1 2018-04-30 Richard Biener <rguenther@suse.de>
4 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
6 2018-04-30 Richard Biener <rguenther@suse.de>
8 PR tree-optimization/28364
9 PR tree-optimization/85275
10 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
11 copying first exit test.
13 2018-04-28 Mark Wielaard <mark@klomp.org>
15 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
17 (dwarf_AT): Handle DW_AT_addr_base.
18 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
20 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
23 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
24 (*ashl<dwi>3_doubleword_mask_1): Ditto.
25 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
26 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
28 2018-04-28 Richard Biener <rguenther@suse.de>
30 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
31 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
32 to reflect use. Only add interesting stmts.
34 2018-04-27 Martin Jambor <mjambor@suse.cz>
37 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
38 the jump function allows for passing through aggregate values.
40 2018-04-27 David Malcolm <dmalcolm@redhat.com>
42 * input.h (in_system_header_at): Convert from macro to inline
44 (from_macro_expansion_at): Likewise.
45 (from_macro_definition_at): Likewise.
47 2018-04-27 Jeff Law <law@redhat.com>
49 * config.gcc: Mark tile* targets as deprecated/obsolete.
51 2018-04-27 Richard Biener <rguenther@suse.de>
53 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
56 2018-04-27 Richard Biener <rguenther@suse.de>
58 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
60 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
62 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
63 with Yd constraint. Set "preferred_for_speed" attribute from
64 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
66 (*movdi_internal): Ditto.
67 (movti_interunit splitters): Remove
68 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
69 (movdi_interunit splitters): Ditto.
70 * config/i386/constraints.md (Ye): Remove.
71 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
73 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
76 * config/aarch64/constraints.md (Usg): Limit to 31.
79 2018-04-27 Jakub Jelinek <jakub@redhat.com>
81 PR tree-optimization/85529
82 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
83 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
84 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
85 zero extension or masking of the MSB bit.
86 (optimize_range_tests): Add FIRST_BB argument, pass it through
87 to optimize_range_tests_var_bound.
88 (maybe_optimize_range_tests, reassociate_bb): Adjust
89 optimize_range_tests callers.
91 2018-04-26 Richard Biener <rguenther@suse.de>
92 Jakub Jelinek <jakub@redhat.com>
94 * cgraph.h (symbol_table): Just declare debug method here.
95 * symtab.c (symbol_table::debug): Define.
97 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
99 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
101 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
103 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
104 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
105 (*movdi_internal): Substitute Yi and Yj constraint with x
106 and Ym and Yn constraint with y constraint. Update "isa"
107 attribute and set "preferred_for_speed" attribute from
108 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
109 (*movsi_internal): Ditto.
110 (*movdf_internal): Ditto.
111 (*movsf_internal): Ditto.
112 (*zero_extendsidi2): Ditto.
113 * config/i386/sse.md (vec_set<mode>_0): Ditto.
114 (sse2_loadld): Ditto.
115 (*vec_extract<ssevecmodelower>_0): Ditto.
116 (*vec_extractv4si_0_zext_sse4): Ditto.
117 (vec_concatv2di): Ditto.
118 (*vec_dup<mode>): Ditto.
119 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
120 * config/i386/constraints.md (Yi): Remove.
125 2018-04-26 Nathan Sidwell <nathan@acm.org>
127 * dumpfile.c (dump_open): New.
128 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
129 (dump_finish): Detect stdio/stderr by value not name.
131 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
133 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
135 2018-04-26 Tom de Vries <tom@codesourcery.com>
138 * config/nvptx/nvptx.c (verify_neutering_jumps)
139 (verify_neutering_labels): New function
140 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
142 2018-04-26 Tom de Vries <tom@codesourcery.com>
145 * config/nvptx/nvptx.c (needs_neutering_p): New function.
146 (nvptx_single): Use needs_neutering_p to skip over insns that do not
149 2018-04-26 Richard Biener <rguenther@suse.de>
150 Tom de Vries <tom@codesourcery.com>
153 * lto-streamer-out.c (output_function): Fixup loops if required to match
154 discovery done in the reader.
156 2018-04-26 Richard Biener <rguenther@suse.de>
158 PR tree-optimization/85116
159 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
160 have a loop exit from the single latch predecessor. Remove
161 case of header with just condition.
162 (ch_base::copy_headers): Exclude infinite loops from any
164 (pass_ch::execute): Record exits.
166 2018-04-26 Richard Biener <rguenther@suse.de>
168 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
169 prologue cost vector and pass it to vect_get_load_cost.
170 (vect_get_peeling_costs_all_drs): Likewise.
171 (vect_peeling_hash_get_lowest_cost): Likewise.
172 (vect_enhance_data_refs_alignment): Likewise.
174 2018-04-26 Richard Biener <rguenther@suse.de>
177 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
178 checking of integer<->pointer conversions.
179 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
180 sign-/zero-extending pointer types.
181 (expand_omp_for_static_chunk): Likewise.
183 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
184 Jean Lee <xiaoyur347@gmail.com>
186 * config/mips/mips.c (mips_asan_shadow_offset): New function.
187 (TARGET_ASAN_SHADOW_OFFSET): Define.
188 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
189 true for -fsanitize=address.
191 2018-04-25 Mark Wielaard <mark@klomp.org>
193 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
196 2018-04-25 Jakub Jelinek <jakub@redhat.com>
198 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
199 than "alu", remove explicit "memory" and "imm_disp" attributes.
200 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
203 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
204 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
207 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
210 * config/i386/i386.c (ix86_expand_builtin): Change memory
211 operand to XI, extend p0 to Pmode.
212 * config/i386/i386.md: Change unspec volatile and operand
213 1 mode to XI, change operand 0 mode to P.
215 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
217 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
218 GET_MODE_MASK before any checking.
219 (nds32_can_use_bset_p): Likewise.
220 (nds32_can_use_btgl_p): Likewise.
222 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
224 * config/nds32/nds32-doubleword.md: New define_split pattern for
225 illegal register number.
227 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
229 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
231 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
233 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
235 2018-04-25 Richard Biener <rguenther@suse.de>
237 * lto-streamer.h (LTO_major_version): Bump to 8.
239 2018-04-25 Jakub Jelinek <jakub@redhat.com>
241 * BASE-VER: Set to 9.0.0.
243 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
245 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
246 in __abskf2 and __powikf2.
248 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
251 * config/aarch64/constraints.md (Usg, Usj): New constraints.
252 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
253 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
254 Use the above on operand 2. Reindent.
255 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
257 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
260 * common/config/i386/i386-common.c (ix86_handle_option): Don't
262 * config/i386/i386.opt (mcet): Removed.
263 * doc/install.texi: Remove -mcet documentation.
264 * doc/invoke.texi: Likewise.
266 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
269 * doc/install.texi: Remove -mcet from bootstrap-cet.
271 2018-04-24 Jakub Jelinek <jakub@redhat.com>
274 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
275 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
279 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
280 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
281 containing a CONST_VECTOR.
283 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
285 * doc/install.texi: Update newlib dependency for nvptx.
287 2018-04-24 Jakub Jelinek <jakub@redhat.com>
290 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
291 instead of INTVAL when shifting x left.
293 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
295 PR tree-optimization/85478
296 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
297 vect_grouped_store_supported for single element vectors.
299 2018-04-24 Richard Biener <rguenther@suse.de>
302 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
303 load cost increase to the case of non-constant step.
305 2018-04-24 Jakub Jelinek <jakub@redhat.com>
308 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
309 destination if any_malformed_asm.
311 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
314 * expr.c (store_field): In the bitfield case, if the value comes from
315 a function call and is returned in registers by means of a PARALLEL,
316 do not change the mode of the temporary unless BLKmode and VOIDmode.
318 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
320 PR rtl-optimization/85423
321 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
322 dependencies to debug insns when the previous insn is non-debug.
324 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
326 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
327 enums into a single definition.
328 (fls): Fix predicates and printing.
331 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
333 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
334 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
335 and short u6 immediate.
336 (check_if_valid_sleep_operand): Remove.
337 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
339 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
341 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
342 flag_always_save_lp condition.
343 * config/nds32/nds32.opt (malways-save-lp): New option.
345 2018-04-22 Shiva Chen <shiva0217@gmail.com>
347 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
348 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
349 * config/nds32/nds32.h
350 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
351 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
353 2018-04-22 Shiva Chen <shiva0217@gmail.com>
355 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
356 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
358 2018-04-22 Shiva Chen <shiva0217@gmail.com>
359 Chung-Ju Wu <jasonwucj@gmail.com>
361 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
363 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
364 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
366 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
368 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
370 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
372 * config/nds32/nds32-protos.h (nds32_data_alignment,
373 nds32_local_alignment): Declare.
374 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
375 nds32_local_alignment): New functions.
376 (TARGET_CONSTANT_ALIGNMENT): Define.
377 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
379 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
381 * config/nds32/nds32.c
382 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
383 (TARGET_MODES_TIEABLE_P): Likewise.
385 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
387 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
390 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
391 Chung-Ju Wu <jasonwucj@gmail.com>
393 * config/nds32/constants.md (unspec_volatile_element): Add enum values
394 for unaligned access.
395 * config/nds32/nds32-intrinsic.c: Implementation of expanding
397 * config/nds32/nds32-intrinsic.md: Likewise.
398 * config/nds32/nds32_intrinsic.h: Likewise.
399 * config/nds32/nds32.h (nds32_builtins): Likewise.
400 * config/nds32/nds32.opt (munaligned-access): New option.
401 * config/nds32/nds32.c (nds32_asm_file_start): Display
402 flag_unaligned_access status.
404 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
406 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
407 -mno-relax is present.
408 * config/riscv/linux.h (LINK_SPEC): Ditto.
410 2018-04-20 Martin Sebor <msebor@redhat.com>
413 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
415 (gimple_fold_builtin_stxcpy_chk): Same.
416 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
418 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
421 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
422 __powikf2 when long double is IEEE 128-bit.
424 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
426 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
427 step to make sure stack always aligned.
429 2018-04-20 Carl Love <cel@us.ibm.com>
432 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
435 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
436 Tom de Vries <tom@codesourcery.com>
439 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
440 Emit insns for calls too.
441 (nvptx_find_par): Always look for worker-level predecessor insn.
442 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
444 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
445 (nvptx_process_pars): Propagate frames for calls.
447 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
450 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
452 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
453 (ix86_handle_option): Don't handle OPT_mibt.
454 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
456 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
458 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
459 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
460 (ix86_target_macros): Define __CET__ with flag_cf_protection
462 * config/i386/i386.c (isa2_opts): Remove -mibt.
463 * config/i386/i386.h (TARGET_IBT): Removed.
464 (TARGET_IBT_P): Likewise.
465 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
466 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
467 * config/i386/i386.opt (mcet): Update help message.
470 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
471 -mcet as an alias for -mshstk.
473 2018-04-20 Richard Biener <rguenther@suse.de>
476 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
477 complexity by forcing a single use of the multiply operand.
479 2018-04-20 Martin Jambor <mjambor@suse.cz>
482 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
483 recursion dependency to only apply to non-clones.
485 2018-04-20 Martin Jambor <mjambor@suse.cz>
488 * ipa-cp.c (create_specialized_node): Check that clones of
489 self-recursive edges exist during IPA-CP.
491 2018-04-19 Toon Moene <toon@moene.org>
493 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
496 2018-04-19 Jakub Jelinek <jakub@redhat.com>
498 PR tree-optimization/85467
499 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
500 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
501 VECTOR_CST element to type.
503 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
506 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
507 * config/i386/i386.md (builtin_setjmp_setup): Removed.
508 (builtin_longjmp): Likewise.
509 (save_stack_nonlocal): New pattern.
510 (restore_stack_nonlocal): Likewise.
512 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
515 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
516 Replace ASM_OUTPUT_LABEL with fprintf.
518 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
521 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
522 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
523 * config/i386/i386-c.c (ix86_target_macros_internal): Also
524 define __IBT__ and __SHSTK__ for -fcf-protection.
525 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
527 (ix86_trampoline_init): Likewise.
528 (x86_output_mi_thunk): Likewise.
529 (ix86_notrack_prefixed_insn_p): Likewise.
530 (ix86_option_override_internal): Don't disallow -fcf-protection.
531 * config/i386/i386.md (rdssp<mode>): Also enable for
533 (incssp<mode>): Likewise.
534 (nop_endbr): Likewise.
535 * config/i386/i386.opt (mcet): Change help message to built-in
539 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
540 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
541 enable CET built-in functions.
543 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
545 * common/config/i386/i386-common.c
546 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
547 OPTION_MASK_ISA_MOVDIRI_UNSET,
548 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
549 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
550 * config.gcc (movdirintrin.h): New header.
551 * config/i386/cpuid.h (bit_MOVDIRI,
552 bit_MOVDIR64B): New bits.
553 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
555 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
556 (VOID, PVOID, PCVOID)): New function types.
557 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
558 __builtin_ia32_directstoreu_u64,
559 __builtin_ia32_movdir64b): New builtins.
560 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
561 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
563 (ix86_valid_target_attribute_inner_p): Ditto.
564 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
565 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
566 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
567 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
568 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
569 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
570 (movdiri<mode>, movdir64b_<mode>): New.
571 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
572 * config/i386/immintrin.h: Include movdirintrin.h.
573 * config/i386/movdirintrin.h: New file.
574 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
576 2018-04-19 Richard Biener <rguenther@suse.de>
579 * cfg.c (clear_bb_flags): When loop state says we have
580 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
582 2018-04-19 Richard Biener <rguenther@suse.de>
584 PR tree-optimization/84737
585 * tree-vect-data-refs.c (vect_copy_ref_info): New function
586 copying restrict info.
587 (vect_setup_realignment): Use it.
588 * tree-vectorizer.h (vect_copy_ref_info): Declare.
589 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
590 the first DR to all generated stores.
591 (vectorizable_load): Likewise for loads.
593 2018-04-19 Jakub Jelinek <jakub@redhat.com>
595 PR tree-optimization/85446
596 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
597 the integral and pointer types to have the same precision.
599 * doc/install.texi: Document --disable-cet being the default and
602 2018-04-18 Martin Liska <mliska@suse.cz>
604 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
607 2018-04-18 Martin Liska <mliska@suse.cz>
610 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
613 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
614 arguments if they are comparable.
616 2018-04-18 Martin Liska <mliska@suse.cz>
619 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
622 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
625 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
628 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
629 ENDBR after calling __morestack.
631 2018-04-18 David Malcolm <dmalcolm@redhat.com>
634 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
635 by using gcc_base_ver to generate a gcc_driver_version, and use
636 it when generating GCC_DRIVER_NAME.
637 * configure: Regenerate.
639 2018-04-18 Jakub Jelinek <jakub@redhat.com>
642 * config.gcc: Obsolete powerpc*-*-*spe*.
644 2018-04-17 Jakub Jelinek <jakub@redhat.com>
647 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
648 (stabstr_D): Change type of unum from unsigned int to
649 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
652 2018-04-17 Jim Wilson <jimw@sifive.com>
655 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
656 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
657 Set arg_pointer_offset after using pretend_args_size.
659 2018-04-17 Jakub Jelinek <jakub@redhat.com>
661 PR rtl-optimization/85431
662 * dse.c (record_store): Ignore zero width stores.
665 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
666 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
667 __builtin_stack_restore rather than after it.
668 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
669 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
670 argument instead of virtual_dynamic_stack_rtx.
672 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
674 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
676 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
677 Add note to error message to explain internal mapping of overloaded
678 built-in function name to non-overloaded built-in function name.
679 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
682 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
685 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
686 where the inputs overlap with the output.
688 2018-04-17 Jakub Jelinek <jakub@redhat.com>
690 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
691 (=v, v) alternative and explicit "memory" attribute.
692 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
693 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
695 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
696 "sselog1" type instead of "sselog".
697 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
698 "sselog". Remove explicit "memory" attribute.
699 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
700 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
702 (vec_extract_hi_v32hi): Merge all alternatives into one, use
703 "sselog1" type instead of "sselog". Remove explicit "memory"
705 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
706 use "sselog1" type instead of "sselog". Remove explicit "memory"
708 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
709 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
711 (vec_extract_hi_v64qi): Merge all alternatives into one, use
712 "sselog1" type instead of "sselog". Remove explicit "memory"
714 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
715 use "sselog1" type instead of "sselog". Remove explicit "memory"
719 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
722 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
725 2018-04-17 Martin Jambor <mjambor@suse.cz>
728 * ipa-cp.c (create_specialized_node): Call
729 expand_all_artificial_thunks if necessary.
731 2018-04-17 Martin Liska <mliska@suse.cz>
734 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
735 in message, remote space in between '_G' and '('.
737 2018-04-17 Jakub Jelinek <jakub@redhat.com>
740 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
741 avx512f_vmcmp<mode>3<round_saeonly_name>,
742 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
743 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
744 avx512f_rndscale<mode><round_saeonly_name>,
745 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
746 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
747 Use %<iptr>2 instead of %2 for -masm=intel.
748 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
749 avx512f_vcvttss2usi<round_saeonly_name>,
750 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
752 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
753 avx512f_vcvttsd2usi<round_saeonly_name>,
754 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
755 Use %q1 instead of %1 for -masm=intel.
756 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
757 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
758 of %3 for -masm=intel.
759 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
761 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
763 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
765 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
766 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
768 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
770 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
771 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
772 %g1 and one with %0 and %1.
773 (avx512er_vmrcp28<mode><round_saeonly_name>,
774 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
776 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
777 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
778 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
779 of %0 and %{%4%} for -masm=intel.
780 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
781 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
782 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
783 order of %0 and %{%5%}%{z%} for -masm=intel.
785 2018-04-17 Jan Hubicka <jh@suse.cz>
788 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
790 2018-04-17 Martin Liska <mliska@suse.cz>
793 * multiple_target.c (create_dispatcher_calls): Set apostrophes
794 for target_clone error message. Make default implementation
795 clone to be a local declaration.
796 (separate_attrs): Add new argument and check for an empty
798 (expand_target_clones): Handle it.
799 (ipa_target_clone): Make redirection just for target_clones
802 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
803 Tom de Vries <tom@codesourcery.com>
806 * omp-expand.c (expand_oacc_for): Add dummy false branch for
807 tiled basic blocks without omp continue statements.
809 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
812 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
813 vec_extract expression as having side effects to make sure it gets
816 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
819 * config/i386/i386.c (get_builtin_code_for_version): Check
822 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
825 * gcc/config.gcc: Support "skylake".
826 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
828 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
829 (processor_target_table): Add "skylake".
830 (ix86_option_override_internal): Add "skylake".
831 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
832 PROCESSOR_CANNONLAKE.
833 (get_builtin_code_for_version): Fix priority for
834 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
835 PROCESSOR_SKYLAKE-AVX512.
836 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
837 (processor_type): Add PROCESSOR_SKYLAKE.
839 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
840 Jason Merrill <jason@redhat.com>
843 * convert.c (convert_to_integer_1): Use direct recursion for
844 enumeral types and types with a precision less than the number
845 of bits in their mode.
847 2018-04-16 Julia Koval <julia.koval@intel.com>
850 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
851 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
853 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
856 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
857 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
858 and -mno-direct-move.
860 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
863 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
864 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
865 Ensure negative shifts result in {0}.
867 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
869 PR rtl-optimization/79916
870 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
871 regs (if any) to define how to gnerate SD moves when LRA is in
874 2018-04-13 Jakub Jelinek <jakub@redhat.com>
876 PR rtl-optimization/85393
877 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
878 * except.c (expand_dw2_landing_pad_for_region): Make static.
879 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
880 a label and unconditional jump to old_bb, rather than
881 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
884 PR rtl-optimization/85376
885 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
886 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
887 instead of a specific value.
889 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
890 Bin Cheng <bin.cheng@arm.com>
892 PR tree-optimization/82965
893 PR tree-optimization/83991
894 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
895 by_profile_only parameter.
896 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
897 information if the loop was predicted to iterate too many times.
898 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
900 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
903 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
906 2018-04-13 Martin Liska <mliska@suse.cz>
907 Jakub Jelinek <jakub@redhat.com>
910 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
911 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
912 * builtins.c (expand_builtin_memory_copy_args): Use
913 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
914 handle dest_addr == pc_rtx.
916 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
919 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
920 asked to not generate direct moves.
921 (fix_trunc<mode>si2_stfiwx): Similar.
922 (fix_trunc<mode>si2_internal): Similar.
924 2018-04-12 Jakub Jelinek <jakub@redhat.com>
927 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
928 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
929 lookup if dest in some wider mode is known to be const0_rtx and
930 if so, record permanent equivalence for it to be ZERO_EXTEND of
931 the narrower mode destination.
933 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
935 * lto-streamer-out.c (output_function): Revert 259346.
936 * omp-expand.c (expand_oacc_for): Likewise.
938 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
940 PR rtl-optimization/85354
941 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
942 * sel-sched.c (sel_global_init): ... here.
944 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
947 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
948 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
949 mode for PE-COFF targets.
950 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
951 (i386_pe_asm_lto_end): Likewise.
952 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
953 (TARGET_ASM_LTO_END): Likewise.
954 * config/i386/winnt.c (saved_debug_info_level): New static variable.
955 (i386_pe_asm_lto_start): New function.
956 (i386_pe_asm_lto_end): Likewise.
958 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
959 Richard Biener <rguenther@suse.de>
962 * lto-streamer-out.c (output_function): Fix CFG loop state before
964 * omp-expand.c (expand_oacc_for): Handle calls to internal
965 functions like regular functions.
967 2018-04-12 Richard Biener <rguenther@suse.de>
970 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
971 for the early LTO debug to properly generate references to it
972 during DIE emission. Do not re-use that for the skeleton for
974 (dwarf2out_early_finish): Likewise.
976 2018-04-12 Jakub Jelinek <jakub@redhat.com>
980 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
981 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
982 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
983 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
984 and output is a reg, avoid creating invalid lowpart subreg, but
985 instead split into a 512-bit move. Don't split if not AVX512VL,
986 input is xmm16+ reg and output is a mem.
987 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
988 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
989 xmm16+ reg and output is a mem.
991 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
993 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
994 also for flag_dwarf2_cfi_asm.
996 2018-04-12 Jakub Jelinek <jakub@redhat.com>
998 PR rtl-optimization/85342
999 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
1000 a bool scalar var inside of the loop instead. Don't try to update
1001 recog_data.operand after failed apply_change_group.
1003 2018-04-12 Tom de Vries <tom@codesourcery.com>
1006 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
1007 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
1008 array with flexible array member as array without given dimension.
1009 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
1010 argument for undefined param to true.
1012 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
1015 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1016 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
1017 from PowerPC section.
1018 * config/rs6000/sysv4.opt (mcall-): Improve help text.
1019 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
1020 help text that is too long.
1021 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
1022 help text that is too long.
1023 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
1024 help text that is too long.
1026 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
1028 * config/alpha/alpha.md (stack_probe_internal): Rename
1029 from "probe_stack". Update all callers.
1031 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1033 PR rtl-optimization/84566
1034 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
1035 sched_macro_fuse_insns.
1037 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1040 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
1041 (compute_block_dependences): ... from here.
1043 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1045 PR tree-optimization/85331
1046 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
1047 from int to HOST_WIDE_INT.
1049 2018-04-11 Martin Jambor <mjambor@suse.cz>
1052 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
1053 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
1054 not the same as the source val.
1055 (cgraph_edge_brings_value_p): New parameter.
1056 (gather_edges_for_value): Pass destination value to
1057 cgraph_edge_brings_value_p.
1058 (perhaps_add_new_callers): Likewise.
1059 (get_info_about_necessary_edges): Likewise and exclude values brought
1060 only by self-recursive edges.
1061 (create_specialized_node): Redirect only clones of self-calling edges.
1062 (+self_recursive_pass_through_p): New function.
1063 (find_more_scalar_values_for_callers_subset): Use it.
1064 (find_aggregate_values_for_callers_subset): Likewise.
1065 (known_aggs_to_agg_replacement_list): Removed.
1066 (decide_whether_version_node): Re-calculate known constants for all
1067 remaining context clones.
1069 2018-04-11 Richard Biener <rguenther@suse.de>
1072 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
1073 from early DWARF output.
1074 (dwarf2out_early_finish): Output line info unconditionally into
1075 early DWARF and add reference to it.
1077 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1080 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
1081 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
1082 other than V2DFmode using iptr mode attribute.
1083 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
1085 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1087 PR rtl-optimization/84659
1088 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
1090 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1093 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
1095 (output_loc_list): Pass address of a dummy size variable even in the
1096 locview handling loop.
1097 (index_location_lists): Add comment on why skip_loc_list_entry can't
1100 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
1103 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
1106 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
1109 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1110 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
1111 and -mstring-compare-inline-limit.
1113 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1116 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
1117 for stack clash protection in a register whenever we need it to be in
1120 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1122 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
1123 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
1125 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1128 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
1130 (mlong-double-): Ditto.
1131 * config/rs6000/sysv4.opt (msdata=): Ditto.
1132 (mtls-size=): Ditto.
1134 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
1136 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1137 erroneous entries for
1138 "vector int vec_ldl (int, long int *)", and
1139 "vector unsigned int vec_ldl (int, unsigned long int *)".
