substitute_pseudo_within_insn.
6683 2018-01-19 Tom de Vries <tom@codesourcery.com>
6684 Cesar Philippidis <cesar@codesourcery.com>
6687 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
6689 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
6692 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
6693 spaces for function labels.
6695 2018-01-19 Martin Liska <mliska@suse.cz>
6697 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
6698 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
6699 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
6700 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
6701 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
6702 (PRED_CONST_RETURN): Change from 69 to 65.
6703 (PRED_NULL_RETURN): Change from 91 to 71.
6704 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
6705 (PRED_LOOP_GUARD): Change from 66 to 73.
6707 2018-01-19 Martin Liska <mliska@suse.cz>
6709 * predict.c (predict_insn_def): Add new assert.
6710 (struct branch_predictor): Change type to signed integer.
6711 (test_prediction_value_range): Amend test to cover
6713 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
6714 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
6715 (PRED_LOOP_ITERATIONS_MAX): Likewise.
6716 (PRED_LOOP_IV_COMPARE): Likewise.
6717 * predict.h (PROB_UNINITIALIZED): Define new constant.
6719 2018-01-19 Martin Liska <mliska@suse.cz>
6721 * predict.c (dump_prediction): Add new format for
6722 analyze_brprob.py script which is enabled with -details
6724 * profile-count.h (precise_p): New function.
6726 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6728 PR tree-optimization/83922
6729 * tree-vect-loop.c (vect_verify_full_masking): Return false if
6730 there are no statements that need masking.
6731 (vect_active_double_reduction_p): New function.
6732 (vect_analyze_loop_operations): Use it when handling phis that
6733 are not in the loop header.
6735 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6737 PR tree-optimization/83914
6738 * tree-vect-loop.c (vectorizable_induction): Don't convert
6739 init_expr or apply the peeling adjustment for inductions
6740 that are nested within the vectorized loop.
6742 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6744 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
6747 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6751 * function.h (gimplify_parameters): Add gimple_seq * argument.
6752 * function.c: Include gimple.h and options.h.
6753 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
6754 for the added local temporaries if needed.
6755 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
6756 if there are any parameter cleanups, wrap whole body into a
6757 try/finally with the cleanups.
6759 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
6762 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
6763 Use GET_MODE_CLASS for scalar floating point.
6765 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6769 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
6770 Fix call of call_cgraph_insertion_hooks.
6772 2018-01-18 Martin Sebor <msebor@redhat.com>
6774 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
6776 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6779 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
6782 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
6785 * common.opt: (-ffile-prefix-map): New option.
6786 * opts.c (common_handle_option): Defer it.
6787 * opts-global.c (handle_common_deferred_options): Handle it.
6788 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
6789 * file-prefix-map.h: New file.
6790 (remap_debug_filename, add_debug_prefix_map): ...here.
6791 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
6792 * final.c (debug_prefix_map, add_debug_prefix_map
6793 remap_debug_filename): Move to...
6794 * file-prefix-map.c: New file.
6795 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
6796 generalize, get rid of alloca(), use strrchr() instead of strchr().
6797 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
6798 Implement in terms of add_prefix_map().
6799 (remap_macro_filename, remap_debug_filename): Implement in term of
6801 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
6802 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
6803 * dbxout.c: Include file-prefix-map.h.
6804 * varasm.c: Likewise.
6805 * vmsdbgout.c: Likewise.
6806 * xcoffout.c: Likewise.
6807 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
6808 * doc/cppopts.texi (-fmacro-prefix-map): Document.
6809 * doc/invoke.texi (-ffile-prefix-map): Document.
6810 (-fdebug-prefix-map): Update description.
6812 2018-01-18 Martin Liska <mliska@suse.cz>
6814 * config/i386/i386.c (indirect_thunk_name): Document that also
6816 (output_indirect_thunk): Document why both instructions
6817 (pause and lfence) are generated.
6819 2018-01-18 Richard Biener <rguenther@suse.de>
6821 PR tree-optimization/83887
6822 * graphite-scop-detection.c
6823 (scop_detection::get_nearest_dom_with_single_entry): Remove.
6824 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
6825 (scop_detection::merge_sese): Re-implement with a flood-fill
6826 algorithm that properly finds a SESE region if it exists.
6828 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6831 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
6832 pointer_diff optimizations use view_convert instead of convert.
6834 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6836 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6837 Generate different code for -mno-speculate-indirect-jumps.
6838 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6839 (*call_indirect_aix<mode>): Disable for
6840 -mno-speculate-indirect-jumps.
6841 (*call_indirect_aix<mode>_nospec): New define_insn.
6842 (*call_value_indirect_aix<mode>): Disable for
6843 -mno-speculate-indirect-jumps.
6844 (*call_value_indirect_aix<mode>_nospec): New define_insn.
6845 (*sibcall_nonlocal_sysv<mode>): Generate different code for
6846 -mno-speculate-indirect-jumps.
6847 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6849 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6851 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
6852 long double type, set the flags for noting the default long double
6853 type, even if we don't pass or return a long double type.
6855 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
6858 * ipa-inline.c (flatten_function): Do not overwrite final inlining
6861 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
6863 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
6864 support for merge[hl].
6865 (fold_mergehl_helper): New helper function.
6866 (tree-vector-builder.h): New #include for tree_vector_builder usage.
6867 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
6868 (altivec_vmrglw_direct): Add xxmrglw insn.
6870 2018-01-17 Andrew Waterman <andrew@sifive.com>
6872 * config/riscv/riscv.c (riscv_conditional_register_usage): If
6873 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
6875 2018-01-17 David Malcolm <dmalcolm@redhat.com>
6878 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
6879 call the lto_location_cache before reading the
6880 DECL_SOURCE_LOCATION of the types.
6882 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
6883 Richard Sandiford <richard.sandiford@linaro.org>
6885 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
6886 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
6887 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
6888 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
6889 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
6891 * config/aarch64/constraints.md (aarch64_movti_operand):
6893 * config/aarch64/predicates.md (Uti): Add new constraint.
6895 2018-01-17 Carl Love <cel@us.ibm.com>
6897 * config/rs6000/vsx.md (define_expand xl_len_r,
6898 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
6899 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
6901 (define_expand, define_insn): Move the shift left from the
6902 define_insn to the define_expand for lxvl and stxvl instructions.
6903 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
6904 and XL_LEN_R definitions to PURE.
6906 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
6908 * config/i386/i386.c (indirect_thunk_name): Declare regno
6909 as unsigned int. Compare regno with INVALID_REGNUM.
6910 (output_indirect_thunk): Ditto.
6911 (output_indirect_thunk_function): Ditto.
6912 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
6913 in the call to output_indirect_thunk_function.
6915 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
6918 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
6919 rather than the size of inner_type to determine the stack slot size
6920 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
6922 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
6925 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
6928 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
6930 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
6931 endian Linux systems to optionally enable multilibs for selecting
6932 the long double type if the user configured an explicit type.
6933 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
6934 have no long double multilibs if not defined.
6935 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
6936 warn if the user used -mabi={ieee,ibm}longdouble and we built
6937 multilibs for long double.
6938 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
6939 appropriate multilib option.
6940 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
6942 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
6943 for building long double multilibs.
6944 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
6946 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
6948 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
6951 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
6953 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
6956 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
6959 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
6962 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6964 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
6965 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
6967 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
6969 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
6970 different rtl trees depending on TARGET_64BIT.
6971 (rs6000_gen_lvx): Likewise.
6973 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6975 * config/visium/visium.md (nop): Tweak comment.
6976 (hazard_nop): Likewise.
6978 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6980 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
6981 -mspeculate-indirect-jumps.
6982 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
6983 for -mno-speculate-indirect-jumps.
6984 (*call_indirect_elfv2<mode>_nospec): New define_insn.
6985 (*call_value_indirect_elfv2<mode>): Disable for
6986 -mno-speculate-indirect-jumps.
6987 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
6988 (indirect_jump): Emit different RTL for
6989 -mno-speculate-indirect-jumps.
6990 (*indirect_jump<mode>): Disable for
6991 -mno-speculate-indirect-jumps.
6992 (*indirect_jump<mode>_nospec): New define_insn.
6993 (tablejump): Emit different RTL for
6994 -mno-speculate-indirect-jumps.
6995 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
6996 (tablejumpsi_nospec): New define_expand.
6997 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
6998 (tablejumpdi_nospec): New define_expand.
6999 (*tablejump<mode>_internal1): Disable for
7000 -mno-speculate-indirect-jumps.
7001 (*tablejump<mode>_internal1_nospec): New define_insn.
7002 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
7005 2018-01-16 Artyom Skrobov tyomitch@gmail.com
7007 * caller-save.c (insert_save): Drop unnecessary parameter. All
7010 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7011 Richard Biener <rguenth@suse.de>
7014 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
7015 return early, inline manually is_gimple_sizepos. Make sure if we
7016 call gimplify_expr we don't end up with a gimple constant.
7017 * tree.c (variably_modified_type_p): Don't return true for
7018 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
7019 * gimplify.h (is_gimple_sizepos): Remove.
7021 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7023 PR tree-optimization/83857
7024 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
7025 vectorizable_live_operation for pure SLP statements.
7026 (vectorizable_live_operation): Handle PHIs.
7028 2018-01-16 Richard Biener <rguenther@suse.de>
7030 PR tree-optimization/83867
7031 * tree-vect-stmts.c (vect_transform_stmt): Precompute
7032 nested_in_vect_loop_p since the scalar stmt may get invalidated.
7034 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7037 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
7038 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
7039 If off is not INTEGER_CST, issue a may not be aligned warning
7040 rather than isn't aligned. Use isn%'t rather than isn't.
7041 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
7043 <case MULT_EXPR>: Improve the case when bottom and one of the
7044 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
7045 operand, in that case check if the other operand is multiple of
7046 bottom divided by the INTEGER_CST operand.
7048 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7051 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
7052 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
7053 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
7054 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
7055 * config/pa/pa.c (pa_function_arg_advance): Likewise.
7056 (pa_function_arg, pa_arg_partial_bytes): Likewise.
7057 (pa_function_arg_size): New function.
7059 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7061 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
7062 in a separate statement.
7064 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7066 PR tree-optimization/83847
7067 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
7068 group gathers and scatters.
7070 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7072 PR rtl-optimization/86620
7073 * params.def (max-sched-ready-insns): Bump minimum value to 1.
7075 PR rtl-optimization/83213
7076 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
7077 to last if both are JUMP_INSNs.
7079 PR tree-optimization/83843
7080 * gimple-ssa-store-merging.c
7081 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
7082 store_immediate_info for bswap/nop orig_stores.
7084 2018-01-15 Andrew Waterman <andrew@sifive.com>
7086 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
7088 <UDIV>: Increase cost if !TARGET_DIV.
7090 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
7092 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
7093 (define_attr "cr_logical_3op"): New.
7094 (cceq_ior_compare): Adjust.
7095 (cceq_ior_compare_complement): Adjust.
7096 (*cceq_rev_compare): Adjust.
7097 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
7098 (is_cracked_insn): Adjust.
7099 (insn_must_be_first_in_group): Adjust.
7100 * config/rs6000/40x.md: Adjust.
7101 * config/rs6000/440.md: Adjust.
7102 * config/rs6000/476.md: Adjust.
7103 * config/rs6000/601.md: Adjust.
7104 * config/rs6000/603.md: Adjust.
7105 * config/rs6000/6xx.md: Adjust.
7106 * config/rs6000/7450.md: Adjust.
7107 * config/rs6000/7xx.md: Adjust.
7108 * config/rs6000/8540.md: Adjust.
7109 * config/rs6000/cell.md: Adjust.
7110 * config/rs6000/e300c2c3.md: Adjust.
7111 * config/rs6000/e500mc.md: Adjust.
7112 * config/rs6000/e500mc64.md: Adjust.
7113 * config/rs6000/e5500.md: Adjust.
7114 * config/rs6000/e6500.md: Adjust.
7115 * config/rs6000/mpc.md: Adjust.
7116 * config/rs6000/power4.md: Adjust.
7117 * config/rs6000/power5.md: Adjust.
7118 * config/rs6000/power6.md: Adjust.
7119 * config/rs6000/power7.md: Adjust.
7120 * config/rs6000/power8.md: Adjust.
7121 * config/rs6000/power9.md: Adjust.
7122 * config/rs6000/rs64.md: Adjust.
7123 * config/rs6000/titan.md: Adjust.
7125 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7127 * config/i386/predicates.md (indirect_branch_operand): Rewrite
7128 ix86_indirect_branch_register logic.
7130 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7132 * config/i386/constraints.md (Bs): Update
7133 ix86_indirect_branch_register check. Don't check
7134 ix86_indirect_branch_register with GOT_memory_operand.
7136 * config/i386/predicates.md (GOT_memory_operand): Don't check
7137 ix86_indirect_branch_register here.
7138 (GOT32_symbol_operand): Likewise.
7140 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7142 * config/i386/predicates.md (constant_call_address_operand):
7143 Rewrite ix86_indirect_branch_register logic.
7144 (sibcall_insn_operand): Likewise.
7146 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7148 * config/i386/constraints.md (Bs): Replace
7149 ix86_indirect_branch_thunk_register with
7150 ix86_indirect_branch_register.
7152 * config/i386/i386.md (indirect_jump): Likewise.
7153 (tablejump): Likewise.
7154 (*sibcall_memory): Likewise.
7155 (*sibcall_value_memory): Likewise.
7156 Peepholes of indirect call and jump via memory: Likewise.
7157 * config/i386/i386.opt: Likewise.
7158 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7159 (GOT_memory_operand): Likewise.
7160 (call_insn_operand): Likewise.
7161 (sibcall_insn_operand): Likewise.
7162 (GOT32_symbol_operand): Likewise.
7164 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7167 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
7168 type rather than type addr's type points to.
7169 (expand_omp_atomic_mutex): Likewise.
7170 (expand_omp_atomic): Likewise.
7172 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7175 * config/i386/i386.c (output_indirect_thunk_function): Use
7176 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
7177 for __x86_return_thunk.
7179 2018-01-15 Richard Biener <rguenther@suse.de>
7182 * expmed.c (extract_bit_field_1): Fix typo.
7184 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7187 * config/arm/iterators.md (VF): New mode iterator.
7188 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
7189 Remove integer-related logic from pattern.
7190 (neon_vabd<mode>_3): Likewise.
7192 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7195 * common.opt (fstrict-overflow): No longer an alias.
7196 (fwrapv-pointer): New option.
7197 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
7198 also for pointer types based on flag_wrapv_pointer.
7199 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
7200 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
7201 opts->x_flag_wrapv got set.
7202 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
7203 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
7204 POINTER_TYPE_OVERFLOW_UNDEFINED.
7205 * match.pd: Likewise in address comparison pattern.
7206 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
7208 2018-01-15 Richard Biener <rguenther@suse.de>
7211 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
7212 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
7213 Reset type names to their identifier if their TYPE_DECL doesn't
7214 have linkage (and thus is used for ODR and devirt).
7215 (save_debug_info_for_decl): Remove.
7216 (save_debug_info_for_type): Likewise.
7217 (add_tree_to_fld_list): Adjust.
7218 * tree-pretty-print.c (dump_generic_node): Make dumping of
7219 type names more robust.
7221 2018-01-15 Richard Biener <rguenther@suse.de>
7223 * BASE-VER: Bump to 8.0.1.
7225 2018-01-14 Martin Sebor <msebor@redhat.com>
7228 * builtins.c (check_access): Avoid warning when the no-warning bit
7231 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
7233 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
7234 * ira-color (allocno_hard_regs_compare): Likewise.
7236 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
7239 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
7240 Use .pushsection/.popsection.
7242 2018-01-14 Martin Sebor <msebor@redhat.com>
7245 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
7247 2018-01-14 Jakub Jelinek <jakub@redhat.com>
7249 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
7250 entry from extra_headers.
7251 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
7252 extra_headers, make the list bitwise identical to the i?86-*-* one.
7254 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7256 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
7257 -mcmodel=large with -mindirect-branch=thunk,
7258 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
7259 -mfunction-return=thunk-extern.
7260 * doc/invoke.texi: Document -mcmodel=large is incompatible with
7261 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
7262 -mfunction-return=thunk and -mfunction-return=thunk-extern.
7264 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7266 * config/i386/i386.c (print_reg): Print the name of the full
7267 integer register without '%'.
7268 (ix86_print_operand): Handle 'V'.
7269 * doc/extend.texi: Document 'V' modifier.
7271 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7273 * config/i386/constraints.md (Bs): Disallow memory operand for
7274 -mindirect-branch-register.
7276 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7277 (GOT_memory_operand): Likewise.
7278 (call_insn_operand): Likewise.
7279 (sibcall_insn_operand): Likewise.
7280 (GOT32_symbol_operand): Likewise.
7281 * config/i386/i386.md (indirect_jump): Call convert_memory_address
7282 for -mindirect-branch-register.
7283 (tablejump): Likewise.
7284 (*sibcall_memory): Likewise.
7285 (*sibcall_value_memory): Likewise.
7286 Disallow peepholes of indirect call and jump via memory for
7287 -mindirect-branch-register.
7288 (*call_pop): Replace m with Bw.
7289 (*call_value_pop): Likewise.
7290 (*sibcall_pop_memory): Replace m with Bs.
7291 * config/i386/i386.opt (mindirect-branch-register): New option.
7292 * doc/invoke.texi: Document -mindirect-branch-register option.
7294 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7296 * config/i386/i386-protos.h (ix86_output_function_return): New.
7297 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
7298 set function_return_type.
7299 (indirect_thunk_name): Add ret_p to indicate thunk for function
7301 (output_indirect_thunk_function): Pass false to
7302 indirect_thunk_name.
7303 (ix86_output_indirect_branch_via_reg): Likewise.
7304 (ix86_output_indirect_branch_via_push): Likewise.
7305 (output_indirect_thunk_function): Create alias for function
7306 return thunk if regno < 0.
7307 (ix86_output_function_return): New function.
7308 (ix86_handle_fndecl_attribute): Handle function_return.
7309 (ix86_attribute_table): Add function_return.
7310 * config/i386/i386.h (machine_function): Add
7311 function_return_type.
7312 * config/i386/i386.md (simple_return_internal): Use
7313 ix86_output_function_return.
7314 (simple_return_internal_long): Likewise.
7315 * config/i386/i386.opt (mfunction-return=): New option.
7316 (indirect_branch): Mention -mfunction-return=.
7317 * doc/extend.texi: Document function_return function attribute.
7318 * doc/invoke.texi: Document -mfunction-return= option.
7320 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7322 * config/i386/i386-opts.h (indirect_branch): New.
7323 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
7324 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
7325 with local indirect jump when converting indirect call and jump.
7326 (ix86_set_indirect_branch_type): New.
7327 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
7328 (indirectlabelno): New.
7329 (indirect_thunk_needed): Likewise.
7330 (indirect_thunk_bnd_needed): Likewise.
7331 (indirect_thunks_used): Likewise.
7332 (indirect_thunks_bnd_used): Likewise.
7333 (INDIRECT_LABEL): Likewise.
7334 (indirect_thunk_name): Likewise.
7335 (output_indirect_thunk): Likewise.
7336 (output_indirect_thunk_function): Likewise.
7337 (ix86_output_indirect_branch_via_reg): Likewise.
7338 (ix86_output_indirect_branch_via_push): Likewise.
7339 (ix86_output_indirect_branch): Likewise.
7340 (ix86_output_indirect_jmp): Likewise.
7341 (ix86_code_end): Call output_indirect_thunk_function if needed.
7342 (ix86_output_call_insn): Call ix86_output_indirect_branch if
7344 (ix86_handle_fndecl_attribute): Handle indirect_branch.
7345 (ix86_attribute_table): Add indirect_branch.
7346 * config/i386/i386.h (machine_function): Add indirect_branch_type
7347 and has_local_indirect_jump.
7348 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
7350 (tablejump): Likewise.
7351 (*indirect_jump): Use ix86_output_indirect_jmp.
7352 (*tablejump_1): Likewise.
7353 (simple_return_indirect_internal): Likewise.
7354 * config/i386/i386.opt (mindirect-branch=): New option.
7355 (indirect_branch): New.
7358 (thunk-inline): Likewise.
7359 (thunk-extern): Likewise.
7360 * doc/extend.texi: Document indirect_branch function attribute.
7361 * doc/invoke.texi: Document -mindirect-branch= option.
7363 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
7366 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
7368 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
7370 * ipa-inline.c (want_inline_small_function_p): Return false if
7371 inlining has already failed with CIF_FINAL_ERROR.
7372 (update_caller_keys): Call want_inline_small_function_p before
7374 (update_callee_keys): Likewise.
7376 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7378 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
7380 (rs6000_quadword_masked_address_p): Likewise.
7381 (quad_aligned_load_p): Likewise.
7382 (quad_aligned_store_p): Likewise.
7383 (const_load_sequence_p): Add comment to describe the outer-most loop.
7384 (mimic_memory_attributes_and_flags): New function.
7385 (rs6000_gen_stvx): Likewise.
7386 (replace_swapped_aligned_store): Likewise.
7387 (rs6000_gen_lvx): Likewise.
7388 (replace_swapped_aligned_load): Likewise.
7389 (replace_swapped_load_constant): Capitalize argument name in
7390 comment describing this function.
7391 (rs6000_analyze_swaps): Add a third pass to search for vector loads
7392 and stores that access quad-word aligned addresses and replace
7393 with stvx or lvx instructions when appropriate.
7394 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
7395 New function prototype.
7396 (rs6000_quadword_masked_address_p): Likewise.
7397 (rs6000_gen_lvx): Likewise.
7398 (rs6000_gen_stvx): Likewise.
7399 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
7400 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
7401 when memory address is aligned.
7402 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
7403 this split to select lvx instruction when memory address is aligned.
7404 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
7405 instruction when memory address is aligned.
7406 (*vsx_le_perm_load_v16qi): Likewise.
7407 (four unnamed splitters): Modify to select the stvx instruction
7408 when memory is aligned.
7410 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
7412 * predict.c (determine_unlikely_bbs): Handle correctly BBs
7413 which appears in the queue multiple times.
7415 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7416 Alan Hayward <alan.hayward@arm.com>
7417 David Sherwood <david.sherwood@arm.com>
7419 * tree-vectorizer.h (vec_lower_bound): New structure.
7420 (_loop_vec_info): Add check_nonzero and lower_bounds.
7421 (LOOP_VINFO_CHECK_NONZERO): New macro.
7422 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
7423 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
7424 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
7425 fields. Make seg_len the distance travelled, not including the
7427 (dr_direction_indicator): Declare.
7428 (dr_zero_step_indicator): Likewise.
7429 (dr_known_forward_stride_p): Likewise.
7430 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
7432 (runtime_alias_check_p): Allow runtime alias checks with
7434 (operator ==): Compare access_size and align.
7435 (prune_runtime_alias_test_list): Rework for new distinction between
7436 the access_size and seg_len.
7437 (create_intersect_range_checks_index): Likewise. Cope with polynomial
7439 (get_segment_min_max): New function.
7440 (create_intersect_range_checks): Use it.
7441 (dr_step_indicator): New function.
7442 (dr_direction_indicator): Likewise.
7443 (dr_zero_step_indicator): Likewise.
7444 (dr_known_forward_stride_p): Likewise.
7445 * tree-loop-distribution.c (data_ref_segment_size): Return
7446 DR_STEP * (niters - 1).
7447 (compute_alias_check_pairs): Update call to the dr_with_seg_len
7449 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
7450 (vect_preserves_scalar_order_p): New function, split out from...
7451 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
7452 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
7453 (vect_vfa_access_size): New function.
7454 (vect_vfa_align): Likewise.
7455 (vect_compile_time_alias): Take access_size_a and access_b arguments.
7456 (dump_lower_bound): New function.
7457 (vect_check_lower_bound): Likewise.
7458 (vect_small_gap_p): Likewise.
7459 (vectorizable_with_step_bound_p): Likewise.
7460 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
7461 depencies if the vectorization factor is 1. Convert the checks
7462 for nonzero steps into checks on the bounds of DR_STEP. Try using
7463 a bunds check for variable steps if the minimum required step is
7464 relatively small. Update calls to the dr_with_seg_len
7465 constructor and to vect_compile_time_alias.
7466 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
7468 (vect_loop_versioning): Call it.
7469 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
7471 (vect_estimate_min_profitable_iters): Account for any bounds checks.
7473 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7474 Alan Hayward <alan.hayward@arm.com>
7475 David Sherwood <david.sherwood@arm.com>
7477 * doc/sourcebuild.texi (vect_scatter_store): Document.
7478 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
7480 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
7482 * genopinit.c (main): Add supports_vec_scatter_store and
7483 supports_vec_scatter_store_cached to target_optabs.
7484 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
7485 IFN_MASK_SCATTER_STORE.
7486 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
7488 * internal-fn.h (internal_store_fn_p): Declare.
7489 (internal_fn_stored_value_index): Likewise.
7490 * internal-fn.c (scatter_store_direct): New macro.
7491 (expand_scatter_store_optab_fn): New function.
7492 (direct_scatter_store_optab_supported_p): New macro.
7493 (internal_store_fn_p): New function.
7494 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
7495 IFN_MASK_SCATTER_STORE.
7496 (internal_fn_mask_index): Likewise.
7497 (internal_fn_stored_value_index): New function.
7498 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
7500 * optabs-query.h (supports_vec_scatter_store_p): Declare.
7501 * optabs-query.c (supports_vec_scatter_store_p): New function.
7502 * tree-vectorizer.h (vect_get_store_rhs): Declare.
7503 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
7504 true for scatter stores.
7505 (vect_gather_scatter_fn_p): Handle scatter stores too.
7506 (vect_check_gather_scatter): Consider using scatter stores if
7507 supports_vec_scatter_store_p.
7508 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
7510 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7511 internal_fn_stored_value_index.
7512 (check_load_store_masking): Handle scatter stores too.
7513 (vect_get_store_rhs): Make public.
7514 (vectorizable_call): Use internal_store_fn_p.
7515 (vectorizable_store): Handle scatter store internal functions.
7516 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
7517 when deciding whether the end of the group has been reached.
7518 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
7519 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
7520 (mask_scatter_store<mode>): New insns.
7522 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7523 Alan Hayward <alan.hayward@arm.com>
7524 David Sherwood <david.sherwood@arm.com>
7526 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
7527 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
7528 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
7530 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
7531 Use vect_truncate_gather_scatter_offset if we can't treat the
7532 operation as a normal gather load or scatter store.
7533 (get_group_load_store_type): Take the gather_scatter_info
7534 as argument. Try using a gather load or scatter store for
7535 single-element groups.
7536 (get_load_store_type): Update calls to get_group_load_store_type
7537 and vect_use_strided_gather_scatters_p.
7539 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7540 Alan Hayward <alan.hayward@arm.com>
7541 David Sherwood <david.sherwood@arm.com>
7543 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
7544 optional tree argument.
7545 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
7547 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
7548 but continue to use the current value as a fallback.
7549 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
7550 to compare the updates.
7551 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
7552 (get_load_store_type): Use it when handling a strided access.
7553 (vect_get_strided_load_store_ops): New function.
7554 (vect_get_data_ptr_increment): Likewise.
7555 (vectorizable_load): Handle strided gather loads. Always pass
7556 a step to vect_create_data_ref_ptr and bump_vector_ptr.
7558 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7559 Alan Hayward <alan.hayward@arm.com>
7560 David Sherwood <david.sherwood@arm.com>
7562 * doc/md.texi (gather_load@var{m}): Document.
7563 (mask_gather_load@var{m}): Likewise.
7564 * genopinit.c (main): Add supports_vec_gather_load and
7565 supports_vec_gather_load_cached to target_optabs.
7566 * optabs-tree.c (init_tree_optimization_optabs): Use
7567 ggc_cleared_alloc to allocate target_optabs.
7568 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
7569 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
7571 * internal-fn.h (internal_load_fn_p): Declare.
7572 (internal_gather_scatter_fn_p): Likewise.
7573 (internal_fn_mask_index): Likewise.
7574 (internal_gather_scatter_fn_supported_p): Likewise.
7575 * internal-fn.c (gather_load_direct): New macro.
7576 (expand_gather_load_optab_fn): New function.
7577 (direct_gather_load_optab_supported_p): New macro.
7578 (direct_internal_fn_optab): New function.
7579 (internal_load_fn_p): Likewise.
7580 (internal_gather_scatter_fn_p): Likewise.
7581 (internal_fn_mask_index): Likewise.
7582 (internal_gather_scatter_fn_supported_p): Likewise.
7583 * optabs-query.c (supports_at_least_one_mode_p): New function.
7584 (supports_vec_gather_load_p): Likewise.
7585 * optabs-query.h (supports_vec_gather_load_p): Declare.
7586 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
7587 and memory_type field.
7588 (NUM_PATTERNS): Bump to 15.
7589 * tree-vect-data-refs.c: Include internal-fn.h.
7590 (vect_gather_scatter_fn_p): New function.
7591 (vect_describe_gather_scatter_call): Likewise.
7592 (vect_check_gather_scatter): Try using internal functions for
7593 gather loads. Recognize existing calls to a gather load function.
7594 (vect_analyze_data_refs): Consider using gather loads if
7595 supports_vec_gather_load_p.
7596 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
7597 (vect_get_gather_scatter_offset_type): Likewise.
7598 (vect_convert_mask_for_vectype): Likewise.
7599 (vect_add_conversion_to_patterm): Likewise.
7600 (vect_try_gather_scatter_pattern): Likewise.
7601 (vect_recog_gather_scatter_pattern): New pattern recognizer.
7602 (vect_vect_recog_func_ptrs): Add it.
7603 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7604 internal_fn_mask_index and internal_gather_scatter_fn_p.
7605 (check_load_store_masking): Take the gather_scatter_info as an
7606 argument and handle gather loads.
7607 (vect_get_gather_scatter_ops): New function.
7608 (vectorizable_call): Check internal_load_fn_p.
7609 (vectorizable_load): Likewise. Handle gather load internal
7611 (vectorizable_store): Update call to check_load_store_masking.
7612 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
7613 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
7614 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
7615 (aarch64_gather_scale_operand_d): New predicates.
7616 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
7617 (mask_gather_load<mode>): New insns.
7619 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7620 Alan Hayward <alan.hayward@arm.com>
7621 David Sherwood <david.sherwood@arm.com>
7623 * optabs.def (fold_left_plus_optab): New optab.
7624 * doc/md.texi (fold_left_plus_@var{m}): Document.
7625 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
7626 * internal-fn.c (fold_left_direct): Define.
7627 (expand_fold_left_optab_fn): Likewise.
7628 (direct_fold_left_optab_supported_p): Likewise.
7629 * fold-const-call.c (fold_const_fold_left): New function.
7630 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
7631 * tree-parloops.c (valid_reduction_p): New function.
7632 (gather_scalar_reductions): Use it.
7633 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
7634 (vect_finish_replace_stmt): Declare.
7635 * tree-vect-loop.c (fold_left_reduction_fn): New function.
7636 (needs_fold_left_reduction_p): New function, split out from...
7637 (vect_is_simple_reduction): ...here. Accept reductions that
7638 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
7639 (vect_force_simple_reduction): Also store the reduction type in
7640 the assignment's STMT_VINFO_REDUC_TYPE.
7641 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
7642 (merge_with_identity): New function.
7643 (vect_expand_fold_left): Likewise.
7644 (vectorize_fold_left_reduction): Likewise.
7645 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
7646 scalar phi in place for it. Check for target support and reject
7647 cases that would reassociate the operation. Defer the transform
7648 phase to vectorize_fold_left_reduction.
7649 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
7650 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
7651 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
7653 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7655 * tree-if-conv.c (predicate_mem_writes): Remove redundant
7656 call to ifc_temp_var.
7658 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7659 Alan Hayward <alan.hayward@arm.com>
7660 David Sherwood <david.sherwood@arm.com>
7662 * target.def (legitimize_address_displacement): Take the original
7663 offset as a poly_int.
7664 * targhooks.h (default_legitimize_address_displacement): Update
7666 * targhooks.c (default_legitimize_address_displacement): Likewise.
7667 * doc/tm.texi: Regenerate.
7668 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
7669 as an argument, moving assert of ad->disp == ad->disp_term to...
7670 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
7671 Try calling targetm.legitimize_address_displacement before expanding
7672 the address rather than afterwards, and adjust for the new interface.
7673 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
7674 Match the new hook interface. Handle SVE addresses.
7675 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
7678 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7680 * Makefile.in (OBJS): Add early-remat.o.
7681 * target.def (select_early_remat_modes): New hook.
7682 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
7683 * doc/tm.texi: Regenerate.
7684 * targhooks.h (default_select_early_remat_modes): Declare.
7685 * targhooks.c (default_select_early_remat_modes): New function.
7686 * timevar.def (TV_EARLY_REMAT): New timevar.
7687 * passes.def (pass_early_remat): New pass.
7688 * tree-pass.h (make_pass_early_remat): Declare.
7689 * early-remat.c: New file.
7690 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
7692 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
7694 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7695 Alan Hayward <alan.hayward@arm.com>
7696 David Sherwood <david.sherwood@arm.com>
7698 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
7699 vfm1 with a bound_epilog parameter.
7700 (vect_do_peeling): Update calls accordingly, and move the prologue
7701 call earlier in the function. Treat the base bound_epilog as 0 for
7702 fully-masked loops and retain vf - 1 for other loops. Add 1 to
7703 this base when peeling for gaps.
7704 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
7705 with fully-masked loops.
7706 (vect_estimate_min_profitable_iters): Handle the single peeled
7707 iteration in that case.
7709 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7710 Alan Hayward <alan.hayward@arm.com>
7711 David Sherwood <david.sherwood@arm.com>
7713 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
7714 single-element interleaving even if the size is not a power of 2.
7715 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
7716 accesses for single-element interleaving if the group size is
7719 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7720 Alan Hayward <alan.hayward@arm.com>
7721 David Sherwood <david.sherwood@arm.com>
7723 * doc/md.texi (fold_extract_last_@var{m}): Document.
7724 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
7725 * optabs.def (fold_extract_last_optab): New optab.
7726 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
7727 * internal-fn.c (fold_extract_direct): New macro.
7728 (expand_fold_extract_optab_fn): Likewise.
7729 (direct_fold_extract_optab_supported_p): Likewise.
7730 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
7731 * tree-vect-loop.c (vect_model_reduction_cost): Handle
7732 EXTRACT_LAST_REDUCTION.
7733 (get_initial_def_for_reduction): Do not create an initial vector
7734 for EXTRACT_LAST_REDUCTION reductions.
7735 (vectorizable_reduction): Leave the scalar phi in place for
7736 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
7737 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
7738 epilogue code for EXTRACT_LAST_REDUCTION and defer the
7739 transform phase to vectorizable_condition.
7740 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
7742 (vect_finish_stmt_generation): ...here.
7743 (vect_finish_replace_stmt): New function.
7744 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
7745 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
7747 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
7749 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7750 Alan Hayward <alan.hayward@arm.com>
7751 David Sherwood <david.sherwood@arm.com>
7753 * doc/md.texi (extract_last_@var{m}): Document.
7754 * optabs.def (extract_last_optab): New optab.
7755 * internal-fn.def (EXTRACT_LAST): New internal function.
7756 * internal-fn.c (cond_unary_direct): New macro.
7757 (expand_cond_unary_optab_fn): Likewise.
7758 (direct_cond_unary_optab_supported_p): Likewise.
7759 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
7760 loops using EXTRACT_LAST.
7761 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
7762 (extract_last_<mode>): ...this optab.
7763 (vec_extract<mode><Vel>): Update accordingly.
7765 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7766 Alan Hayward <alan.hayward@arm.com>
7767 David Sherwood <david.sherwood@arm.com>
7769 * target.def (empty_mask_is_expensive): New hook.
7770 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
7771 * doc/tm.texi: Regenerate.
7772 * targhooks.h (default_empty_mask_is_expensive): Declare.
7773 * targhooks.c (default_empty_mask_is_expensive): New function.
7774 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
7775 if the target says that empty masks are expensive.
7776 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
7778 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
7780 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7781 Alan Hayward <alan.hayward@arm.com>
7782 David Sherwood <david.sherwood@arm.com>
7784 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
7785 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
7786 (vect_use_loop_mask_for_alignment_p): New function.
7787 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
7788 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
7789 niters_skip argument. Make sure that the first niters_skip elements
7790 of the first iteration are inactive.
7791 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
7792 Update call to vect_set_loop_masks_directly.
7793 (get_misalign_in_elems): New function, split out from...
7794 (vect_gen_prolog_loop_niters): ...here.
7795 (vect_update_init_of_dr): Take a code argument that specifies whether
7796 the adjustment should be added or subtracted.
7797 (vect_update_init_of_drs): Likewise.
7798 (vect_prepare_for_masked_peels): New function.
7799 (vect_do_peeling): Skip prologue peeling if we're using a mask
7800 instead. Update call to vect_update_inits_of_drs.
7801 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7803 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
7804 alignment. Do not include the number of peeled iterations in
7805 the minimum threshold in that case.
7806 (vectorizable_induction): Adjust the start value down by
7807 LOOP_VINFO_MASK_SKIP_NITERS iterations.
7808 (vect_transform_loop): Call vect_prepare_for_masked_peels.
7809 Take the number of skipped iterations into account when calculating
7811 * tree-vect-stmts.c (vect_gen_while_not): New function.
7813 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7814 Alan Hayward <alan.hayward@arm.com>
7815 David Sherwood <david.sherwood@arm.com>
7817 * doc/sourcebuild.texi (vect_fully_masked): Document.
7818 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
7820 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
7822 (vect_analyze_loop_2): ...here. Don't check the vectorization
7823 factor against the number of loop iterations if the loop is
7826 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7827 Alan Hayward <alan.hayward@arm.com>
7828 David Sherwood <david.sherwood@arm.com>
7830 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
7831 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
7832 (dump_groups): Update accordingly.
7833 (iv_use::mem_type): New member variable.
7834 (address_p): New function.
7835 (record_use): Add a mem_type argument and initialize the new
7837 (record_group_use): Add a mem_type argument. Use address_p.
7838 Remove obsolete null checks of base_object. Update call to record_use.
7839 (find_interesting_uses_op): Update call to record_group_use.
7840 (find_interesting_uses_cond): Likewise.
7841 (find_interesting_uses_address): Likewise.
7842 (get_mem_type_for_internal_fn): New function.
7843 (find_address_like_use): Likewise.
7844 (find_interesting_uses_stmt): Try find_address_like_use before
7845 calling find_interesting_uses_op.
7846 (addr_offset_valid_p): Use the iv mem_type field as the type
7847 of the addressed memory.
7848 (add_autoinc_candidates): Likewise.
7849 (get_address_cost): Likewise.
7850 (split_small_address_groups_p): Use address_p.
7851 (split_address_groups): Likewise.
7852 (add_iv_candidate_for_use): Likewise.
7853 (autoinc_possible_for_pair): Likewise.
7854 (rewrite_groups): Likewise.
7855 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
7856 (determine_group_iv_cost): Update after split of USE_ADDRESS.
7857 (get_alias_ptr_type_for_ptr_address): New function.
7858 (rewrite_use_address): Rewrite address uses in calls that were
7859 identified by find_address_like_use.
7861 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7862 Alan Hayward <alan.hayward@arm.com>
7863 David Sherwood <david.sherwood@arm.com>
7865 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
7867 * gimple-expr.h (is_gimple_addressable: Likewise.