1140 Add comments and entries for
1141 "vector bool char vec_ldl (int, bool char *)",
1142 "vector bool short vec_ldl (int, bool short *)",
1143 "vector bool int vec_ldl (int, bool int *)",
1144 "vector bool long long vec_ldl (int, bool long long *)",
1145 "vector pixel vec_ldl (int, pixel *)",
1146 "vector long long vec_ldl (int, long long *)",
1147 "vector unsigned long long vec_ldl (int, unsigned long long *)".
1148 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
1149 type tree bool_long_long_type_node and correct definition of
1150 bool_V2DI_type_node to make reference to this new type tree.
1151 (rs6000_mangle_type): Replace erroneous reference to
1152 bool_long_type_node with bool_long_long_type_node.
1153 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
1154 comments to emphasize sign distinctions for char and int types and
1155 replace RS6000_BTI_bool_long constant with
1156 RS6000_BTI_bool_long_long constant. Also add comment to restrict
1157 use of RS6000_BTI_pixel.
1158 (bool_long_type_node): Remove this macro definition.
1159 (bool_long_long_type_node): New macro definition
1161 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1163 PR rtl-optimization/85300
1164 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
1165 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
1166 simplify_unary_operation fails.
1168 2018-04-10 Martin Liska <mliska@suse.cz>
1170 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
1171 cgraph_edge and ipa_ref.
1173 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1177 * config/i386/sse.md
1178 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
1179 computation of the VEC_MERGE selector from mask.
1180 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
1181 Fix decoding of the VEC_MERGE selector into mask.
1183 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1185 PR tree-optimization/85286
1186 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
1188 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1190 * final.c (final_1): Set insn_last_address as well as
1191 insn_current_address.
1193 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1196 * explow.c (emit_stack_probe): Call validize_mem on memory location
1197 before passing it to gen_probe_stack. Create address operand and
1198 legitimize it for the probe_stack_address case.
1200 2018-04-09 Jan Hubicka <jh@suse.cz>
1203 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
1204 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
1205 * tree.c (free_lang_data_in_type): Fix handling of binfos;
1207 (free_lang_data): Rebuild type inheritance graph.
1209 2018-04-09 Martin Sebor <msebor@redhat.com>
1211 * invoke.texi (-finline-small-functions): Mention other optimization
1213 (-findirect-inlining, -fpartial-inlining): Same.
1214 (-finline-functions-called-once): Same.
1215 (-freorder-blocks-and-partition): Same.
1217 2018-04-09 Jan Hubicka <jh@suse.cz>
1220 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
1221 jumps; choose last target that matches the criteria (i.e.
1222 no partition changes for non-crossing jumps).
1223 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
1224 support for redirecting crossing jumps to non-crossing.
1226 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
1228 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
1229 also for naked functions.
1231 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
1233 * config/arc/arc.md (add_shift): New pattern.
1234 (add_shift2): Likewise.
1235 (sub_shift): Likewise.
1236 (sub_shift_cmp0_noout): Likewise.
1237 (compare_si_ashiftsi): Likewise.
1238 (xbfu_cmp0_noout): New combine pattern.
1239 (xbfu_cmp0"): Likewise.
1240 (movsi_set_cc_insn): Place the predicable variant first.
1241 (commutative_binary_cmp0_noout): Remove clobber.
1242 (commutative_binary_cmp0): New pattern.
1243 (noncommutative_binary_cmp0): Likewise.
1244 (noncommutative_binary_cmp0_noout): Likewise.
1245 (noncommutative_binary_comparison_result_used): Removed.
1246 (rsub_cmp0): New pattern.
1247 (rsub_cmp0_noout): Likewise.
1248 (extzvsi): Changed, keep only meaningful variants.
1249 (SQH, SEZ): New iterators.
1250 (SQH_postfix): New mode attribute.
1251 (SEZ_prefix): New code attribute.
1252 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
1253 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
1254 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
1256 (noncommutative_operator): Check the availability of barrel
1259 2018-04-09 Richard Biener <rguenther@suse.de>
1261 PR tree-optimization/85284
1262 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
1263 Only use the niter constraining form of simple_iv when the exit
1266 2018-04-09 Tom de Vries <tom@codesourcery.com>
1269 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1270 (define_expand "*memory_barrier"): New define_expand.
1271 (define_insn "memory_barrier"): New insn.
1273 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1275 PR rtl-optimization/80463
1276 PR rtl-optimization/83972
1277 PR rtl-optimization/83480
1279 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
1280 correct producer for the insn.
1281 (tidy_control_flow): Fixup seqnos in case of debug insns.
1283 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1285 PR rtl-optimization/83913
1287 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
1288 different sched-times when merging exprs.
1290 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1292 PR rtl-optimization/83962
1294 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
1295 tidy_fallthru_edge and tidy_control_flow.
1297 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1299 PR rtl-optimization/83530
1301 * sel-sched.c (force_next_insn): New global variable.
1302 (remove_insn_for_debug): When force_next_insn is true, also leave only
1303 next insn in the ready list.
1304 (sel_sched_region): When the region wasn't scheduled, make another pass
1305 over it with force_next_insn set to 1.
1307 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1309 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
1311 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1312 for interrupt control.
1313 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1314 functions for interrupt control.
1315 * config/nds32/nds32-intrinsic.md: Likewise.
1316 * config/nds32/nds32_intrinsic.h: Likewise.
1317 * config/nds32/nds32.h (nds32_builtins): Likewise.
1319 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1321 * config/nds32/nds32.c (nds32_init_machine_status,
1322 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
1323 strict_aligned_p field.
1324 (nds32_expand_to_rtl_hook): New function.
1325 (TARGET_EXPAND_TO_RTL_HOOK): Define.
1326 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
1328 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1329 Chung-Ju Wu <jasonwucj@gmail.com>
1331 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
1332 * config/nds32/nds32-n7.md: New file.
1333 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
1334 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
1336 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
1337 * config/nds32/nds32.md (pipeline_model): Add n7.
1338 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
1339 * config/nds32/pipelines.md: Include n7 settings.
1341 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1342 Chung-Ju Wu <jasonwucj@gmail.com>
1344 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
1345 * config/nds32/nds32-e8.md: New file.
1346 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
1347 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
1349 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
1350 * config/nds32/nds32.md (pipeline_model): Add e8.
1351 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
1352 * config/nds32/pipelines.md: Include e8 settings.
1354 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1355 Chung-Ju Wu <jasonwucj@gmail.com>
1357 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
1358 * config/nds32/nds32-n8.md: New file.
1359 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
1360 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
1362 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
1363 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
1364 * config/nds32/nds32.md (pipeline_model): Add n8.
1365 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
1366 * config/nds32/pipelines.md: Include n8 settings.
1368 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1369 Chung-Ju Wu <jasonwucj@gmail.com>
1371 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
1372 * config/nds32/nds32-n9-2r1w.md: New file.
1373 * config/nds32/nds32-n9-3r2w.md: New file.
1374 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
1375 nds32_register_ports): New or modify for cpu n9.
1376 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
1378 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
1379 * config/nds32/nds32-utils.c: New file.
1380 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
1381 TARGET_MUL_SLOW): Define.
1382 * config/nds32/nds32.md (pipeline_model): New attribute.
1383 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
1384 New options that support cpu n9.
1385 * config/nds32/pipelines.md: Include n9 settings.
1386 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
1388 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1390 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
1391 information if necessary.
1392 (output_cond_branch_compare_zero): Likewise.
1393 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
1394 (nds32_target_alignment): Refine for alignment.
1395 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
1396 (FUNCTION_BOUNDARY): Modify.
1397 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
1399 * config/nds32/nds32.opt (malways-align, malign-functions): New.
1401 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1403 * config/nds32/constants.md (unspec_volatile_element): Add values for
1404 TLB operation and data prefetch.
1405 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1406 functions for TLB operation and data prefetch.
1407 * config/nds32/nds32-intrinsic.md: Likewise.
1408 * config/nds32/nds32_intrinsic.h: Likewise.
1409 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
1410 (nds32_print_operand): Likewise.
1411 * config/nds32/nds32.h (nds32_builtins): Likewise.
1413 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
1414 Andrew Pinski <pinsika@gcc.gnu.org>
1417 * match.pd: Use constant_boolean_node of correct type instead of
1418 boolean_true_node or boolean_false_node for simplifying
1419 pointer comparisons to zero.
1421 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1423 PR tree-optimization/80021
1424 * tree.c (verify_type_variant): Make error call in verify_variant_match
1425 translatable and remove final full stop.
1427 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1429 * config/nds32/constants.md (unspec_volatile_element): Add
1430 UNSPEC_VOLATILE_EH_RETURN.
1431 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
1432 nds32_output_stack_pop): Support dwarf exception handling process.
1433 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
1434 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
1435 exception handling process.
1436 (nds32_compute_stack_frame): Likewise.
1437 (nds32_return_addr_rtx): Likewise.
1438 (nds32_initial_elimination_offset): Likewise.
1439 (nds32_expand_prologue): Likewise.
1440 (nds32_expand_epilogue): Likewise.
1441 (nds32_dynamic_chain_address): New function.
1442 * config/nds32/nds32.h (machine_function): Add fields for dwarf
1444 (DYNAMIC_CHAIN_ADDRESS): Define.
1445 (EH_RETURN_DATA_REGNO): Define.
1446 (EH_RETURN_STACKADJ_RTX): Define.
1447 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
1448 patterns for dwarf exception handling.
1450 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1452 * config/nds32/nds32.h: Clean up obsolete macros.
1454 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1456 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1457 Add enum values for particular instructions.
1458 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1459 particular intrinsic functions.
1460 * config/nds32/nds32-intrinsic.md: Likewise.
1461 * config/nds32/nds32_intrinsic.h: Likewise.
1462 * config/nds32/nds32.h (nds32_builtins): Likewise.
1463 * config/nds32/nds32.md (type): Add pbsad and pbsada.
1464 (btst, ave): New patterns for particular instructions.
1466 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1468 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1469 Add enum values for atomic load/store and memory sync.
1470 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
1472 * config/nds32/nds32-intrinsic.md: Likewise.
1473 * config/nds32/nds32_intrinsic.h: Likewise.
1474 * config/nds32/nds32.h (nds32_builtins): Likewise.
1476 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1478 PR tree-optimization/85257
1479 * fold-const.c (native_encode_vector): If not all elts could fit
1480 and off is -1, return 0 rather than offset.
1481 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
1482 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
1483 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
1484 adjust buffer in native_interpret_expr call.
1486 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1488 * config/nds32/constants.md (unspec_volatile_element): Add cache
1489 control enum values.
1490 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
1491 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
1492 * config/nds32/nds32.c (nds32_cctl_names): New.
1493 (nds32_print_operand): Handle cache control register names.
1494 * config/nds32/nds32.h (nds32_builtins): New enum values.
1495 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
1497 * config/nds32/nds32.md (type): Add mmu.
1498 * config/nds32/pipelines.md (simple_insn): Add mmu.
1500 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1502 * config/nds32/nds32.md (type): Remove call.
1503 * config/nds32/pipelines.md (simple_insn): Likewise.
1505 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1507 * config/nds32/constants.md (unspec_volatile_element): Add
1508 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
1509 UNSPEC_VOLATILE_FMFCFG.
1510 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
1511 description for fmfcfg and fmfcsr.
1512 (bdesc_1arg): Add fmtcsr.
1513 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
1514 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
1515 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
1516 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
1517 unspec_fmfcfg): New patterns.
1518 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
1519 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
1520 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
1521 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
1522 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
1523 __nds32__fmfcfg): Define.
1525 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1527 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
1528 intrinsic register names.
1529 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
1530 intrinsic register enum values and macros.
1532 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1534 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
1535 for load/store addressing form.
1536 (nds32_print_operand_address): Likewise.
1538 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
1541 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
1542 based on LABEL_REF. Remove useless assertion.
1543 (pic_address_needs_scratch): Fix formatting.
1544 (sparc_legitimize_pic_address): Minor tweaks.
1545 (sparc_delegitimize_address): Adjust assertion accordingly.
1546 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
1547 into symbolic_operand.
1548 (movsi_high_pic_label_ref): Likewise.
1549 (movsi_lo_sum_pic_label_ref): Likewise.
1550 (movdi_pic_label_ref): Likewise.
1551 (movdi_high_pic_label_ref): Likewise.
1552 (movdi_lo_sum_pic_label_ref): Likewise.
1554 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
1556 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
1557 custom LIB_SPEC setup.
1559 2018-04-06 Ruslan Bukin <br@bsdpad.com>
1560 Kito Cheng <kito.cheng@gmail.com>
1562 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
1563 * config/riscv/freebsd.h: New.
1565 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1567 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
1568 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
1571 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1572 Kito Cheng <kito.cheng@gmail.com>
1574 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
1575 nds32_output_call, nds32_symbol_binds_local_p): New functions.
1576 * config/nds32/nds32-protos.h (nds32_output_call,
1577 nds32_output_return): Declare.
1578 * config/nds32/nds32.md: Refine all the call and return patterns.
1580 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1583 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
1584 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
1586 PR rtl-optimization/84872
1587 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
1588 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
1591 2018-04-06 Tamar Christina <tamar.christina@arm.com>
1593 * expr.c (copy_blkmode_to_reg): Revert 254862.
1594 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
1596 2018-04-06 Richard Biener <rguenther@suse.de>
1599 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
1600 after seeing a component reference with an adjacent field. Treat
1601 refs to arrays at struct end of external decls similar to
1602 refs to unconstrained commons.
1604 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1607 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
1608 look through SAVE_EXPRs with non-side-effects argument. Adjust
1610 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
1613 2018-04-06 Richard Biener <rguenther@suse.de>
1616 * alias.c (find_base_term): New wrapper around find_base_term
1617 unwinding CSELIB_VAL_PTR changes.
1618 (find_base_term): Do not restore CSELIB_VAL_PTR during the
1621 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1623 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
1625 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
1626 constant definitions.
1627 ("nop"): lr 0,0 -> nopr r0
1628 ("nop_lr0", "nop_lr1"): New insn definitions.
1630 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1632 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
1633 NDS32_V3PUSH_AVAILABLE_P macro.
1635 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
1636 Chung-Ju Wu <jasonwucj@gmail.com>
1638 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
1639 (nds32*-*-*): Add float and fpu_config into supported_defaults.
1640 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
1641 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
1642 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
1643 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
1644 * config/nds32/constraints.md: New constraints and checking for hard
1645 float configuration.
1646 * config/nds32/iterators.md: New mode iterator and attribute for hard
1647 float configuration.
1648 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
1650 * config/nds32/nds32-fpu.md: New file.
1651 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
1652 deal with hard float code generation.
1653 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
1655 (abi_type, float_reg_number): New enum type.
1656 * config/nds32/nds32-predicates.c: New predicates for hard float.
1657 * config/nds32/nds32-protos.h: Declare functions for hard float.
1658 * config/nds32/nds32.c: Implementation for hard float configuration.
1659 * config/nds32/nds32.h: Definitions for hard float configuration.
1660 * config/nds32/nds32.md: Include hard float machine description and
1661 modify patterns for hard float configuration.
1662 * config/nds32/nds32.opt: New options for hard float configuration.
1663 * config/nds32/predicates.md: New predicates for hard float
1666 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1668 * common/config/nds32/nds32-common.c
1669 (nds32_option_optimization_table): Enable -mreleax-hint by default.
1671 2018-04-05 Jakub Jelinek <jakub@redhat.com>
1674 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
1675 CONSTRUCTOR_ELT (ctor, ...)->value.
1677 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
1680 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
1682 2018-04-05 Tom de Vries <tom@codesourcery.com>
1685 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
1688 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1689 Kito Cheng <kito.cheng@gmail.com>
1691 * config/nds32/constraints.md (U33): Fine-tune checking condition.
1692 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
1693 * config/nds32/nds32.h (nds32_16bit_address_type): Add
1694 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
1696 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1697 Kito Cheng <kito.cheng@gmail.com>
1699 * config/nds32/constraints.md (Ufe): New memory constraint.
1700 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
1701 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
1702 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
1704 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
1705 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
1707 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1709 * config/nds32/nds32.md: Use optimize_size in the condition for
1710 alu-shift instructions.
1712 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1714 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
1716 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1718 * config/nds32/nds32.md (negsi2): Refine pattern.
1720 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1721 Chung-Ju Wu <jasonwucj@gmail.com>
1723 * config/nds32/iterators.md (shift_rotate): New code iterator.
1724 (shift): New code attribute.
1725 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
1726 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
1727 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
1728 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
1729 bit-wise operations.
1730 (andsi3, *andsi3): Ditto.
1731 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
1732 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
1733 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
1734 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
1735 nds32_ior_operand, nds32_xor_operand): New predicates.
1737 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1739 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
1740 (addsi3, subsi3): ... this.
1742 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1744 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
1746 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1748 * config/nds32/nds32.md: Adjust indention.
1750 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1752 * config/nds32/nds32.md (feature): New attribute.
1754 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1756 * config/nds32/nds32.md (subtype): New attribute.
1758 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
1761 * config/arm/arm-builtins.c (arm_expand_builtin): Change
1762 expansion to perform a bitwise AND of the argument followed by a
1763 boolean negation of the result.
1765 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
1767 PR rtl-optimization/84878
1768 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
1769 the basic block. Assert the use reference is not artificial and that
1770 it has an associated insn.
1772 2018-04-04 Michael Matz <matz@suse.de>
1774 * builtins.c (compute_objsize): Pass correct operand
1775 to array_at_struct_end_p.
1777 2018-04-04 Richard Biener <rguenther@suse.de>
1780 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
1781 from contexts for DINFO_LEVEL_TERSE and below.
1783 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1785 * config/nds32/nds32-doubleword.md (move_<mode>): Require
1786 resiter_operand condition.
1787 * config/nds32/nds32.md (*move<mode>): Ditto.
1789 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1790 Monk Chiang <sh.chiang04@gmail.com>
1792 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
1794 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1796 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
1798 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1799 Kito Cheng <kito.cheng@gmail.com>
1801 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
1802 nds32_cond_code_str, output_cond_branch,
1803 output_cond_branch_compare_zero, nds32_expand_cbranch,
1804 nds32_expand_cstore, nds32_expand_movcc,
1805 nds32_output_cbranchsi4_equality_zero,
1806 nds32_output_cbranchsi4_equality_reg,
1807 nds32_output_cbranchsi4_equality_reg_or_const_int,
1808 nds32_output_cbranchsi4_greater_less_zero: New functions.
1809 * config/nds32/nds32-protos.h (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): Declare.
1815 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
1816 nds32_rimm11s_operand): New predicates.
1817 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
1818 * config/nds32/nds32.md: Rewrite all the branch and conditional move
1821 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1823 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
1824 * config/nds32/nds32.md: Ditto.
1825 * config/nds32/pipelines.md: Ditto.
1827 2018-04-04 Richard Biener <rguenther@suse.de>
1829 PR tree-optimization/85168
1830 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
1831 propagating abnormals.
1833 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1835 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
1837 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1838 Kito Cheng <kito.cheng@gmail.com>
1840 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
1841 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
1842 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
1843 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
1844 * config/nds32/nds32.md (sibcall_internal): New.
1845 (sibcall_register): Remove.
1846 (sibcall_immediate): Remove.
1847 (sibcall_value_internal): New.
1848 (sibcall_value_register): Remove.
1849 (sibcall_value_immediate): Remove.
1850 * config/nds32/predicates.md (nds32_general_register_operand): New.
1851 (nds32_call_address_operand): New.
1853 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1855 PR rtl-optimization/85167
1856 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
1857 bb_defs if *split_p, instead preinitialize it to NULL.
1859 PR tree-optimization/85156
1860 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
1861 evaluating the argument multiple times.
1863 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
1865 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
1867 (_mm_cvtpd_ps): Likewise.
1868 (_mm_cvttpd_epi32): Likewise.
1869 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
1870 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
1871 vector, pixel, and bool following altivec.h include.
1873 2018-04-03 Martin Sebor <msebor@redhat.com>
1875 * doc/extend.texi (Common Function Attributes): Clarify.
1876 (const attribute): Likewise.
1877 (pure attribute): Likewise.
1879 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1882 * config/i386/i386.c (ix86_expand_vector_set): Use
1883 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
1885 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
1887 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
1888 instructions when changing rounding bits to preserve precision bits
1889 in the x87 control word.
1891 2018-04-03 Martin Liska <mliska@suse.cz>
1893 PR tree-optimization/82491
1894 * rtl.h (strip_offset_and_add): Replace += suboffset with
1895 poly_uint64 () + suboffset.
1897 2018-03-29 Martin Liska <mliska@suse.cz>
1898 Martin Jambor <mjambor@suse.cz>
1901 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
1902 param_type is not an integral or pointer type.
1904 2018-04-03 Richard Biener <rguenther@suse.de>
1906 * sese.h (recompute_all_dominators): Remove.
1908 2018-04-02 Martin Sebor <msebor@redhat.com>
1910 * doc/invoke.texi (-Wrestrict): Fix typos.
1912 2018-04-02 Jim Wilson <jimw@sifive.com>
1914 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
1915 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
1916 (<optab>di3, <optab>si3_extend): Likewise.
1917 (<optab>si3_mask, <optab>si3_mask_1): New.
1918 (<optab>di3_mask, <optab>di3_mask_1): New.
1919 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
1920 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
1921 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
1923 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
1925 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
1928 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
1930 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
1931 (nds32_canonicalize_comparison): New function.
1933 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1934 Kito Cheng <kito.cheng@gmail.com>
1935 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1937 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
1938 * config/nds32/constants.md (unspec_volatile_element): Add
1939 UNSPEC_VOLATILE_RELAX_GROUP.
1940 * config/nds32/nds32-relax-opt.c: New file.
1941 * config/nds32/nds32-predicates.c
1942 (nds32_symbol_load_store_p): New function.
1943 * config/nds32/nds32-protos.h
1944 (nds32_symbol_load_store_p): Declare function.
1945 (make_pass_nds32_relax_opt): Declare new rtl pass function.
1946 * config/nds32/nds32.c
1947 (nds32_register_pass): New function to register pass.
1948 (nds32_register_passes): New function to register passes.
1949 * config/nds32/nds32.md (relax_group): New pattern.
1950 * config/nds32/nds32.opt (mrelax-hint): New option.
1951 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
1953 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
1955 * config/nds32/t-nds32: Modify files dependency.
1957 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1959 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
1960 (PROFILE_HOOK): Define its implementation.
1962 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1964 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
1965 type and 32-bit size.
1967 2018-04-01 Jakub Jelinek <jakub@redhat.com>
1970 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
1971 (V_128_256): New mode iterator.
1972 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
1973 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
1974 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
1976 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
1977 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
1979 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
1982 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
1983 NaN inputs correctly.
1985 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
1988 * config/rs6000/vsx.md (??r): New mode attribute.
1989 (*vsx_mov<mode>_64bit): Use it.
1990 (*vsx_mov<mode>_32bit): Likewise.
1992 2018-03-30 Martin Sebor <msebor@redhat.com>
1994 PR tree-optimization/84818
1995 * builtins.c (check_access): Use warning_n.
1997 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2000 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
2002 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
2005 2018-03-30 Julia Koval <julia.koval@intel.com>
2008 * x86-tune.def (movx, partial_reg_dependency): Enable for
2011 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
2014 * lra-constraints.c (process_alt_operands): Move setting
2015 this_alternative_matches below.
2017 2018-03-29 Martin Liska <mliska@suse.cz>
2020 * doc/invoke.texi: Document how LTO works with debug info.
2021 Describe auto-load support of binutils. Mention 'x86-64'
2022 as valid option value of -march option.
2024 2018-03-29 Jakub Jelinek <jakub@redhat.com>
2026 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
2029 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
2030 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
2031 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
2034 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2037 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
2038 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
2039 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
2040 for RS6000_BTM_POWERPC64.
2041 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
2042 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
2043 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
2048 2018-03-28 Carl Love <cel@us.ibm.com>
2051 2017-09-27 Carl Love <cel@us.ibm.com>
2053 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
2054 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
2055 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
2058 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2060 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
2061 instead of __vector bool.
2062 (_mm_max_pu8): Likewise.
2063 (_mm_min_pi16): Likewise.
2065 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2068 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
2069 (DIVWEUO): Likewise.
2071 (DIVDEUO): Likewise.
2072 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
2073 DIVWEUO and DIVDEUO.
2074 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
2075 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
2076 (div_extend): Likewise.
2077 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
2079 (__builtin_divweuo): Likewise.
2080 (__builtin_divdeo): Likewise.
2081 (__builtin_divdeuo): Likewise.
2083 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2086 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
2087 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
2089 PR tree-optimization/82004
2090 * gimple-match-head.c (optimize_pow_to_exp): New function.
2091 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
2092 Don't fold to exp if optimize_pow_to_exp is false.
2094 2018-03-28 Martin Liska <mliska@suse.cz>
2097 * calls.c (initialize_argument_information): Fix trailing space.
2098 * common.opt: Fix typo and provide better explanation for
2099 -fsanitize-coverage option.
2100 * config/i386/i386.opt: Fix typo.