7868 * gimple-expr.c (is_gimple_address): Likewise.
7869 * internal-fn.c (expand_call_mem_ref): New function.
7870 (expand_mask_load_optab_fn): Use it.
7871 (expand_mask_store_optab_fn): Likewise.
7873 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7874 Alan Hayward <alan.hayward@arm.com>
7875 David Sherwood <david.sherwood@arm.com>
7877 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
7878 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
7879 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
7880 (cond_umax@var{mode}): Document.
7881 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
7882 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
7883 (cond_umin_optab, cond_umax_optab): New optabs.
7884 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
7885 (COND_IOR, COND_XOR): New internal functions.
7886 * internal-fn.h (get_conditional_internal_fn): Declare.
7887 * internal-fn.c (cond_binary_direct): New macro.
7888 (expand_cond_binary_optab_fn): Likewise.
7889 (direct_cond_binary_optab_supported_p): Likewise.
7890 (get_conditional_internal_fn): New function.
7891 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
7892 Cope with reduction statements that are vectorized as calls rather
7894 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
7895 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
7896 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
7897 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
7898 (UNSPEC_COND_EOR): New unspecs.
7899 (optab): Add mappings for them.
7900 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
7901 (sve_int_op, sve_fp_op): New int attributes.
7903 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7904 Alan Hayward <alan.hayward@arm.com>
7905 David Sherwood <david.sherwood@arm.com>
7907 * optabs.def (while_ult_optab): New optab.
7908 * doc/md.texi (while_ult@var{m}@var{n}): Document.
7909 * internal-fn.def (WHILE_ULT): New internal function.
7910 * internal-fn.h (direct_internal_fn_supported_p): New override
7911 that takes two types as argument.
7912 * internal-fn.c (while_direct): New macro.
7913 (expand_while_optab_fn): New function.
7914 (convert_optab_supported_p): Likewise.
7915 (direct_while_optab_supported_p): New macro.
7916 * wide-int.h (wi::udiv_ceil): New function.
7917 * tree-vectorizer.h (rgroup_masks): New structure.
7918 (vec_loop_masks): New typedef.
7919 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
7921 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
7922 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
7923 (vect_max_vf): New function.
7924 (slpeel_make_loop_iterate_ntimes): Delete.
7925 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
7926 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
7927 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
7928 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
7929 internal-fn.h, stor-layout.h and optabs-query.h.
7930 (vect_set_loop_mask): New function.
7931 (add_preheader_seq): Likewise.
7932 (add_header_seq): Likewise.
7933 (interleave_supported_p): Likewise.
7934 (vect_maybe_permute_loop_masks): Likewise.
7935 (vect_set_loop_masks_directly): Likewise.
7936 (vect_set_loop_condition_masked): Likewise.
7937 (vect_set_loop_condition_unmasked): New function, split out from
7938 slpeel_make_loop_iterate_ntimes.
7939 (slpeel_make_loop_iterate_ntimes): Rename to..
7940 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
7941 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
7942 (vect_do_peeling): Update call accordingly.
7943 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
7945 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7946 mask_compare_type, can_fully_mask_p and fully_masked_p.
7947 (release_vec_loop_masks): New function.
7948 (_loop_vec_info): Use it to free the loop masks.
7949 (can_produce_all_loop_masks_p): New function.
7950 (vect_get_max_nscalars_per_iter): Likewise.
7951 (vect_verify_full_masking): Likewise.
7952 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
7953 retries, and free the mask rgroups before retrying. Check loop-wide
7954 reasons for disallowing fully-masked loops. Make the final decision
7955 about whether use a fully-masked loop or not.
7956 (vect_estimate_min_profitable_iters): Do not assume that peeling
7957 for the number of iterations will be needed for fully-masked loops.
7958 (vectorizable_reduction): Disable fully-masked loops.
7959 (vectorizable_live_operation): Likewise.
7960 (vect_halve_mask_nunits): New function.
7961 (vect_double_mask_nunits): Likewise.
7962 (vect_record_loop_mask): Likewise.
7963 (vect_get_loop_mask): Likewise.
7964 (vect_transform_loop): Handle the case in which the final loop
7965 iteration might handle a partial vector. Call vect_set_loop_condition
7966 instead of slpeel_make_loop_iterate_ntimes.
7967 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
7968 (check_load_store_masking): New function.
7969 (prepare_load_store_mask): Likewise.
7970 (vectorizable_store): Handle fully-masked loops.
7971 (vectorizable_load): Likewise.
7972 (supportable_widening_operation): Use vect_halve_mask_nunits for
7974 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
7975 (vect_gen_while): New function.
7976 * config/aarch64/aarch64.md (umax<mode>3): New expander.
7977 (aarch64_uqdec<mode>): New insn.
7979 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7980 Alan Hayward <alan.hayward@arm.com>
7981 David Sherwood <david.sherwood@arm.com>
7983 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
7984 (reduc_xor_scal_optab): New optabs.
7985 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
7986 (reduc_xor_scal_@var{m}): Document.
7987 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
7988 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
7990 * fold-const-call.c (fold_const_call): Handle them.
7991 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
7992 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
7993 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
7994 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
7995 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
7996 (UNSPEC_XORV): New unspecs.
7997 (optab): Add entries for them.
7998 (BITWISEV): New int iterator.
7999 (bit_reduc_op): New int attributes.
8001 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8002 Alan Hayward <alan.hayward@arm.com>
8003 David Sherwood <david.sherwood@arm.com>
8005 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
8006 * internal-fn.def (VEC_SHL_INSERT): New internal function.
8007 * optabs.def (vec_shl_insert_optab): New optab.
8008 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
8009 (duplicate_and_interleave): Likewise.
8010 * tree-vect-loop.c: Include internal-fn.h.
8011 (neutral_op_for_slp_reduction): New function, split out from
8012 get_initial_defs_for_reduction.
8013 (get_initial_def_for_reduction): Handle option 2 for variable-length
8014 vectors by loading the neutral value into a vector and then shifting
8015 the initial value into element 0.
8016 (get_initial_defs_for_reduction): Replace the code argument with
8017 the neutral value calculated by neutral_op_for_slp_reduction.
8018 Use gimple_build_vector for constant-length vectors.
8019 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
8020 but the first group_size elements have a neutral value.
8021 Use duplicate_and_interleave otherwise.
8022 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
8023 Update call to get_initial_defs_for_reduction. Handle SLP
8024 reductions for variable-length vectors by creating one vector
8025 result for each scalar result, with the elements associated
8026 with other scalar results stubbed out with the neutral value.
8027 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
8028 Require IFN_VEC_SHL_INSERT for double reductions on
8029 variable-length vectors, or SLP reductions that have
8030 a neutral value. Require can_duplicate_and_interleave_p
8031 support for variable-length unchained SLP reductions if there
8032 is no neutral value, such as for MIN/MAX reductions. Also require
8033 the number of vector elements to be a multiple of the number of
8034 SLP statements when doing variable-length unchained SLP reductions.
8035 Update call to vect_create_epilog_for_reduction.
8036 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
8037 and remove initial values.
8038 (duplicate_and_interleave): Make public.
8039 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
8040 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
8042 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8043 Alan Hayward <alan.hayward@arm.com>
8044 David Sherwood <david.sherwood@arm.com>
8046 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
8047 (can_duplicate_and_interleave_p): New function.
8048 (vect_get_and_check_slp_defs): Take the vector of statements
8049 rather than just the current one. Remove excess parentheses.
8050 Restriction rejectinon of vect_constant_def and vect_external_def
8051 for variable-length vectors to boolean types, or types for which
8052 can_duplicate_and_interleave_p is false.
8053 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
8054 (duplicate_and_interleave): New function.
8055 (vect_get_constant_vectors): Use gimple_build_vector for
8056 constant-length vectors and suitable variable-length constant
8057 vectors. Use duplicate_and_interleave for other variable-length
8058 vectors. Don't defer the update when inserting new statements.
8060 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8061 Alan Hayward <alan.hayward@arm.com>
8062 David Sherwood <david.sherwood@arm.com>
8064 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
8065 min_profitable_iters doesn't go negative.
8067 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8068 Alan Hayward <alan.hayward@arm.com>
8069 David Sherwood <david.sherwood@arm.com>
8071 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
8072 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
8073 * optabs.def (vec_mask_load_lanes_optab): New optab.
8074 (vec_mask_store_lanes_optab): Likewise.
8075 * internal-fn.def (MASK_LOAD_LANES): New internal function.
8076 (MASK_STORE_LANES): Likewise.
8077 * internal-fn.c (mask_load_lanes_direct): New macro.
8078 (mask_store_lanes_direct): Likewise.
8079 (expand_mask_load_optab_fn): Handle masked operations.
8080 (expand_mask_load_lanes_optab_fn): New macro.
8081 (expand_mask_store_optab_fn): Handle masked operations.
8082 (expand_mask_store_lanes_optab_fn): New macro.
8083 (direct_mask_load_lanes_optab_supported_p): Likewise.
8084 (direct_mask_store_lanes_optab_supported_p): Likewise.
8085 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
8087 (vect_load_lanes_supported): Likewise.
8088 * tree-vect-data-refs.c (strip_conversion): New function.
8089 (can_group_stmts_p): Likewise.
8090 (vect_analyze_data_ref_accesses): Use it instead of checking
8091 for a pair of assignments.
8092 (vect_store_lanes_supported): Take a masked_p parameter.
8093 (vect_load_lanes_supported): Likewise.
8094 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
8095 vect_store_lanes_supported and vect_load_lanes_supported.
8096 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8097 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
8098 parameter. Don't allow gaps for masked accesses.
8099 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
8100 and vect_load_lanes_supported.
8101 (get_load_store_type): Take a masked_p parameter and update
8102 call to get_group_load_store_type.
8103 (vectorizable_store): Update call to get_load_store_type.
8104 Handle IFN_MASK_STORE_LANES.
8105 (vectorizable_load): Update call to get_load_store_type.
8106 Handle IFN_MASK_LOAD_LANES.
8108 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8109 Alan Hayward <alan.hayward@arm.com>
8110 David Sherwood <david.sherwood@arm.com>
8112 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
8114 * config/aarch64/aarch64-protos.h
8115 (aarch64_sve_struct_memory_operand_p): Declare.
8116 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
8117 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
8118 (VPRED, vpred): Handle SVE structure modes.
8119 * config/aarch64/constraints.md (Utx): New constraint.
8120 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
8121 (aarch64_sve_struct_nonimmediate_operand): New predicates.
8122 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
8123 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
8124 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
8125 structure modes. Split into pieces after RA.
8126 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
8127 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
8129 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
8130 SVE structure modes.
8131 (aarch64_classify_address): Likewise.
8132 (sizetochar): Move earlier in file.
8133 (aarch64_print_operand): Handle SVE register lists.
8134 (aarch64_array_mode): New function.
8135 (aarch64_sve_struct_memory_operand_p): Likewise.
8136 (TARGET_ARRAY_MODE): Redefine.
8138 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8139 Alan Hayward <alan.hayward@arm.com>
8140 David Sherwood <david.sherwood@arm.com>
8142 * target.def (array_mode): New target hook.
8143 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
8144 * doc/tm.texi: Regenerate.
8145 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
8146 * hooks.c (hook_optmode_mode_uhwi_none): New function.
8147 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
8149 * stor-layout.c (mode_for_array): Likewise. Support polynomial
8152 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8153 Alan Hayward <alan.hayward@arm.com>
8154 David Sherwood <david.sherwood@arm.com>
8156 * fold-const.c (fold_binary_loc): Check the argument types
8157 rather than the result type when testing for a vector operation.
8159 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8161 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
8162 * doc/tm.texi: Regenerate.
8164 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8165 Alan Hayward <alan.hayward@arm.com>
8166 David Sherwood <david.sherwood@arm.com>
8168 * doc/invoke.texi (-msve-vector-bits=): Document new option.
8169 (sve): Document new AArch64 extension.
8170 * doc/md.texi (w): Extend the description of the AArch64
8171 constraint to include SVE vectors.
8172 (Upl, Upa): Document new AArch64 predicate constraints.
8173 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
8175 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
8176 (msve-vector-bits=): New option.
8177 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
8178 SVE when these are disabled.
8179 (sve): New extension.
8180 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
8181 modes. Adjust their number of units based on aarch64_sve_vg.
8182 (MAX_BITSIZE_MODE_ANY_MODE): Define.
8183 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
8184 aarch64_addr_query_type.
8185 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
8186 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
8187 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
8188 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
8189 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
8190 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
8191 (aarch64_simd_imm_zero_p): Delete.
8192 (aarch64_check_zero_based_sve_index_immediate): Declare.
8193 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8194 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8195 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8196 (aarch64_sve_float_mul_immediate_p): Likewise.
8197 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8199 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
8200 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
8201 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
8202 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
8203 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
8204 (aarch64_regmode_natural_size): Likewise.
8205 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
8206 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
8208 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
8209 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
8210 for VG and the SVE predicate registers.
8211 (V_ALIASES): Add a "z"-prefixed alias.
8212 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
8213 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
8214 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
8215 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
8216 (REG_CLASS_NAMES): Add entries for them.
8217 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
8218 and the predicate registers.
8219 (aarch64_sve_vg): Declare.
8220 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
8221 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
8222 (REGMODE_NATURAL_SIZE): Define.
8223 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
8225 * config/aarch64/aarch64.c: Include cfgrtl.h.
8226 (simd_immediate_info): Add a constructor for series vectors,
8227 and an associated step field.
8228 (aarch64_sve_vg): New variable.
8229 (aarch64_dbx_register_number): Handle VG and the predicate registers.
8230 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
8231 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
8232 (VEC_ANY_DATA, VEC_STRUCT): New constants.
8233 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
8234 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
8235 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
8236 (aarch64_get_mask_mode): New functions.
8237 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
8238 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
8239 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
8240 predicate modes and predicate registers. Explicitly restrict
8241 GPRs to modes of 16 bytes or smaller. Only allow FP registers
8242 to store a vector mode if it is recognized by
8243 aarch64_classify_vector_mode.
8244 (aarch64_regmode_natural_size): New function.
8245 (aarch64_hard_regno_caller_save_mode): Return the original mode
8247 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
8248 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
8249 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
8250 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
8252 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
8253 does not overlap dest if the function is frame-related. Handle
8255 (aarch64_split_add_offset): New function.
8256 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
8257 them aarch64_add_offset.
8258 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
8259 and update call to aarch64_sub_sp.
8260 (aarch64_add_cfa_expression): New function.
8261 (aarch64_expand_prologue): Pass extra temporary registers to the
8262 functions above. Handle the case in which we need to emit new
8263 DW_CFA_expressions for registers that were originally saved
8264 relative to the stack pointer, but now have to be expressed
8265 relative to the frame pointer.
8266 (aarch64_output_mi_thunk): Pass extra temporary registers to the
8268 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
8269 IP0 and IP1 values for SVE frames.
8270 (aarch64_expand_vec_series): New function.
8271 (aarch64_expand_sve_widened_duplicate): Likewise.
8272 (aarch64_expand_sve_const_vector): Likewise.
8273 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
8274 Handle SVE constants. Use emit_move_insn to move a force_const_mem
8275 into the register, rather than emitting a SET directly.
8276 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
8277 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
8278 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
8279 (offset_9bit_signed_scaled_p): New functions.
8280 (aarch64_replicate_bitmask_imm): New function.
8281 (aarch64_bitmask_imm): Use it.
8282 (aarch64_cannot_force_const_mem): Reject expressions involving
8283 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
8284 (aarch64_classify_index): Handle SVE indices, by requiring
8285 a plain register index with a scale that matches the element size.
8286 (aarch64_classify_address): Handle SVE addresses. Assert that
8287 the mode of the address is VOIDmode or an integer mode.
8288 Update call to aarch64_classify_symbol.
8289 (aarch64_classify_symbolic_expression): Update call to
8290 aarch64_classify_symbol.
8291 (aarch64_const_vec_all_in_range_p): New function.
8292 (aarch64_print_vector_float_operand): Likewise.
8293 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
8294 "vN" for FP registers with SVE modes. Handle (const ...) vectors
8295 and the FP immediates 1.0 and 0.5.
8296 (aarch64_print_address_internal): Handle SVE addresses.
8297 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
8298 (aarch64_regno_regclass): Handle predicate registers.
8299 (aarch64_secondary_reload): Handle big-endian reloads of SVE
8301 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
8302 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
8303 (aarch64_convert_sve_vector_bits): New function.
8304 (aarch64_override_options): Use it to handle -msve-vector-bits=.
8305 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8307 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
8308 Handle SVE vector and predicate modes. Accept VL-based constants
8309 that need only one temporary register, and VL offsets that require
8310 no temporary registers.
8311 (aarch64_conditional_register_usage): Mark the predicate registers
8312 as fixed if SVE isn't available.
8313 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
8314 Return true for SVE vector and predicate modes.
8315 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
8316 rather than an unsigned int. Handle SVE modes.
8317 (aarch64_preferred_simd_mode): Update call accordingly. Handle
8319 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
8321 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8322 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8323 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8324 (aarch64_sve_float_mul_immediate_p): New functions.
8325 (aarch64_sve_valid_immediate): New function.
8326 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
8327 Explicitly reject structure modes. Check for INDEX constants.
8328 Handle PTRUE and PFALSE constants.
8329 (aarch64_check_zero_based_sve_index_immediate): New function.
8330 (aarch64_simd_imm_zero_p): Delete.
8331 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
8332 vector modes. Accept constants in the range of CNT[BHWD].
8333 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
8334 ask for an Advanced SIMD mode.
8335 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
8336 (aarch64_simd_vector_alignment): Handle SVE predicates.
8337 (aarch64_vectorize_preferred_vector_alignment): New function.
8338 (aarch64_simd_vector_alignment_reachable): Use it instead of
8340 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
8341 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
8343 (MAX_VECT_LEN): Delete.
8344 (expand_vec_perm_d): Add a vec_flags field.
8345 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
8346 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
8347 (aarch64_evpc_ext): Don't apply a big-endian lane correction
8349 (aarch64_evpc_rev): Rename to...
8350 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
8351 (aarch64_evpc_rev_global): New function.
8352 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
8353 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
8355 (aarch64_evpc_sve_tbl): New function.
8356 (aarch64_expand_vec_perm_const_1): Update after rename of
8357 aarch64_evpc_rev. Handle SVE permutes too, trying
8358 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
8359 than aarch64_evpc_tbl.
8360 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
8361 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
8362 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
8363 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
8364 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
8365 (aarch64_expand_sve_vcond): New functions.
8366 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
8367 of aarch64_vector_mode_p.
8368 (aarch64_dwarf_poly_indeterminate_value): New function.
8369 (aarch64_compute_pressure_classes): Likewise.
8370 (aarch64_can_change_mode_class): Likewise.
8371 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
8372 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
8373 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
8374 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
8375 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
8376 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
8377 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
8378 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
8380 (Dn, Dl, Dr): Accept const as well as const_vector.
8381 (Dz): Likewise. Compare against CONST0_RTX.
8382 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
8383 of "vector" where appropriate.
8384 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
8385 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
8386 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
8387 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
8388 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
8389 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
8390 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
8391 (v_int_equiv): Extend to SVE modes.
8392 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
8394 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
8395 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
8396 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
8397 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
8398 (SVE_COND_FP_CMP): New int iterators.
8399 (perm_hilo): Handle the new unpack unspecs.
8400 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
8402 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
8403 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
8404 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
8405 (aarch64_equality_operator, aarch64_constant_vector_operand)
8406 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
8407 (aarch64_sve_nonimmediate_operand): Likewise.
8408 (aarch64_sve_general_operand): Likewise.
8409 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
8410 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
8411 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
8412 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
8413 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
8414 (aarch64_sve_float_arith_immediate): Likewise.