2102 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2103 Martin Liska <mliska@suse.cz>
2106 * gimplify.c (asan_poison_variable): Don't do the check for
2107 gimplify_omp_ctxp here.
2108 (gimplify_decl_expr): Do it here.
2109 (gimplify_target_expr): Likewise.
2111 2018-03-28 Martin Liska <mliska@suse.cz>
2114 * config/i386/i386.c (ix86_function_arg_advance): Do not call
2115 chkp_type_bounds_count if MPX is not enabled.
2117 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
2119 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
2121 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
2124 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
2125 function to create the function decl for complex long double
2126 multiply and divide for -mabi=ieeelongdouble.
2127 (init_float128_ieee): Call it.
2129 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
2132 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
2133 -fcf-protection=branch -mibt.
2134 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
2136 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2139 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
2141 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
2144 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
2145 extern array declarations.
2147 2018-03-27 Richard Biener <rguenther@suse.de>
2150 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
2151 explicit single_use checks.
2153 2018-03-27 Richard Biener <rguenther@suse.de>
2155 PR tree-optimization/85082
2156 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
2159 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2161 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
2162 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
2163 Turn on fasynchronous-unwind-tables and funwind-tables.
2165 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
2168 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
2169 (*bmi_blsr_<mode>_ccz): Ditto.
2171 2018-03-26 Tom de Vries <tom@codesourcery.com>
2173 PR tree-optimization/85063
2174 * omp-general.c (offloading_function_p): New function. Factor out
2176 * omp-offload.c (pass_omp_target_link::gate): ... here.
2177 * omp-general.h (offloading_function_p): Declare.
2178 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
2179 with attribute omp declare target for offloading functions.
2181 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
2183 PR tree-optimization/84005
2184 * tree-data-ref.h (get_base_for_alignment): Declare.
2185 * tree-data-ref.c (get_base_for_alignment_1): New function.
2186 (get_base_for_alignment): Likewise.
2187 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
2188 get_base_for_alignment to find a suitable base object, instead
2189 of always using drb->base_address.
2191 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2194 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
2195 known size by default.
2197 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
2200 * lra-constraints.c (process_alt_operands): Don't match BLKmode
2201 and non BLKmode operands.
2203 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2206 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
2207 Clean up attributes.
2209 2018-03-23 Richard Biener <rguenther@suse.de>
2212 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
2213 we are going to emit early debug for LTO.
2215 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2218 * function.c (match_asm_constraints_1): Don't optimize if input
2219 doesn't satisfy general_operand predicate for output's mode.
2222 * alias.c (write_dependence_p): Don't require for x_canonicalized
2223 non-VOIDmode if x has VOIDmode.
2226 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
2227 just don't try to optimize it rather than assert it never happens.
2229 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2231 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
2232 macro expansions for definition of ST_INTERNAL_<mode> and
2233 LD_INTERNAL_<mode> builtins.
2234 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
2236 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
2238 (altivec_expand_st_builtin): Likewise.
2239 (altivec_expand_builtin): Remove calls to deleted functions.
2240 (rs6000_address_for_altivec): Delete this function.
2241 * config/rs6000/vector.md: Remove expands for
2242 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
2244 2018-03-22 Sudakshina Das <sudi.das@arm.com>
2247 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
2248 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
2249 re-computing once computed.
2250 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
2251 (arm_init_machine_status): Initialize
2252 machine->static_chain_stack_bytes.
2254 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2257 * doc/extend.texi: Add four new prototypes for vec_ld.
2258 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
2259 definitions for more logical presentation.
2260 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
2261 entries for V1TI variants of __builtin_altivec_ld builtin.
2262 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
2263 handling of V1TI variant of LVX icode pattern.
2264 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
2265 (rs6000_gimple_fold_builtin): Likewise.
2266 (altivec_init_builtins): Add code to define
2267 __builtin_altivec_lvx_v1ti function.
2269 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2272 * function.c (match_asm_constraints_1): Don't do the optimization
2273 if input isn't a REG, SUBREG, MEM or constant.
2275 2018-03-22 Tom de Vries <tom@codesourcery.com>
2277 PR tree-optimization/84956
2278 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
2279 bb_has_abnormal_pred.
2281 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2284 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
2285 DECL_INITIAL (decl) to decl at the end.
2286 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
2289 2018-03-21 Joseph Myers <joseph@codesourcery.com>
2291 * doc/extend.texi (__builtin_tgmath): Document when complex
2292 integer types are treated as _Complex _Float64.
2294 2018-03-21 Tom de Vries <tom@codesourcery.com>
2296 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
2298 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2300 PR tree-optimization/84960
2301 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
2302 if it is ENTRY block, move them into single succ of ENTRY in that case.
2304 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
2306 PR tree-optimization/84811
2307 * poly-int.h (poly_span_traits): Remove the T3 parameter and
2308 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
2309 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
2310 (known_subrange_p): Update accordingly. Cast each value involved
2311 in the size comparison, rather than casting the result of the
2314 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2316 PR tree-optimization/84982
2317 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
2318 by flipping the least significant bit rather than all bits from
2319 bitpos to bitpos + bitsize - 1.
2321 2018-03-21 Nathan Sidwell <nathan@acm.org>
2323 * doc/extend.texi (Deprecated Features): Remove mention of
2324 long-deleted deprecations.
2326 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2329 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
2330 * configure: Regenerate.
2332 2018-03-21 Tom de Vries <tom@codesourcery.com>
2334 PR tree-optimization/83126
2335 * tree-parloops.c (num_phis): New function.
2336 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
2338 2018-03-21 Nathan Sidwell <nathan@acm.org>
2340 * doc/extend.texi (Deprecated Features): Update deprecated flags,
2341 mention anon-struct/union members and trailing attributes.
2343 2018-03-21 Bin Cheng <bin.cheng@arm.com>
2345 PR tree-optimization/84969
2346 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
2347 builtin memset partitions if they set different rhs values.
2349 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2351 PR rtl-optimization/84989
2352 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
2353 VEC_DUPLICATE with scalar result mode.
2355 2018-03-21 Martin Liska <mliska@suse.cz>
2358 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
2359 not intended return statement.
2361 2018-03-21 Martin Liska <mliska@suse.cz>
2364 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
2365 (chkp_find_bound_slots_1): Limit number of iterations.
2367 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
2370 * Minor grammar fixes for x86 options.
2372 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2375 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
2376 holding REG_CFA_RESTORE notes, instead turn them into a USE.
2378 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
2381 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
2382 (altivec_lvx_<mode>_1op): Likewise.
2383 (altivec_stvx_<mode>_2op): Likewise.
2384 (altivec_stvx_<mode>_1op): Likewise.
2385 (altivec_lvx_<VM2:mode>): New define_expand.
2386 (altivec_stvx_<VM2:mode>): Likewise.
2387 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
2388 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2389 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
2390 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2391 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
2392 (rs6000_gen_lvx): Likewise.
2393 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
2394 (altivec_expand_stv_builtin): Likewise.
2395 (altivec_expand_builtin): Likewise.
2396 * config/rs6000/vector.md: Likewise.
2398 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2401 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
2404 2018-03-20 Richard Biener <rguenther@suse.de>
2407 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
2408 sign-conversions as zero, fall back to standard scalar_stmt
2411 2018-03-20 Martin Liska <mliska@suse.cz>
2414 * predict.c (rebuild_frequencies): Handle case when we have
2415 PROFILE_ABSENT, but flag_guess_branch_prob is false.
2417 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2420 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
2421 flag_section_anchors.
2422 * varasm.c (use_blocks_for_decl_p): Remove hack for
2423 dw2_force_const_mem.
2426 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
2428 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
2429 be created, use lowpart_subreg of operands[0] rather than operands[0]
2431 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
2432 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
2433 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
2434 and n constraint instead of aarch64_shift_imm_di and Usd.
2435 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
2436 (*aarch64_<optab>_reg_minus<mode>3): ... this.
2438 2018-03-20 Sudakshina Das <sudi.das@arm.com>
2441 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
2442 to favor GPR over NEON registers.
2443 (<shift>di3_neon): Likewise.
2445 2018-03-20 Tom de Vries <tom@codesourcery.com>
2448 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
2449 (nvptx_process_pars): Emit bar.sync asap and alap.
2451 2018-03-20 Tom de Vries <tom@codesourcery.com>
2454 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
2455 seen_label if seen_label is already set.
2457 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2460 * config/i386/i386.c (fold_builtin_cpu): For features above 31
2461 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
2462 Use 1U instead of 1. Formatting fixes.
2465 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
2466 instead of TREE_TYPE (s1) for the return value.
2468 2018-03-19 Jakub Jelinek <jakub@redhat.com>
2470 PR tree-optimization/84946
2471 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
2472 bitsize + bitsize in poly_uint64 rather than poly_int64.
2475 * dwarf2asm.c: Include fold-const.c.
2476 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
2477 of decl rather than decl itself.
2479 PR rtl-optimization/84643
2480 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
2482 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
2485 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
2486 calling assemble_variable.
2488 2018-03-19 Sudakshina Das <sudi.das@arm.com>
2491 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
2492 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
2494 2018-03-19 Jim Wilson <jimw@sifive.com>
2497 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
2498 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
2499 (riscv_first_stack_step): Likewise.
2500 (riscv_option_override): Use STACK_BOUNDARY instead of
2502 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
2504 (BIGGEST_ALIGNMENT): Set to 128.
2505 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
2506 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
2509 2018-03-19 Richard Biener <rguenther@suse.de>
2511 PR tree-optimization/84933
2512 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
2513 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
2515 2018-03-19 Richard Biener <rguenther@suse.de>
2517 PR tree-optimization/84859
2518 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
2519 (cond_if_else_store_replacement): Perform sinking operation on
2520 single-store BBs regardless of MAX_STORES_TO_SINK setting.
2521 Generalize what a BB with a single eligible store is.
2523 2018-03-19 Richard Biener <rguenther@suse.de>
2525 PR tree-optimization/84929
2526 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
2527 chrec_is_positive against non-chrec arg.
2529 2018-03-19 Tamar Christina <tamar.christina@arm.com>
2532 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
2534 2018-03-18 Martin Liska <mliska@suse.cz>
2536 PR rtl-optimization/84635
2537 * regrename.c (build_def_use): Use matches_mode only when
2540 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
2542 PR tree-optimization/84913
2543 * tree-vect-loop.c (vectorizable_reduction): Don't try to
2544 vectorize chains of COND_EXPRs.
2546 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2548 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
2550 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2552 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
2554 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2556 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
2558 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2559 Kito Cheng <kito.cheng@gmail.com>
2561 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
2562 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
2563 (nds32_adjust_reg_alloc_order): New function.
2564 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
2566 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
2568 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
2569 nds32_print_operand, nds32_print_operand_address): Use
2570 HOST_WIDE_INT_PRINT_DEC instead.
2572 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2574 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
2576 2018-03-17 Jakub Jelinek <jakub@redhat.com>
2579 * config/i386/i386.c (initial_ix86_tune_features,
2580 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
2582 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
2583 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
2584 rather than 1u << ix86_tune. Formatting fix.
2585 (ix86_option_override_internal): Change ix86_arch_mask from
2586 unsigned int to unsigned HOST_WIDE_INT, initialize to
2587 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
2588 (ix86_function_specific_restore): Likewise.
2590 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2593 * postreload.c (reload_combine_recognize_pattern): Perform
2594 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
2595 truncate_int_for_mode the result for the destination's mode.
2598 * hsa-gen.c (mem_type_for_type): Fix comment typo.
2599 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
2601 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2604 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
2607 * lra-assigns.c (lra_split_hard_reg_for): Don't use
2608 regno_allocno_class_array and sorted_pseudos.
2609 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
2610 insns where regno is used.
2612 2018-03-16 Martin Liska <mliska@suse.cz>
2615 * multiple_target.c (create_dispatcher_calls): Redirect
2616 reference in the symbol table.
2618 2018-03-16 Martin Liska <mliska@suse.cz>
2621 * multiple_target.c (create_dispatcher_calls): Redirect also
2624 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2628 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
2629 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
2630 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
2632 2018-03-16 Julia Koval <julia.koval@intel.com>
2634 * doc/invoke.texi (Skylake Server): Add CLWB.
2635 Cannonlake): Remove CLWB.
2637 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2639 PR tree-optimization/84841
2640 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
2642 (FLOAT_ONE_CONST_TYPE): Define.
2643 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
2644 (sort_by_operand_rank): Put entries with higher constant_type last
2645 rather than first to match comments.
2647 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
2649 * config/nios2/nios2.md (movsi_internal): Fix thinko in
2652 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2655 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
2656 check and use address of target always.
2658 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
2661 * config/i386/i386.c (indirect_thunk_needed): Update comments.
2662 (indirect_thunk_bnd_needed): Likewise.
2663 (indirect_thunks_used): Likewise.
2664 (indirect_thunks_bnd_used): Likewise.
2665 (indirect_return_needed): New.
2666 (indirect_return_bnd_needed): Likewise.
2667 (output_indirect_thunk_function): Add a bool argument for
2669 (output_indirect_thunk_function): Don't generate alias for
2670 function return thunk.
2671 (ix86_code_end): Call output_indirect_thunk_function to generate
2672 function return thunks.
2673 (ix86_output_function_return): Set indirect_return_bnd_needed
2674 and indirect_return_needed instead of indirect_thunk_bnd_needed
2675 and indirect_thunk_needed.
2677 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
2679 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
2680 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
2681 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
2683 2018-03-15 David Malcolm <dmalcolm@redhat.com>
2684 Paul Hua <paul.hua.gm@gmail.com>
2687 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
2689 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
2691 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
2692 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
2695 2018-03-15 Tamar Christina <tamar.christina@arm.com>
2698 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
2699 instead of GET_MODE_SIZE when comparing Units.
2701 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2704 * varasm.c (hash_section): Return an unchangeble hash value
2705 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
2706 Return !aarch64_can_use_per_function_literal_pools_p ().
2708 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2711 * optabs.c (emit_conditional_move): Pass address of cmode's copy
2712 rather than address of cmode as last argument to prepare_cmp_insn.
2714 2018-03-15 Julia Koval <julia.koval@intel.com>
2716 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
2717 F_AVX512VNNI, F_AVX512BITALG): New.
2719 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
2722 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
2723 insn for floating-point loads and stores.
2725 2018-03-14 Carl Love <cel@us.ibm.com>
2727 * config/rs6000/rs6000-c.c: Add macro definitions for
2728 ALTIVEC_BUILTIN_VEC_PERMXOR.
2729 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
2730 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
2731 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
2732 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
2734 * config/doc/extend.texi: Add prototypes for vec_permxor.
2736 2018-03-14 David Malcolm <dmalcolm@redhat.com>
2739 * diagnostic-show-locus.c (class layout_point): Convert m_line
2740 from int to linenum_type.
2741 (line_span::comparator): Use linenum "compare" function when
2742 comparing line numbers.
2743 (test_line_span): New function.
2744 (layout_range::contains_point): Convert param "row" from int to
2746 (layout_range::intersects_line_p): Likewise.
2747 (layout::will_show_line_p): Likewise.
2748 (layout::print_source_line): Likewise.
2749 (layout::should_print_annotation_line_p): Likewise.
2750 (layout::print_annotation_line): Likewise.
2751 (layout::print_leading_fixits): Likewise.
2752 (layout::annotation_line_showed_range_p): Likewise.
2753 (struct line_corrections): Likewise for field m_row.
2754 (line_corrections::line_corrections): Likewise for param "row".
2755 (layout::print_trailing_fixits): Likewise.
2756 (layout::get_state_at_point): Likewise.
2757 (layout::get_x_bound_for_row): Likewise.
2758 (layout::print_line): Likewise.
2759 (diagnostic_show_locus): Likewise for locals "last_line" and
2761 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
2762 * input.c (selftest::test_linenum_comparisons): New function.
2763 (selftest::input_c_tests): Call it.
2764 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
2765 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
2766 * selftest.h (ASSERT_GT): New macro.
2767 (ASSERT_GT_AT): New macro.
2768 (ASSERT_LT): New macro.
2769 (ASSERT_LT_AT): New macro.
2771 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
2773 PR rtl-optimization/84780
2774 * combine.c (distribute_links): Don't make a link based on pc_rtx.
2776 2018-03-14 Martin Liska <mliska@suse.cz>
2778 * tree.c (record_node_allocation_statistics): Use
2779 get_stats_node_kind.
2780 (get_stats_node_kind): New function extracted from
2781 record_node_allocation_statistics.
2782 (free_node): Use get_stats_node_kind.
2784 2018-03-14 Richard Biener <rguenther@suse.de>
2786 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
2787 that the value-set of ANTIC_IN doesn't grow.
2790 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2792 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2793 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2795 2018-03-14 Julia Koval <julia.koval@intel.com>
2797 * config.gcc (icelake-client, icelake-server): New.
2799 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
2800 (initial_ix86_arch_features): Ditto.
2801 (PTA_SKYLAKE): Add SGX.
2802 (PTA_ICELAKE): Remove.
2803 (PTA_ICELAKE_CLIENT): New.
2804 (PTA_ICELAKE_SERVER): New.
2805 (ix86_option_override_internal): Split up icelake on icelake client and
2807 (get_builtin_code_for_version): Ditto.
2808 (fold_builtin_cpu): Ditto.
2809 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
2810 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
2811 * config/i386/i386.h (processor_type): Ditto.
2812 * doc/invoke.texi: Ditto.
2814 2018-03-14 Jakub Jelinek <jakub@redhat.com>
2817 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
2818 INTEGER_CST offset, add it together with bitpos / 8 and
2819 sign extend based on POINTER_SIZE.
2823 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
2826 * config/i386/constraints.md (Yc): New register constraint.
2827 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
2828 Use Yc constraint for alternative 2 of operand 0. Remove
2829 preferred_for_speed attribute.
2831 2018-03-14 Richard Biener <rguenther@suse.de>
2833 PR tree-optimization/84830
2834 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
2835 with the old one to avoid oscillations.
2837 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
2840 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
2842 (assign_by_spills): Return a flag of reload assignment failure.
2843 Do not process the reload assignment failures. Do not spill other
2844 reload pseudos if they has the same reg class. Update n if
2846 (lra_assign): Add a return arg. Set up from the result of
2847 assign_by_spills call.
2848 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2849 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2850 usage_insns if it is not NULL.
2851 (spill_hard_reg_in_range): New function.
2852 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2853 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2854 function prototypes.
2855 (lra_assign): Change prototype.
2856 * lra.c (lra): Add code to deal with fails by splitting hard reg
2859 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
2861 * config/riscv/riscv.opt (mrelax): New option.
2862 * config/riscv/riscv.c (riscv_file_start): Emit ".option
2863 "norelax" when riscv_mrelax is disabled.
2864 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
2866 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2869 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
2870 reassociation for int modes.
2872 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2874 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
2875 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
2877 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
2878 * config/aarch64/aarch64-sve.md
2879 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
2880 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
2881 (*extend<mode><Vwide>2): Rename to...
2882 (aarch64_sve_extend<mode><Vwide>2): ...this.
2883 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
2884 renaming the old pattern to...
2885 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
2887 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
2888 define_expand, renaming the old pattern to...
2889 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
2890 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
2891 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
2892 account when deciding which SVE instruction the optab should use.
2893 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
2895 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2897 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
2898 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
2899 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
2900 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
2901 (tlsdesc_small_<mode>): Turn a define_expand and use
2902 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
2903 (tlsdesc_small_advsimd_<mode>): ...this.
2904 (tlsdesc_small_sve_<mode>): New pattern.
2906 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2908 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
2909 (UNSPEC_UMUL_HIGHPART): New constants.
2910 (MUL_HIGHPART): New int iteraor.
2911 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
2912 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
2914 (*<su>mul<mode>3_highpart): New define_insn.
2916 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2919 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2922 2018-03-13 Martin Liska <mliska@suse.cz>
2925 * (sem_item_optimizer::sem_item_optimizer): Initialize new
2927 (sem_item_optimizer::~sem_item_optimizer): Release it.
2928 (sem_item_optimizer::merge_classes): Register variable aliases.
2929 (sem_item_optimizer::fixup_pt_set): New function.
2930 (sem_item_optimizer::fixup_points_to_sets): Likewise.
2931 * ipa-icf.h: Declare new variables and functions.
2933 2018-03-13 Jakub Jelinek <jakub@redhat.com>
2936 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
2937 integer_pow2p@2 and test integer_pow2p in condition.
2938 (A < 0 ? C : 0): Similarly for @1.
2941 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
2942 characters starting at p contain '\0' character, don't look beyond
2946 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
2947 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
2950 * reg-stack.c (change_stack): Change update_end var from int to
2951 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
2952 also call set_block_for_insn on the newly added insns and rescan.
2955 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
2956 on the last operand.
2959 * tree.c (stabilize_reference_1): Return save_expr (e) for
2960 STATEMENT_LIST even if it doesn't have side-effects.
2962 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
2964 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
2966 2018-03-12 Renlin Li <renlin.li@arm.com>
2968 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
2969 aarch64_output_scalar_simd_mov_immediate.
2971 2018-03-12 Martin Sebor <msebor@redhat.com>
2973 PR tree-optimization/83456
2974 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
2975 for perfectly overlapping calls to memcpy.
2976 (gimple_fold_builtin_memory_chk): Same.
2977 (gimple_fold_builtin_strcpy): Handle no-warning.
2978 (gimple_fold_builtin_stxcpy_chk): Same.
2979 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
2981 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2983 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
2984 parameter. Use it for SFmode.
2985 (rs6000_function_arg_advance_1): Adjust.
2986 (rs6000_function_arg): Adjust.
2987 (rs6000_gimplify_va_arg): Pass false for that new parameter.
2989 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2991 PR rtl-optimization/84169
2992 PR rtl-optimization/84780
2993 * combine.c (can_combine_p): Check for a 2-insn combination whether
2994 the destination register is used between the two insns, too.
2996 2018-03-12 Richard Biener <rguenther@suse.de>
2998 PR tree-optimization/84803
2999 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
3000 for refs DR analysis didn't process.
3002 2018-03-12 Richard Biener <rguenther@suse.de>
3004 PR tree-optimization/84777
3005 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
3006 force-vectorize loops ignore whether we are optimizing for size.
3008 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
3010 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
3011 (TARGET_MD_ASM_ADJUST): Define.
3013 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
3014 Kito Cheng <kito.cheng@gmail.com>
3015 Chung-Ju Wu <jasonwucj@gmail.com>
3017 * config/nds32/nds32.c (nds32_compute_stack_frame,
3018 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
3019 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
3020 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
3021 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
3022 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
3023 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
3024 * config/nds32/nds32.md (prologue, epilogue): Use macro
3025 NDS32_V3PUSH_AVAILABLE_P to do checking.
3027 2018-03-11 Jakub Jelinek <jakub@redhat.com>
3030 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
3031 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
3032 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
3033 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
3034 addition of most attributes on !orig_type_die or the attribute not
3035 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
3037 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3038 Chung-Ju Wu <jasonwucj@gmail.com>
3040 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
3042 * config/nds32/nds32.opt (mvh): New option.
3044 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3045 Chung-Ju Wu <jasonwucj@gmail.com>
3047 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
3049 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
3050 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
3053 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3054 Chung-Ju Wu <jasonwucj@gmail.com>
3056 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
3058 * config/nds32/nds32-multiple.md (strlensi): New pattern.
3059 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
3061 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
3062 Kito Cheng <kito.cheng@gmail.com>
3063 Chung-Ju Wu <jasonwucj@gmail.com>
3065 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
3066 UNSPEC_FFMISM and UNSPEC_FLMISM.
3067 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
3068 for ffb, ffmism and flmism.
3069 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
3070 (unspec_ffmism): Ditto.
3071 (unspec_flmism): Ditto.
3072 (nds32_expand_builtin_impl): Check if string extension is available.
3073 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
3074 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
3076 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
3079 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3082 * lra-assigns.c (assign_by_spills): Return a flag of reload
3083 assignment failure. Do not process the reload assignment
3084 failures. Do not spill other reload pseudos if they has the same
3086 (lra_assign): Add a return arg. Set up from the result of
3087 assign_by_spills call.
3088 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3089 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3090 usage_insns if it is not NULL.
3091 (spill_hard_reg_in_range): New function.
3092 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3093 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3094 function prototypes.
3095 (lra_assign): Change prototype.
3096 * lra.c (lra): Add code to deal with fails by splitting hard reg
3099 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
3102 * config/i386/i386.opt: Replace Enforcment with Enforcement.
3104 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
3107 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
3108 (dw_val_node): Add val_symbolic_view.
3109 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
3110 (symview_upper_bound): New.
3111 (new_line_info_table): Initialize symviews_since_reset.
3112 (dwarf2out_source_line): Count symviews_since_reset and set
3113 symview_upper_bound.
3114 (dw_val_equal_p): Handle symview.
3115 (add_AT_symview): New.
3116 (print_dw_val): Handle symview.
3117 (attr_checksum, attr_checksum_ordered): Likewise.
3118 (same_dw_val_p, size_of_die): Likewise.
3119 (value_format, output_die): Likewise.
3120 (add_high_low_attributes): Use add_AT_symview for entry_view.
3121 (dwarf2out_finish): Reset symview_upper_bound, clear
3124 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
3127 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
3128 Add strict argument and use it.
3129 (rs6000_split_multireg_move): Update for new strict argument.
3130 (mem_operand_gpr): Disallow all non-offsettable addresses.
3131 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
3133 2018-03-09 Jakub Jelinek <jakub@redhat.com>
3136 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
3137 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
3138 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
3141 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
3142 decl, use remap_type if we want to use the type.
3144 2018-03-09 Martin Sebor <msebor@redhat.com>
3146 PR tree-optimization/84526
3147 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3149 (builtin_access::generic_overlap): Be prepared to handle non-array
3152 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
3154 PR rtl-optimization/84682
3155 * lra-constraints.c (process_address_1): Check is_address flag
3156 for address constraints.
3157 (process_alt_operands): Likewise.
3158 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
3159 preprocess_constraints.
3160 * recog.h (preprocess_constraints): Add oploc parameter.
3162 * recog.c (preprocess_constraints): Test address_operand for
3163 CT_ADDRESS constraints.
3165 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3168 * lra-assigns.c (assign_by_spills): Return a flag of reload
3169 assignment failure. Do not process the reload assignment
3170 failures. Do not spill other reload pseudos if they has the same
3172 (lra_assign): Add a return arg. Set up from the result of
3173 assign_by_spills call.
3174 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3175 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3176 usage_insns if it is not NULL.
3177 (spill_hard_reg_in_range): New function.
3178 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3179 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3180 function prototypes.
3181 (lra_assign): Change prototype.
3182 * lra.c (lra): Add code to deal with fails by splitting hard reg
3185 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3188 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
3189 Accept complain bool parameter. Only emit errors if it is true.
3190 (arm_parse_cpu_option_name): Likewise.
3191 (arm_target_thumb_only): Adjust callers of the above.
3192 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
3193 prototype to take a default true bool parameter.
3194 (arm_parse_arch_option_name): Likewise.
3196 2018-03-09 David Malcolm <dmalcolm@redhat.com>
3197 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
3201 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
3202 * configure: Regenerate.
3203 * configure.ac ("linker --version-script option"): New.
3204 ("linker soname option"): New.
3206 2018-03-09 Richard Biener <rguenther@suse.de>
3208 PR tree-optimization/84775
3209 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
3210 immediate uses of predicate stmts and mark them modified.
3213 PR tree-optimization/84178
3214 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3216 (version_loop_for_if_conversion): Delay update_ssa call.
3217 (tree_if_conversion): Delay update_ssa until after predicate
3220 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
3223 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
3224 when the function accesses prior frames.
3226 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3229 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
3230 gen_llsym, otherwise call maybe_gen_llsym.
3233 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
3234 has ',' character inside of it.
3236 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3239 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
3240 as clobbering CC_REGNUM.
3242 2018-03-08 Richard Biener <rguenther@suse.de>
3245 * tree-scalar-evolution.c: Include tree-into-ssa.h.
3246 (follow_copies_to_constant): Do not follow SSA names registered
3249 2018-03-08 Richard Biener <rguenther@suse.de>
3251 PR tree-optimization/84178
3252 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3254 (version_loop_for_if_conversion): Delay update_ssa call.
3255 (tree_if_conversion): Delay update_ssa until after predicate
3258 2018-03-08 David Malcolm <dmalcolm@redhat.com>
3260 PR tree-optimization/84178
3261 * tree-if-conv.c (release_bb_predicate): Remove the
3262 the assertion that the stmts have NULL use_ops.
3263 Discard the statements, asserting that they haven't
3264 yet been added to a BB.
3266 2018-03-08 Richard Biener <rguenther@suse.de>
3268 PR tree-optimization/84746
3269 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
3270 (phi_translate): Pass in destination ANTIC_OUT set.
3271 (phi_translate_1): Likewise. For a simplified result lookup
3272 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
3273 (phi_translate_set): Adjust.
3274 (do_pre_regular_insertion): Likewise.
3275 (do_pre_partial_partial_insertion): Likewise.
3277 2018-03-08 Martin Liska <mliska@suse.cz>
3279 PR gcov-profile/84735
3280 * doc/gcov.texi: Document usage of profile files.
3281 * gcov-io.h: Document changes in the format.
3283 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
3287 * dwarf2out.c (struct dw_line_info_table): Update comments for
3289 (FORCE_RESET_NEXT_VIEW): New.
3290 (FORCE_RESETTING_VIEW_P): New.
3291 (RESETTING_VIEW_P): Check for -1 too.
3292 (ZERO_VIEW_P): Likewise.
3293 (new_line_info_table): Force-reset next view.
3294 (dwarf2out_begin_function): Likewise.
3295 (dwarf2out_source_line): Simplify zero_view_p initialization.
3296 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
3297 view directly. Omit view when omitting .loc at line 0.
3299 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3301 PR tree-optimization/84740
3302 * tree-switch-conversion.c (process_switch): Call build_constructors
3303 only if info.phi_count is non-zero.
3305 PR tree-optimization/84739
3306 * tree-tailcall.c (find_tail_calls): Check call arguments against
3307 DECL_ARGUMENTS (current_function_decl) rather than
3308 DECL_ARGUMENTS (func) when checking for tail recursion.
3310 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3312 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
3313 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
3314 Volker Reichelt's entry and add entries for people that perform
3315 GCC fuzzy testing and report numerous bugs.
3317 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
3320 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
3321 readonly data in sdata, if that is disabled.
3322 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
3323 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3324 -mreadonly-in-sdata option.
3326 2018-03-07 Martin Sebor <msebor@redhat.com>
3328 PR tree-optimization/84468
3329 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
3330 basic block when looking for nul assignment.
3332 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
3335 * except.h (output_function_exception_table): Adjust prototype.
3336 * except.c (output_function_exception_table): Remove FNNAME parameter
3337 and add SECTION parameter. Ouput one part of the table at a time.
3338 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
3339 the first part of the exception table and emit unwind directives.
3340 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
3341 (i386_pe_seh_cold_init): Likewise.
3342 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
3343 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
3344 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
3345 (ix86_output_call_insn): Emit a nop in one more case for SEH.
3346 * config/i386/winnt.c: Include except.h.
3347 (struct seh_frame_state): Add reg_offset, after_prologue and
3348 in_cold_section fields.
3349 (i386_pe_seh_end_prologue): Set seh->after_prologue.
3350 (i386_pe_seh_cold_init): New function.
3351 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
3352 to seh->in_cold_section.
3353 (seh_emit_push): Record the offset of the push.
3354 (seh_emit_save): Record the offet of the save.
3355 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
3356 Test seh->after_prologue to disregard the epilogue.
3357 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
3358 (i386_pe_end_cold_function): New function.
3360 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3363 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
3364 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
3367 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
3369 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
3370 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
3373 * multiple_target.c: Include tree-inline.h and intl.h.
3374 (expand_target_clones): Diagnose and fail if node->definition and
3375 !tree_versionable_function_p (node->decl).
3377 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
3379 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
3381 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
3382 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
3383 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
3385 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3388 * combine.c (try_combine): Use reg_or_subregno instead of handling
3389 just paradoxical SUBREGs and REGs.
3391 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
3393 * config/arc/arc.c (arc_finalize_pic): Remove function.
3394 (arc_must_save_register): We use single base PIC register, remove
3395 checks to save/restore the PIC register.
3396 (arc_expand_prologue): Likewise.
3397 * config/arc/arc-protos.h (arc_set_default_type_attributes):
3399 (arc_verify_short): Likewise.
3400 (arc_attr_type): Likewise.
3401 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
3402 (walk_stores): Likewise.
3403 (arc_address_cost): Make it static.
3404 (arc_verify_short): Likewise.
3405 (branch_dest): Likewise.
3406 (arc_attr_type): Likewise.
3407 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
3408 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
3409 (arc_final_prescan_insn): Remove inserting the nops due to
3410 hardware hazards. It is done in reorg step.
3411 (insn_length_variant_t): Remove.
3412 (insn_length_parameters_t): Likewise.
3413 (arc_insn_length_parameters): Likewise.
3414 (arc_get_insn_variants): Likewise.
3415 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
3417 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3420 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
3423 PR tree-optimization/84687
3424 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
3426 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
3428 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3430 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
3431 Rename to ppc_speculation_barrier.
3432 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
3433 __builtin_ppc_speculation_barrier.
3435 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3438 * combine.c (combine_simplify_rtx): Don't try to simplify if
3439 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
3442 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
3444 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
3445 to 32 bytes when compiling for POWER9.
3447 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3450 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
3451 regparm >= 3 with no arg reg available also for calls with
3452 flag_force_indirect_call. Pass decl to ix86_function_regparm.
3455 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
3457 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
3459 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
3462 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
3464 2018-03-05 Richard Biener <rguenther@suse.de>
3466 PR tree-optimization/84486
3467 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
3468 When inserting a __builtin_assume_aligned call set the LHS
3469 SSA name alignment info accordingly.
3471 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
3473 PR tree-optimization/84114
3474 * config/aarch64/aarch64.c (aarch64_reassociation_width)
3475 Avoid reassociation of FLOAT_MODE addition.
3477 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
3479 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
3480 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
3481 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
3482 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
3483 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
3484 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
3485 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
3487 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
3488 __builtin_ia32_wbinvd): New builtins.
3489 (SPECIAL_ARGS2): New.
3490 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
3491 (SPECIAL_ARGS2): New.
3492 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
3493 (ix86_valid_target_attribute_inner_p): Ditto.
3494 (ix86_init_mmx_sse_builtins): Add special_args2.
3495 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
3496 TARGET_WBNOINVD_P): New.
3497 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
3498 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
3499 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
3500 * config/i386/immintrin.h (_wbinvd): New intrinsic.
3501 * config/i386/pconfigintrin.h: New file.
3502 * config/i386/wbnoinvdintrin.h: Ditto.
3503 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
3504 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
3506 2018-03-05 Richard Biener <rguenther@suse.de>
3508 PR tree-optimization/84670
3509 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3511 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3512 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3513 (compute_antic_aux): Only assert the number of values in ANTIC_IN
3514 doesn't grow if all successors (recursively) were visited at least
3517 2018-03-05 Richard Biener <rguenther@suse.de>
3519 PR tree-optimization/84650
3520 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
3521 if executed in the loop pipeline.
3523 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
3525 * doc/configfiles.texi (Configuration Files): Move info about
3526 conditionalizing $target-protos.h to...
3527 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
3528 differs from $target-protos.h.
3530 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
3531 Chung-Ju Wu <jasonwucj@gmail.com>
3533 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
3534 * config/nds32/nds32-multiple.md (setmemsi): Define.
3535 * config/nds32/nds32-memory-manipulation.c
3536 (nds32_gen_dup_4_byte_to_word_value): New.
3537 (emit_setmem_word_loop): New.
3538 (emit_setmem_byte_loop): New.
3539 (nds32_expand_setmem_loop): New.
3540 (nds32_expand_setmem_loop_v3m): New.
3541 (nds32_expand_setmem_unroll): New.
3542 (nds32_expand_setmem): New.
3544 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3545 Chung-Ju Wu <jasonwucj@gmail.com>
3547 * config/nds32/nds32-memory-manipulation.c
3548 (nds32_emit_load_store): New.
3549 (nds32_emit_post_inc_load_store): New.
3550 (nds32_emit_mem_move): New.
3551 (nds32_emit_mem_move_block): New.
3552 (nds32_expand_movmemsi_loop_unknown_size): New.
3553 (nds32_expand_movmemsi_loop_known_size): New.
3554 (nds32_expand_movmemsi_loop): New.
3555 (nds32_expand_movmemsi_unroll): New.
3556 (nds32_expand_movmemqi): Rename ...
3557 (nds32_expand_movmemsi): ... to this.
3558 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
3559 (movmemsi): ... to this.
3560 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
3561 (nds32_expand_movmemsi): ... to this.
3563 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3564 Monk Chiang <sh.chiang04@gmail.com>
3565 Chung-Ju Wu <jasonwucj@gmail.com>
3567 * config/nds32/nds32-protos.h
3568 (nds32_expand_load_multiple): New arguments.
3569 (nds32_expand_store_multiple): Ditto.
3570 (nds32_valid_multiple_load_store): Rename ...
3571 (nds32_valid_multiple_load_store_p): ... to this.
3572 * config/nds32/nds32-memory-manipulation.c
3573 (nds32_expand_load_multiple): Refine implementation.
3574 (nds32_expand_store_multiple): Ditto.
3575 * config/nds32/nds32-multiple.md
3576 (load_multiple): Update nds32_expand_load_multiple interface.
3577 (store_multiple): Update nds32_expand_store_multiple interface.
3578 * config/nds32/nds32-predicates.c
3579 (nds32_valid_multiple_load_store): Rename ...
3580 (nds32_valid_multiple_load_store_p): ... to this and refine
3582 * config/nds32/predicates.md
3583 (nds32_load_multiple_and_update_address_operation): New predicate.
3584 (nds32_store_multiple_and_update_address_operation): New predicate.
3586 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3587 Chung-Ju Wu <jasonwucj@gmail.com>
3589 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
3590 (combo): New attribute.
3591 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
3593 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
3595 * config/nds32/nds32.opt: Change -mcmodel= default value.
3597 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
3598 Monk Chiang <sh.chiang04@gmail.com>
3599 Chung-Ju Wu <jasonwucj@gmail.com>
3601 * config/nds32/constants.md (unspec_element): New enum.
3602 * config/nds32/constraints.md (Umw): New constraint.
3603 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
3604 * config/nds32/nds32-intrinsic.md: Likewise.
3605 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
3606 (nds32_valid_smw_lwm_base_p): New.
3607 (nds32_output_smw_single_word): New.
3608 (nds32_output_lmw_single_word): New.
3609 (nds32_expand_unaligned_load): New.
3610 (nds32_expand_unaligned_store): New.
3611 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
3612 (nds32_output_smw_single_word): Declare.
3613 (nds32_output_lmw_single_word): Declare.
3614 (nds32_expand_unaligned_load): Declare.
3615 (nds32_expand_unaligned_store): Declare.
3616 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
3617 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
3618 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
3619 NDS32_BUILTIN_UASTORE_DW.
3620 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
3623 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
3624 Kito Cheng <kito.cheng@gmail.com>
3625 Chung-Ju Wu <jasonwucj@gmail.com>
3627 * config/nds32/nds32-intrinsic.c
3628 (nds32_expand_builtin_null_ftype_reg): Delete.
3629 (nds32_expand_builtin_reg_ftype_imm): Ditto.
3630 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
3631 (nds32_read_argument): New.
3632 (nds32_legitimize_target): Ditto.
3633 (nds32_legitimize_argument): Ditto.
3634 (nds32_check_constant_argument): Ditto.
3635 (nds32_expand_unop_builtin): Ditto.
3636 (nds32_expand_unopimm_builtin): Ditto.
3637 (nds32_expand_binop_builtin): Ditto.
3638 (nds32_builtin_decl_impl): Ditto.
3639 (builtin_description): Ditto.
3640 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
3641 (nds32_init_builtins_impl): Ditto.
3642 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
3643 (nds32_builtin_decl): New.
3644 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
3645 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
3647 2018-03-02 Jeff Law <law@redhat.com>
3649 * reorg.c (stop_search_p): Handle DEBUG_INSN.
3650 (redundant_insn, fill_simple_delay_slots): Likewise.
3651 (fill_slots_from_thread): Likewise.
3652 * resource.c (mark_referenced_resources): Likewise.
3653 (mark_set_resources, find_dead_or_set_registers): Likewise.
3655 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3657 * substring-locations.h (format_warning_va): Formatting fix for
3659 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
3661 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
3662 * substring-locations.c: Include intl.h.
3663 (format_warning_va): Turned into small wrapper around
3664 format_warning_n_va, renamed to ...
3665 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
3666 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
3668 (format_warning_at_substring_n): New function.
3669 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
3670 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
3671 format_warning_at_substring with just a shorter name instead of
3672 const function pointer.
3673 (fmtwarn_n): New function.
3674 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
3675 appropriate, get rid of all the fmtstr temporaries, move conditionals
3676 with G_() wrapped string literals directly into fmtwarn arguments,
3677 cast dir.len to (int), formatting fixes.
3679 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
3681 * doc/invoke.texi: Remove "Cilk Plus" references.
3683 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3684 Richard Biener <rguenther@suse.de>
3687 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
3688 for error or warning attributes if CALL_FROM_THUNK_P is set.
3691 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3694 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
3695 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
3698 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
3699 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
3701 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
3702 instead of -1U in last predictors element's probability member.
3704 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3707 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3708 arguments if they are comparable.
3710 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3712 PR tree-optimization/84634
3713 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
3714 masks and masked_loop_p with a single loop_masks, making sure it's
3715 null for bb vectorization.
3717 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3719 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3720 (vect_analyze_data_ref_access): Use loop->safe_len rather than
3721 loop->force_vectorize to check whether there is no alias.
3723 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3726 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
3728 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
3730 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
3731 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
3732 instead of a loop around prev_real_insn.
3733 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
3737 * config/i386/i386.c (ix86_print_operand): Use conditional
3738 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
3741 2018-03-02 Richard Biener <rguenther@suse.de>
3743 PR tree-optimization/84427
3744 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
3745 (bitmap_set_subtract_values): Rewrite to handle multiple
3748 (prune_clobbered_mems): Likewise.
3749 (phi_translate): Take edge instead of pred/phiblock.
3750 (phi_translate_1): Likewise.
3751 (phi_translate_set): Likewise. Insert all translated
3752 exprs for a value into the set, keeping possibly multiple
3753 expressions per value.
3754 (compute_antic_aux): Adjust for phi_translate changes.
3755 When intersecting union the expressions and prune those
3756 not in the final value set, keeping possibly multiple
3757 expressions per value. Do not use value-insertion
3758 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
3759 all expressions. Add verification that the value-sets
3760 only shrink during iteration.
3761 (compute_partial_antic_aux): Adjust for the phi_translate changes.
3762 (do_pre_regular_insertion): Likewise.
3763 (do_pre_partial_partial_insertion): Likewise.
3765 2018-03-02 Richard Biener <rguenther@suse.de>
3768 * config/darwin.c (saved_debug_info_level): New static global.
3769 (darwin_asm_lto_start): Disable debug info generation for LTO out.
3770 (darwin_asm_lto_end): Restore debug info generation settings.
3772 2018-03-01 Martin Liska <mliska@suse.cz>
3775 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
3778 2018-03-01 Richard Biener <rguenther@suse.de>
3781 * dwarf2out.c (gen_variable_die): Properly handle late VLA
3782 type annotation with LTO when debug was disabled at compile-time.
3784 2018-03-01 Matthew Fortune <mfortune@gmail.com>
3786 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
3788 (mips_final_postscan_insn): Likewise.
3790 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
3792 PR rtl-optimization/84528
3793 * alias.c (init_alias_target): Add commentary.
3794 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
3795 a unique base value if the frame pointer is not eliminated
3796 to the stack pointer.
3798 2018-03-01 Tom de Vries <tom@codesourcery.com>
3800 PR rtl-optimization/83327
3801 * lra-int.h (hard_regs_spilled_into): Declare.
3802 * lra.c (hard_regs_spilled_into): Define.
3803 (init_reg_info): Init hard_regs_spilled_into.
3804 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
3805 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
3806 (process_bb_lives): Handle hard_regs_spilled_into.
3807 (lra_create_live_ranges_1): Before doing liveness propagation, clear
3808 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
3810 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
3812 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
3813 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
3814 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
3815 * config/rs6000/aix72.h: New file.
3817 2018-02-28 Jakub Jelinek <jakub@redhat.com>
3819 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
3820 instead of warning_at with conditional singular and plural messages
3824 * stor-layout.c (update_alignment_for_field): For
3825 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
3826 && !DECL_PACKED (field), do the alignment update, just use
3827 only desired_align instead of MAX (type_align, desired_align)
3829 (place_field): Don't do known_align < desired_align handling
3830 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
3831 is non-NULL, instead do it after rli->prev_field handling and
3832 only if not within a bitfield word. For DECL_PACKED (field)
3833 use type_align of BITS_PER_UNIT.
3835 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
3837 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
3838 superfluous parentheses and trailing spaces.
3840 2018-02-28 Richard Biener <rguenther@suse.de>
3842 PR tree-optimization/84584
3843 * graphite-scop-detection.c (scop_detection::add_scop): Discard
3844 SCoPs with fake exit edge.
3846 2018-02-28 Martin Liska <mliska@suse.cz>
3849 * timevar.c (timer::print): Fix format to properly print 100%
3852 2018-02-28 Richard Biener <rguenther@suse.de>
3855 * genmatch.c (capture_info::walk_match): Do not mark
3856 captured expressions without operands as expr_p given
3857 they act more like predicates and should be subject to
3858 "lost tail" side-effect preserving.
3860 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
3862 PR rtl-optimization/81611
3863 * auto-inc-dec.c (attempt_change): Move dead note from
3864 mem_insn if it's the next use of regno
3865 (find_address): Take address use of reg holding
3866 non-incremented value. Add parm to limit search to the named
3868 (merge_in_block): Attempt to use a mem insn that is the next
3869 use of the original regno.
3871 2018-02-27 Martin Sebor <msebor@redhat.com>
3874 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
3875 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
3877 2018-02-27 Martin Sebor <msebor@redhat.com>
3879 PR translation/84207
3880 * diagnostic-core.h (warning_n, error_n, inform_n): Change
3881 n argument to unsigned HOST_WIDE_INT.
3882 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
3883 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
3884 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
3885 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
3887 2018-02-27 Richard Biener <rguenther@suse.de>
3889 PR tree-optimization/84512
3890 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
3891 Do not use the estimate returned from record_stmt_cost for
3892 the scalar iteration cost but sum properly using add_stmt_cost.
3894 2018-02-27 Richard Biener <rguenther@suse.de>
3896 PR tree-optimization/84466
3897 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3898 Adjust last change to less strictly validate use operands.
3900 2018-02-27 Martin Liska <mliska@suse.cz>
3902 PR gcov-profile/84548
3903 * gcov.c (process_file): Allow partial overlap and consider it
3904 also as group functions.
3905 (output_lines): Properly calculate range of lines for a group.
3907 2018-02-27 Martin Liska <mliska@suse.cz>
3909 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
3910 'ggc' suffixes. Change first column width.
3911 (timer::print): Fix formatting of the column.
3913 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
3915 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
3916 preserve inline entry blocks for the sake of debug inline
3917 entry point markers alone.
3918 (remove_unused_locals): Suggest in comments a better place to
3919 force the preservation of inline entry blocks that are
3920 otherwise unused, but do not preserve them.
3922 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3924 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
3926 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3929 * config/i386/constraints.md (Bs): Replace
3930 ix86_indirect_branch_register with
3931 TARGET_INDIRECT_BRANCH_REGISTER.
3933 * config/i386/i386.md (indirect_jump): Likewise.
3934 (tablejump): Likewise.
3935 (*sibcall_memory): Likewise.
3936 (*sibcall_value_memory): Likewise.
3937 Peepholes of indirect call and jump via memory: Likewise.
3938 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
3939 (*sibcall_value_GOT_32): Likewise.
3940 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3941 (GOT_memory_operand): Likewise.
3942 (call_insn_operand): Likewise.
3943 (sibcall_insn_operand): Likewise.
3944 (GOT32_symbol_operand): Likewise.
3945 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
3947 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3949 PR rtl-optimization/83496
3950 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
3951 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
3952 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
3953 redundant insn, if any.
3954 (relax_delay_slots): Likewise.
3955 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
3957 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
3959 PR tree-optimization/83965
3960 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
3961 that grouped statements are part of a reduction chain. Return
3962 true if the statement is not marked as a reduction itself but
3964 (vect_recog_dot_prod_pattern): Don't check whether the statement
3965 is part of a group here.
3966 (vect_recog_sad_pattern): Likewise.
3967 (vect_recog_widen_sum_pattern): Likewise.
3969 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3972 * final.c (rest_of_clean_state): Also look for calls inside sequences.
3974 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3977 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
3979 (ix86_output_indirect_function_return): New prototype.
3980 (ix86_split_simple_return_pop_internal): Likewise.
3981 * config/i386/i386.c (indirect_return_via_cx): New.
3982 (indirect_return_via_cx_bnd): Likewise.
3983 (indirect_thunk_name): Handle return va CX_REG.
3984 (output_indirect_thunk_function): Create alias for
3985 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
3986 (ix86_output_indirect_jmp): Remove the bool argument.
3987 (ix86_output_indirect_function_return): New function.
3988 (ix86_split_simple_return_pop_internal): Likewise.
3989 * config/i386/i386.md (*indirect_jump): Don't pass false
3990 to ix86_output_indirect_jmp.
3991 (*tablejump_1): Likewise.
3992 (simple_return_pop_internal): Change it to define_insn_and_split.
3993 Call ix86_split_simple_return_pop_internal to split it for
3995 (simple_return_indirect_internal): Call
3996 ix86_output_indirect_function_return instead of
3997 ix86_output_indirect_jmp.
3999 2018-02-26 Jakub Jelinek <jakub@redhat.com>
4002 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
4003 memset and value initialization afterwards.