8415 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
8416 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
8417 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
8418 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
8419 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
8420 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
8421 (aarch64_sve_float_arith_operand): Likewise.
8422 (aarch64_sve_float_arith_with_sub_operand): Likewise.
8423 (aarch64_sve_float_mul_operand): Likewise.
8424 (aarch64_sve_vec_perm_operand): Likewise.
8425 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
8426 (aarch64_mov_operand): Accept const_poly_int and const_vector.
8427 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
8428 as well as const_vector.
8429 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
8430 in file. Use CONST0_RTX and CONSTM1_RTX.
8431 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
8432 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
8433 Use aarch64_simd_imm_zero.
8434 * config/aarch64/aarch64-sve.md: New file.
8435 * config/aarch64/aarch64.md: Include it.
8436 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
8437 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
8438 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
8439 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
8440 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
8441 (sve): New attribute.
8442 (enabled): Disable instructions with the sve attribute unless
8444 (movqi, movhi): Pass CONST_POLY_INT operaneds through
8445 aarch64_expand_mov_immediate.
8446 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
8447 CNT[BHSD] immediates.
8448 (movti): Split CONST_POLY_INT moves into two halves.
8449 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
8450 Split additions that need a temporary here if the destination
8451 is the stack pointer.
8452 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
8453 (*add<mode>3_poly_1): New instruction.
8454 (set_clobber_cc): New expander.
8456 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8458 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
8459 parameter and use it instead of GET_MODE_SIZE (innermode). Use
8460 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
8461 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
8462 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
8463 Change innermode from fixed_mode_size to machine_mode.
8464 (simplify_subreg): Update call accordingly. Handle a constant-sized
8465 subreg of a variable-length CONST_VECTOR.
8467 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8468 Alan Hayward <alan.hayward@arm.com>
8469 David Sherwood <david.sherwood@arm.com>
8471 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
8472 (add_offset_to_base): New function, split out from...
8473 (create_mem_ref): ...here. When handling a scale other than 1,
8474 check first whether the address is valid without the offset.
8475 Add it into the base if so, leaving the index and scale as-is.
8477 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8480 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
8481 fold_for_warn before checking if arg2 is INTEGER_CST.
8483 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
8485 * config/rs6000/predicates.md (load_multiple_operation): Delete.
8486 (store_multiple_operation): Delete.
8487 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
8488 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
8489 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
8490 guarded by TARGET_STRING.
8491 (rs6000_output_load_multiple): Delete.
8492 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
8493 OPTION_MASK_STRING / TARGET_STRING handling.
8494 (print_operand) <'N', 'O'>: Add comment that these are unused now.
8495 (const rs6000_opt_masks) <"string">: Change mask to 0.
8496 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
8497 (MASK_STRING): Delete.
8498 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
8500 (load_multiple): Delete.
8507 (store_multiple): Delete.
8514 (movmemsi_8reg): Delete.
8515 (corresponding unnamed define_insn): Delete.
8516 (movmemsi_6reg): Delete.
8517 (corresponding unnamed define_insn): Delete.
8518 (movmemsi_4reg): Delete.
8519 (corresponding unnamed define_insn): Delete.
8520 (movmemsi_2reg): Delete.
8521 (corresponding unnamed define_insn): Delete.
8522 (movmemsi_1reg): Delete.
8523 (corresponding unnamed define_insn): Delete.
8524 * config/rs6000/rs6000.opt (mno-string): New.
8525 (mstring): Replace by deprecation warning stub.
8526 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
8528 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8530 * regrename.c (regrename_do_replace): If replacing the same
8531 reg multiple times, try to reuse last created gen_raw_REG.
8534 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
8535 main to workaround a bug in GDB.
8537 2018-01-12 Tom de Vries <tom@codesourcery.com>
8540 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
8542 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
8544 PR rtl-optimization/80481
8545 * ira-color.c (get_cap_member): New function.
8546 (allocnos_conflict_by_live_ranges_p): Use it.
8547 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
8548 (setup_slot_coalesced_allocno_live_ranges): Ditto.
8550 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
8553 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
8554 (*saddl_se_1): Ditto.
8556 (*ssubl_se_1): Ditto.
8558 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8560 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
8561 rather than wi::to_widest for DR_INITs.
8562 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
8563 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
8564 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
8566 (vect_analyze_group_access_1): Note that here.
8568 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8570 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
8571 polynomial type sizes.
8573 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8575 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
8576 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
8577 (gimple_add_tmp_var): Likewise.
8579 2018-01-12 Martin Liska <mliska@suse.cz>
8581 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
8582 (gimple_alloc_sizes): Likewise.
8583 (dump_gimple_statistics): Use PRIu64 in printf format.
8584 * gimple.h: Change uint64_t to int.
8586 2018-01-12 Martin Liska <mliska@suse.cz>
8588 * tree-core.h: Use uint64_t instead of int.
8589 * tree.c (tree_node_counts): Likewise.
8590 (tree_node_sizes): Likewise.
8591 (dump_tree_statistics): Use PRIu64 in printf format.
8593 2018-01-12 Martin Liska <mliska@suse.cz>
8595 * Makefile.in: As qsort_chk is implemented in vec.c, add
8596 vec.o to linkage of gencfn-macros.
8597 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
8598 passing the info to record_node_allocation_statistics.
8599 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
8601 * ggc-common.c (struct ggc_usage): Add operator== and use
8602 it in operator< and compare function.
8603 * mem-stats.h (struct mem_usage): Likewise.
8604 * vec.c (struct vec_usage): Remove operator< and compare
8605 function. Can be simply inherited.
8607 2018-01-12 Martin Jambor <mjambor@suse.cz>
8610 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
8611 * tree-ssa-math-opts.c: Include domwalk.h.
8612 (convert_mult_to_fma_1): New function.
8613 (fma_transformation_info): New type.
8614 (fma_deferring_state): Likewise.
8615 (cancel_fma_deferring): New function.
8616 (result_of_phi): Likewise.
8617 (last_fma_candidate_feeds_initial_phi): Likewise.
8618 (convert_mult_to_fma): Added deferring logic, split actual
8619 transformation to convert_mult_to_fma_1.
8620 (math_opts_dom_walker): New type.
8621 (math_opts_dom_walker::after_dom_children): New method, body moved
8622 here from pass_optimize_widening_mul::execute, added deferring logic
8624 (pass_optimize_widening_mul::execute): Moved most of code to
8625 math_opts_dom_walker::after_dom_children.
8626 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
8627 * config/i386/i386.c (ix86_option_override_internal): Added
8628 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
8630 2018-01-12 Richard Biener <rguenther@suse.de>
8633 * dwarf2out.c (gen_variable_die): Do not reset old_die for
8634 inline instance vars.
8636 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
8639 * config/rx/rx.c (rx_is_restricted_memory_address):
8642 2018-01-12 Richard Biener <rguenther@suse.de>
8644 PR tree-optimization/80846
8645 * target.def (split_reduction): New target hook.
8646 * targhooks.c (default_split_reduction): New function.
8647 * targhooks.h (default_split_reduction): Declare.
8648 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
8649 target requests first reduce vectors by combining low and high
8651 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
8652 (get_vectype_for_scalar_type_and_size): Export.
8653 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
8654 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
8655 * doc/tm.texi: Regenerate.
8656 * config/i386/i386.c (ix86_split_reduction): Implement
8657 TARGET_VECTORIZE_SPLIT_REDUCTION.
8659 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8662 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
8663 in PIC mode except for TARGET_VXWORKS_RTP.
8664 * config/sparc/sparc.c: Include cfgrtl.h.
8665 (TARGET_INIT_PIC_REG): Define.
8666 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
8667 (sparc_pic_register_p): New predicate.
8668 (sparc_legitimate_address_p): Use it.
8669 (sparc_legitimize_pic_address): Likewise.
8670 (sparc_delegitimize_address): Likewise.
8671 (sparc_mode_dependent_address_p): Likewise.
8672 (gen_load_pcrel_sym): Remove 4th parameter.
8673 (load_got_register): Adjust call to above. Remove obsolete stuff.
8674 (sparc_expand_prologue): Do not call load_got_register here.
8675 (sparc_flat_expand_prologue): Likewise.
8676 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
8677 (sparc_use_pseudo_pic_reg): New function.
8678 (sparc_init_pic_reg): Likewise.
8679 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8680 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
8682 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
8684 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
8685 Add item for branch_cost.
8687 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8689 PR rtl-optimization/83565
8690 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
8691 not extend the result to a larger mode for rotate operations.
8692 (num_sign_bit_copies1): Likewise.
8694 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8697 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
8699 Use values-Xc.o for -pedantic.
8700 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
8702 2018-01-12 Martin Liska <mliska@suse.cz>
8705 * ipa-devirt.c (final_warning_record::grow_type_warnings):
8707 (possible_polymorphic_call_targets): Use it.
8708 (ipa_devirt): Likewise.
8710 2018-01-12 Martin Liska <mliska@suse.cz>
8712 * profile-count.h (enum profile_quality): Use 0 as invalid
8713 enum value of profile_quality.
8715 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
8717 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
8718 -mext-string options.
8720 2018-01-12 Richard Biener <rguenther@suse.de>
8722 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
8723 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
8724 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
8726 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
8728 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
8730 * configure.ac (--with-long-double-format): Add support for the
8731 configuration option to change the default long double format on
8733 * config.gcc (powerpc*-linux*-*): Likewise.
8734 * configure: Regenerate.
8735 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
8736 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
8737 used without modification.
8739 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8741 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
8742 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
8743 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
8744 MISC_BUILTIN_SPEC_BARRIER.
8745 (rs6000_init_builtins): Likewise.
8746 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
8748 (speculation_barrier): New define_insn.
8749 * doc/extend.texi: Document __builtin_speculation_barrier.
8751 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8754 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
8755 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
8756 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
8758 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
8759 integral modes instead of "ss" and "sd".
8760 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
8761 vectors with 32-bit and 64-bit elements.
8762 (vecdupssescalarmodesuffix): New mode attribute.
8763 (vec_dup<mode>): Use it.
8765 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
8768 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
8769 frame if argument is passed on stack.
8771 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8774 * ree.c (combine_reaching_defs): Optimize also
8775 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
8776 reg2=any_extend(exp); reg1=reg2;, formatting fix.
8778 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8781 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
8783 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8786 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
8787 after they are computed.
8789 2018-01-11 Bin Cheng <bin.cheng@arm.com>
8791 PR tree-optimization/83695
8792 * gimple-loop-linterchange.cc
8793 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
8794 reset cached scev information after interchange.
8795 (pass_linterchange::execute): Remove call to scev_reset_htab.
8797 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8799 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
8800 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
8801 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
8802 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
8803 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
8804 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
8805 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
8806 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
8807 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
8808 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
8809 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
8810 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
8811 (V_lane_reg): Likewise.
8812 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
8814 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
8815 (vfmal_lane_low<mode>_intrinsic,
8816 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
8817 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
8818 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
8819 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
8820 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
8821 vfmsl_lane_high<mode>_intrinsic): New define_insns.
8823 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8825 * config/arm/arm-cpus.in (fp16fml): New feature.
8826 (ALL_SIMD): Add fp16fml.
8827 (armv8.2-a): Add fp16fml as an option.
8828 (armv8.3-a): Likewise.
8829 (armv8.4-a): Add fp16fml as part of fp16.
8830 * config/arm/arm.h (TARGET_FP16FML): Define.
8831 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
8833 * config/arm/arm-modes.def (V2HF): Define.
8834 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
8835 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
8836 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
8837 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
8838 vfmsl_low, vfmsl_high): New set of builtins.
8839 * config/arm/iterators.md (PLUSMINUS): New code iterator.
8840 (vfml_op): New code attribute.
8841 (VFMLHALVES): New int iterator.
8842 (VFML, VFMLSEL): New mode attributes.
8843 (V_reg): Define mapping for V2HF.
8844 (V_hi, V_lo): New mode attributes.
8845 (VF_constraint): Likewise.
8846 (vfml_half, vfml_half_selector): New int attributes.
8847 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
8849 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
8850 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
8852 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
8853 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
8854 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
8855 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
8857 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
8858 Document new effective target and option set.
8860 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8862 * config/arm/arm-cpus.in (armv8_4): New feature.
8863 (ARMv8_4a): New fgroup.
8864 (armv8.4-a): New arch.
8865 * config/arm/arm-tables.opt: Regenerate.
8866 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
8867 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
8868 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
8869 Add matching rules for -march=armv8.4-a and extensions.
8870 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
8872 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
8875 * config/rx/rx.md (BW): New mode attribute.
8876 (sync_lock_test_and_setsi): Add mode suffix to insn output.
8878 2018-01-11 Richard Biener <rguenther@suse.de>
8880 PR tree-optimization/83435
8881 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
8882 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
8883 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
8885 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8886 Alan Hayward <alan.hayward@arm.com>
8887 David Sherwood <david.sherwood@arm.com>
8889 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
8891 (aarch64_classify_address): Initialize it. Track polynomial offsets.
8892 (aarch64_print_address_internal): Use it to check for a zero offset.
8894 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8895 Alan Hayward <alan.hayward@arm.com>
8896 David Sherwood <david.sherwood@arm.com>
8898 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
8899 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
8900 Return a poly_int64 rather than a HOST_WIDE_INT.
8901 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
8902 rather than a HOST_WIDE_INT.
8903 * config/aarch64/aarch64.h (aarch64_frame): Protect with
8904 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
8905 hard_fp_offset, frame_size, initial_adjust, callee_offset and
8906 final_offset from HOST_WIDE_INT to poly_int64.
8907 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
8908 to_constant when getting the number of units in an Advanced SIMD
8910 (aarch64_builtin_vectorized_function): Check for a constant number
8912 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
8914 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
8915 attribute instead of GET_MODE_NUNITS.
8916 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
8917 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
8918 GET_MODE_SIZE for fixed-size registers.
8919 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
8920 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
8921 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
8922 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
8923 (aarch64_print_operand, aarch64_print_address_internal)
8924 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
8925 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
8926 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
8927 Handle polynomial GET_MODE_SIZE.
8928 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
8929 wider than SImode without modification.
8930 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
8931 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
8932 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
8933 passing and returning SVE modes.
8934 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
8935 rather than GEN_INT.
8936 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
8937 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
8938 (aarch64_allocate_and_probe_stack_space): Likewise.
8939 (aarch64_layout_frame): Cope with polynomial offsets.
8940 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
8941 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
8943 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
8944 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
8945 poly_int64 rather than a HOST_WIDE_INT.
8946 (aarch64_get_separate_components, aarch64_process_components)
8947 (aarch64_expand_prologue, aarch64_expand_epilogue)
8948 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
8949 (aarch64_anchor_offset): New function, split out from...
8950 (aarch64_legitimize_address): ...here.
8951 (aarch64_builtin_vectorization_cost): Handle polynomial
8952 TYPE_VECTOR_SUBPARTS.
8953 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
8955 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
8956 number of elements from the PARALLEL rather than the mode.
8957 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
8958 rather than GET_MODE_BITSIZE.
8959 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
8960 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
8961 (aarch64_expand_vec_perm_const_1): Handle polynomial
8962 d->perm.length () and d->perm elements.
8963 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
8964 Apply to_constant to d->perm elements.
8965 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
8966 polynomial CONST_VECTOR_NUNITS.
8967 (aarch64_move_pointer): Take amount as a poly_int64 rather
8969 (aarch64_progress_pointer): Avoid temporary variable.
8970 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
8971 the mode attribute instead of GET_MODE.
8973 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8974 Alan Hayward <alan.hayward@arm.com>
8975 David Sherwood <david.sherwood@arm.com>
8977 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
8978 x exists before using it.
8979 (aarch64_add_constant_internal): Rename to...
8980 (aarch64_add_offset_1): ...this. Replace regnum with separate
8981 src and dest rtxes. Handle the case in which they're different,
8982 including when the offset is zero. Replace scratchreg with an rtx.
8983 Use 2 additions if there is no spare register into which we can
8984 move a 16-bit constant.
8985 (aarch64_add_constant): Delete.
8986 (aarch64_add_offset): Replace reg with separate src and dest
8987 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
8988 Use aarch64_add_offset_1.
8989 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
8990 an rtx rather than an int. Take the delta as a poly_int64
8991 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
8992 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
8993 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
8994 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
8995 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
8997 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
8998 aarch64_add_constant.
9000 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9002 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
9003 Use scalar_float_mode.
9005 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9007 * config/aarch64/aarch64-simd.md
9008 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
9009 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
9010 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
9011 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
9012 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
9013 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
9014 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
9015 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
9016 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
9017 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
9019 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9022 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
9023 targ_options->x_arm_arch_string is non NULL.
9025 2018-01-11 Tamar Christina <tamar.christina@arm.com>
9027 * config/aarch64/aarch64.h
9028 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
9030 2018-01-11 Sudakshina Das <sudi.das@arm.com>
9033 * expmed.c (emit_store_flag_force): Swap if const op0
9034 and change VOIDmode to mode of op0.
9036 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9038 PR rtl-optimization/83761
9039 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
9040 than bytes to mode_for_size.
9042 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9045 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
9046 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
9049 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9052 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
9053 when in layout mode.
9054 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
9055 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
9058 2018-01-10 Michael Collison <michael.collison@arm.com>
9060 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
9061 * config/aarch64/aarch64-option-extension.def: Add
9062 AARCH64_OPT_EXTENSION of 'fp16fml'.
9063 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9064 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
9065 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
9066 * config/aarch64/constraints.md (Ui7): New constraint.
9067 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
9068 (VFMLA_SEL_W): Ditto.
9071 (VFMLA16_LOW): New int iterator.
9072 (VFMLA16_HIGH): Ditto.
9073 (UNSPEC_FMLAL): New unspec.
9074 (UNSPEC_FMLSL): Ditto.
9075 (UNSPEC_FMLAL2): Ditto.
9076 (UNSPEC_FMLSL2): Ditto.
9077 (f16mac): New code attribute.
9078 * config/aarch64/aarch64-simd-builtins.def
9079 (aarch64_fmlal_lowv2sf): Ditto.
9080 (aarch64_fmlsl_lowv2sf): Ditto.
9081 (aarch64_fmlalq_lowv4sf): Ditto.
9082 (aarch64_fmlslq_lowv4sf): Ditto.
9083 (aarch64_fmlal_highv2sf): Ditto.
9084 (aarch64_fmlsl_highv2sf): Ditto.
9085 (aarch64_fmlalq_highv4sf): Ditto.
9086 (aarch64_fmlslq_highv4sf): Ditto.
9087 (aarch64_fmlal_lane_lowv2sf): Ditto.
9088 (aarch64_fmlsl_lane_lowv2sf): Ditto.
9089 (aarch64_fmlal_laneq_lowv2sf): Ditto.
9090 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
9091 (aarch64_fmlalq_lane_lowv4sf): Ditto.
9092 (aarch64_fmlsl_lane_lowv4sf): Ditto.
9093 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
9094 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
9095 (aarch64_fmlal_lane_highv2sf): Ditto.
9096 (aarch64_fmlsl_lane_highv2sf): Ditto.
9097 (aarch64_fmlal_laneq_highv2sf): Ditto.
9098 (aarch64_fmlsl_laneq_highv2sf): Ditto.
9099 (aarch64_fmlalq_lane_highv4sf): Ditto.
9100 (aarch64_fmlsl_lane_highv4sf): Ditto.
9101 (aarch64_fmlalq_laneq_highv4sf): Ditto.
9102 (aarch64_fmlsl_laneq_highv4sf): Ditto.
9103 * config/aarch64/aarch64-simd.md:
9104 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
9105 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9106 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
9107 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9108 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
9109 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
9110 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
9111 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
9112 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
9113 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
9114 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
9115 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
9116 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
9117 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
9118 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
9119 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
9120 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
9121 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
9122 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
9123 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
9124 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
9125 (vfmlsl_low_u32): Ditto.