4005 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
4007 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
4009 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4012 * common/config/aarch64/aarch64-common.c
4013 (aarch_option_optimization_table[]): Switch
4014 off fomit-frame-pointer
4016 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4017 Chung-Ju Wu <jasonwucj@gmail.com>
4019 * config/nds32/nds32-multiple.md (load_multiple): Disallow
4021 (store_multiple): Ditto.
4023 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4025 * config.gcc: Add --with-cpu support for nds32 target.
4026 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
4027 * config/nds32/nds32.opt: Add -mcpu= option.
4029 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
4031 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
4032 isel=yes): Warn for these deprecated options.
4034 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
4036 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
4037 ISA_2_5_MASKS_EMBEDDED.
4039 2018-02-23 Jakub Jelinek <jakub@redhat.com>
4041 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
4042 p->max as pointers rather than using iterative_hash_expr.
4044 2018-02-23 Carl Love <cel@us.ibm.com>
4046 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
4047 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
4049 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
4050 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
4051 P8V_BUILTIN_VEC_VUNSIGNED2.
4053 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
4056 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
4057 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
4059 * lra-constraints.c (curr_insn_transform): Set up
4060 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
4062 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
4063 * lra-eliminations.c (spill_pseudos): Ditto.
4064 (process_insn_for_elimination): Ditto.
4065 * lra-lives.c (reg_early_clobber_p): Use the new macros.
4066 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
4067 LRA_NON_CLOBBERED_ALT.
4069 2018-02-22 Martin Sebor <msebor@redhat.com>
4071 PR tree-optimization/84480
4072 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
4073 to maybe_diag_stxncpy_trunc. Call it.
4074 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
4075 from gimple_fold_builtin_strcpy. Print inlining stack.
4076 (handle_builtin_stxncpy): Print inlining stack.
4077 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
4079 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
4082 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
4083 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
4084 and -fcheck-pointer-bounds are used together.
4085 (indirect_thunk_prefix): New enum.
4086 (indirect_thunk_need_prefix): New function.
4087 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
4088 "_nt" instead of "_bnd" for NOTRACK prefix.
4089 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
4090 (output_indirect_thunk_function): Likewise.
4092 (ix86_code_end): Update output_indirect_thunk_function calls.
4093 (ix86_output_indirect_branch_via_reg): Replace
4094 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
4095 (ix86_output_indirect_branch_via_push): Likewise.
4096 (ix86_output_function_return): Likewise.
4097 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
4098 incompatible with -fcf-protection=branch and
4099 -fcheck-pointer-bounds.
4101 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4104 * config/aarch64/aarch64.c (aarch64_print_address_internal):
4105 Change gcc_assert call to output_operand_lossage.
4107 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4109 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
4111 2018-02-22 DJ Delorie <dj@redhat.com>
4112 Sebastian Perta <sebastian.perta@renesas.com>
4113 Oleg Endo <olegendo@gcc.gnu.org>
4115 * config/rx/rx.c (rx_rtx_costs): New function.
4116 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
4118 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
4120 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
4122 2018-02-22 Martin Liska <mliska@suse.cz>
4125 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
4126 Add "native" as a possible value.
4128 2018-02-22 Martin Liska <mliska@suse.cz>
4131 * config/i386/i386.c (ix86_option_override_internal):
4132 Add "native" as a possible value for -march and -mtune.
4134 2018-02-22 Jakub Jelinek <jakub@redhat.com>
4137 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
4138 to all type variants.
4140 PR tree-optimization/84503
4141 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
4142 width as info->bitpos + info->bitsize - start.
4143 (merged_store_group::merge_overlapping): Simplify width computation.
4144 (check_no_overlap): New function.
4145 (imm_store_chain_info::try_coalesce_bswap): Compute expected
4146 start + width and last_order of the group, fail if check_no_overlap
4148 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
4149 to group if check_no_overlap fails.
4151 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4153 * config/rs6000/altivec.md: Delete contraint arguments to
4154 define_expand, define_split, and define_peephole2, and in
4155 define_insn_and_split if always unused.
4156 * config/rs6000/darwin.md: Ditto.
4157 * config/rs6000/dfp.md: Ditto.
4158 * config/rs6000/rs6000.md: Ditto.
4159 * config/rs6000/sync.md: Ditto.
4160 * config/rs6000/vector.md: Ditto.
4161 * config/rs6000/vsx.md: Ditto.
4163 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4165 * config/rs6000/altivec.md: Write output control strings as braced
4166 blocks instead of double-quoted strings.
4167 * config/rs6000/darwin.md: Ditto.
4168 * config/rs6000/rs6000.md: Ditto.
4169 * config/rs6000/vector.md: Ditto.
4170 * config/rs6000/vsx.md: Ditto.
4172 2018-02-21 Jason Merrill <jason@redhat.com>
4174 PR c++/84314 - ICE with templates and fastcall attribute.
4175 * attribs.c (build_type_attribute_qual_variant): Remove assert.
4177 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4179 * ipa-cp.c (determine_versionability): Fix comment typos.
4181 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4184 * ipa-cp.c (determine_versionability): Do not version functions caling
4187 2018-02-21 Martin Liska <mliska@suse.cz>
4190 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
4191 Add "native" as a possible value.
4192 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
4193 the macro when native cpu detection is available.
4195 2018-02-21 Martin Liska <mliska@suse.cz>
4198 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
4199 Add "native" as a possible value.
4200 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
4201 when native cpu detection is available.
4203 2018-02-21 Jakub Jelinek <jakub@redhat.com>
4204 Martin Sebor <msebor@redhat.com>
4206 PR tree-optimization/84478
4207 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
4209 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
4210 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
4211 support which is conservatively correct, for 2 only stay conservative
4212 for maxlen. Formatting and comment capitalization fixes. Add STRICT
4213 argument to the 2 argument get_range_strlen, adjust 6 arg
4214 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
4216 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
4217 (gimple_fold_builtin_strlen): Pass true as last argument to
4220 2018-02-20 Martin Sebor <msebor@redhat.com>
4223 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
4224 (builtin_memref::set_base_and_offset): Same. Handle inner references.
4225 (builtin_memref::builtin_memref): Factor out parts into
4226 set_base_and_offset and call it.
4228 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4231 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
4232 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
4233 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
4234 search at the associated MODE_INT.
4236 2018-02-20 Jeff Law <law@redhat.com>
4239 PR tree-optimization/81592
4241 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
4242 for range data rather than using global data.
4243 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
4244 range data rather than using global data.
4245 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
4246 pass it to children as needed.
4247 (struct directive::fmtresult): Similarly.
4248 (struct directive::set_width): Similarly.
4249 (struct directive::set_precision): Similarly.
4250 (format_integer, format_directive, parse_directive): Similarly.
4251 (format_none): Accept unnamed vr_values parameter.
4252 (format_percent, format_floating, format_character): Similarly.
4253 (format_string, format_plain): Similarly.
4254 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
4255 the EVRP range analyzer for range data rather than using global data.
4256 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
4257 gimple-ssa-evrp-analyze.h
4258 (class sprintf_dom_walker): Add after_dom_children member function.
4259 Add evrp_range_analyzer member.
4260 (sprintf_dom_walker::before_dom_children): Call into the EVRP
4261 range analyzer as needed.
4262 (sprintf_dom_walker::after_dom_children): New member function.
4263 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
4265 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
4266 (evrp_range_analyzer::pop_to_marker): Likewise.
4268 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4270 PR tree-optimization/84419
4271 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
4272 with the required type if its current type is compatible but
4275 2018-02-20 Jakub Jelinek <jakub@redhat.com>
4278 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
4279 after vectorization.
4281 2018-02-20 Martin Liska <mliska@suse.cz>
4284 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
4285 possible values if we don't have a hint.
4287 2018-02-20 Martin Liska <mliska@suse.cz>
4291 * final.c (shorten_branches): Build align_tab array with one
4293 * opts.c (finish_options): Add alignment option limit check.
4294 (MAX_CODE_ALIGN): Likewise.
4295 (MAX_CODE_ALIGN_VALUE): Likewise.
4296 * doc/invoke.texi: Document maximum allowed option value for
4297 all -falign-* options.
4299 2018-02-19 Jakub Jelinek <jakub@redhat.com>
4302 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
4303 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
4304 * var-tracking.c (emit_note_insn_var_location): Remove all references
4305 to NOTE_INSN_CALL_ARG_LOCATION.
4306 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
4307 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
4308 Use copy_rtx_if_shared.
4309 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
4310 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
4311 (dwarf2out_var_location): Remove handling of
4312 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
4314 * final.c (final_scan_insn): Remove all references to
4315 NOTE_INSN_CALL_ARG_LOCATION.
4316 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
4317 before dumping final insns.
4318 * except.c (emit_note_eh_region_end): Remove all references to
4319 NOTE_INSN_CALL_ARG_LOCATION.
4320 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
4321 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
4322 * config/arc/arc.c (hwloop_optimize): Likewise.
4323 * config/arm/arm.c (create_fix_barrier): Likewise.
4324 * config/s390/s390.c (s390_chunkify_start): Likewise.
4325 * config/sh/sh.c (find_barrier): Likewise.
4326 * config/i386/i386.c (rest_of_insert_endbranch,
4327 ix86_seh_fixup_eh_fallthru): Likewise.
4328 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
4329 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
4330 * config/frv/frv.c (frv_function_prologue): Likewise.
4331 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
4333 (note_outside_basic_block_p): Remove all references to
4334 NOTE_INSN_CALL_ARG_LOCATION.
4335 * gengtype.c (adjust_field_rtx_def): Likewise.
4336 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
4338 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
4339 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
4342 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
4345 PR tree-optimization/84452
4346 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
4347 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
4350 2018-02-19 Martin Liska <mliska@suse.cz>
4353 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
4355 2018-02-19 Martin Liska <mliska@suse.cz>
4356 Richard Sandiford <richard.sandiford@linaro.org>
4358 PR tree-optimization/82491
4359 * gimple-fold.c (get_base_constructor): Make earlier bail out
4362 2018-02-19 Carl Love <cel@us.ibm.com>
4364 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
4365 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
4367 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
4368 P8V_BUILTIN_VEC_NEG.
4370 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
4372 * config/rl78/rl78.md (movdf): New define expand.
4374 2018-02-19 Martin Liska <mliska@suse.cz>
4377 * doc/invoke.texi: Fix typo.
4378 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
4380 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
4382 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
4383 handle rs6000_single_float and rs6000_double_float specially for
4386 2018-02-16 Jeff Law <law@redhat.com>
4388 * config/rx/rx.c (add_pop_cfi_notes): New function.;
4391 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4394 * ipa-inline.c (inline_small_functions): Fix a typo.
4396 2018-02-16 Nathan Sidwell <nathan@acm.org>
4398 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
4400 2018-02-16 Carl Love <cel@us.ibm.com>
4402 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
4403 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
4404 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
4405 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
4406 expansion to P8V_BUILTIN_VEC_FLOAT2.
4408 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
4410 PR rtl-optimization/70023
4411 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
4412 src_regno into account.
4414 2018-02-16 Carl Love <cel@us.ibm.com>
4416 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
4417 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
4418 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
4419 * config/rs6000/rs6000.c: Remove case statements for
4420 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
4421 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
4422 and P9V_BUILTIN_VEC_VINSERT4B.
4423 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
4424 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
4425 * config/rs6000/vsx.md:
4426 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
4429 2018-02-16 Carl Love <cel@us.ibm.com>
4431 * config/rs6000/altivec.h: Add builtin names vec_extract4b
4433 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
4435 * config/rs6000/rs6000-c.c: Add the definitions for
4436 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
4437 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
4438 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
4439 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
4440 definition for insert4b and define insn *insert3b_internal.
4441 * doc/extend.texi: Add documentation for vec_extract4b.
4443 2018-02-16 Nathan Sidwell <nathan@acm.org>
4445 * doc/extend.texi (Backwards Compatibility): Mention friend
4446 injection. Note for-scope is deprecated.
4447 * doc/invoke.texi (-ffriend-injection): Deprecate.
4449 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
4451 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
4452 that moved to I2, also allow destinations that are a paradoxical
4453 subreg (instead of a normal reg).
4455 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
4458 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
4461 2018-02-16 Richard Biener <rguenther@suse.de>
4463 PR tree-optimization/84037
4464 PR tree-optimization/84016
4466 * config/i386/i386.c (ix86_builtin_vectorization_cost):
4467 Adjust vec_construct for the fact we need additional higher latency
4468 128bit inserts for AVX256 and AVX512 vector builds.
4469 (ix86_add_stmt_cost): Scale vector construction cost for
4472 2018-02-16 Richard Biener <rguenther@suse.de>
4474 PR tree-optimization/84417
4475 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
4476 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
4477 (non_rewritable_lvalue_p): Likewise, use poly-ints.
4479 2018-02-16 Martin Liska <mliska@suse.cz>
4482 * internal-fn.def (ASAN_CHECK): Set proper flags.
4483 (ASAN_MARK): Likewise.
4485 2018-02-16 Julia Koval <julia.koval@intel.com>
4487 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
4488 from PTA_CANNONLAKE.
4490 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4493 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
4494 Use ++iter rather than iter++ for std::list iterators.
4495 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
4496 defer deleting them until all nodes in the forest are processed. Do
4497 free even leaf nodes. Change to_process into auto_vec.
4500 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
4501 * vec.h (vec_default_construct): Use memset instead of placement new
4502 if BROKEN_VALUE_INITIALIZATION is defined.
4503 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
4504 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
4507 PR rtl-optimization/83723
4508 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
4509 * lra.c (lra_substitute_pseudo): Likewise. If true, use
4510 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
4512 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
4514 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
4516 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
4518 PR rtl-optimization/81443
4519 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
4520 from inner REGs to paradoxical SUBREGs.
4522 2018-02-16 Richard Biener <rguenther@suse.de>
4524 PR tree-optimization/84399
4525 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4526 For operands we can analyze at their definition make sure we can
4527 analyze them at each use as well.
4529 2018-02-16 Richard Biener <rguenther@suse.de>
4531 PR tree-optimization/84190
4532 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
4533 volatile accesses if the decl isn't volatile.
4535 2018-02-15 Jason Merrill <jason@redhat.com>
4537 PR c++/84314 - ICE with templates and fastcall attribute.
4538 * attribs.c (build_type_attribute_qual_variant): Don't clobber
4539 TYPE_CANONICAL on an existing type.
4541 2018-02-15 Jakub Jelinek <jakub@redhat.com>
4543 PR tree-optimization/84383
4544 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
4545 dstoff nor call operand_equal_p if dstbase is NULL.
4547 PR tree-optimization/84334
4548 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
4549 also a CONSTANT_CLASS_P, punt.
4551 2018-02-14 Jim Wilson <jimw@sifive.com>
4553 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
4554 first SMALL_OPERAND check. New local min_second_step. Move assert
4555 to where locals are set. Add TARGET_RVC support.
4556 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
4558 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
4560 * doc/invoke.texi: Correct -Wformat-overflow code sample.
4562 2018-02-14 Martin Sebor <msebor@redhat.com>
4564 PR tree-optimization/83698
4565 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
4566 arrays constrain the offset range to their bounds.
4567 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
4568 (builtin_access::overlap): Avoid setting the size of overlap if it's
4570 (maybe_diag_overlap): Also consider arrays when deciding what values
4571 of offsets to include in diagnostics.
4573 2018-02-14 Martin Sebor <msebor@redhat.com>
4576 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
4577 that correspond to the kind of a declaration.
4579 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
4582 * config/pa/pa.md: Load address of PIC label using the linkage table
4583 if the label is nonlocal.
4585 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
4587 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
4588 warning message if user requests -maltivec=be.
4589 * doc/invoke.texi: Document deprecation of -maltivec=be.
4591 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
4594 * config/rs6000/rs6000-c.c: Update definitions for
4595 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
4596 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
4598 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4601 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
4602 add _get_ssp intrinsics. Remove argument from
4603 __builtin_ia32_rdssp[d|q].
4604 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
4605 * config/i386/i386-builtin.def: Remove argument from
4606 __builtin_ia32_rdssp[d|q].
4607 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
4608 ix86_expand_special_args_builtin for _rdssp[d|q].
4609 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
4610 Clear register before usage.
4611 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
4612 Add documentation for new _get_ssp and _inc_ssp intrinsics.
4614 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
4616 PR tree-optimization/84357
4617 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
4618 operand 1 of an ARRAY_REF too.
4620 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
4623 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
4624 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4626 (set_of_reg): New struct.
4627 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
4628 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
4629 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4631 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
4632 Split into bitclr, bitset, bitinvert patterns if appropriate.
4633 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
4634 use rx_fuse_in_memory_bitop.
4635 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
4636 to named insn, correct maximum insn length.
4638 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
4641 * machmode.def: Define a complex mode for PARTIAL_INT.
4642 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
4644 * doc/rtl.texi: Document CSPImode.
4645 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
4647 (msp430_hard_regno_nregs_with_padding): Likewise.
4649 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
4652 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
4654 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
4656 PR rtl-optimization/84169
4657 * combine.c (try_combine): New variable split_i2i3. Set it to true if
4658 we generated a parallel as new i3 and we split that to new i2 and i3
4659 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
4660 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
4661 those to i2, not i1. Partially rewrite this scan code.
4663 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4666 * stor-layout.c (place_field): For variable length fields, adjust
4667 offset_align afterwards not just based on the field's alignment,
4668 but also on the size.
4671 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
4672 of exps and logs in the use_exp2 case.
4674 2018-02-13 Jeff Law <law@redhat.com>
4676 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
4679 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
4682 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
4686 * common.opt (gas-loc-support, gas-locview-support): New.
4687 (ginline-points, ginternal-reset-location-views): New.
4688 * doc/invoke.texi: Document them. Use @itemx where intended.
4689 (gvariable-location-views): Adjust.
4690 * target.def (reset_location_view): New.
4691 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
4692 (TARGET_RESET_LOCATION_VIEW): New.
4693 * doc/tm.texi: Rebuilt.
4694 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
4695 (dwarf2out_default_as_locview_support): New.
4696 (output_asm_line_debug_info): Use option variables.
4697 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
4698 (output_loc_list): Likewise.
4699 (add_high_low_attributes): Check option variables.
4700 Don't output entry view attribute in strict mode.
4701 (gen_inlined_subroutine_die): Check option variables.
4702 (dwarf2out_inline_entry): Likewise.
4703 (init_sections_and_labels): Likewise.
4704 (dwarf2out_early_finish): Likewise.
4705 (maybe_reset_location_view): New, from...
4706 (dwarf2out_var_location): ... here. Call it.
4707 * debug.h (dwarf2out_default_as_loc_support): Declare.
4708 (dwarf2out_default_as_locview_support): Declare.
4709 * hooks.c (hook_int_rtx_insn_0): New.
4710 * hooks.h (hook_int_rtx_insn_0): Declare.
4711 * toplev.c (process_options): Take -gas-loc-support and
4712 -gas-locview-support from dwarf2out. Enable
4713 -gvariable-location-views by default only with locview
4714 assembler support. Enable -ginternal-reset-location-views by
4715 default only if the target defines the corresponding hook.
4716 Enable -ginline-points by default if location views are
4717 enabled; force it disabled if statement frontiers are
4719 * tree-inline.c (expand_call_inline): Check option variables.
4720 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
4722 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
4724 PR tree-optimization/84321
4725 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
4726 handling. Also check whether the anti-range contains any values
4727 that satisfy the mask; switch to a VR_RANGE if not.
4729 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
4732 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
4734 2018-02-13 Martin Jambor <mjambor@suse.cz>
4737 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
4738 of call statements, also set location of a load to a temporary.
4740 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
4742 * config/rl78/rl78.c (add_vector_labels): New function.
4743 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
4744 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
4745 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
4746 which checks that no arguments are passed.
4747 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
4748 * doc/extend.texi: Documentation for the new attribute.
4750 2018-02-13 Andreas Schwab <schwab@suse.de>
4752 * config/riscv/linux.h (CPP_SPEC): Define.
4754 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4757 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
4758 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
4759 OPTION_MASK_ISA_AES as first argument to def_builtin_const
4760 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
4761 instead of OPTION_MASK_ISA_PCLMUL as first argument to
4762 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
4763 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
4764 temporarily for AES and PCLMUL builtins.
4766 PR tree-optimization/84339
4767 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
4768 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
4772 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
4773 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
4774 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
4776 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
4778 * omp-simd-clone.h: New file.
4779 * omp-simd-clone.c: Include omp-simd-clone.h.
4780 (expand_simd_clones): No longer static.
4781 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
4782 cgraph.h and omp-simd-clone.h.
4783 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
4784 (vect_recog_widen_shift_pattern): Formatting fix.
4785 (vect_pattern_recog_1): Don't check optab for calls.
4788 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
4789 operands[2] into a REG before using gen_lowpart on it.
4791 2018-02-12 Jeff Law <law@redhat.com>
4794 * config/sh/sh.c (find_barrier): Consider a sibling call
4797 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
4798 successfully back substituting a reg.
4800 2018-02-12 Richard Biener <rguenther@suse.de>
4802 PR tree-optimization/84037
4803 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
4804 parameter, move visited init to caller.
4805 (vect_slp_analyze_operations): Separate cost from validity
4806 check, initialize visited once for all instances.
4807 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
4809 * tree-vect-stmts.c (vect_model_simple_cost): Make early
4811 (vect_model_promotion_demotion_cost): Likewise.
4812 (vectorizable_bswap): Guard cost modeling with !slp_node
4813 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
4815 (vectorizable_call): Likewise.
4816 (vectorizable_conversion): Likewise.
4817 (vectorizable_assignment): Likewise.
4818 (vectorizable_shift): Likewise.
4819 (vectorizable_operation): Likewise.
4820 (vectorizable_store): Likewise.
4821 (vectorizable_load): Likewise.
4822 (vectorizable_condition): Likewise.
4823 (vectorizable_comparison): Likewise.
4825 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
4828 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
4829 (ASAN_MARK): Fix fnspec to account for return value, change pointer
4830 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
4832 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4835 * params.def (inline-min-speedup): Increase from 8 to 15.
4836 (max-inline-insns-auto): Decrease from 40 to 30.
4837 * ipa-split.c (consider_split): Add some buffer for function to
4838 be considered inlining candidate.
4839 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
4842 2018-02-12 Richard Biener <rguenther@suse.de>
4844 PR tree-optimization/84037
4845 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
4846 matched stmts if we cannot swap the non-matched ones.
4848 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4850 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
4851 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
4852 _mm_maskz_scalef_round_ss): New intrinsics.
4853 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
4854 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
4855 __builtin_ia32_scalefss_round): Remove.
4856 (__builtin_ia32_scalefsd_mask_round,
4857 __builtin_ia32_scalefss_mask_round): New intrinsics.
4858 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
4859 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
4860 ((match_operand:VF_128 2 "<round_nimm_predicate>"
4861 "<round_constraint>")): Changed to ...
4862 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
4863 "<round_scalar_constraint>")): ... this.
4864 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
4865 %0, %1, %2<round_op3>}"): Changed to ...
4866 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
4867 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
4868 %2<round_scalar_mask_op3>}"): ... this.
4869 * config/i386/subst.md (round_scalar_nimm_predicate): New.
4871 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4873 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
4874 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
4875 (_mm_maskz_sqrt_round_ss): New intrinsics.
4876 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
4877 (__builtin_ia32_sqrtsd_mask_round)
4878 (__builtin_ia32_sqrtss_mask_round): New builtins.
4879 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
4880 (__builtin_ia32_sqrtss_round): Remove.
4881 (__builtin_ia32_sqrtsd_mask_round)
4882 (__builtin_ia32_sqrtss_mask_round): New builtins.
4883 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
4884 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
4885 ((match_operand:VF_128 1 "vector_operand"
4886 "xBm,<round_constraint>")): Changed to ...
4887 ((match_operand:VF_128 1 "vector_operand"
4888 "xBm,<round_scalar_constraint>")): ... this.
4889 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
4890 %0, %2, %<iptr>1<round_op3>}): Changed to ...
4891 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
4892 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
4893 %<iptr>1<round_scalar_mask_op3>}): ... this.
4894 ((set_attr "prefix" "<round_prefix>")): Changed to ...
4895 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
4897 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
4900 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
4901 Cast vec_cmpeq result to correct type.
4902 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
4903 Cast vec_cmpgt result to correct type.
4905 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
4907 * final.c (final_scan_insn_1): Renamed from...
4908 (final_scan_insn): ... this. New wrapper, to recover
4909 seen from the outermost call in recursive ones.
4910 * config/sparc/sparc.c (output_return): Drop seen from call.
4911 (output_sibcall): Likewise.
4912 * config/visium/visium.c (output_branch): Likewise.
4914 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
4916 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
4919 2018-02-10 Alan Modra <amodra@gmail.com>
4922 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
4923 Specify LR as an input.
4925 2018-02-10 Jakub Jelinek <jakub@redhat.com>
4928 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
4929 remove_member_access_dummy_vars): New functions.
4930 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
4931 lower_omp_1, execute_lower_omp): Use them.
4933 PR rtl-optimization/84308
4934 * shrink-wrap.c (spread_components): Release todo vector.
4936 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
4938 PR rtl-optimization/57193
4939 * ira-color.c (struct allocno_color_data): Add member
4940 conflict_allocno_hard_prefs.
4941 (update_conflict_allocno_hard_prefs): New.
4942 (bucket_allocno_compare_func): Add a preference based on
4943 conflict_allocno_hard_prefs.
4944 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
4945 (color_allocnos): Remove a dead code. Initiate
4946 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
4948 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4951 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
4952 constraint from =wa to wa. Avoid a subreg on the output operand,
4953 instead use a pseudo and subreg it in a move.
4954 (p9_xxbrd_<mode>): Changed to ...
4955 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
4956 (p9_xxbrd_v2df): New expander.
4957 (p9_xxbrw_<mode>): Changed to ...
4958 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
4959 (p9_xxbrw_v4sf): New expander.
4961 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4963 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
4965 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
4968 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
4969 multiply in 32-bit mode.
4970 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
4971 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
4974 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4976 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
4977 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
4978 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
4979 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
4981 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
4984 * dwarf2out.c (is_trivial_indirect_ref): New function.
4985 (dwarf2out_late_global_decl): Do not generate a location
4986 attribute for variables that have a non-trivial DECL_VALUE_EXPR
4987 and that are not defined in the current unit.
4989 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4991 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
4992 instead of a libcall for UNORDERED.
4994 2018-02-09 Tamar Christina <tamar.christina@arm.com>
4997 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
4998 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
5000 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5003 * config/s390/s390.c (s390_set_current_function): Invoke
5004 s390_indirect_branch_settings also if fndecl didn't change.
5006 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5008 * config/rs6000/rs6000.md (blockage): Set length to zero.
5010 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5012 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
5014 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5017 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
5018 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
5022 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
5023 PARALLEL incoming that failed vt_get_decl_and_offset check.
5026 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
5027 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
5029 (get_variable_section): For decls in named .bss* sections pass true as
5030 second argument to bss_initializer_p.
5032 2018-02-09 Marek Polacek <polacek@redhat.com>
5033 Jakub Jelinek <jakub@redhat.com>
5036 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
5037 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
5038 Sync some changes from cxx_fold_indirect_ref.
5040 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5042 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
5044 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
5045 (BLOCK_INLINE_ENTRY_LABEL): New.
5046 (dwarf2out_var_location): Disregard inline entry markers.
5047 (inline_entry_data): New struct.
5048 (inline_entry_data_hasher): New hashtable type.
5049 (inline_entry_data_hasher::hash): New.
5050 (inline_entry_data_hasher::equal): New.
5051 (inline_entry_data_table): New variable.
5052 (add_high_low_attributes): Add DW_AT_entry_pc and
5053 DW_AT_GNU_entry_view attributes if a pending entry is found
5054 in inline_entry_data_table. Add old entry_pc attribute only
5055 if debug nonbinding markers are disabled.
5056 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
5057 markers are enabled.
5058 (block_within_block_p, dwarf2out_inline_entry): New.
5059 (dwarf2out_finish): Check that no entries remained in
5060 inline_entry_data_table.
5061 * final.c (reemit_insn_block_notes): Handle inline entry notes.
5062 (final_scan_insn, notice_source_line): Likewise.
5063 (rest_of_clean_state): Skip inline entry markers.
5064 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
5066 * gimple.c (gimple_build_debug_inline_entry): New.
5067 * gimple.h (enum gimple_debug_subcode): Add
5068 GIMPLE_DEBUG_INLINE_ENTRY.
5069 (gimple_build_debug_inline_entry): Declare.
5070 (gimple_debug_inline_entry_p): New.
5071 (gimple_debug_nonbind_marker_p): Adjust.
5072 * insn-notes.def (INLINE_ENTRY): New.
5073 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
5074 inline entry marker notes.
5075 (print_insn): Likewise.
5076 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
5077 (INSN_DEBUG_MARKER_KIND): Likewise.
5078 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
5079 * tree-inline.c (expand_call_inline): Build and insert
5080 debug_inline_entry stmt.
5081 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
5082 inline entry blocks early, if nonbind markers are enabled.
5083 (dump_scope_block): Dump fragment info.
5084 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
5085 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
5086 (gimple_build_debug_inline_entry): New.
5087 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
5088 Enable/disable inline entry points too.
5089 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
5090 (DEBUG_INSN): Describe inline entry markers.
5092 * common.opt (gvariable-location-views): New.
5093 (gvariable-location-views=incompat5): New.
5094 * config.in: Rebuilt.
5095 * configure: Rebuilt.
5096 * configure.ac: Test assembler for view support.
5097 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
5098 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
5099 * dwarf2out.c (var_loc_view): New typedef.
5100 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
5101 (dwarf2out_locviews_in_attribute): New.
5102 (dwarf2out_locviews_in_loclist): New.
5103 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
5104 (enum dw_line_info_opcode): Add LI_adv_address.
5105 (struct dw_line_info_table): Add view.
5106 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
5107 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
5108 (zero_view_p): New variable.
5109 (ZERO_VIEW_P): New macro.
5110 (output_asm_line_debug_info): New.
5111 (struct var_loc_node): Add view.
5112 (add_AT_view_list, AT_loc_list): New.
5113 (add_var_loc_to_decl): Add view param. Test it against last.
5114 (new_loc_list): Add view params. Record them.
5115 (AT_loc_list_ptr): Handle loc and view lists.
5116 (view_list_to_loc_list_val_node): New.
5117 (print_dw_val): Handle dw_val_class_view_list.
5118 (size_of_die): Likewise.
5119 (value_format): Likewise.
5120 (loc_list_has_views): New.
5121 (gen_llsym): Set vl_symbol too.
5122 (maybe_gen_llsym, skip_loc_list_entry): New.
5123 (dwarf2out_maybe_output_loclist_view_pair): New.
5124 (output_loc_list): Output view list or entries too.
5125 (output_view_list_offset): New.
5126 (output_die): Handle dw_val_class_view_list.
5127 (output_dwarf_version): New.
5128 (output_compilation_unit_header): Use it.
5129 (output_skeleton_debug_sections): Likewise.
5130 (output_rnglists, output_line_info): Likewise.
5131 (output_pubnames, output_aranges): Update version comments.
5132 (output_one_line_info_table): Output view numbers in asm comments.
5133 (dw_loc_list): Determine current endview, pass it to new_loc_list.
5134 Call maybe_gen_llsym.
5135 (loc_list_from_tree_1): Adjust.
5136 (add_AT_location_description): Create view list attribute if
5137 needed, check it's absent otherwise.
5138 (convert_cfa_to_fb_loc_list): Adjust.
5139 (maybe_emit_file): Call output_asm_line_debug_info for test.
5140 (dwarf2out_var_location): Reset views as needed. Precompute
5141 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
5142 attribute. Set view.
5143 (new_line_info_table): Reset next view.
5144 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
5145 (dwarf2out_source_line): Likewise. Output view resets and labels to
5146 the assembler, or select appropriate line info opcodes.
5147 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
5148 (optimize_string_length): Catch it. Adjust.
5149 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
5150 dw_val_class_view_list, and remove it if no longer needed.
5151 (hash_loc_list): Hash view numbers.
5152 (loc_list_hasher::equal): Compare them.
5153 (optimize_location_lists): Check whether a view list symbol is
5154 needed, and whether the locview attribute is present, and
5155 whether they match. Remove the locview attribute if no longer
5157 (index_location_lists): Call skip_loc_list_entry for test.
5158 (dwarf2out_finish): Call output_asm_line_debug_info for test.
5159 Use output_dwarf_version.
5160 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
5161 (struct dw_val_node): Add val_view_list.
5162 * final.c (SEEN_NEXT_VIEW): New.
5163 (set_next_view_needed): New.
5164 (clear_next_view_needed): New.
5165 (maybe_output_next_view): New.
5166 (final_start_function): Rename to...
5167 (final_start_function_1): ... this. Take pointer to FIRST,
5168 add SEEN parameter. Emit param bindings in the initial view.
5169 (final_start_function): Reintroduce SEEN-less interface.
5170 (final): Rename to...
5171 (final_1): ... this. Take SEEN parameter. Output final pending
5172 next view at the end.
5173 (final): Reintroduce seen-less interface.
5174 (final_scan_insn): Output pending next view before switching
5175 sections or ending a block. Mark the next view as needed when
5176 outputting variable locations. Notify debug backend of section
5177 changes, and of location view changes.
5178 (rest_of_handle_final): Adjust.
5179 * toplev.c (process_options): Autodetect value for debug variable
5180 location views option. Warn on incompat5 without -gdwarf-5.
5181 * doc/invoke.texi (gvariable-location-views): New.
5182 (gvariable-location-views=incompat5): New.
5183 (gno-variable-location-views): New.
5185 2018-02-08 David Malcolm <dmalcolm@redhat.com>
5187 PR tree-optimization/84136
5188 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
5189 that the result of find_edge is non-NULL.
5191 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5194 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
5195 storing integer register in SImode. Fix cost of 256 and 512
5196 byte aligned SSE register store.
5198 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5200 * config/i386/i386.c (ix86_multiplication_cost): Fix
5201 multiplication cost for TARGET_AVX512DQ.
5203 2018-02-08 Marek Polacek <polacek@redhat.com>
5205 PR tree-optimization/84238
5206 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
5209 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5211 PR tree-optimization/84265
5212 * tree-vect-stmts.c (vectorizable_store): Don't treat
5213 VMAT_CONTIGUOUS accesses as grouped.
5214 (vectorizable_load): Likewise.
5216 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5218 PR tree-optimization/81635
5219 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
5220 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
5221 (test_round_for_mask): New functions.
5222 (wide_int_cc_tests): Call test_round_for_mask.
5223 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
5224 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
5225 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
5226 range returned by get_range_info.
5228 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5231 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
5232 * symtab.c: Include builtins.h
5233 (symtab_node::output_to_lto_symbol_table_p): Move here
5234 from lto-streamer-out.c:output_symbol_p.
5235 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
5236 (output_symbol_p): Move all logic to symtab.c
5237 (produce_symtab): Update.
5239 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5241 * config/s390/s390-opts.h (enum indirect_branch): Define.
5242 * config/s390/s390-protos.h (s390_return_addr_from_memory)
5243 (s390_indirect_branch_via_thunk)
5244 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
5245 (enum s390_indirect_branch_type): Define.
5246 * config/s390/s390.c (struct s390_frame_layout, struct
5247 machine_function): Remove.
5248 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
5249 (indirect_branch_table_label_no, indirect_branch_table_name):
5251 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
5252 (enum s390_indirect_branch_option): Define.
5253 (s390_return_addr_from_memory): New function.
5254 (s390_handle_string_attribute): New function.
5255 (s390_attribute_table): Add new attribute handler.
5256 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
5257 (s390_indirect_branch_via_thunk): New function.
5258 (s390_indirect_branch_via_inline_thunk): New function.
5259 (s390_function_ok_for_sibcall): When jumping via thunk disallow
5260 sibling call optimization for non z10 compiles.
5261 (s390_emit_call): Force indirect branch target to be a single
5262 register. Add r1 clobber for non-z10 compiles.
5263 (s390_emit_epilogue): Emit return jump via return_use expander.
5264 (s390_reorg): Handle JUMP_INSNs as execute targets.
5265 (s390_option_override_internal): Perform validity checks for the
5266 new command line options.
5267 (s390_indirect_branch_attrvalue): New function.
5268 (s390_indirect_branch_settings): New function.
5269 (s390_set_current_function): Invoke s390_indirect_branch_settings.
5270 (s390_output_indirect_thunk_function): New function.
5271 (s390_code_end): Implement target hook.
5272 (s390_case_values_threshold): Implement target hook.
5273 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
5275 * config/s390/s390.h (struct s390_frame_layout)
5276 (struct machine_function): Move here from s390.c.
5277 (TARGET_INDIRECT_BRANCH_NOBP_RET)
5278 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
5279 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
5280 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
5281 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
5282 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
5283 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
5284 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
5285 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
5286 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
5287 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
5288 (mnemonic attribute): Add values which aren't recognized
5290 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
5291 pattern for branch conversion. Fix mnemonic attribute.
5292 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
5293 indirect branch via thunk if requested.
5294 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
5295 ("*indirect_jump"): Disable for branch conversion using out of
5297 ("indirect_jump_via_thunk<mode>_z10")
5298 ("indirect_jump_via_thunk<mode>")
5299 ("indirect_jump_via_inlinethunk<mode>_z10")
5300 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
5301 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
5302 ("casesi_jump_via_inlinethunk<mode>_z10")
5303 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
5304 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
5305 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
5306 ("*indirect2_jump"): Disable for branch conversion.
5307 ("casesi_jump"): Turn into expander and expand patterns for branch
5309 ("return_use"): New expander.
5310 ("*return"): Emit return via thunk and rename it to ...
5311 ("*return<mode>"): ... this one.
5312 * config/s390/s390.opt: Add new options and and enum for the
5315 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5317 * lra-constraints.c (match_reload): Unconditionally use
5318 gen_lowpart_SUBREG, rather than selecting between that
5319 and equivalent gen_rtx_SUBREG code.
5321 2018-02-08 Richard Biener <rguenther@suse.de>
5323 PR tree-optimization/84233
5324 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
5325 changed flag instead of boguously re-using phi_inserted.
5327 2018-02-08 Martin Jambor <mjambor@suse.cz>
5329 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
5330 static local variables.
5332 2018-02-08 Richard Biener <rguenther@suse.de>
5334 PR tree-optimization/84278
5335 * tree-vect-stmts.c (vectorizable_store): When looking for
5336 smaller vector types to perform grouped strided loads/stores
5337 make sure the mode is supported by the target.
5338 (vectorizable_load): Likewise.
5340 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5342 * config/aarch64/aarch64.c (aarch64_components_for_bb):
5343 Increase LDP/STP opportunities by adding adjacent callee-saves.
5345 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5347 PR rtl-optimization/84068
5348 PR rtl-optimization/83459
5349 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
5351 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
5353 PR tree-optimization/84224
5354 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
5355 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
5358 2018-02-07 Iain Sandoe <iain@codesourcery.com>
5361 * config/rs6000/altivec.md (*restore_world): Remove LR use.
5362 * config/rs6000/predicates.md (restore_world_operation): Adjust op
5363 count, remove one USE.
5365 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
5367 * doc/install.texi (Configuration): Document the
5368 --with-long-double-format={ibm,ieee} PowerPC configuration
5372 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
5373 Convert from define_expand to be define_insn_and_split. Rework
5374 float/double/_Float128 conversions to QI/HI/SImode to work with
5375 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
5376 conversions to QI/HImode types did a store and then a load to
5377 truncate the value. For conversions to VSX registers, don't split
5378 the insn, instead emit the code directly. Use the code iterator
5379 any_fix to combine signed and unsigned conversions.
5380 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
5381 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
5382 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5383 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
5384 (fix_<mode>di2_hw): Likewise.
5385 (fixuns_<mode>di2_hw): Likewise.
5386 (fix_<mode>si2_hw): Likewise.
5387 (fixuns_<mode>si2_hw): Likewise.
5388 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
5389 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5390 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
5391 fix<uns>_trunc<SFDF:mode>si2_p8.
5392 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
5394 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
5395 (fix<uns>_<mode>_mem): Likewise.
5396 (fctiw<u>z_<mode>_mem): Likewise.
5397 (fix<uns>_<mode>_mem): Likewise.
5398 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
5399 the register allocator from doing a direct move to the GPRs to do
5400 a store, and instead use the ISA 3.0 store byte/half-word from
5401 vector register instruction. For IEEE 128-bit floating point,
5402 also optimize stores of 32-bit ints.
5403 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
5405 2018-02-07 Alan Hayward <alan.hayward@arm.com>
5407 * genextract.c (push_pathstr_operand): New function to support
5409 (walk_rtx): Call push_pathstr_operand.
5410 (print_path): Support [a-zA-Z].
5412 2018-02-07 Richard Biener <rguenther@suse.de>
5414 PR tree-optimization/84037
5415 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
5416 (cse_and_gimplify_to_preheader): Declare.
5417 (vect_get_place_in_interleaving_chain): Likewise.
5418 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5420 (_loop_vec_info::~_loop_vec_info): Delete it.
5421 (cse_and_gimplify_to_preheader): New function.
5422 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
5423 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
5424 (vectorizable_load): Likewise. For grouped stores always base
5425 the IV on the first element.
5426 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
5427 condition before gimplifying.
5429 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5431 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
5432 *DIV_EXPR and *MOD_EXPR.
5434 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
5437 * config/i386/i386.c (ix86_option_override_internal): Mask out
5438 the CF_SET bit when checking -fcf-protection.
5440 2018-02-07 Tom de Vries <tom@codesourcery.com>
5443 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
5446 2018-02-07 Richard Biener <rguenther@suse.de>
5448 PR tree-optimization/84204
5449 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
5452 PR tree-optimization/84205
5453 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
5454 special-case isl_ast_op_zdiv_r.
5456 PR tree-optimization/84223
5457 * graphite-scop-detection.c (gather_bbs::before_dom_children):
5458 Only add conditions from within the region.
5459 (gather_bbs::after_dom_children): Adjust.
5461 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
5464 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
5465 * config/avr/avr.md: Only post-reload split REG-REG moves if
5466 either register is GENERAL_REG_P.
5468 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5470 PR tree-optimization/84235
5471 * tree-ssa-scopedtables.c
5472 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
5473 if the subtraction is performed in floating point type where NaNs are
5474 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
5475 build 1. Formatting fix.
5477 2018-02-06 Jakub Jelinek <jakub@redhat.com>
5480 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
5481 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
5482 and skip it regardless of bb boundaries. Use CALL_P macro,
5483 don't test INSN_P (insn) together with CALL_P or JUMP_P check
5484 unnecessarily, formatting fix.
5486 2018-02-06 Michael Collison <michael.collison@arm.com>
5488 * config/arm/thumb2.md:
5489 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
5490 (*thumb_mov_notscc): Ditto.
5492 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
5495 * config/rs6000/rs6000.md (su code attribute): Use "u" for
5496 unsigned_fix, not "s".
5498 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5500 * configure.ac (gcc_fn_eh_frame_ro): New function.
5501 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
5502 correct .eh_frame permissions.
5503 * configure: Regenerate.
5505 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
5507 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
5510 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5512 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5513 Display warning message for -mno-speculate-indirect-jumps.
5515 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
5517 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
5519 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
5521 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
5523 PR tree-optimization/84225
5524 * tree-eh.c (find_trapping_overflow): Only call
5525 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
5527 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5530 * config/i386/i386.c: Reimplement the check of possible options
5531 -mibt/-mshstk conbination. Change error messages.
5532 * doc/invoke.texi: Fix a typo: remove extra '='.
5534 2018-02-06 Marek Polacek <polacek@redhat.com>
5536 PR tree-optimization/84228
5537 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
5539 2018-02-06 Tamar Christina <tamar.christina@arm.com>
5542 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
5543 emitted arch directives.
5544 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
5545 __ARM_FEATURE_COPROC before changing architectures.
5547 2018-02-06 Richard Biener <rguenther@suse.de>
5549 * config/i386/i386.c (print_reg): Fix typo.
5550 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
5552 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
5554 * configure: Regenerate.
5556 2018-02-05 Martin Sebor <msebor@redhat.com>
5558 PR tree-optimization/83369
5559 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
5562 2018-02-05 Martin Liska <mliska@suse.cz>
5564 * doc/invoke.texi: Cherry-pick upstream r323995.
5566 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
5568 * ira.c (ira_init_register_move_cost): Adjust comment.
5570 2018-02-05 Martin Liska <mliska@suse.cz>
5572 PR gcov-profile/84137
5573 * doc/gcov.texi: Fix typo in documentation.
5575 2018-02-05 Martin Liska <mliska@suse.cz>
5577 PR gcov-profile/83879
5578 * doc/gcov.texi: Document necessity of --dynamic-list-data when
5579 using dlopen functionality.
5581 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
5583 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
5584 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
5585 _mm_maskz_range_ss, _mm_mask_range_round_ss,
5586 _mm_maskz_range_round_ss): New intrinsics.
5587 (__builtin_ia32_rangesd128_round)
5588 (__builtin_ia32_rangess128_round): Remove.
5589 (__builtin_ia32_rangesd128_mask_round,
5590 __builtin_ia32_rangess128_mask_round): New builtins.
5591 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
5592 __builtin_ia32_rangess128_round): Remove.
5593 (__builtin_ia32_rangesd128_mask_round,
5594 __builtin_ia32_rangess128_mask_round): New builtins.
5595 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
5596 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
5597 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
5598 "<round_saeonly_constraint>")): Changed to ...
5599 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
5600 "<round_saeonly_scalar_constraint>")): ... this.
5601 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
5602 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
5603 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
5604 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
5605 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
5607 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
5609 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
5611 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
5612 Remove all values except native, 8540 and 8548.
5614 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
5616 * config/i386/i386.c (ix86_output_function_return): Pass
5617 INVALID_REGNUM, instead of -1, as invalid register number to
5618 indirect_thunk_name and output_indirect_thunk.
5620 2018-02-02 Julia Koval <julia.koval@intel.com>
5622 * config.gcc: Add -march=icelake.
5623 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
5624 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
5625 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
5626 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
5627 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
5628 (processor_target_table): Add icelake.
5629 (ix86_option_override_internal): Handle new PTAs.
5630 (get_builtin_code_for_version): Handle icelake.
5631 (M_INTEL_COREI7_ICELAKE): New.
5632 (fold_builtin_cpu): Handle icelake.
5633 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
5634 * doc/invoke.texi: Add -march=icelake.
5636 2018-02-02 Julia Koval <julia.koval@intel.com>
5638 * config/i386/i386.c (ix86_option_override_internal): Change flags type
5639 to wide_int_bitmask.
5640 * wide-int-bitmask.h: New.
5642 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5645 * config/i386/i386.md: Replace Pmode with word_mode in
5646 builtin_setjmp_setup and builtin_longjmp to support x32.
5648 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
5652 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
5654 (rs6000_supported_cpu_names): New static variable.
5655 (linux_cpu_translation_table): Likewise.
5656 (elf_platform) <cpu>: Define new static variable and use it.
5657 Translate kernel AT_PLATFORM name to canonical name if needed.
5658 Error if platform name is unknown.
5660 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
5663 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
5665 2018-02-01 Jeff Law <law@redhat.com>
5668 * config/i386/i386.c (release_scratch_register_on_entry): Add new
5669 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
5670 the scratch if RELEASE_VIA_POP is false.
5671 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
5672 If we have to save a temporary register, decrement SIZE appropriately.
5673 Pass new arguments to release_scratch_register_on_entry.
5674 (ix86_adjust_stack_and_probe): Likewise.
5675 (ix86_emit_probe_stack_range): Pass new arguments to
5676 release_scratch_register_on_entry.
5678 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
5680 PR rtl-optimization/84157
5681 * combine.c (change_zero_ext): Use REG_P predicate in
5682 front of HARD_REGISTER_P predicate.
5684 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
5686 * config/avr/avr.c (avr_option_override): Move disabling of
5687 -fdelete-null-pointer-checks to...
5688 * common/config/avr/avr-common.c (avr_option_optimization_table):
5691 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5693 PR tree-optimization/81635
5694 * tree-data-ref.c (split_constant_offset_1): For types that
5695 wrap on overflow, try to use range info to prove that wrapping
5698 2018-02-01 Renlin Li <renlin.li@arm.com>
5701 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
5703 (aarch64_register_move_cost): Likewise.
5704 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
5706 (REG_CLASS_NAMES): Likewise.
5707 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
5708 TAILCALL_ADDR_REGS. Remove IP registers.
5709 * config/aarch64/aarch64.md (Ucs): Update register constraint.
5711 2018-02-01 Richard Biener <rguenther@suse.de>
5713 * domwalk.h (dom_walker::dom_walker): Add additional constructor
5714 for specifying RPO order and allow NULL for that.
5715 * domwalk.c (dom_walker::dom_walker): Likewise.
5716 (dom_walker::walk): Handle NULL RPO order.
5717 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
5719 (rewrite_update_dom_walker): Likewise.
5720 (mark_def_dom_walker): Likewise.
5722 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5724 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
5725 (aarch64_maybe_expand_sve_subreg_move): Declare.
5726 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
5727 * config/aarch64/predicates.md (aarch64_any_register_operand): New
5729 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
5730 that are semantically a reverse operation.
5731 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
5732 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
5733 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
5735 (aarch64_can_change_mode_class): For big-endian, forbid changes
5736 between two SVE modes if they have different element sizes.
5738 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5740 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
5741 the TImode handling for big-endian targets.
5743 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5745 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
5746 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
5748 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
5749 Remove BSWAP handing for big-endian targets and use the form of
5750 LD1RQ appropariate for the mode.
5752 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5754 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
5755 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
5758 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5761 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
5762 check for operands that need to go through aarch64_sve_reload_be.
5764 2018-02-01 Jakub Jelinek <jakub@redhat.com>
5766 PR tree-optimization/81661
5767 PR tree-optimization/84117
5768 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
5769 * tree-eh.c: Include gimplify.h.
5770 (find_trapping_overflow, replace_trapping_overflow,
5771 rewrite_to_non_trapping_overflow): New functions.
5772 * tree-vect-loop.c: Include tree-eh.h.