9126 (vfmlalq_low_u32): Ditto.
9127 (vfmlslq_low_u32): Ditto.
9128 (vfmlal_high_u32): Ditto.
9129 (vfmlsl_high_u32): Ditto.
9130 (vfmlalq_high_u32): Ditto.
9131 (vfmlslq_high_u32): Ditto.
9132 (vfmlal_lane_low_u32): Ditto.
9133 (vfmlsl_lane_low_u32): Ditto.
9134 (vfmlal_laneq_low_u32): Ditto.
9135 (vfmlsl_laneq_low_u32): Ditto.
9136 (vfmlalq_lane_low_u32): Ditto.
9137 (vfmlslq_lane_low_u32): Ditto.
9138 (vfmlalq_laneq_low_u32): Ditto.
9139 (vfmlslq_laneq_low_u32): Ditto.
9140 (vfmlal_lane_high_u32): Ditto.
9141 (vfmlsl_lane_high_u32): Ditto.
9142 (vfmlal_laneq_high_u32): Ditto.
9143 (vfmlsl_laneq_high_u32): Ditto.
9144 (vfmlalq_lane_high_u32): Ditto.
9145 (vfmlslq_lane_high_u32): Ditto.
9146 (vfmlalq_laneq_high_u32): Ditto.
9147 (vfmlslq_laneq_high_u32): Ditto.
9148 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
9149 (AARCH64_FL_FOR_ARCH8_4): New.
9150 (AARCH64_ISA_F16FML): New ISA flag.
9151 (TARGET_F16FML): New feature flag for fp16fml.
9152 (doc/invoke.texi): Document new fp16fml option.
9154 2018-01-10 Michael Collison <michael.collison@arm.com>
9156 * config/aarch64/aarch64-builtins.c:
9157 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
9158 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9159 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
9160 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
9161 (AARCH64_ISA_SHA3): New ISA flag.
9162 (TARGET_SHA3): New feature flag for sha3.
9163 * config/aarch64/iterators.md (sha512_op): New int attribute.
9164 (CRYPTO_SHA512): New int iterator.
9165 (UNSPEC_SHA512H): New unspec.
9166 (UNSPEC_SHA512H2): Ditto.
9167 (UNSPEC_SHA512SU0): Ditto.
9168 (UNSPEC_SHA512SU1): Ditto.
9169 * config/aarch64/aarch64-simd-builtins.def
9170 (aarch64_crypto_sha512hqv2di): New builtin.
9171 (aarch64_crypto_sha512h2qv2di): Ditto.
9172 (aarch64_crypto_sha512su0qv2di): Ditto.
9173 (aarch64_crypto_sha512su1qv2di): Ditto.
9174 (aarch64_eor3qv8hi): Ditto.
9175 (aarch64_rax1qv2di): Ditto.
9176 (aarch64_xarqv2di): Ditto.
9177 (aarch64_bcaxqv8hi): Ditto.
9178 * config/aarch64/aarch64-simd.md:
9179 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
9180 (aarch64_crypto_sha512su0qv2di): Ditto.
9181 (aarch64_crypto_sha512su1qv2di): Ditto.
9182 (aarch64_eor3qv8hi): Ditto.
9183 (aarch64_rax1qv2di): Ditto.
9184 (aarch64_xarqv2di): Ditto.
9185 (aarch64_bcaxqv8hi): Ditto.
9186 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
9187 (vsha512h2q_u64): Ditto.
9188 (vsha512su0q_u64): Ditto.
9189 (vsha512su1q_u64): Ditto.
9190 (veor3q_u16): Ditto.
9191 (vrax1q_u64): Ditto.
9193 (vbcaxq_u16): Ditto.
9194 * config/arm/types.md (crypto_sha512): New type attribute.
9195 (crypto_sha3): Ditto.
9196 (doc/invoke.texi): Document new sha3 option.
9198 2018-01-10 Michael Collison <michael.collison@arm.com>
9200 * config/aarch64/aarch64-builtins.c:
9201 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
9202 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9203 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
9204 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
9205 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
9206 (AARCH64_ISA_SM4): New ISA flag.
9207 (TARGET_SM4): New feature flag for sm4.
9208 * config/aarch64/aarch64-simd-builtins.def
9209 (aarch64_sm3ss1qv4si): Ditto.
9210 (aarch64_sm3tt1aq4si): Ditto.
9211 (aarch64_sm3tt1bq4si): Ditto.
9212 (aarch64_sm3tt2aq4si): Ditto.
9213 (aarch64_sm3tt2bq4si): Ditto.
9214 (aarch64_sm3partw1qv4si): Ditto.
9215 (aarch64_sm3partw2qv4si): Ditto.
9216 (aarch64_sm4eqv4si): Ditto.
9217 (aarch64_sm4ekeyqv4si): Ditto.
9218 * config/aarch64/aarch64-simd.md:
9219 (aarch64_sm3ss1qv4si): Ditto.
9220 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
9221 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
9222 (aarch64_sm4eqv4si): Ditto.
9223 (aarch64_sm4ekeyqv4si): Ditto.
9224 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
9225 (sm3part_op): Ditto.
9226 (CRYPTO_SM3TT): Ditto.
9227 (CRYPTO_SM3PART): Ditto.
9228 (UNSPEC_SM3SS1): New unspec.
9229 (UNSPEC_SM3TT1A): Ditto.
9230 (UNSPEC_SM3TT1B): Ditto.
9231 (UNSPEC_SM3TT2A): Ditto.
9232 (UNSPEC_SM3TT2B): Ditto.
9233 (UNSPEC_SM3PARTW1): Ditto.
9234 (UNSPEC_SM3PARTW2): Ditto.
9235 (UNSPEC_SM4E): Ditto.
9236 (UNSPEC_SM4EKEY): Ditto.
9237 * config/aarch64/constraints.md (Ui2): New constraint.
9238 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
9239 * config/arm/types.md (crypto_sm3): New type attribute.
9240 (crypto_sm4): Ditto.
9241 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
9242 (vsm3tt1aq_u32): Ditto.
9243 (vsm3tt1bq_u32): Ditto.
9244 (vsm3tt2aq_u32): Ditto.
9245 (vsm3tt2bq_u32): Ditto.
9246 (vsm3partw1q_u32): Ditto.
9247 (vsm3partw2q_u32): Ditto.
9248 (vsm4eq_u32): Ditto.
9249 (vsm4ekeyq_u32): Ditto.
9250 (doc/invoke.texi): Document new sm4 option.
9252 2018-01-10 Michael Collison <michael.collison@arm.com>
9254 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
9255 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
9256 (AARCH64_FL_FOR_ARCH8_4): New.
9257 (AARCH64_FL_V8_4): New flag.
9258 (doc/invoke.texi): Document new armv8.4-a option.
9260 2018-01-10 Michael Collison <michael.collison@arm.com>
9262 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9263 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
9264 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
9265 * config/aarch64/aarch64-option-extension.def: Add
9266 AARCH64_OPT_EXTENSION of 'sha2'.
9267 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
9268 (crypto): Disable sha2 and aes if crypto disabled.
9269 (crypto): Enable aes and sha2 if enabled.
9270 (simd): Disable sha2 and aes if simd disabled.
9271 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
9273 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
9274 (TARGET_SHA2): New feature flag for sha2.
9275 (TARGET_AES): New feature flag for aes.
9276 * config/aarch64/aarch64-simd.md:
9277 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
9278 conditional on TARGET_AES.
9279 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
9280 (aarch64_crypto_sha1hsi): Make pattern conditional
9282 (aarch64_crypto_sha1hv4si): Ditto.
9283 (aarch64_be_crypto_sha1hv4si): Ditto.
9284 (aarch64_crypto_sha1su1v4si): Ditto.
9285 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
9286 (aarch64_crypto_sha1su0v4si): Ditto.
9287 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
9288 (aarch64_crypto_sha256su0v4si): Ditto.
9289 (aarch64_crypto_sha256su1v4si): Ditto.
9290 (doc/invoke.texi): Document new aes and sha2 options.
9292 2018-01-10 Martin Sebor <msebor@redhat.com>
9294 PR tree-optimization/83781
9295 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
9298 2018-01-11 Martin Sebor <msebor@gmail.com>
9299 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9301 PR tree-optimization/83501
9302 PR tree-optimization/81703
9304 * tree-ssa-strlen.c (get_string_cst): Rename...
9305 (get_string_len): ...to this. Handle global constants.
9306 (handle_char_store): Adjust.
9308 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
9309 Jim Wilson <jimw@sifive.com>
9311 * config/riscv/riscv-protos.h (riscv_output_return): New.
9312 * config/riscv/riscv.c (struct machine_function): New naked_p field.
9313 (riscv_attribute_table, riscv_output_return),
9314 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
9315 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
9316 (riscv_compute_frame_info): Only compute frame->mask if not a naked
9318 (riscv_expand_prologue): Add early return for naked function.
9319 (riscv_expand_epilogue): Likewise.
9320 (riscv_function_ok_for_sibcall): Return false for naked function.
9321 (riscv_set_current_function): New.
9322 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
9323 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
9324 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
9325 * doc/extend.texi (RISC-V Function Attributes): New.
9327 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
9329 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
9330 check for 128-bit long double before checking TCmode.
9331 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
9332 128-bit long doubles before checking TFmode or TCmode.
9333 (FLOAT128_IBM_P): Likewise.
9335 2018-01-10 Martin Sebor <msebor@redhat.com>
9337 PR tree-optimization/83671
9338 * builtins.c (c_strlen): Unconditionally return zero for the empty
9340 Use -Warray-bounds for warnings.
9341 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
9342 for non-constant array indices with COMPONENT_REF, arrays of
9343 arrays, and pointers to arrays.
9344 (gimple_fold_builtin_strlen): Determine and set length range for
9345 non-constant character arrays.
9347 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
9350 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
9353 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
9355 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
9357 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9360 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
9361 VECTOR_MEM_ALTIVEC_OR_VSX_P.
9362 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
9363 indexed_or_indirect_operand predicate.
9364 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
9365 (*vsx_le_perm_load_v8hi): Likewise.
9366 (*vsx_le_perm_load_v16qi): Likewise.
9367 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
9368 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
9369 (*vsx_le_perm_store_v8hi): Likewise.
9370 (*vsx_le_perm_store_v16qi): Likewise.
9371 (eight unnamed splitters): Likewise.
9373 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9375 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
9376 * config/rs6000/emmintrin.h: Likewise.
9377 * config/rs6000/mmintrin.h: Likewise.
9378 * config/rs6000/xmmintrin.h: Likewise.
9380 2018-01-10 David Malcolm <dmalcolm@redhat.com>
9383 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
9385 * tree.c (tree_nop_conversion): Return true for location wrapper
9387 (maybe_wrap_with_location): New function.
9388 (selftest::check_strip_nops): New function.
9389 (selftest::test_location_wrappers): New function.
9390 (selftest::tree_c_tests): Call it.
9391 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
9392 (maybe_wrap_with_location): New decl.
9393 (EXPR_LOCATION_WRAPPER_P): New macro.
9394 (location_wrapper_p): New inline function.
9395 (tree_strip_any_location_wrapper): New inline function.
9397 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
9400 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
9401 stack_realign_offset for the largest alignment of stack slot
9403 (ix86_find_max_used_stack_alignment): New function.
9404 (ix86_finalize_stack_frame_flags): Use it. Set
9405 max_used_stack_alignment if we don't realign stack.
9406 * config/i386/i386.h (machine_function): Add
9407 max_used_stack_alignment.
9409 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
9411 * config/arm/arm.opt (-mbranch-cost): New option.
9412 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
9415 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
9418 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
9419 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
9421 2018-01-10 Richard Biener <rguenther@suse.de>
9424 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
9425 early out so it also covers the case where we have a non-NULL
9428 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9430 PR tree-optimization/83753
9431 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
9432 for non-strided grouped accesses if the number of elements is 1.
9434 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9437 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
9438 * i386.h (TARGET_USE_GATHER): Define.
9439 * x86-tune.def (X86_TUNE_USE_GATHER): New.
9441 2018-01-10 Martin Liska <mliska@suse.cz>
9444 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
9445 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
9447 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
9448 CLEANUP_NO_PARTITIONING is not set.
9450 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9452 * doc/rtl.texi: Remove documentation of (const ...) wrappers
9453 for vectors, as a partial revert of r254296.
9454 * rtl.h (const_vec_p): Delete.
9455 (const_vec_duplicate_p): Don't test for vector CONSTs.
9456 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
9457 * expmed.c (make_tree): Likewise.
9460 * common.md (E, F): Use CONSTANT_P instead of checking for
9462 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9463 checking for CONST_VECTOR.
9465 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9468 * predict.c (force_edge_cold): Handle in more sane way edges
9471 2018-01-09 Carl Love <cel@us.ibm.com>
9473 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
9475 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
9476 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
9477 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
9478 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
9479 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
9480 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
9481 * config/rs6000/rs6000-protos.h: Add extern defition for
9482 rs6000_generate_float2_double_code.
9483 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
9485 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
9486 (float2_v2df): Add define_expand.
9488 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
9491 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
9492 op_mode in the force_to_mode call.
9494 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9496 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
9497 instead of checking each element individually.
9498 (aarch64_evpc_uzp): Likewise.
9499 (aarch64_evpc_zip): Likewise.
9500 (aarch64_evpc_ext): Likewise.
9501 (aarch64_evpc_rev): Likewise.
9502 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
9503 instead of checking each element individually. Return true without
9505 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
9506 whether all selected elements come from the same input, instead of
9507 checking each element individually. Remove calls to gen_rtx_REG,
9508 start_sequence and end_sequence and instead assert that no rtl is
9511 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9513 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
9514 order of HIGH and CONST checks.
9516 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9518 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
9519 if the destination isn't an SSA_NAME.
9521 2018-01-09 Richard Biener <rguenther@suse.de>
9523 PR tree-optimization/83668
9524 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
9526 (canonicalize_loop_form): ... here, renamed from ...
9527 (canonicalize_loop_closed_ssa_form): ... this and amended to
9528 swap successor edges for loop exit blocks to make us use
9529 the RPO order we need for initial schedule generation.
9531 2018-01-09 Joseph Myers <joseph@codesourcery.com>
9533 PR tree-optimization/64811
9534 * match.pd: When optimizing comparisons with Inf, avoid
9535 introducing or losing exceptions from comparisons with NaN.
9537 2018-01-09 Martin Liska <mliska@suse.cz>
9540 * asan.c (shadow_mem_size): Add gcc_assert.
9542 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
9544 Don't save registers in main().
9547 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
9548 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
9549 * config/avr/avr.c (avr_set_current_function): Don't error if
9550 naked, OS_task or OS_main are specified at the same time.
9551 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
9553 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
9555 * common/config/avr/avr-common.c (avr_option_optimization_table):
9556 Switch on -mmain-is-OS_task for optimizing compilations.
9558 2018-01-09 Richard Biener <rguenther@suse.de>
9560 PR tree-optimization/83572
9561 * graphite.c: Include cfganal.h.
9562 (graphite_transform_loops): Connect infinite loops to exit
9563 and remove fake edges at the end.
9565 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9567 * ipa-inline.c (edge_badness): Revert accidental checkin.
9569 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9572 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
9573 symbols; not inline clones.
9575 2018-01-09 Jakub Jelinek <jakub@redhat.com>
9578 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
9579 hard registers. Formatting fixes.
9581 PR preprocessor/83722
9582 * gcc.c (try_generate_repro): Pass
9583 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
9584 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
9587 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
9588 Kito Cheng <kito.cheng@gmail.com>
9590 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
9591 (riscv_leaf_function_p): Delete.
9592 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
9594 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9596 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
9598 (do_ifelse): New function.
9599 (do_isel): New function.
9600 (do_sub3): New function.
9601 (do_add3): New function.
9602 (do_load_mask_compare): New function.
9603 (do_overlap_load_compare): New function.
9604 (expand_compare_loop): New function.
9605 (expand_block_compare): Call expand_compare_loop() when appropriate.
9606 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
9608 (-mblock-compare-inline-loop-limit): New option.
9610 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9613 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
9614 Reverse order of second and third operands in first alternative.
9615 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
9616 of first and second elements in UNSPEC_VPERMR vector.
9617 (altivec_expand_vec_perm_le): Likewise.
9619 2018-01-08 Jeff Law <law@redhat.com>
9621 PR rtl-optimizatin/81308
9622 * tree-switch-conversion.c (cfg_altered): New file scoped static.
9623 (process_switch): If group_case_labels makes a change, then set
9625 (pass_convert_switch::execute): If a switch is converted, then
9626 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
9628 PR rtl-optimization/81308
9629 * recog.c (split_all_insns): Conditionally cleanup the CFG after
9632 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
9634 PR target/83663 - Revert r255946
9635 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
9636 generation for cases where splatting a value is not useful.
9637 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
9638 across a vec_duplicate and a paradoxical subreg forming a vector
9639 mode to a vec_concat.
9641 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9643 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
9644 -march=armv8.3-a variants.
9645 * config/arm/t-multilib: Likewise.
9646 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
9648 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9650 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
9652 (cceq_ior_compare_complement): Give it a name so I can use it, and
9653 change boolean_or_operator predicate to boolean_operator so it can
9654 be used to generate a crand.
9655 (eqne): New code iterator.
9656 (bd/bd_neg): New code_attrs.
9657 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
9658 a single define_insn.
9659 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
9660 decrement (bdnzt/bdnzf/bdzt/bdzf).
9661 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
9662 with the new names of the branch decrement patterns, and added the
9663 names of the branch decrement conditional patterns.
9665 2018-01-08 Richard Biener <rguenther@suse.de>
9667 PR tree-optimization/83563
9668 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
9671 2018-01-08 Richard Biener <rguenther@suse.de>
9674 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
9676 2018-01-08 Richard Biener <rguenther@suse.de>
9678 PR tree-optimization/83685
9679 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
9680 references to abnormals.
9682 2018-01-08 Richard Biener <rguenther@suse.de>
9685 * dwarf2out.c (output_indirect_strings): Handle empty
9686 skeleton_debug_str_hash.
9687 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
9689 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9691 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
9692 (emit_store_direct): Likewise.
9693 (arc_trampoline_adjust_address): Likewise.
9694 (arc_asm_trampoline_template): New function.
9695 (arc_initialize_trampoline): Use asm_trampoline_template.
9696 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
9697 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
9698 * config/arc/arc.md (flush_icache): Delete pattern.
9700 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9702 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
9703 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
9706 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9709 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
9710 by not USED_FOR_TARGET.
9711 (make_pass_resolve_sw_modes): Likewise.
9713 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9715 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
9718 2018-01-08 Richard Biener <rguenther@suse.de>
9721 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
9723 2018-01-08 Richard Biener <rguenther@suse.de>
9726 * match.pd ((t * 2) / 2) -> t): Add missing :c.
9728 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
9731 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
9732 basic blocks with a small number of successors.
9733 (convert_control_dep_chain_into_preds): Improve handling of
9735 (dump_predicates): Split apart into...
9736 (dump_pred_chain): ...here...
9737 (dump_pred_info): ...and here.
9738 (can_one_predicate_be_invalidated_p): Add debugging printfs.
9739 (can_chain_union_be_invalidated_p): Improve check for invalidation
9741 (uninit_uses_cannot_happen): Avoid unnecessary if
9742 convert_control_dep_chain_into_preds yielded nothing.
9744 2018-01-06 Martin Sebor <msebor@redhat.com>
9746 PR tree-optimization/83640
9747 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
9748 subtracting negative offset from size.
9749 (builtin_access::overlap): Adjust offset bounds of the access to fall
9750 within the size of the object if possible.
9752 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
9754 PR rtl-optimization/83699
9755 * expmed.c (extract_bit_field_1): Restrict the vector usage of
9756 extract_bit_field_as_subreg to cases in which the extracted
9757 value is also a vector.
9759 * lra-constraints.c (process_alt_operands): Test for the equivalence
9760 substitutions when detecting a possible reload cycle.
9762 2018-01-06 Jakub Jelinek <jakub@redhat.com>
9765 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
9766 by default if flag_selective_schedling{,2}. Formatting fixes.
9768 PR rtl-optimization/83682
9769 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
9770 if it has non-VECTOR_MODE element mode.
9771 (vec_duplicate_p): Likewise.
9774 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
9775 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
9777 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9780 * config/i386/i386-builtin.def
9781 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
9782 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
9783 Require also OPTION_MASK_ISA_AVX512F in addition to
9784 OPTION_MASK_ISA_GFNI.
9785 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
9786 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
9787 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
9788 to OPTION_MASK_ISA_GFNI.
9789 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
9790 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
9791 OPTION_MASK_ISA_AVX512BW.
9792 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
9793 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
9794 addition to OPTION_MASK_ISA_GFNI.
9795 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
9796 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
9797 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
9798 to OPTION_MASK_ISA_GFNI.
9799 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
9800 a requirement for all ISAs rather than any of them with a few
9802 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
9804 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
9805 bitmasks to be enabled with 3 exceptions, instead of requiring any
9806 enabled ISA with lots of exceptions.
9807 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
9808 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
9809 Change avx512bw in isa attribute to avx512f.
9810 * config/i386/sgxintrin.h: Add license boilerplate.
9811 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
9812 to __AVX512F__ and __AVX512VL to __AVX512VL__.
9813 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
9814 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
9816 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
9817 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
9818 temporarily sse2 rather than sse if not enabled already.
9821 * config/i386/sse.md (VI248_VLBW): Rename to ...
9822 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
9823 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
9824 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
9825 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
9826 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
9827 mode iterator instead of VI248_VLBW.
9829 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
9831 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
9832 (record_modified): Skip clobbers; add debug output.
9833 (param_change_prob): Use sreal frequencies.
9835 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9837 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
9838 punt for user-aligned variables.
9840 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9842 * tree-chrec.c (chrec_contains_symbols): Return true for
9845 2018-01-05 Sudakshina Das <sudi.das@arm.com>
9848 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
9849 of (x|y) == x for BICS pattern.
9851 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9853 PR tree-optimization/83605
9854 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
9855 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
9858 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
9860 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
9861 * config/epiphany/rtems.h: New file.
9863 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9864 Uros Bizjak <ubizjak@gmail.com>
9867 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
9868 QIreg_operand instead of register_operand predicate.
9869 * config/i386/i386.c (ix86_rop_should_change_byte_p,
9870 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
9871 comments instead of -fmitigate[-_]rop.
9873 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9876 * cgraphunit.c (symbol_table::compile): Switch to text_section
9877 before calling assembly_start debug hook.
9878 * run-rtl-passes.c (run_rtl_passes): Likewise.
9881 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9883 * tree-vrp.c (extract_range_from_binary_expr_1): Check
9884 range_int_cst_p rather than !symbolic_range_p before calling
9885 extract_range_from_multiplicative_op_1.
9887 2018-01-04 Jeff Law <law@redhat.com>
9889 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
9890 redundant test in assertion.
9892 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9894 * doc/rtl.texi: Document machine_mode wrapper classes.
9896 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9898 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
9901 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9903 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
9904 the VEC_PERM_EXPR fold to fail.
9906 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9909 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
9910 to switched_sections.
9912 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9915 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
9918 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
9921 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
9922 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
9924 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9927 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
9928 is BLKmode and bitpos not zero or mode change is needed.
9930 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9933 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
9936 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
9939 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
9940 instead of MULT rtx. Update all corresponding splitters.
9942 (*ssub<modesuffix>): Ditto.
9944 (*cmp_sadd_di): Update split patterns.
9945 (*cmp_sadd_si): Ditto.
9946 (*cmp_sadd_sidi): Ditto.
9947 (*cmp_ssub_di): Ditto.
9948 (*cmp_ssub_si): Ditto.
9949 (*cmp_ssub_sidi): Ditto.
9950 * config/alpha/predicates.md (const23_operand): New predicate.
9951 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
9952 Look for ASHIFT, not MULT inner operand.
9953 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
9955 2018-01-04 Martin Liska <mliska@suse.cz>
9957 PR gcov-profile/83669
9958 * gcov.c (output_intermediate_file): Add version to intermediate
9960 * doc/gcov.texi: Document new field 'version' in intermediate
9961 file format. Fix location of '-k' option of gcov command.
9963 2018-01-04 Martin Liska <mliska@suse.cz>
9966 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
9968 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9970 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
9972 2018-01-03 Martin Sebor <msebor@redhat.com>
9974 PR tree-optimization/83655
9975 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
9976 checking calls with invalid arguments.
9978 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9980 * tree-vect-stmts.c (vect_get_store_rhs): New function.
9981 (vectorizable_mask_load_store): Delete.
9982 (vectorizable_call): Return false for masked loads and stores.
9983 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
9984 instead of gimple_assign_rhs1.
9985 (vectorizable_load): Handle IFN_MASK_LOAD.
9986 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
9988 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9990 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
9992 (vectorizable_mask_load_store): ...here.
9993 (vectorizable_load): ...and here.
9995 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9997 * tree-vect-stmts.c (vect_build_all_ones_mask)
9998 (vect_build_zero_merge_argument): New functions, split out from...
9999 (vectorizable_load): ...here.
10001 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10003 * tree-vect-stmts.c (vect_check_store_rhs): New function,
10005 (vectorizable_mask_load_store): ...here.
10006 (vectorizable_store): ...and here.
10008 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10010 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
10012 (vectorizable_mask_load_store): ...here.
10014 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10016 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
10017 (vect_model_store_cost): Take a vec_load_store_type instead of a
10019 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
10020 (vect_model_store_cost): Take a vec_load_store_type instead of a
10022 (vectorizable_mask_load_store): Update accordingly.
10023 (vectorizable_store): Likewise.
10024 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
10026 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10028 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
10029 IFN_MASK_LOAD calls here rather than...
10030 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
10032 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10033 Alan Hayward <alan.hayward@arm.com>
10034 David Sherwood <david.sherwood@arm.com>
10036 * expmed.c (extract_bit_field_1): For vector extracts,
10037 fall back to extract_bit_field_as_subreg if vec_extract
10040 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10041 Alan Hayward <alan.hayward@arm.com>
10042 David Sherwood <david.sherwood@arm.com>
10044 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
10045 they are variable or constant sized.
10046 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
10047 slots for constant-sized data.
10049 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10050 Alan Hayward <alan.hayward@arm.com>
10051 David Sherwood <david.sherwood@arm.com>
10053 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
10054 handling COND_EXPRs with boolean comparisons, try to find a better
10055 basis for the mask type than the boolean itself.
10057 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10059 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
10060 is calculated and how it can be overridden.
10061 * genmodes.c (max_bitsize_mode_any_mode): New variable.
10062 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
10064 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
10067 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10068 Alan Hayward <alan.hayward@arm.com>
10069 David Sherwood <david.sherwood@arm.com>
10071 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
10072 Remove the mode argument.
10073 (aarch64_simd_valid_immediate): Remove the mode and inverse
10075 * config/aarch64/iterators.md (bitsize): New iterator.
10076 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
10077 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
10078 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
10079 aarch64_simd_valid_immediate.
10080 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
10081 (aarch64_reg_or_bic_imm): Likewise.
10082 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
10083 with an insn_type enum and msl with a modifier_type enum.
10084 Replace element_width with a scalar_mode. Change the shift
10085 to unsigned int. Add constructors for scalar_float_mode and
10086 scalar_int_mode elements.
10087 (aarch64_vect_float_const_representable_p): Delete.
10088 (aarch64_can_const_movi_rtx_p)
10089 (aarch64_simd_scalar_immediate_valid_for_move)
10090 (aarch64_simd_make_constant): Update call to
10091 aarch64_simd_valid_immediate.
10092 (aarch64_advsimd_valid_immediate_hs): New function.
10093 (aarch64_advsimd_valid_immediate): Likewise.
10094 (aarch64_simd_valid_immediate): Remove mode and inverse
10095 arguments. Rewrite to use the above. Use const_vec_duplicate_p
10096 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
10097 and aarch64_float_const_representable_p on the result.
10098 (aarch64_output_simd_mov_immediate): Remove mode argument.
10099 Update call to aarch64_simd_valid_immediate and use of
10100 simd_immediate_info.
10101 (aarch64_output_scalar_simd_mov_immediate): Update call
10104 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10105 Alan Hayward <alan.hayward@arm.com>
10106 David Sherwood <david.sherwood@arm.com>
10108 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
10109 (mode_nunits): Likewise CONST_MODE_NUNITS.
10110 * machmode.def (ADJUST_NUNITS): Document.
10111 * genmodes.c (mode_data::need_nunits_adj): New field.
10112 (blank_mode): Update accordingly.
10113 (adj_nunits): New variable.
10114 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
10116 (emit_mode_size_inline): Set need_bytesize_adj for all modes
10117 listed in adj_nunits.
10118 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
10119 listed in adj_nunits. Don't emit case statements for such modes.
10120 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
10121 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
10122 nothing if adj_nunits is nonnull.
10123 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
10124 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
10125 (emit_mode_fbit): Update use of print_maybe_const_decl.
10126 (emit_move_size): Likewise. Treat the array as non-const
10128 (emit_mode_adjustments): Handle adj_nunits.
10130 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10132 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
10133 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
10134 (VECTOR_MODES): Use it.
10135 (make_vector_modes): Take the prefix as an argument.
10137 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10138 Alan Hayward <alan.hayward@arm.com>
10139 David Sherwood <david.sherwood@arm.com>
10141 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
10142 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
10143 for MODE_VECTOR_BOOL.
10144 * machmode.def (VECTOR_BOOL_MODE): Document.
10145 * genmodes.c (VECTOR_BOOL_MODE): New macro.
10146 (make_vector_bool_mode): New function.
10147 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
10149 * lto-streamer-in.c (lto_input_mode_table): Likewise.
10150 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
10152 * stor-layout.c (int_mode_for_mode): Likewise.
10153 * tree.c (build_vector_type_for_mode): Likewise.
10154 * varasm.c (output_constant_pool_2): Likewise.
10155 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
10156 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
10157 for MODE_VECTOR_BOOL.
10158 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
10159 of mode class checks.
10160 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
10161 instead of a list of mode class checks.
10162 (expand_vector_scalar_condition): Likewise.
10163 (type_for_widest_vector_mode): Handle BImode as an inner mode.
10165 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10166 Alan Hayward <alan.hayward@arm.com>
10167 David Sherwood <david.sherwood@arm.com>
10169 * machmode.h (mode_size): Change from unsigned short to
10171 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
10172 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10173 or if measurement_type is not polynomial.
10174 (fixed_size_mode::includes_p): Check for constant-sized modes.
10175 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
10176 return a poly_uint16 rather than an unsigned short.
10177 (emit_mode_size): Change the type of mode_size from unsigned short
10178 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
10179 (emit_mode_adjustments): Cope with polynomial vector sizes.
10180 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10182 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10184 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
10185 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
10186 * caller-save.c (setup_save_areas): Likewise.
10187 (replace_reg_with_saved_mem): Likewise.
10188 * calls.c (emit_library_call_value_1): Likewise.
10189 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
10190 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
10191 (gen_lowpart_for_combine): Likewise.
10192 * convert.c (convert_to_integer_1): Likewise.
10193 * cse.c (equiv_constant, cse_insn): Likewise.
10194 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
10195 (cselib_subst_to_values): Likewise.
10196 * dce.c (word_dce_process_block): Likewise.
10197 * df-problems.c (df_word_lr_mark_ref): Likewise.
10198 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
10199 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
10200 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
10201 (rtl_for_decl_location): Likewise.
10202 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
10203 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
10204 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
10205 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
10206 (expand_expr_real_1): Likewise.
10207 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
10208 (pad_below): Likewise.
10209 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10210 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
10211 * ira.c (get_subreg_tracking_sizes): Likewise.
10212 * ira-build.c (ira_create_allocno_objects): Likewise.
10213 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
10214 (ira_sort_regnos_for_alter_reg): Likewise.
10215 * ira-costs.c (record_operand_costs): Likewise.
10216 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
10217 (resolve_simple_move): Likewise.
10218 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
10219 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
10220 (lra_constraints): Likewise.
10221 (CONST_POOL_OK_P): Reject variable-sized modes.
10222 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
10223 (add_pseudo_to_slot, lra_spill): Likewise.
10224 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10225 * optabs-query.c (get_best_extraction_insn): Likewise.
10226 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10227 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
10228 (expand_mult_highpart, valid_multiword_target_p): Likewise.
10229 * recog.c (offsettable_address_addr_space_p): Likewise.
10230 * regcprop.c (maybe_mode_change): Likewise.
10231 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
10232 * regrename.c (build_def_use): Likewise.
10233 * regstat.c (dump_reg_info): Likewise.
10234 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
10235 (find_reloads, find_reloads_subreg_address): Likewise.
10236 * reload1.c (eliminate_regs_1): Likewise.
10237 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
10238 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
10239 (simplify_binary_operation_1, simplify_subreg): Likewise.
10240 * targhooks.c (default_function_arg_padding): Likewise.
10241 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
10242 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
10243 (verify_gimple_assign_ternary): Likewise.
10244 * tree-inline.c (estimate_move_cost): Likewise.
10245 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10246 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
10247 (get_address_cost_ainc): Likewise.
10248 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
10249 (vect_supportable_dr_alignment): Likewise.
10250 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10251 (vectorizable_reduction): Likewise.
10252 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
10253 (vectorizable_operation, vectorizable_load): Likewise.
10254 * tree.c (build_same_sized_truth_vector_type): Likewise.
10255 * valtrack.c (cleanup_auto_inc_dec): Likewise.
10256 * var-tracking.c (emit_note_insn_var_location): Likewise.
10257 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
10258 (ADDR_VEC_ALIGN): Likewise.
10260 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10261 Alan Hayward <alan.hayward@arm.com>
10262 David Sherwood <david.sherwood@arm.com>
10264 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
10266 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10267 or if measurement_type is polynomial.
10268 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
10269 * combine.c (make_extraction): Likewise.
10270 * dse.c (find_shift_sequence): Likewise.
10271 * dwarf2out.c (mem_loc_descriptor): Likewise.
10272 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
10273 (extract_bit_field, extract_low_bits): Likewise.
10274 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
10275 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
10276 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
10277 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
10278 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10279 * reload.c (find_reloads): Likewise.
10280 * reload1.c (alter_reg): Likewise.
10281 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
10282 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
10283 * tree-if-conv.c (predicate_mem_writes): Likewise.
10284 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
10285 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
10286 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
10287 * valtrack.c (dead_debug_insert_temp): Likewise.
10288 * varasm.c (mergeable_constant_section): Likewise.
10289 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
10291 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10292 Alan Hayward <alan.hayward@arm.com>
10293 David Sherwood <david.sherwood@arm.com>
10295 * expr.c (expand_assignment): Cope with polynomial mode sizes
10296 when assigning to a CONCAT.
10298 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10299 Alan Hayward <alan.hayward@arm.com>
10300 David Sherwood <david.sherwood@arm.com>
10302 * machmode.h (mode_precision): Change from unsigned short to
10304 (mode_to_precision): Return a poly_uint16 rather than an unsigned
10306 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
10307 or if measurement_type is not polynomial.
10308 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
10309 in which the mode is already known to be a scalar_int_mode.
10310 * genmodes.c (emit_mode_precision): Change the type of mode_precision
10311 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
10313 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10314 for GET_MODE_PRECISION.
10315 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10316 for GET_MODE_PRECISION.
10317 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
10319 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
10320 (expand_field_assignment, make_extraction): Likewise.
10321 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
10322 (get_last_value): Likewise.
10323 * convert.c (convert_to_integer_1): Likewise.
10324 * cse.c (cse_insn): Likewise.
10325 * expr.c (expand_expr_real_1): Likewise.
10326 * lra-constraints.c (simplify_operand_subreg): Likewise.
10327 * optabs-query.c (can_atomic_load_p): Likewise.
10328 * optabs.c (expand_atomic_load): Likewise.
10329 (expand_atomic_store): Likewise.
10330 * ree.c (combine_reaching_defs): Likewise.
10331 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
10332 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
10333 * tree.h (type_has_mode_precision_p): Likewise.
10334 * ubsan.c (instrument_si_overflow): Likewise.
10336 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10337 Alan Hayward <alan.hayward@arm.com>
10338 David Sherwood <david.sherwood@arm.com>
10340 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
10341 polynomial numbers of units.
10342 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
10343 (valid_vector_subparts_p): New function.
10344 (build_vector_type): Remove temporary shim and take the number
10345 of units as a poly_uint64 rather than an int.
10346 (build_opaque_vector_type): Take the number of units as a
10347 poly_uint64 rather than an int.
10348 * tree.c (build_vector_from_ctor): Handle polynomial
10349 TYPE_VECTOR_SUBPARTS.
10350 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
10351 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
10352 (build_vector_from_val): If the number of units is variable,
10353 use build_vec_duplicate_cst for constant operands and
10354 VEC_DUPLICATE_EXPR otherwise.
10355 (make_vector_type): Remove temporary is_constant ().
10356 (build_vector_type, build_opaque_vector_type): Take the number of
10357 units as a poly_uint64 rather than an int.
10358 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
10360 * cfgexpand.c (expand_debug_expr): Likewise.
10361 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
10362 (store_constructor, expand_expr_real_1): Likewise.
10363 (const_scalar_mask_from_tree): Likewise.
10364 * fold-const-call.c (fold_const_reduction): Likewise.
10365 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
10366 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
10367 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
10368 (fold_relational_const): Likewise.
10369 (native_interpret_vector): Likewise. Change the size from an
10370 int to an unsigned int.
10371 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
10372 TYPE_VECTOR_SUBPARTS.
10373 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
10374 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
10375 duplicating a non-constant operand into a variable-length vector.
10376 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
10377 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
10378 * ipa-icf.c (sem_variable::equals): Likewise.
10379 * match.pd: Likewise.
10380 * omp-simd-clone.c (simd_clone_subparts): Likewise.
10381 * print-tree.c (print_node): Likewise.
10382 * stor-layout.c (layout_type): Likewise.
10383 * targhooks.c (default_builtin_vectorization_cost): Likewise.
10384 * tree-cfg.c (verify_gimple_comparison): Likewise.
10385 (verify_gimple_assign_binary): Likewise.
10386 (verify_gimple_assign_ternary): Likewise.
10387 (verify_gimple_assign_single): Likewise.
10388 * tree-pretty-print.c (dump_generic_node): Likewise.
10389 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10390 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
10391 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
10392 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
10393 (vect_shift_permute_load_chain): Likewise.
10394 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
10395 (expand_vector_condition, optimize_vector_constructor): Likewise.
10396 (lower_vec_perm, get_compute_type): Likewise.
10397 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10398 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
10399 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
10400 (vect_recog_mask_conversion_pattern): Likewise.
10401 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
10402 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
10403 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10404 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
10405 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
10406 (vectorizable_shift, vectorizable_operation, vectorizable_store)
10407 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
10408 (supportable_widening_operation): Likewise.
10409 (supportable_narrowing_operation): Likewise.
10410 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
10412 * varasm.c (output_constant): Likewise.
10414 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10415 Alan Hayward <alan.hayward@arm.com>
10416 David Sherwood <david.sherwood@arm.com>
10418 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
10419 so that both the length == 3 and length != 3 cases set up their
10420 own permute vectors. Add comments explaining why we know the
10421 number of elements is constant.
10422 (vect_permute_load_chain): Likewise.