5773 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
5774 * tree-data-ref.c: Include tree-eh.h.
5775 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
5777 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
5779 PR rtl-optimization/84123
5780 * combine.c (change_zero_ext): Check if hard register satisfies
5781 can_change_dest_mode before calling gen_lowpart_SUBREG.
5783 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
5786 * ira.c (ira_init_register_move_cost): Remove assert.
5788 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5790 PR rtl-optimization/84071
5791 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
5792 * doc/tm.texi: Regenerate.
5794 2018-01-31 Richard Biener <rguenther@suse.de>
5796 PR tree-optimization/84132
5797 * tree-data-ref.c (analyze_miv_subscript): Properly
5798 check whether evolution_function_is_affine_multivariate_p
5799 before calling gcd_of_steps_may_divide_p.
5801 2018-01-31 Julia Koval <julia.koval@intel.com>
5804 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
5805 * config/i386/i386.md (rdpid_rex64) New.
5806 (rdpid): Make 32bit only.
5808 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
5811 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
5812 an IDENTIFIER_NODE for FUNCTION_TYPE's.
5814 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5817 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5819 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5821 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5823 PR rtl-optimization/84071
5824 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
5825 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
5827 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5829 * config/arc/arc.c (arc_handle_aux_attribute): New function.
5830 (arc_attribute_table): Add 'aux' attribute.
5831 (arc_in_small_data_p): Consider aux like variables.
5832 (arc_is_aux_reg_p): New function.
5833 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
5834 (arc_get_aux_arg): New function.
5835 (prepare_move_operands): Handle aux-register access.
5836 (arc_handle_aux_attribute): New function.
5837 * doc/extend.texi (ARC Variable attributes): Add subsection.
5839 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5841 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
5842 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
5843 (arc_attribute_table): Add 'uncached' attribute.
5844 (arc_print_operand): Print '.di' flag for uncached memory
5846 (arc_in_small_data_p): Do not consider for small data the uncached
5848 (arc_is_uncached_mem_p): New function.
5849 * config/arc/predicates.md (compact_store_memory_operand): Check
5850 for uncached memory accesses.
5851 (nonvol_nonimm_operand): Likewise.
5852 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
5854 2018-01-31 Jakub Jelinek <jakub@redhat.com>
5857 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
5858 falign-loops=): Add Optimization flag.
5860 2018-01-30 Jeff Law <law@redhat.com>
5863 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
5864 INT_REGISTERS_SAVED. Check it prior to calling
5865 get_scratch_register_on_entry.
5866 (ix86_adjust_stack_and_probe): Similarly.
5867 (ix86_emit_probe_stack_range): Similarly.
5868 (ix86_expand_prologue): Corresponding changes.
5870 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5873 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
5874 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
5876 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
5879 * lra-constraints.c (curr_insn_transform): Process AND in the
5882 2018-01-30 Jakub Jelinek <jakub@redhat.com>
5884 PR rtl-optimization/83986
5885 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
5886 dependence against last_pending_memory_flush in addition to
5889 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
5891 PR tree-optimization/81611
5892 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
5895 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5898 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
5901 2018-01-30 Richard Biener <rguenther@suse.de>
5902 Jakub Jelinek <jakub@redhat.com>
5904 PR tree-optimization/84111
5905 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
5906 inner loops added during recursion, as they don't have up-to-date
5909 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5912 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
5913 (can_inline_edge_by_limits_p): ... here.
5914 (can_early_inline_edge_p, check_callers,
5915 update_caller_keys, update_callee_keys, recursive_inlining,
5916 add_new_edges_to_heap, speculation_useful_p,
5917 inline_small_functions,
5918 inline_small_functions, flatten_function,
5919 inline_to_all_callers_1): Update.
5921 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5923 * profile-count.c (profile_count::combine_with_ipa_count): Handle
5926 2018-01-30 Richard Biener <rguenther@suse.de>
5928 PR tree-optimization/83008
5929 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
5930 invariant and constant vector uses in stmts when they need
5933 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5936 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
5937 * configure: Regenerate.
5939 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5941 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
5943 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
5944 Use gen_rtx_REG rather than gen_lowpart.
5946 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5948 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
5949 rather than 0 when creating partial subregs.
5951 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5953 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
5956 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
5959 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
5960 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
5961 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
5962 flags. This restores the settings used before the 2017-07-24.
5963 Turning off pre increment/decrement/modify allows IVOPTS to
5964 optimize DF/SF loops where the index is an int.
5966 2018-01-29 Richard Biener <rguenther@suse.de>
5967 Kelvin Nilsen <kelvin@gcc.gnu.org>
5970 * tree-vect-stmts.c (vectorizable_call): Don't call
5971 targetm.vectorize_builtin_md_vectorized_function if callee is
5974 2018-01-22 Carl Love <cel@us.ibm.com>
5976 * doc/extend.tex: Fix typo in second arg in
5977 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
5979 2018-01-29 Richard Biener <rguenther@suse.de>
5981 PR tree-optimization/84086
5982 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
5983 (flush_ssaname_freelist): When SSA names were released reset
5984 the SCEV hash table.
5986 2018-01-29 Richard Biener <rguenther@suse.de>
5988 PR tree-optimization/84057
5989 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
5990 removed paths when removing edges.
5992 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
5994 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
5995 -mfunction-return=@var{choice}.
5997 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
6000 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
6002 (layout::print_source_line): Likewise.
6003 (test_get_line_width_without_trailing_whitespace): Add test cases.
6005 2018-01-27 Jakub Jelinek <jakub@redhat.com>
6008 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
6011 2018-01-26 Jim Wilson <jimw@sifive.com>
6013 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
6015 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
6018 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6020 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
6021 and CMP + SUB-immediate -> SUBS.
6023 2018-01-26 Martin Sebor <msebor@redhat.com>
6025 PR tree-optimization/83896
6026 * tree-ssa-strlen.c (get_string_len): Rename...
6027 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
6028 Avoid assuming length is constant.
6029 (handle_char_store): Use HOST_WIDE_INT for string length.
6031 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
6034 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
6035 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
6037 2018-01-26 Richard Biener <rguenther@suse.de>
6039 PR rtl-optimization/84003
6040 * dse.c (record_store): Only record redundant stores when
6041 the earlier store aliases at least all accesses the later one does.
6043 2018-01-26 Jakub Jelinek <jakub@redhat.com>
6045 PR rtl-optimization/83985
6046 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
6047 REG_CFA_RESTORE insns.
6048 (delete_unmarked_insns): Don't ignore separate shrink wrapping
6049 REG_CFA_RESTORE insns here.
6052 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
6053 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
6055 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6057 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
6058 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
6059 (arc_init): Likewise.
6060 (arc_override_options): Likewise.
6061 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
6063 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
6065 * config/arc/arc.h (TARGET_DBNZ): Define.
6066 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
6067 properly set the tune attribute.
6068 (dbnz): Use TARGET_DBNZ guard.
6069 * config/arc/arc.opt (mtune): Add core3 option.
6071 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6073 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
6074 recognize new pic like addresses.
6075 (arc_delegitimize_address): Clean up.
6077 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6079 * config/arc/arc-arches.def: Option mrf16 valid for all
6081 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
6082 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
6083 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
6084 * config/arc/arc-tables.opt: Regenerate.
6085 * config/arc/arc.c (arc_conditional_register_usage): Handle
6086 reduced register file case.
6087 (arc_file_start): Set must have build attributes.
6088 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
6090 * config/arc/arc.opt (mrf16): Add new option.
6091 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
6092 * config/arc/genmultilib.awk: Handle new mrf16 option.
6093 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
6094 * config/arc/t-multilib: Regenerate.
6095 * doc/invoke.texi (ARC Options): Document mrf16 option.
6097 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6099 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
6100 * config/arc/arc.c (arc_handle_secure_attribute): New function.
6101 (arc_attribute_table): Add 'secure_call' attribute.
6102 (arc_print_operand): Print secure call operand.
6103 (arc_function_ok_for_sibcall): Don't optimize tail calls when
6105 (arc_is_secure_call_p): New function. * config/arc/arc.md
6106 (call_i): Add support for sjli instruction.
6107 (call_value_i): Likewise.
6108 * config/arc/constraints.md (Csc): New constraint.
6110 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6111 John Eric Martin <John.Martin@emmicro-us.com>
6113 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
6114 * config/arc/arc.c (_arc_jli_section): New struct.
6115 (arc_jli_section): New type.
6116 (rc_jli_sections): New static variable.
6117 (arc_handle_jli_attribute): New function.
6118 (arc_attribute_table): Add jli_always and jli_fixed attribute.
6119 (arc_file_end): New function.
6120 (TARGET_ASM_FILE_END): Define.
6121 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
6122 (arc_add_jli_section): New function.
6123 (jli_call_scan): Likewise.
6124 (arc_reorg): Call jli_call_scan.
6125 (arc_output_addsi): Remove 'S' from printing asm operand.
6126 (arc_is_jli_call_p): New function.
6127 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
6129 (movhi_insn): Likewise.
6130 (movsi_insn): Likewise.
6131 (movsi_set_cc_insn): Likewise.
6132 (loadqi_update): Likewise.
6133 (load_zeroextendqisi_update): Likewise.
6134 (load_signextendqisi_update): Likewise.
6135 (loadhi_update): Likewise.
6136 (load_zeroextendhisi_update): Likewise.
6137 (load_signextendhisi_update): Likewise.
6138 (loadsi_update): Likewise.
6139 (loadsf_update): Likewise.
6140 (movsicc_insn): Likewise.
6141 (bset_insn): Likewise.
6142 (bxor_insn): Likewise.
6143 (bclr_insn): Likewise.
6144 (bmsk_insn): Likewise.
6145 (bicsi3_insn): Likewise.
6146 (cmpsi_cc_c_insn): Likewise.
6147 (movsi_ne): Likewise.
6148 (movsi_cond_exec): Likewise.
6149 (clrsbsi2): Likewise.
6160 (call_i): Remove 'S' asm letter, add jli instruction.
6161 (call_value_i): Likewise.
6162 * config/arc/arc.op (mjli-always): New option.
6163 * config/arc/constraints.md (Cji): New constraint.
6164 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
6166 (subsf3_fpx): Likewise.
6167 (mulsf3_fpx): Likewise.
6168 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
6170 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
6172 * doc/invoke.texi (ARC): Document mjli-always option.
6174 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
6176 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
6177 avoid addition with 0 and use incw and decw where possible.
6179 2018-01-26 Richard Biener <rguenther@suse.de>
6181 PR tree-optimization/81082
6182 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
6183 association if it requires casting to unsigned.
6184 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
6185 from fold_plusminus_mult_expr to catch important cases late when
6186 range info is available.
6188 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6190 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
6191 * configure.ac (hidden_linkonce): New test.
6192 * configure: Regenerate.
6193 * config.in: Regenerate.
6195 2018-01-26 Julia Koval <julia.koval@intel.com>
6197 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
6198 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
6199 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
6200 _mm_mask_bitshuffle_epi64_mask): Fix type.
6201 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
6202 USI_FTYPE_V4DI_V4DI_USI): Remove.
6203 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
6204 __builtin_ia32_vpshufbitqmb256_mask,
6205 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
6206 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
6207 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
6209 2018-01-26 Alan Modra <amodra@gmail.com>
6212 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
6213 UNSPEC_VBPERMQ. Sort other unspecs.
6215 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
6217 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
6219 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
6222 * predict.c (drop_profile): Do not push/pop cfun; update also
6224 (handle_missing_profiles): Fix logic looking for zero profiles.
6226 2018-01-25 Jakub Jelinek <jakub@redhat.com>
6229 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
6230 on functions with #pragma omp declare simd or functions with simd
6232 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
6233 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
6234 Remove trailing \n from warning_at calls.
6236 2018-01-25 Tom de Vries <tom@codesourcery.com>
6239 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6240 for neutered workers.
6242 2018-01-24 Joseph Myers <joseph@codesourcery.com>
6245 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
6246 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
6248 2018-01-24 Jeff Law <law@redhat.com>
6251 * i386.c (get_probe_interval): Move to earlier point.
6252 (ix86_compute_frame_layout): If -fstack-clash-protection and
6253 the frame is larger than the probe interval, then use pushes
6254 to save registers rather than reg->mem moves.
6255 (ix86_expand_prologue): Remove conditional for int_registers_saved
6258 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
6261 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
6262 min/max for never referenced object.
6264 2018-01-24 Jakub Jelinek <jakub@redhat.com>
6267 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
6269 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
6270 attributes from DECL_ATTRIBUTES (decl) without affecting
6271 DECL_ATTRIBUTES (current_function_decl).
6272 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
6273 functions with non-NULL DECL_ABSTRACT_ORIGIN.
6275 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
6277 PR tree-optimization/83979
6278 * fold-const.c (fold_comparison): Use constant_boolean_node
6279 instead of boolean_{true,false}_node.
6281 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
6283 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
6286 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6288 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6289 Simplify the clause that sets the length attribute.
6290 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6291 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
6292 clause that sets the length attribute.
6293 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6295 2018-01-24 Tom de Vries <tom@codesourcery.com>
6298 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
6299 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
6300 Add strict parameter.
6301 (prevent_branch_around_nothing): Insert dummy insn between branch to
6302 label and label with no ptx insn inbetween.
6303 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
6305 2018-01-24 Tom de Vries <tom@codesourcery.com>
6308 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6309 for neutered threads in warp.
6310 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
6312 2018-01-24 Richard Biener <rguenther@suse.de>
6314 PR tree-optimization/83176
6315 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
6318 2018-01-24 Richard Biener <rguenther@suse.de>
6320 PR tree-optimization/82819
6321 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
6322 code generating pluses that are no-ops in the target precision.
6324 2018-01-24 Richard Biener <rguenther@suse.de>
6327 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
6329 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6331 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
6332 to merge probabilities.
6333 * predict.c (probably_never_executed): Also mark as cold functions
6334 with global 0 profile and guessed local profile.
6335 * profile-count.c (profile_probability::combine_with_count): New
6337 * profile-count.h (profile_probability::operator*,
6338 profile_probability::operator*=, profile_probability::operator/,
6339 profile_probability::operator/=): Reduce precision to adjusted
6340 and set value to guessed on contradictory divisions.
6341 (profile_probability::combine_with_freq): Remove.
6342 (profile_probability::combine_wiht_count): Declare.
6343 (profile_count::force_nonzero):: Set to adjusted.
6344 (profile_count::probability_in):: Set quality to adjusted.
6345 * tree-ssa-tail-merge.c (replace_block_by): Use
6348 2018-01-23 Andrew Waterman <andrew@sifive.com>
6349 Jim Wilson <jimw@sifive.com>
6351 * config/riscv/riscv.c (riscv_stack_boundary): New.
6352 (riscv_option_override): Set riscv_stack_boundary. Handle
6353 riscv_preferred_stack_boundary_arg.
6354 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
6355 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
6356 (STACK_BOUNDARY): Set to riscv_stack_boundary.
6357 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
6358 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
6359 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
6361 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
6364 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
6365 of struct ix86_frame.
6366 (ix86_expand_epilogue): Likewise. Add a local variable for
6367 the reg_save_offset field in struct ix86_frame.
6369 2018-01-23 Bin Cheng <bin.cheng@arm.com>
6371 PR tree-optimization/82604
6372 * tree-loop-distribution.c (enum partition_kind): New enum item
6373 PKIND_PARTIAL_MEMSET.
6374 (partition_builtin_p): Support above new enum item.
6375 (generate_code_for_partition): Ditto.
6376 (compute_access_range): Differentiate cases that equality can be
6377 proven at all loops, the innermost loops or no loops.
6378 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
6379 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
6380 (finalize_partitions, distribute_loop): Don't fuse partition of
6381 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
6382 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
6385 2018-01-23 Martin Liska <mliska@suse.cz>
6387 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
6388 order to ignore the predictor.
6389 (PRED_POLYMORPHIC_CALL): Likewise.
6390 (PRED_RECURSIVE_CALL): Likewise.
6392 2018-01-23 Martin Liska <mliska@suse.cz>
6394 * tree-profile.c (tree_profiling): Print function header to
6395 aware reader which function we are working on.
6396 * value-prof.c (gimple_find_values_to_profile): Do not print
6397 not interesting value histograms.
6399 2018-01-23 Martin Liska <mliska@suse.cz>
6401 * profile-count.h (enum profile_quality): Add
6402 profile_uninitialized as the first value. Do not number values
6403 as they are zero based.
6404 (profile_count::verify): Update sanity check.
6405 (profile_probability::verify): Likewise.
6407 2018-01-23 Nathan Sidwell <nathan@acm.org>
6409 * doc/invoke.texi (ffor-scope): Deprecate.
6411 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6413 PR tree-optimization/83510
6414 * domwalk.c (set_all_edges_as_executable): New function.
6415 (dom_walker::dom_walker): Convert bool param
6416 "skip_unreachable_blocks" to enum reachability. Move setup of
6417 edge flags to set_all_edges_as_executable and only do it when
6418 reachability is REACHABLE_BLOCKS.
6419 * domwalk.h (enum dom_walker::reachability): New enum.
6420 (dom_walker::dom_walker): Convert bool param
6421 "skip_unreachable_blocks" to enum reachability.
6422 (set_all_edges_as_executable): New decl.
6423 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
6424 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
6426 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
6427 but converting true to REACHABLE_BLOCKS.
6428 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
6430 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
6431 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
6432 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
6434 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
6435 if check_all_array_refs will be called.
6437 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6439 * tree.c (selftest::test_location_wrappers): Add more test
6442 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6444 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
6445 (selftest::test_bit_in_range): Likewise.
6447 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6450 * doc/sourcebuild.texi (vect_float): Say that the selector
6451 only describes the situation when -funsafe-math-optimizations is on.
6452 (vect_float_strict): Document.
6454 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6456 PR tree-optimization/83965
6457 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
6458 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
6459 instead of checking only for a reduction.
6460 (vect_recog_widen_sum_pattern): Likewise.
6462 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6464 * predict.c (probably_never_executed): Only use precise profile info.
6465 (compute_function_frequency): Skip after inlining hack since we now
6466 have quality checking.
6468 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6470 * profile-count.h (profile_probability::very_unlikely,
6471 profile_probability::unlikely, profile_probability::even): Set
6472 precision to guessed.
6474 2018-01-23 Richard Biener <rguenther@suse.de>
6476 PR tree-optimization/83963
6477 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
6478 Properly terminate dominator walk when crossing the exit edge not
6479 when visiting its source block.
6481 2018-01-23 Jakub Jelinek <jakub@redhat.com>
6484 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
6485 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
6487 2018-01-22 Jakub Jelinek <jakub@redhat.com>
6489 PR tree-optimization/83957
6490 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
6491 semicolon after for body surrounded by braces.
6493 PR tree-optimization/83081
6494 * profile-count.h (profile_probability::split): New method.
6495 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
6496 Use profile_probability::split.
6497 (do_compare_rtx_and_jump): Fix adjustment of probabilities
6498 when splitting a single conditional jump into 2.
6500 2018-01-22 David Malcolm <dmalcolm@redhat.com>
6502 PR tree-optimization/69452
6503 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
6506 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6508 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
6509 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
6510 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
6512 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6514 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
6515 * config/rl78/rl78.md (movdi): New define_expand.
6516 * config/rl78/rl78.c (rl78_split_movdi): New function.
6518 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
6521 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
6523 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
6524 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
6525 128-bit to produce an UNSPEC move to get the double word with the
6526 signbit and then a shift directly to do signbit.
6527 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
6528 implementation with a new version that just does either a direct
6529 move or a regular move. Move memory interface to separate insns.
6530 Move insns so they are next to the expander.
6531 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
6532 with signbit move. Split big and little endian case.
6533 (signbit<mode>2_dm_mem_le): Likewise.
6534 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
6535 (signbit<mode>2_dm2): Likewise.
6537 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6539 * config/rl78/rl78.md (anddi3): New define_expand.
6541 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6543 * config/rl78/rl78.md (umindi3): New define_expand.
6545 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6547 * config/rl78/rl78.md (smindi3): New define_expand.
6549 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6551 * config/rl78/rl78.md (smaxdi3): New define_expand.
6553 2018-01-22 Carl Love <cel@us.ibm.com>
6555 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
6556 LVX_V1TI): Add macro expansion.
6557 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
6558 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
6559 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
6560 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
6561 Change check to determine if the instruction is a byte reversing
6562 entry. Fix typo in comment.
6563 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
6564 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
6565 Add def_builtin calls for new builtins.
6566 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
6567 Add define_insn expansion.
6569 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6571 * config/rl78/rl78.md (umaxdi3): New define_expand.
6573 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6575 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
6576 for non-QImode registers.
6578 2018-01-22 Richard Biener <rguenther@suse.de>
6580 PR tree-optimization/83963
6581 * graphite-scop-detection.c (scop_detection::get_sese): Delay
6582 including the loop exit block.
6583 (scop_detection::merge_sese): Likewise.
6584 (scop_detection::add_scop): Do it here instead.
6586 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6588 * doc/sourcebuild.texi (arm_softfloat): Document.
6590 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
6593 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
6594 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
6595 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
6597 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6598 David Edelsohn <dje.gcc@gmail.com>
6601 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6602 Change "crset eq" to "crset 2".
6603 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6604 (*call_indirect_aix<mode>_nospec): Likewise.
6605 (*call_value_indirect_aix<mode>_nospec): Likewise.
6606 (*call_indirect_elfv2<mode>_nospec): Likewise.
6607 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
6608 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
6609 change assembly output from . to $.
6610 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6611 (indirect_jump<mode>_nospec): Change assembly output from . to $.
6612 (*tablejump<mode>_internal1_nospec): Likewise.
6614 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
6617 * config/sh/sh_optimize_sett_clrt.cc:
6618 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
6620 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
6622 PR tree-optimization/83940
6623 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
6624 offset_dt to vect_constant_def rather than vect_unknown_def_type.
6625 (vect_check_load_store_mask): Add a mask_dt_out parameter and
6626 use it to pass back the definition type.
6627 (vect_check_store_rhs): Likewise rhs_dt_out.
6628 (vect_build_gather_load_calls): Add a mask_dt argument and use
6629 it instead of a call to vect_is_simple_use.
6630 (vectorizable_store): Update calls to vect_check_load_store_mask
6631 and vect_check_store_rhs. Use the dt returned by the latter instead
6632 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
6633 instead of calls to vect_is_simple_use. Pass the scalar rather
6634 than the vector operand to vect_is_simple_use when handling
6635 second and subsequent copies of an rhs value.
6636 (vectorizable_load): Update calls to vect_check_load_store_mask
6637 and vect_build_gather_load_calls. Use the cached mask_dt and
6638 gs_info.offset_dt instead of calls to vect_is_simple_use.
6640 2018-01-20 Jakub Jelinek <jakub@redhat.com>
6643 * tree-emutls.c: Include gimplify.h.
6644 (lower_emutls_2): New function.
6645 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
6646 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
6647 it before further processing.
6650 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
6651 UINTVAL (trueop1) instead of INTVAL (op1).
6653 2018-01-19 Jakub Jelinek <jakub@redhat.com>
6657 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
6658 INCOMING_FRAME_SP_OFFSET if not defined.
6659 (scan_trace): Add ENTRY argument. If true and
6660 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
6661 emit a note to adjust the CFA offset.
6662 (create_cfi_notes): Adjust scan_trace callers.
6663 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
6664 INCOMING_FRAME_SP_OFFSET in the CIE.
6665 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
6666 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
6668 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
6669 * doc/tm.texi: Regenerated.
6671 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6673 PR rtl-optimization/83147
6674 * lra-constraints.c (remove_inheritance_pseudos): Use
6675 lra_substitute_pseudo_within_insn.
6677 2018-01-19 Tom de Vries <tom@codesourcery.com>
6678 Cesar Philippidis <cesar@codesourcery.com>
6681 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
6683 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
6686 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
6687 spaces for function labels.
6689 2018-01-19 Martin Liska <mliska@suse.cz>
6691 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
6692 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
6693 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
6694 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
6695 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
6696 (PRED_CONST_RETURN): Change from 69 to 65.
6697 (PRED_NULL_RETURN): Change from 91 to 71.
6698 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
6699 (PRED_LOOP_GUARD): Change from 66 to 73.
6701 2018-01-19 Martin Liska <mliska@suse.cz>
6703 * predict.c (predict_insn_def): Add new assert.
6704 (struct branch_predictor): Change type to signed integer.
6705 (test_prediction_value_range): Amend test to cover
6707 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
6708 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
6709 (PRED_LOOP_ITERATIONS_MAX): Likewise.
6710 (PRED_LOOP_IV_COMPARE): Likewise.
6711 * predict.h (PROB_UNINITIALIZED): Define new constant.
6713 2018-01-19 Martin Liska <mliska@suse.cz>
6715 * predict.c (dump_prediction): Add new format for
6716 analyze_brprob.py script which is enabled with -details
6718 * profile-count.h (precise_p): New function.
6720 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6722 PR tree-optimization/83922
6723 * tree-vect-loop.c (vect_verify_full_masking): Return false if
6724 there are no statements that need masking.
6725 (vect_active_double_reduction_p): New function.