10424 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10425 Alan Hayward <alan.hayward@arm.com>
10426 David Sherwood <david.sherwood@arm.com>
10428 * machmode.h (mode_nunits): Change from unsigned char to
10430 (ONLY_FIXED_SIZE_MODES): New macro.
10431 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
10432 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
10433 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
10435 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
10436 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
10437 or if measurement_type is not polynomial.
10438 * genmodes.c (ZERO_COEFFS): New macro.
10439 (emit_mode_nunits_inline): Make mode_nunits_inline return a
10441 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
10442 Use ZERO_COEFFS when emitting initializers.
10443 * data-streamer.h (bp_pack_poly_value): New function.
10444 (bp_unpack_poly_value): Likewise.
10445 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10446 for GET_MODE_NUNITS.
10447 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10448 for GET_MODE_NUNITS.
10449 * tree.c (make_vector_type): Remove temporary shim and make
10450 the real function take the number of units as a poly_uint64
10451 rather than an int.
10452 (build_vector_type_for_mode): Handle polynomial nunits.
10453 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
10454 * emit-rtl.c (const_vec_series_p_1): Likewise.
10455 (gen_rtx_CONST_VECTOR): Likewise.
10456 * fold-const.c (test_vec_duplicate_folding): Likewise.
10457 * genrecog.c (validate_pattern): Likewise.
10458 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
10459 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10460 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
10461 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
10462 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
10463 * rtlanal.c (subreg_get_info): Likewise.
10464 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10465 (vect_grouped_load_supported): Likewise.
10466 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
10467 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
10468 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
10469 (simplify_const_unary_operation, simplify_binary_operation_1)
10470 (simplify_const_binary_operation, simplify_ternary_operation)
10471 (test_vector_ops_duplicate, test_vector_ops): Likewise.
10472 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
10473 instead of CONST_VECTOR_NUNITS.
10474 * varasm.c (output_constant_pool_2): Likewise.
10475 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
10476 explicit-encoded elements in the XVEC for variable-length vectors.
10478 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10480 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
10482 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10483 Alan Hayward <alan.hayward@arm.com>
10484 David Sherwood <david.sherwood@arm.com>
10486 * coretypes.h (fixed_size_mode): Declare.
10487 (fixed_size_mode_pod): New typedef.
10488 * builtins.h (target_builtins::x_apply_args_mode)
10489 (target_builtins::x_apply_result_mode): Change type to
10490 fixed_size_mode_pod.
10491 * builtins.c (apply_args_size, apply_result_size, result_vector)
10492 (expand_builtin_apply_args_1, expand_builtin_apply)
10493 (expand_builtin_return): Update accordingly.
10495 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10497 * cse.c (hash_rtx_cb): Hash only the encoded elements.
10498 * cselib.c (cselib_hash_rtx): Likewise.
10499 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
10500 CONST_VECTOR encoding.
10502 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10503 Jeff Law <law@redhat.com>
10506 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
10507 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
10508 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
10509 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
10512 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
10513 explicitly probe *sp in a noreturn function if there were any callee
10514 register saves or frame pointer is needed.
10516 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10519 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
10520 BLKmode for ternary, binary or unary expressions.
10523 * var-tracking.c (delete_vta_debug_insn): New inline function.
10524 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
10525 insns from get_insns () to NULL instead of each bb separately.
10526 Use delete_vta_debug_insn. No longer static.
10527 (vt_debug_insns_local, variable_tracking_main_1): Adjust
10528 delete_vta_debug_insns callers.
10529 * rtl.h (delete_vta_debug_insns): Declare.
10530 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
10531 instead of variable_tracking_main.
10533 2018-01-03 Martin Sebor <msebor@redhat.com>
10535 PR tree-optimization/83603
10536 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
10537 arguments past the endof the argument list in functions declared
10538 without a prototype.
10539 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
10540 Avoid checking when arguments are null.
10542 2018-01-03 Martin Sebor <msebor@redhat.com>
10545 * doc/extend.texi (attribute const): Fix a typo.
10546 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
10547 issuing -Wsuggest-attribute for void functions.
10549 2018-01-03 Martin Sebor <msebor@redhat.com>
10551 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
10552 offset_int::from instead of wide_int::to_shwi.
10553 (maybe_diag_overlap): Remove assertion.
10554 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
10555 * gimple-ssa-sprintf.c (format_directive): Same.
10556 (parse_directive): Same.
10557 (sprintf_dom_walker::compute_format_length): Same.
10558 (try_substitute_return_value): Same.
10560 2018-01-03 Jeff Law <law@redhat.com>
10562 PR middle-end/83654
10563 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
10564 non-constant residual for zero at runtime and avoid probing in
10565 that case. Reorganize code for trailing problem to mirror handling
10568 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10570 PR tree-optimization/83501
10571 * tree-ssa-strlen.c (get_string_cst): New.
10572 (handle_char_store): Call get_string_cst.
10574 2018-01-03 Martin Liska <mliska@suse.cz>
10576 PR tree-optimization/83593
10577 * tree-ssa-strlen.c: Include tree-cfg.h.
10578 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
10579 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
10580 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
10582 (strlen_dom_walker::before_dom_children): Call
10583 gimple_purge_dead_eh_edges. Dump tranformation with details
10585 (strlen_dom_walker::before_dom_children): Update call by adding
10586 new argument cleanup_eh.
10587 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
10589 2018-01-03 Martin Liska <mliska@suse.cz>
10592 * cif-code.def (VARIADIC_THUNK): New enum value.
10593 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
10596 2018-01-03 Jan Beulich <jbeulich@suse.com>
10598 * sse.md (mov<mode>_internal): Tighten condition for when to use
10599 vmovdqu<ssescalarsize> for TI and OI modes.
10601 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10603 Update copyright years.
10605 2018-01-03 Martin Liska <mliska@suse.cz>
10608 * ipa-visibility.c (function_and_variable_visibility): Skip
10609 functions with noipa attribure.
10611 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10613 * gcc.c (process_command): Update copyright notice dates.
10614 * gcov-dump.c (print_version): Ditto.
10615 * gcov.c (print_version): Ditto.
10616 * gcov-tool.c (print_version): Ditto.
10617 * gengtype.c (create_file): Ditto.
10618 * doc/cpp.texi: Bump @copying's copyright year.
10619 * doc/cppinternals.texi: Ditto.
10620 * doc/gcc.texi: Ditto.
10621 * doc/gccint.texi: Ditto.
10622 * doc/gcov.texi: Ditto.
10623 * doc/install.texi: Ditto.
10624 * doc/invoke.texi: Ditto.
10626 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10628 * vector-builder.h (vector_builder::m_full_nelts): Change from
10629 unsigned int to poly_uint64.
10630 (vector_builder::full_nelts): Update prototype accordingly.
10631 (vector_builder::new_vector): Likewise.
10632 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
10633 (vector_builder::operator ==): Likewise.
10634 (vector_builder::finalize): Likewise.
10635 * int-vector-builder.h (int_vector_builder::int_vector_builder):
10636 Take the number of elements as a poly_uint64 rather than an
10638 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
10639 from unsigned int to poly_uint64.
10640 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
10641 (vec_perm_indices::new_vector): Likewise.
10642 (vec_perm_indices::length): Likewise.
10643 (vec_perm_indices::nelts_per_input): Likewise.
10644 (vec_perm_indices::input_nelts): Likewise.
10645 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
10646 number of elements per input as a poly_uint64 rather than an
10647 unsigned int. Use the original encoding for variable-length
10648 vectors, rather than clamping each individual element.
10649 For the second and subsequent elements in each pattern,
10650 clamp the step and base before clamping their sum.
10651 (vec_perm_indices::series_p): Handle polynomial element counts.
10652 (vec_perm_indices::all_in_range_p): Likewise.
10653 (vec_perm_indices_to_tree): Likewise.
10654 (vec_perm_indices_to_rtx): Likewise.
10655 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
10656 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
10657 (tree_vector_builder::new_binary_operation): Handle polynomial
10658 element counts. Return false if we need to know the number
10659 of elements at compile time.
10660 * fold-const.c (fold_vec_perm): Punt if the number of elements
10661 isn't known at compile time.
10663 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10665 * vec-perm-indices.h (vec_perm_builder): Change element type
10666 from HOST_WIDE_INT to poly_int64.
10667 (vec_perm_indices::element_type): Update accordingly.
10668 (vec_perm_indices::clamp): Handle polynomial element_types.
10669 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10670 (vec_perm_indices::all_in_range_p): Likewise.
10671 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
10673 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
10674 polynomial vec_perm_indices element types.
10675 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
10676 * fold-const.c (fold_vec_perm): Likewise.
10677 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
10678 * tree-vect-generic.c (lower_vec_perm): Likewise.
10679 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10680 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
10681 element type to HOST_WIDE_INT.
10683 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10684 Alan Hayward <alan.hayward@arm.com>
10685 David Sherwood <david.sherwood@arm.com>
10687 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
10688 rather than an int. Use plus_constant.
10689 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
10690 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
10692 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10693 Alan Hayward <alan.hayward@arm.com>
10694 David Sherwood <david.sherwood@arm.com>
10696 * calls.c (emit_call_1, expand_call): Change struct_value_size from
10697 a HOST_WIDE_INT to a poly_int64.
10699 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10700 Alan Hayward <alan.hayward@arm.com>
10701 David Sherwood <david.sherwood@arm.com>
10703 * calls.c (load_register_parameters): Cope with polynomial
10704 mode sizes. Require a constant size for BLKmode parameters
10705 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
10706 forces a parameter to be padded at the lsb end in order to
10707 fill a complete number of words, require the parameter size
10708 to be ordered wrt UNITS_PER_WORD.
10710 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10711 Alan Hayward <alan.hayward@arm.com>
10712 David Sherwood <david.sherwood@arm.com>
10714 * reload1.c (spill_stack_slot_width): Change element type
10715 from unsigned int to poly_uint64_pod.
10716 (alter_reg): Treat mode sizes as polynomial.
10718 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10719 Alan Hayward <alan.hayward@arm.com>
10720 David Sherwood <david.sherwood@arm.com>
10722 * reload.c (complex_word_subreg_p): New function.
10723 (reload_inner_reg_of_subreg, push_reload): Use it.
10725 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10726 Alan Hayward <alan.hayward@arm.com>
10727 David Sherwood <david.sherwood@arm.com>
10729 * lra-constraints.c (process_alt_operands): Reject matched
10730 operands whose sizes aren't ordered.
10731 (match_reload): Refer to this check here.
10733 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10734 Alan Hayward <alan.hayward@arm.com>
10735 David Sherwood <david.sherwood@arm.com>
10737 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
10738 that the mode size is in the set {1, 2, 4, 8, 16}.
10740 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10741 Alan Hayward <alan.hayward@arm.com>
10742 David Sherwood <david.sherwood@arm.com>
10744 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
10745 Use plus_constant instead of gen_rtx_PLUS.
10747 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10748 Alan Hayward <alan.hayward@arm.com>
10749 David Sherwood <david.sherwood@arm.com>
10751 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
10752 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
10753 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
10754 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
10755 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
10756 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
10757 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
10758 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
10759 * config/i386/i386.c (ix86_push_rounding): ...this new function.
10760 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
10762 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
10763 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
10764 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
10765 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
10766 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
10767 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
10768 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
10769 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
10770 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
10771 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
10773 * expr.c (emit_move_resolve_push): Treat the input and result
10774 of PUSH_ROUNDING as a poly_int64.
10775 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
10776 (emit_push_insn): Likewise.
10777 * lra-eliminations.c (mark_not_eliminable): Likewise.
10778 * recog.c (push_operand): Likewise.
10779 * reload1.c (elimination_effects): Likewise.
10780 * rtlanal.c (nonzero_bits1): Likewise.
10781 * calls.c (store_one_arg): Likewise. Require the padding to be
10782 known at compile time.
10784 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10785 Alan Hayward <alan.hayward@arm.com>
10786 David Sherwood <david.sherwood@arm.com>
10788 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
10789 Use plus_constant instead of gen_rtx_PLUS.
10791 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10792 Alan Hayward <alan.hayward@arm.com>
10793 David Sherwood <david.sherwood@arm.com>
10795 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
10796 rather than an int.
10798 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10799 Alan Hayward <alan.hayward@arm.com>
10800 David Sherwood <david.sherwood@arm.com>
10802 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
10803 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
10804 via stack temporaries. Treat the mode size as polynomial too.
10806 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10807 Alan Hayward <alan.hayward@arm.com>
10808 David Sherwood <david.sherwood@arm.com>
10810 * expr.c (expand_expr_real_2): When handling conversions involving
10811 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
10812 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
10813 as a poly_uint64 too.
10815 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10816 Alan Hayward <alan.hayward@arm.com>
10817 David Sherwood <david.sherwood@arm.com>
10819 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
10821 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10822 Alan Hayward <alan.hayward@arm.com>
10823 David Sherwood <david.sherwood@arm.com>
10825 * combine.c (can_change_dest_mode): Handle polynomial
10826 REGMODE_NATURAL_SIZE.
10827 * expmed.c (store_bit_field_1): Likewise.
10828 * expr.c (store_constructor): Likewise.
10829 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
10830 and polynomial REGMODE_NATURAL_SIZE.
10831 (gen_lowpart_common): Likewise.
10832 * reginfo.c (record_subregs_of_mode): Likewise.
10833 * rtlanal.c (read_modify_subreg_p): Likewise.
10835 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10836 Alan Hayward <alan.hayward@arm.com>
10837 David Sherwood <david.sherwood@arm.com>
10839 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
10840 numbers of elements.
10842 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10843 Alan Hayward <alan.hayward@arm.com>
10844 David Sherwood <david.sherwood@arm.com>
10846 * match.pd: Cope with polynomial numbers of vector elements.
10848 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10849 Alan Hayward <alan.hayward@arm.com>
10850 David Sherwood <david.sherwood@arm.com>
10852 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
10853 in a POINTER_PLUS_EXPR.
10855 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10856 Alan Hayward <alan.hayward@arm.com>
10857 David Sherwood <david.sherwood@arm.com>
10859 * omp-simd-clone.c (simd_clone_subparts): New function.
10860 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
10861 (ipa_simd_modify_function_body): Likewise.
10863 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10864 Alan Hayward <alan.hayward@arm.com>
10865 David Sherwood <david.sherwood@arm.com>
10867 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
10868 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
10869 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
10870 (expand_vector_condition, vector_element): Likewise.
10871 (subparts_gt): New function.
10872 (get_compute_type): Use subparts_gt.
10873 (count_type_subparts): Delete.
10874 (expand_vector_operations_1): Use subparts_gt instead of
10875 count_type_subparts.
10877 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10878 Alan Hayward <alan.hayward@arm.com>
10879 David Sherwood <david.sherwood@arm.com>
10881 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
10882 (vect_compile_time_alias): ...this new function. Do the calculation
10883 on poly_ints rather than trees.
10884 (vect_prune_runtime_alias_test_list): Update call accordingly.
10886 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10887 Alan Hayward <alan.hayward@arm.com>
10888 David Sherwood <david.sherwood@arm.com>
10890 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
10892 (vect_schedule_slp_instance): Likewise.
10894 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10895 Alan Hayward <alan.hayward@arm.com>
10896 David Sherwood <david.sherwood@arm.com>
10898 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
10899 constant and extern definitions for variable-length vectors.
10900 (vect_get_constant_vectors): Note that the number of units
10901 is known to be constant.
10903 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10904 Alan Hayward <alan.hayward@arm.com>
10905 David Sherwood <david.sherwood@arm.com>
10907 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
10908 of units as polynomial. Choose between WIDE and NARROW based
10911 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10912 Alan Hayward <alan.hayward@arm.com>
10913 David Sherwood <david.sherwood@arm.com>
10915 * tree-vect-stmts.c (simd_clone_subparts): New function.
10916 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
10918 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10919 Alan Hayward <alan.hayward@arm.com>
10920 David Sherwood <david.sherwood@arm.com>
10922 * tree-vect-stmts.c (vectorizable_call): Treat the number of
10923 vectors as polynomial. Use build_index_vector for
10924 IFN_GOMP_SIMD_LANE.
10926 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10927 Alan Hayward <alan.hayward@arm.com>
10928 David Sherwood <david.sherwood@arm.com>
10930 * tree-vect-stmts.c (get_load_store_type): Treat the number of
10931 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
10932 for variable-length vectors.
10933 (vectorizable_mask_load_store): Treat the number of units as
10934 polynomial, asserting that it is constant if the condition has
10935 already been enforced.
10936 (vectorizable_store, vectorizable_load): Likewise.
10938 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10939 Alan Hayward <alan.hayward@arm.com>
10940 David Sherwood <david.sherwood@arm.com>
10942 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
10943 of units as polynomial. Punt if we can't tell at compile time
10944 which vector contains the final result.
10946 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10947 Alan Hayward <alan.hayward@arm.com>
10948 David Sherwood <david.sherwood@arm.com>
10950 * tree-vect-loop.c (vectorizable_induction): Treat the number
10951 of units as polynomial. Punt on SLP inductions. Use an integer
10952 VEC_SERIES_EXPR for variable-length integer reductions. Use a
10953 cast of such a series for variable-length floating-point
10956 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10957 Alan Hayward <alan.hayward@arm.com>
10958 David Sherwood <david.sherwood@arm.com>
10960 * tree.h (build_index_vector): Declare.
10961 * tree.c (build_index_vector): New function.
10962 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
10963 of units as polynomial, forcibly converting it to a constant if
10964 vectorizable_reduction has already enforced the condition.
10965 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
10966 to create a {1,2,3,...} vector.
10967 (vectorizable_reduction): Treat the number of units as polynomial.
10968 Choose vectype_in based on the largest scalar element size rather
10969 than the smallest number of units. Enforce the restrictions
10972 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10973 Alan Hayward <alan.hayward@arm.com>
10974 David Sherwood <david.sherwood@arm.com>
10976 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
10977 number of units as polynomial.
10979 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10980 Alan Hayward <alan.hayward@arm.com>
10981 David Sherwood <david.sherwood@arm.com>
10983 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
10984 * target.def (autovectorize_vector_sizes): Return the vector sizes
10985 by pointer, using vector_sizes rather than a bitmask.
10986 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
10987 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
10988 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
10990 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
10991 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
10992 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
10993 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
10994 * omp-general.c (omp_max_vf): Likewise.
10995 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10996 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
10997 * tree-vect-loop.c (vect_analyze_loop): Likewise.
10998 * tree-vect-slp.c (vect_slp_bb): Likewise.
10999 * doc/tm.texi: Regenerate.
11000 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
11002 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
11003 the vector size as a poly_uint64 rather than an unsigned int.
11004 (current_vector_size): Change from an unsigned int to a poly_uint64.
11005 (get_vectype_for_scalar_type): Update accordingly.
11006 * tree.h (build_truth_vector_type): Take the size and number of
11007 units as a poly_uint64 rather than an unsigned int.
11008 (build_vector_type): Add a temporary overload that takes
11009 the number of units as a poly_uint64 rather than an unsigned int.
11010 * tree.c (make_vector_type): Likewise.
11011 (build_truth_vector_type): Take the number of units as a poly_uint64
11012 rather than an unsigned int.
11014 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11015 Alan Hayward <alan.hayward@arm.com>
11016 David Sherwood <david.sherwood@arm.com>
11018 * target.def (get_mask_mode): Take the number of units and length
11019 as poly_uint64s rather than unsigned ints.
11020 * targhooks.h (default_get_mask_mode): Update accordingly.
11021 * targhooks.c (default_get_mask_mode): Likewise.
11022 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
11023 * doc/tm.texi: Regenerate.
11025 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11026 Alan Hayward <alan.hayward@arm.com>
11027 David Sherwood <david.sherwood@arm.com>
11029 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
11030 * omp-general.c (omp_max_vf): Likewise.
11031 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
11032 (expand_omp_simd): Handle polynomial safelen.
11033 * omp-low.c (omplow_simd_context): Add a default constructor.
11034 (omplow_simd_context::max_vf): Change from int to poly_uint64.
11035 (lower_rec_simd_input_clauses): Update accordingly.
11036 (lower_rec_input_clauses): Likewise.
11038 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11039 Alan Hayward <alan.hayward@arm.com>
11040 David Sherwood <david.sherwood@arm.com>
11042 * tree-vectorizer.h (vect_nunits_for_cost): New function.
11043 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
11044 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
11045 (vect_analyze_slp_cost): Likewise.
11046 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
11047 (vect_model_load_cost): Likewise.
11049 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11050 Alan Hayward <alan.hayward@arm.com>
11051 David Sherwood <david.sherwood@arm.com>
11053 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
11054 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
11055 from an unsigned int * to a poly_uint64_pod *.
11056 (calculate_unrolling_factor): New function.
11057 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
11059 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11060 Alan Hayward <alan.hayward@arm.com>
11061 David Sherwood <david.sherwood@arm.com>
11063 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
11064 from an unsigned int to a poly_uint64.
11065 (_loop_vec_info::slp_unrolling_factor): Likewise.
11066 (_loop_vec_info::vectorization_factor): Change from an int
11068 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
11069 (vect_get_num_vectors): New function.
11070 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
11071 (vect_get_num_copies): Use vect_get_num_vectors.
11072 (vect_analyze_data_ref_dependences): Change max_vf from an int *
11073 to an unsigned int *.
11074 (vect_analyze_data_refs): Change min_vf from an int * to a
11076 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11077 than an unsigned HOST_WIDE_INT.
11078 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
11079 (vect_analyze_data_ref_dependence): Change max_vf from an int *
11080 to an unsigned int *.
11081 (vect_analyze_data_ref_dependences): Likewise.
11082 (vect_compute_data_ref_alignment): Handle polynomial vf.
11083 (vect_enhance_data_refs_alignment): Likewise.
11084 (vect_prune_runtime_alias_test_list): Likewise.
11085 (vect_shift_permute_load_chain): Likewise.
11086 (vect_supportable_dr_alignment): Likewise.
11087 (dependence_distance_ge_vf): Take the vectorization factor as a
11088 poly_uint64 rather than an unsigned HOST_WIDE_INT.
11089 (vect_analyze_data_refs): Change min_vf from an int * to a
11091 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
11092 vfm1 as a poly_uint64 rather than an int. Make the same change
11093 for the returned bound_scalar.
11094 (vect_gen_vector_loop_niters): Handle polynomial vf.
11095 (vect_do_peeling): Likewise. Update call to
11096 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
11097 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
11099 * tree-vect-loop.c (vect_determine_vectorization_factor)
11100 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
11101 (vect_get_known_peeling_cost): Likewise.
11102 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
11103 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
11104 (vect_transform_loop): Likewise. Use the lowest possible VF when
11105 updating the upper bounds of the loop.
11106 (vect_min_worthwhile_factor): Make static. Return an unsigned int
11107 rather than an int.
11108 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
11109 polynomial unroll factors.
11110 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
11111 (vect_make_slp_decision): Likewise.
11112 (vect_supported_load_permutation_p): Likewise, and polynomial
11114 (vect_analyze_slp_cost): Handle polynomial vf.
11115 (vect_slp_analyze_node_operations): Likewise.
11116 (vect_slp_analyze_bb_1): Likewise.
11117 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11118 than an unsigned HOST_WIDE_INT.
11119 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
11120 (vectorizable_load): Handle polynomial vf.
11121 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
11123 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
11125 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11126 Alan Hayward <alan.hayward@arm.com>
11127 David Sherwood <david.sherwood@arm.com>
11129 * match.pd: Handle bit operations involving three constants
11130 and try to fold one pair.
11132 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11134 * tree-vect-loop-manip.c: Include gimple-fold.h.
11135 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
11136 niters_maybe_zero parameters. Handle other cases besides a step of 1.
11137 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
11138 Add a path that uses a step of VF instead of 1, but disable it
11140 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
11141 and niters_no_overflow parameters. Update calls to
11142 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
11143 Create a new SSA name if the latter choses to use a ste other
11144 than zero, and return it via niters_vector_mult_vf_var.
11145 * tree-vect-loop.c (vect_transform_loop): Update calls to
11146 vect_do_peeling, vect_gen_vector_loop_niters and
11147 slpeel_make_loop_iterate_ntimes.
11148 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
11149 (vect_gen_vector_loop_niters): Update declarations after above changes.
11151 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
11153 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
11154 128-bit round to integer instructions.
11155 (ceil<mode>2): Likewise.
11156 (btrunc<mode>2): Likewise.
11157 (round<mode>2): Likewise.
11159 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11161 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
11162 unaligned VSX load/store on P8/P9.
11163 (expand_block_clear): Allow the use of unaligned VSX
11164 load/store on P8/P9.
11166 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11168 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
11170 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
11171 swap associated with both a load and a store.
11173 2018-01-02 Andrew Waterman <andrew@sifive.com>
11175 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
11176 * config/riscv/riscv.md (clear_cache): Use it.
11178 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
11180 * web.c: Remove out-of-date comment.
11182 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11184 * expr.c (fixup_args_size_notes): Check that any existing
11185 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
11186 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
11187 (emit_single_push_insn): ...here.
11189 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11191 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
11192 (const_vector_encoded_nelts): New function.
11193 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
11194 (const_vector_int_elt, const_vector_elt): Declare.
11195 * emit-rtl.c (const_vector_int_elt_1): New function.
11196 (const_vector_elt): Likewise.
11197 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
11198 of CONST_VECTOR_ELT.
11200 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11202 * expr.c: Include rtx-vector-builder.h.
11203 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
11204 directly on the tree encoding.
11205 (const_vector_from_tree): Likewise.
11206 * optabs.c: Include rtx-vector-builder.h.
11207 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
11208 sequence of "u" values.
11209 * vec-perm-indices.c: Include rtx-vector-builder.h.
11210 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
11211 directly on the vec_perm_indices encoding.
11213 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11215 * doc/rtl.texi (const_vector): Describe new encoding scheme.
11216 * Makefile.in (OBJS): Add rtx-vector-builder.o.
11217 * rtx-vector-builder.h: New file.
11218 * rtx-vector-builder.c: Likewise.
11219 * rtl.h (rtx_def::u2): Add a const_vector field.
11220 (CONST_VECTOR_NPATTERNS): New macro.
11221 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
11222 (CONST_VECTOR_DUPLICATE_P): Likewise.
11223 (CONST_VECTOR_STEPPED_P): Likewise.
11224 (CONST_VECTOR_ENCODED_ELT): Likewise.
11225 (const_vec_duplicate_p): Check for a duplicated vector encoding.
11226 (unwrap_const_vec_duplicate): Likewise.
11227 (const_vec_series_p): Check for a non-duplicated vector encoding.
11228 Say that the function only returns true for integer vectors.
11229 * emit-rtl.c: Include rtx-vector-builder.h.
11230 (gen_const_vec_duplicate_1): Delete.
11231 (gen_const_vector): Call gen_const_vec_duplicate instead of
11232 gen_const_vec_duplicate_1.
11233 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
11234 (gen_const_vec_duplicate): Use rtx_vector_builder.
11235 (gen_const_vec_series): Likewise.
11236 (gen_rtx_CONST_VECTOR): Likewise.
11237 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
11238 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11239 Build a new vector rather than modifying a CONST_VECTOR in-place.
11240 (handle_special_swappables): Update call accordingly.
11241 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
11242 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11243 Build a new vector rather than modifying a CONST_VECTOR in-place.
11244 (handle_special_swappables): Update call accordingly.
11246 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11248 * simplify-rtx.c (simplify_const_binary_operation): Use
11249 CONST_VECTOR_ELT instead of XVECEXP.
11251 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11253 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
11254 the selector elements to be different from the data elements
11255 if the selector is a VECTOR_CST.
11256 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
11257 ssizetype for the selector.
11259 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11261 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
11262 before testing each element individually.
11263 * tree-vect-generic.c (lower_vec_perm): Likewise.
11265 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11267 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
11268 * selftest-run-tests.c (selftest::run_tests): Call it.
11269 * vector-builder.h (vector_builder::operator ==): New function.
11270 (vector_builder::operator !=): Likewise.
11271 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
11272 (vec_perm_indices::all_from_input_p): New function.
11273 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11274 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
11275 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
11276 instead of reading the VECTOR_CST directly. Detect whether both
11277 vector inputs are the same before constructing the vec_perm_indices,
11278 and update the number of inputs argument accordingly. Use the
11279 utility functions added above. Only construct sel2 if we need to.
11281 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11283 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
11284 the broadcast of the low byte.
11285 (expand_mult_highpart): Use an explicit encoding for the permutes.
11286 * optabs-query.c (can_mult_highpart_p): Likewise.
11287 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
11288 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11289 (vectorizable_bswap): Likewise.
11290 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
11291 explicit encoding for the power-of-2 permutes.
11292 (vect_permute_store_chain): Likewise.
11293 (vect_grouped_load_supported): Likewise.
11294 (vect_permute_load_chain): Likewise.
11296 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11298 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
11299 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
11300 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
11301 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11302 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
11303 (vect_gen_perm_mask_any): Likewise.
11305 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11307 * int-vector-builder.h: New file.
11308 * vec-perm-indices.h: Include int-vector-builder.h.
11309 (vec_perm_indices): Redefine as an int_vector_builder.
11310 (auto_vec_perm_indices): Delete.
11311 (vec_perm_builder): Redefine as a stand-alone class.
11312 (vec_perm_indices::vec_perm_indices): New function.
11313 (vec_perm_indices::clamp): Likewise.
11314 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
11315 (vec_perm_indices::new_vector): New function.
11316 (vec_perm_indices::new_expanded_vector): Update for new
11317 vec_perm_indices class.
11318 (vec_perm_indices::rotate_inputs): New function.
11319 (vec_perm_indices::all_in_range_p): Operate directly on the
11320 encoded form, without computing elided elements.
11321 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
11322 encoding. Update for new vec_perm_indices class.
11323 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
11324 the given vec_perm_builder.
11325 (expand_vec_perm_var): Update vec_perm_builder constructor.
11326 (expand_mult_highpart): Use vec_perm_builder instead of
11327 auto_vec_perm_indices.
11328 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
11329 vec_perm_indices instead of auto_vec_perm_indices. Use a single
11330 or double series encoding as appropriate.
11331 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
11332 vec_perm_indices instead of auto_vec_perm_indices.
11333 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11334 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11335 (vect_permute_store_chain): Likewise.
11336 (vect_grouped_load_supported): Likewise.
11337 (vect_permute_load_chain): Likewise.
11338 (vect_shift_permute_load_chain): Likewise.
11339 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11340 (vect_transform_slp_perm_load): Likewise.
11341 (vect_schedule_slp_instance): Likewise.
11342 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11343 (vectorizable_mask_load_store): Likewise.
11344 (vectorizable_bswap): Likewise.
11345 (vectorizable_store): Likewise.
11346 (vectorizable_load): Likewise.
11347 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
11348 vec_perm_indices instead of auto_vec_perm_indices. Use
11349 tree_to_vec_perm_builder to read the vector from a tree.
11350 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
11351 vec_perm_builder instead of a vec_perm_indices.
11352 (have_whole_vector_shift): Use vec_perm_builder and
11353 vec_perm_indices instead of auto_vec_perm_indices. Leave the
11354 truncation to calc_vec_perm_mask_for_shift.
11355 (vect_create_epilog_for_reduction): Likewise.
11356 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
11357 from auto_vec_perm_indices to vec_perm_indices.
11358 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11359 instead of changing individual elements.
11360 (aarch64_vectorize_vec_perm_const): Use new_vector to install
11361 the vector in d.perm.
11362 * config/arm/arm.c (expand_vec_perm_d::perm): Change
11363 from auto_vec_perm_indices to vec_perm_indices.
11364 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11365 instead of changing individual elements.
11366 (arm_vectorize_vec_perm_const): Use new_vector to install
11367 the vector in d.perm.
11368 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
11369 Update vec_perm_builder constructor.
11370 (rs6000_expand_interleave): Likewise.
11371 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
11372 (rs6000_expand_interleave): Likewise.
11374 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11376 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
11377 to qimode could truncate the indices.
11378 * optabs.c (expand_vec_perm_var): Likewise.
11380 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11382 * Makefile.in (OBJS): Add vec-perm-indices.o.
11383 * vec-perm-indices.h: New file.
11384 * vec-perm-indices.c: Likewise.
11385 * target.h (vec_perm_indices): Replace with a forward class
11387 (auto_vec_perm_indices): Move to vec-perm-indices.h.
11388 * optabs.h: Include vec-perm-indices.h.
11389 (expand_vec_perm): Delete.
11390 (selector_fits_mode_p, expand_vec_perm_var): Declare.
11391 (expand_vec_perm_const): Declare.
11392 * target.def (vec_perm_const_ok): Replace with...
11393 (vec_perm_const): ...this new hook.
11394 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
11395 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
11396 * doc/tm.texi: Regenerate.
11397 * optabs.def (vec_perm_const): Delete.
11398 * doc/md.texi (vec_perm_const): Likewise.
11399 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
11400 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
11401 expand_vec_perm for constant permutation vectors. Assert that
11402 the mode of variable permutation vectors is the integer equivalent
11403 of the mode that is being permuted.
11404 * optabs-query.h (selector_fits_mode_p): Declare.
11405 * optabs-query.c: Include vec-perm-indices.h.
11406 (selector_fits_mode_p): New function.
11407 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
11408 is defined, instead of checking whether the vec_perm_const_optab
11409 exists. Use targetm.vectorize.vec_perm_const instead of
11410 targetm.vectorize.vec_perm_const_ok. Check whether the indices
11411 fit in the vector mode before using a variable permute.
11412 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
11413 vec_perm_indices instead of an rtx.
11414 (expand_vec_perm): Replace with...
11415 (expand_vec_perm_const): ...this new function. Take the selector
11416 as a vec_perm_indices rather than an rtx. Also take the mode of
11417 the selector. Update call to shift_amt_for_vec_perm_mask.
11418 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
11419 Use vec_perm_indices::new_expanded_vector to expand the original
11420 selector into bytes. Check whether the indices fit in the vector
11421 mode before using a variable permute.
11422 (expand_vec_perm_var): Make global.
11423 (expand_mult_highpart): Use expand_vec_perm_const.
11424 * fold-const.c: Includes vec-perm-indices.h.
11425 * tree-ssa-forwprop.c: Likewise.
11426 * tree-vect-data-refs.c: Likewise.
11427 * tree-vect-generic.c: Likewise.
11428 * tree-vect-loop.c: Likewise.
11429 * tree-vect-slp.c: Likewise.
11430 * tree-vect-stmts.c: Likewise.
11431 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
11433 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
11434 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
11435 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
11436 (aarch64_vectorize_vec_perm_const): ...this new function.
11437 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11438 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11439 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
11440 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
11441 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11442 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11443 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
11445 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
11446 check for NEON modes.
11447 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
11448 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
11449 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
11450 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
11452 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
11453 the old VEC_PERM_CONST conditions.
11454 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
11455 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
11456 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
11457 (ia64_vectorize_vec_perm_const_ok): Merge into...
11458 (ia64_vectorize_vec_perm_const): ...this new function.
11459 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
11460 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
11461 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
11462 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
11463 * config/mips/mips.c (mips_expand_vec_perm_const)
11464 (mips_vectorize_vec_perm_const_ok): Merge into...
11465 (mips_vectorize_vec_perm_const): ...this new function.
11466 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
11467 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
11468 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
11469 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
11470 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
11471 (rs6000_expand_vec_perm_const): Delete.
11472 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
11474 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11475 (altivec_expand_vec_perm_const_le): Take each operand individually.
11476 Operate on constant selectors rather than rtxes.
11477 (altivec_expand_vec_perm_const): Likewise. Update call to
11478 altivec_expand_vec_perm_const_le.
11479 (rs6000_expand_vec_perm_const): Delete.
11480 (rs6000_vectorize_vec_perm_const_ok): Delete.
11481 (rs6000_vectorize_vec_perm_const): New function.
11482 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11483 an element count and rtx array.
11484 (rs6000_expand_extract_even): Update call accordingly.
11485 (rs6000_expand_interleave): Likewise.
11486 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
11487 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
11488 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
11489 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
11490 (rs6000_expand_vec_perm_const): Delete.
11491 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11492 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11493 (altivec_expand_vec_perm_const_le): Take each operand individually.
11494 Operate on constant selectors rather than rtxes.
11495 (altivec_expand_vec_perm_const): Likewise. Update call to
11496 altivec_expand_vec_perm_const_le.
11497 (rs6000_expand_vec_perm_const): Delete.
11498 (rs6000_vectorize_vec_perm_const_ok): Delete.
11499 (rs6000_vectorize_vec_perm_const): New function. Remove stray
11500 reference to the SPE evmerge intructions.
11501 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11502 an element count and rtx array.
11503 (rs6000_expand_extract_even): Update call accordingly.
11504 (rs6000_expand_interleave): Likewise.
11505 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
11506 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
11508 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11510 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11512 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
11513 vector mode and that that mode matches the mode of the data
11515 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
11516 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
11517 directly using expand_vec_perm_1 when forcing selectors into
11519 (expand_vec_perm_var): New function, split out from expand_vec_perm.
11521 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11523 * optabs-query.h (can_vec_perm_p): Delete.
11524 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
11525 * optabs-query.c (can_vec_perm_p): Split into...
11526 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
11527 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
11528 particular selector is valid.
11529 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11530 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11531 (vect_grouped_load_supported): Likewise.
11532 (vect_shift_permute_load_chain): Likewise.
11533 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11534 (vect_transform_slp_perm_load): Likewise.
11535 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11536 (vectorizable_bswap): Likewise.
11537 (vect_gen_perm_mask_checked): Likewise.
11538 * fold-const.c (fold_ternary_loc): Likewise. Don't take
11539 implementations of variable permutation vectors into account
11540 when deciding which selector to use.
11541 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
11542 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
11543 with a false third argument.
11544 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
11545 to test whether the constant selector is valid and can_vec_perm_var_p
11546 to test whether a variable selector is valid.
11548 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11550 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
11551 * optabs-query.c (can_vec_perm_p): Likewise.
11552 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
11553 instead of vec_perm_indices.
11554 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
11555 (vect_gen_perm_mask_checked): Likewise,
11556 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
11557 (vect_gen_perm_mask_checked): Likewise,
11559 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11561 * optabs-query.h (qimode_for_vec_perm): Declare.
11562 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
11563 (qimode_for_vec_perm): ...this new function.
11564 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
11566 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11568 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
11569 does not have a conditional at the top.
11571 2018-01-02 Richard Biener <rguenther@suse.de>
11573 * ipa-inline.c (big_speedup_p): Fix expression.
11575 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11578 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
11581 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11585 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
11586 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
11587 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
11588 cond_taken_branch_cost 3->4.
11590 2018-01-01 Jakub Jelinek <jakub@redhat.com>
11592 PR tree-optimization/83581
11593 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
11594 TODO_cleanup_cfg if any changes have been made.
11596 PR middle-end/83608
11597 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
11598 convert_modes if target mode has the right side, but different mode
11601 PR middle-end/83609
11602 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
11603 last argument when extracting from CONCAT. If either from_real or
11604 from_imag is NULL, use expansion through memory. If result is not
11605 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
11606 the parts directly to inner mode, if even that fails, use expansion
11609 PR middle-end/83623
11610 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
11611 check for bswap in mode rather than HImode and use that in expand_unop
11614 Copyright (C) 2018 Free Software Foundation, Inc.
11616 Copying and distribution of this file, with or without modification,
11617 are permitted in any medium without royalty provided the copyright
11618 notice and this notice are preserved.