6726 (vect_analyze_loop_operations): Use it when handling phis that
6727 are not in the loop header.
6729 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6731 PR tree-optimization/83914
6732 * tree-vect-loop.c (vectorizable_induction): Don't convert
6733 init_expr or apply the peeling adjustment for inductions
6734 that are nested within the vectorized loop.
6736 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6738 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
6741 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6745 * function.h (gimplify_parameters): Add gimple_seq * argument.
6746 * function.c: Include gimple.h and options.h.
6747 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
6748 for the added local temporaries if needed.
6749 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
6750 if there are any parameter cleanups, wrap whole body into a
6751 try/finally with the cleanups.
6753 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
6756 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
6757 Use GET_MODE_CLASS for scalar floating point.
6759 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6763 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
6764 Fix call of call_cgraph_insertion_hooks.
6766 2018-01-18 Martin Sebor <msebor@redhat.com>
6768 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
6770 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6773 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
6776 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
6779 * common.opt: (-ffile-prefix-map): New option.
6780 * opts.c (common_handle_option): Defer it.
6781 * opts-global.c (handle_common_deferred_options): Handle it.
6782 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
6783 * file-prefix-map.h: New file.
6784 (remap_debug_filename, add_debug_prefix_map): ...here.
6785 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
6786 * final.c (debug_prefix_map, add_debug_prefix_map
6787 remap_debug_filename): Move to...
6788 * file-prefix-map.c: New file.
6789 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
6790 generalize, get rid of alloca(), use strrchr() instead of strchr().
6791 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
6792 Implement in terms of add_prefix_map().
6793 (remap_macro_filename, remap_debug_filename): Implement in term of
6795 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
6796 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
6797 * dbxout.c: Include file-prefix-map.h.
6798 * varasm.c: Likewise.
6799 * vmsdbgout.c: Likewise.
6800 * xcoffout.c: Likewise.
6801 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
6802 * doc/cppopts.texi (-fmacro-prefix-map): Document.
6803 * doc/invoke.texi (-ffile-prefix-map): Document.
6804 (-fdebug-prefix-map): Update description.
6806 2018-01-18 Martin Liska <mliska@suse.cz>
6808 * config/i386/i386.c (indirect_thunk_name): Document that also
6810 (output_indirect_thunk): Document why both instructions
6811 (pause and lfence) are generated.
6813 2018-01-18 Richard Biener <rguenther@suse.de>
6815 PR tree-optimization/83887
6816 * graphite-scop-detection.c
6817 (scop_detection::get_nearest_dom_with_single_entry): Remove.
6818 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
6819 (scop_detection::merge_sese): Re-implement with a flood-fill
6820 algorithm that properly finds a SESE region if it exists.
6822 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6825 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
6826 pointer_diff optimizations use view_convert instead of convert.
6828 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6830 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6831 Generate different code for -mno-speculate-indirect-jumps.
6832 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6833 (*call_indirect_aix<mode>): Disable for
6834 -mno-speculate-indirect-jumps.
6835 (*call_indirect_aix<mode>_nospec): New define_insn.
6836 (*call_value_indirect_aix<mode>): Disable for
6837 -mno-speculate-indirect-jumps.
6838 (*call_value_indirect_aix<mode>_nospec): New define_insn.
6839 (*sibcall_nonlocal_sysv<mode>): Generate different code for
6840 -mno-speculate-indirect-jumps.
6841 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6843 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6845 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
6846 long double type, set the flags for noting the default long double
6847 type, even if we don't pass or return a long double type.
6849 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
6852 * ipa-inline.c (flatten_function): Do not overwrite final inlining
6855 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
6857 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
6858 support for merge[hl].
6859 (fold_mergehl_helper): New helper function.
6860 (tree-vector-builder.h): New #include for tree_vector_builder usage.
6861 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
6862 (altivec_vmrglw_direct): Add xxmrglw insn.
6864 2018-01-17 Andrew Waterman <andrew@sifive.com>
6866 * config/riscv/riscv.c (riscv_conditional_register_usage): If
6867 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
6869 2018-01-17 David Malcolm <dmalcolm@redhat.com>
6872 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
6873 call the lto_location_cache before reading the
6874 DECL_SOURCE_LOCATION of the types.
6876 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
6877 Richard Sandiford <richard.sandiford@linaro.org>
6879 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
6880 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
6881 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
6882 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
6883 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
6885 * config/aarch64/constraints.md (aarch64_movti_operand):
6887 * config/aarch64/predicates.md (Uti): Add new constraint.
6889 2018-01-17 Carl Love <cel@us.ibm.com>
6891 * config/rs6000/vsx.md (define_expand xl_len_r,
6892 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
6893 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
6895 (define_expand, define_insn): Move the shift left from the
6896 define_insn to the define_expand for lxvl and stxvl instructions.
6897 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
6898 and XL_LEN_R definitions to PURE.
6900 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
6902 * config/i386/i386.c (indirect_thunk_name): Declare regno
6903 as unsigned int. Compare regno with INVALID_REGNUM.
6904 (output_indirect_thunk): Ditto.
6905 (output_indirect_thunk_function): Ditto.
6906 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
6907 in the call to output_indirect_thunk_function.
6909 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
6912 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
6913 rather than the size of inner_type to determine the stack slot size
6914 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
6916 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
6919 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
6922 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
6924 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
6925 endian Linux systems to optionally enable multilibs for selecting
6926 the long double type if the user configured an explicit type.
6927 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
6928 have no long double multilibs if not defined.
6929 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
6930 warn if the user used -mabi={ieee,ibm}longdouble and we built
6931 multilibs for long double.
6932 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
6933 appropriate multilib option.
6934 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
6936 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
6937 for building long double multilibs.
6938 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
6940 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
6942 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
6945 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
6947 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
6950 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
6953 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
6956 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6958 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
6959 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
6961 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
6963 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
6964 different rtl trees depending on TARGET_64BIT.
6965 (rs6000_gen_lvx): Likewise.
6967 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6969 * config/visium/visium.md (nop): Tweak comment.
6970 (hazard_nop): Likewise.
6972 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6974 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
6975 -mspeculate-indirect-jumps.
6976 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
6977 for -mno-speculate-indirect-jumps.
6978 (*call_indirect_elfv2<mode>_nospec): New define_insn.
6979 (*call_value_indirect_elfv2<mode>): Disable for
6980 -mno-speculate-indirect-jumps.
6981 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
6982 (indirect_jump): Emit different RTL for
6983 -mno-speculate-indirect-jumps.
6984 (*indirect_jump<mode>): Disable for
6985 -mno-speculate-indirect-jumps.
6986 (*indirect_jump<mode>_nospec): New define_insn.
6987 (tablejump): Emit different RTL for
6988 -mno-speculate-indirect-jumps.
6989 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
6990 (tablejumpsi_nospec): New define_expand.
6991 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
6992 (tablejumpdi_nospec): New define_expand.
6993 (*tablejump<mode>_internal1): Disable for
6994 -mno-speculate-indirect-jumps.
6995 (*tablejump<mode>_internal1_nospec): New define_insn.
6996 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
6999 2018-01-16 Artyom Skrobov tyomitch@gmail.com
7001 * caller-save.c (insert_save): Drop unnecessary parameter. All
7004 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7005 Richard Biener <rguenth@suse.de>
7008 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
7009 return early, inline manually is_gimple_sizepos. Make sure if we
7010 call gimplify_expr we don't end up with a gimple constant.
7011 * tree.c (variably_modified_type_p): Don't return true for
7012 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
7013 * gimplify.h (is_gimple_sizepos): Remove.
7015 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7017 PR tree-optimization/83857
7018 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
7019 vectorizable_live_operation for pure SLP statements.
7020 (vectorizable_live_operation): Handle PHIs.
7022 2018-01-16 Richard Biener <rguenther@suse.de>
7024 PR tree-optimization/83867
7025 * tree-vect-stmts.c (vect_transform_stmt): Precompute
7026 nested_in_vect_loop_p since the scalar stmt may get invalidated.
7028 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7031 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
7032 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
7033 If off is not INTEGER_CST, issue a may not be aligned warning
7034 rather than isn't aligned. Use isn%'t rather than isn't.
7035 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
7037 <case MULT_EXPR>: Improve the case when bottom and one of the
7038 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
7039 operand, in that case check if the other operand is multiple of
7040 bottom divided by the INTEGER_CST operand.
7042 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7045 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
7046 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
7047 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
7048 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
7049 * config/pa/pa.c (pa_function_arg_advance): Likewise.
7050 (pa_function_arg, pa_arg_partial_bytes): Likewise.
7051 (pa_function_arg_size): New function.
7053 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7055 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
7056 in a separate statement.
7058 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7060 PR tree-optimization/83847
7061 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
7062 group gathers and scatters.
7064 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7066 PR rtl-optimization/86620
7067 * params.def (max-sched-ready-insns): Bump minimum value to 1.
7069 PR rtl-optimization/83213
7070 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
7071 to last if both are JUMP_INSNs.
7073 PR tree-optimization/83843
7074 * gimple-ssa-store-merging.c
7075 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
7076 store_immediate_info for bswap/nop orig_stores.
7078 2018-01-15 Andrew Waterman <andrew@sifive.com>
7080 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
7082 <UDIV>: Increase cost if !TARGET_DIV.
7084 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
7086 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
7087 (define_attr "cr_logical_3op"): New.
7088 (cceq_ior_compare): Adjust.
7089 (cceq_ior_compare_complement): Adjust.
7090 (*cceq_rev_compare): Adjust.
7091 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
7092 (is_cracked_insn): Adjust.
7093 (insn_must_be_first_in_group): Adjust.
7094 * config/rs6000/40x.md: Adjust.
7095 * config/rs6000/440.md: Adjust.
7096 * config/rs6000/476.md: Adjust.
7097 * config/rs6000/601.md: Adjust.
7098 * config/rs6000/603.md: Adjust.
7099 * config/rs6000/6xx.md: Adjust.
7100 * config/rs6000/7450.md: Adjust.
7101 * config/rs6000/7xx.md: Adjust.
7102 * config/rs6000/8540.md: Adjust.
7103 * config/rs6000/cell.md: Adjust.
7104 * config/rs6000/e300c2c3.md: Adjust.
7105 * config/rs6000/e500mc.md: Adjust.
7106 * config/rs6000/e500mc64.md: Adjust.
7107 * config/rs6000/e5500.md: Adjust.
7108 * config/rs6000/e6500.md: Adjust.
7109 * config/rs6000/mpc.md: Adjust.
7110 * config/rs6000/power4.md: Adjust.
7111 * config/rs6000/power5.md: Adjust.
7112 * config/rs6000/power6.md: Adjust.
7113 * config/rs6000/power7.md: Adjust.
7114 * config/rs6000/power8.md: Adjust.
7115 * config/rs6000/power9.md: Adjust.
7116 * config/rs6000/rs64.md: Adjust.
7117 * config/rs6000/titan.md: Adjust.
7119 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7121 * config/i386/predicates.md (indirect_branch_operand): Rewrite
7122 ix86_indirect_branch_register logic.
7124 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7126 * config/i386/constraints.md (Bs): Update
7127 ix86_indirect_branch_register check. Don't check
7128 ix86_indirect_branch_register with GOT_memory_operand.
7130 * config/i386/predicates.md (GOT_memory_operand): Don't check
7131 ix86_indirect_branch_register here.
7132 (GOT32_symbol_operand): Likewise.
7134 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7136 * config/i386/predicates.md (constant_call_address_operand):
7137 Rewrite ix86_indirect_branch_register logic.
7138 (sibcall_insn_operand): Likewise.
7140 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7142 * config/i386/constraints.md (Bs): Replace
7143 ix86_indirect_branch_thunk_register with
7144 ix86_indirect_branch_register.
7146 * config/i386/i386.md (indirect_jump): Likewise.
7147 (tablejump): Likewise.
7148 (*sibcall_memory): Likewise.
7149 (*sibcall_value_memory): Likewise.
7150 Peepholes of indirect call and jump via memory: Likewise.
7151 * config/i386/i386.opt: Likewise.
7152 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7153 (GOT_memory_operand): Likewise.
7154 (call_insn_operand): Likewise.
7155 (sibcall_insn_operand): Likewise.
7156 (GOT32_symbol_operand): Likewise.
7158 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7161 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
7162 type rather than type addr's type points to.
7163 (expand_omp_atomic_mutex): Likewise.
7164 (expand_omp_atomic): Likewise.
7166 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7169 * config/i386/i386.c (output_indirect_thunk_function): Use
7170 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
7171 for __x86_return_thunk.
7173 2018-01-15 Richard Biener <rguenther@suse.de>
7176 * expmed.c (extract_bit_field_1): Fix typo.
7178 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7181 * config/arm/iterators.md (VF): New mode iterator.
7182 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
7183 Remove integer-related logic from pattern.
7184 (neon_vabd<mode>_3): Likewise.
7186 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7189 * common.opt (fstrict-overflow): No longer an alias.
7190 (fwrapv-pointer): New option.
7191 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
7192 also for pointer types based on flag_wrapv_pointer.
7193 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
7194 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
7195 opts->x_flag_wrapv got set.
7196 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
7197 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
7198 POINTER_TYPE_OVERFLOW_UNDEFINED.
7199 * match.pd: Likewise in address comparison pattern.
7200 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
7202 2018-01-15 Richard Biener <rguenther@suse.de>
7205 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
7206 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
7207 Reset type names to their identifier if their TYPE_DECL doesn't
7208 have linkage (and thus is used for ODR and devirt).
7209 (save_debug_info_for_decl): Remove.
7210 (save_debug_info_for_type): Likewise.
7211 (add_tree_to_fld_list): Adjust.
7212 * tree-pretty-print.c (dump_generic_node): Make dumping of
7213 type names more robust.
7215 2018-01-15 Richard Biener <rguenther@suse.de>
7217 * BASE-VER: Bump to 8.0.1.
7219 2018-01-14 Martin Sebor <msebor@redhat.com>
7222 * builtins.c (check_access): Avoid warning when the no-warning bit
7225 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
7227 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
7228 * ira-color (allocno_hard_regs_compare): Likewise.
7230 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
7233 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
7234 Use .pushsection/.popsection.
7236 2018-01-14 Martin Sebor <msebor@redhat.com>
7239 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
7241 2018-01-14 Jakub Jelinek <jakub@redhat.com>
7243 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
7244 entry from extra_headers.
7245 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
7246 extra_headers, make the list bitwise identical to the i?86-*-* one.
7248 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7250 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
7251 -mcmodel=large with -mindirect-branch=thunk,
7252 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
7253 -mfunction-return=thunk-extern.
7254 * doc/invoke.texi: Document -mcmodel=large is incompatible with
7255 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
7256 -mfunction-return=thunk and -mfunction-return=thunk-extern.
7258 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7260 * config/i386/i386.c (print_reg): Print the name of the full
7261 integer register without '%'.
7262 (ix86_print_operand): Handle 'V'.
7263 * doc/extend.texi: Document 'V' modifier.
7265 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7267 * config/i386/constraints.md (Bs): Disallow memory operand for
7268 -mindirect-branch-register.
7270 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7271 (GOT_memory_operand): Likewise.
7272 (call_insn_operand): Likewise.
7273 (sibcall_insn_operand): Likewise.
7274 (GOT32_symbol_operand): Likewise.
7275 * config/i386/i386.md (indirect_jump): Call convert_memory_address
7276 for -mindirect-branch-register.
7277 (tablejump): Likewise.
7278 (*sibcall_memory): Likewise.
7279 (*sibcall_value_memory): Likewise.
7280 Disallow peepholes of indirect call and jump via memory for
7281 -mindirect-branch-register.
7282 (*call_pop): Replace m with Bw.
7283 (*call_value_pop): Likewise.
7284 (*sibcall_pop_memory): Replace m with Bs.
7285 * config/i386/i386.opt (mindirect-branch-register): New option.
7286 * doc/invoke.texi: Document -mindirect-branch-register option.
7288 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7290 * config/i386/i386-protos.h (ix86_output_function_return): New.
7291 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
7292 set function_return_type.
7293 (indirect_thunk_name): Add ret_p to indicate thunk for function
7295 (output_indirect_thunk_function): Pass false to
7296 indirect_thunk_name.
7297 (ix86_output_indirect_branch_via_reg): Likewise.
7298 (ix86_output_indirect_branch_via_push): Likewise.
7299 (output_indirect_thunk_function): Create alias for function
7300 return thunk if regno < 0.
7301 (ix86_output_function_return): New function.
7302 (ix86_handle_fndecl_attribute): Handle function_return.
7303 (ix86_attribute_table): Add function_return.
7304 * config/i386/i386.h (machine_function): Add
7305 function_return_type.
7306 * config/i386/i386.md (simple_return_internal): Use
7307 ix86_output_function_return.
7308 (simple_return_internal_long): Likewise.
7309 * config/i386/i386.opt (mfunction-return=): New option.
7310 (indirect_branch): Mention -mfunction-return=.
7311 * doc/extend.texi: Document function_return function attribute.
7312 * doc/invoke.texi: Document -mfunction-return= option.
7314 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7316 * config/i386/i386-opts.h (indirect_branch): New.
7317 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
7318 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
7319 with local indirect jump when converting indirect call and jump.
7320 (ix86_set_indirect_branch_type): New.
7321 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
7322 (indirectlabelno): New.
7323 (indirect_thunk_needed): Likewise.
7324 (indirect_thunk_bnd_needed): Likewise.
7325 (indirect_thunks_used): Likewise.
7326 (indirect_thunks_bnd_used): Likewise.
7327 (INDIRECT_LABEL): Likewise.
7328 (indirect_thunk_name): Likewise.
7329 (output_indirect_thunk): Likewise.
7330 (output_indirect_thunk_function): Likewise.
7331 (ix86_output_indirect_branch_via_reg): Likewise.
7332 (ix86_output_indirect_branch_via_push): Likewise.
7333 (ix86_output_indirect_branch): Likewise.
7334 (ix86_output_indirect_jmp): Likewise.
7335 (ix86_code_end): Call output_indirect_thunk_function if needed.
7336 (ix86_output_call_insn): Call ix86_output_indirect_branch if
7338 (ix86_handle_fndecl_attribute): Handle indirect_branch.
7339 (ix86_attribute_table): Add indirect_branch.
7340 * config/i386/i386.h (machine_function): Add indirect_branch_type
7341 and has_local_indirect_jump.
7342 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
7344 (tablejump): Likewise.
7345 (*indirect_jump): Use ix86_output_indirect_jmp.
7346 (*tablejump_1): Likewise.
7347 (simple_return_indirect_internal): Likewise.
7348 * config/i386/i386.opt (mindirect-branch=): New option.
7349 (indirect_branch): New.
7352 (thunk-inline): Likewise.
7353 (thunk-extern): Likewise.
7354 * doc/extend.texi: Document indirect_branch function attribute.
7355 * doc/invoke.texi: Document -mindirect-branch= option.
7357 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
7360 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
7362 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
7364 * ipa-inline.c (want_inline_small_function_p): Return false if
7365 inlining has already failed with CIF_FINAL_ERROR.
7366 (update_caller_keys): Call want_inline_small_function_p before
7368 (update_callee_keys): Likewise.
7370 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7372 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
7374 (rs6000_quadword_masked_address_p): Likewise.
7375 (quad_aligned_load_p): Likewise.
7376 (quad_aligned_store_p): Likewise.
7377 (const_load_sequence_p): Add comment to describe the outer-most loop.
7378 (mimic_memory_attributes_and_flags): New function.
7379 (rs6000_gen_stvx): Likewise.
7380 (replace_swapped_aligned_store): Likewise.
7381 (rs6000_gen_lvx): Likewise.
7382 (replace_swapped_aligned_load): Likewise.
7383 (replace_swapped_load_constant): Capitalize argument name in
7384 comment describing this function.
7385 (rs6000_analyze_swaps): Add a third pass to search for vector loads
7386 and stores that access quad-word aligned addresses and replace
7387 with stvx or lvx instructions when appropriate.
7388 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
7389 New function prototype.
7390 (rs6000_quadword_masked_address_p): Likewise.
7391 (rs6000_gen_lvx): Likewise.
7392 (rs6000_gen_stvx): Likewise.
7393 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
7394 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
7395 when memory address is aligned.
7396 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
7397 this split to select lvx instruction when memory address is aligned.
7398 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
7399 instruction when memory address is aligned.
7400 (*vsx_le_perm_load_v16qi): Likewise.
7401 (four unnamed splitters): Modify to select the stvx instruction
7402 when memory is aligned.
7404 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
7406 * predict.c (determine_unlikely_bbs): Handle correctly BBs
7407 which appears in the queue multiple times.
7409 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7410 Alan Hayward <alan.hayward@arm.com>
7411 David Sherwood <david.sherwood@arm.com>
7413 * tree-vectorizer.h (vec_lower_bound): New structure.
7414 (_loop_vec_info): Add check_nonzero and lower_bounds.
7415 (LOOP_VINFO_CHECK_NONZERO): New macro.
7416 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
7417 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
7418 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
7419 fields. Make seg_len the distance travelled, not including the
7421 (dr_direction_indicator): Declare.
7422 (dr_zero_step_indicator): Likewise.
7423 (dr_known_forward_stride_p): Likewise.
7424 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
7426 (runtime_alias_check_p): Allow runtime alias checks with
7428 (operator ==): Compare access_size and align.
7429 (prune_runtime_alias_test_list): Rework for new distinction between
7430 the access_size and seg_len.
7431 (create_intersect_range_checks_index): Likewise. Cope with polynomial
7433 (get_segment_min_max): New function.
7434 (create_intersect_range_checks): Use it.
7435 (dr_step_indicator): New function.
7436 (dr_direction_indicator): Likewise.
7437 (dr_zero_step_indicator): Likewise.
7438 (dr_known_forward_stride_p): Likewise.
7439 * tree-loop-distribution.c (data_ref_segment_size): Return
7440 DR_STEP * (niters - 1).
7441 (compute_alias_check_pairs): Update call to the dr_with_seg_len
7443 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
7444 (vect_preserves_scalar_order_p): New function, split out from...
7445 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
7446 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
7447 (vect_vfa_access_size): New function.
7448 (vect_vfa_align): Likewise.
7449 (vect_compile_time_alias): Take access_size_a and access_b arguments.
7450 (dump_lower_bound): New function.
7451 (vect_check_lower_bound): Likewise.
7452 (vect_small_gap_p): Likewise.
7453 (vectorizable_with_step_bound_p): Likewise.
7454 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
7455 depencies if the vectorization factor is 1. Convert the checks
7456 for nonzero steps into checks on the bounds of DR_STEP. Try using
7457 a bunds check for variable steps if the minimum required step is
7458 relatively small. Update calls to the dr_with_seg_len
7459 constructor and to vect_compile_time_alias.
7460 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
7462 (vect_loop_versioning): Call it.
7463 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
7465 (vect_estimate_min_profitable_iters): Account for any bounds checks.
7467 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7468 Alan Hayward <alan.hayward@arm.com>
7469 David Sherwood <david.sherwood@arm.com>
7471 * doc/sourcebuild.texi (vect_scatter_store): Document.
7472 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
7474 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
7476 * genopinit.c (main): Add supports_vec_scatter_store and
7477 supports_vec_scatter_store_cached to target_optabs.
7478 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
7479 IFN_MASK_SCATTER_STORE.
7480 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
7482 * internal-fn.h (internal_store_fn_p): Declare.
7483 (internal_fn_stored_value_index): Likewise.
7484 * internal-fn.c (scatter_store_direct): New macro.
7485 (expand_scatter_store_optab_fn): New function.
7486 (direct_scatter_store_optab_supported_p): New macro.
7487 (internal_store_fn_p): New function.
7488 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
7489 IFN_MASK_SCATTER_STORE.
7490 (internal_fn_mask_index): Likewise.
7491 (internal_fn_stored_value_index): New function.
7492 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
7494 * optabs-query.h (supports_vec_scatter_store_p): Declare.
7495 * optabs-query.c (supports_vec_scatter_store_p): New function.
7496 * tree-vectorizer.h (vect_get_store_rhs): Declare.
7497 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
7498 true for scatter stores.
7499 (vect_gather_scatter_fn_p): Handle scatter stores too.
7500 (vect_check_gather_scatter): Consider using scatter stores if
7501 supports_vec_scatter_store_p.
7502 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
7504 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7505 internal_fn_stored_value_index.
7506 (check_load_store_masking): Handle scatter stores too.
7507 (vect_get_store_rhs): Make public.
7508 (vectorizable_call): Use internal_store_fn_p.
7509 (vectorizable_store): Handle scatter store internal functions.
7510 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
7511 when deciding whether the end of the group has been reached.
7512 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
7513 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
7514 (mask_scatter_store<mode>): New insns.
7516 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7517 Alan Hayward <alan.hayward@arm.com>
7518 David Sherwood <david.sherwood@arm.com>
7520 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
7521 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
7522 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
7524 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
7525 Use vect_truncate_gather_scatter_offset if we can't treat the
7526 operation as a normal gather load or scatter store.
7527 (get_group_load_store_type): Take the gather_scatter_info
7528 as argument. Try using a gather load or scatter store for
7529 single-element groups.
7530 (get_load_store_type): Update calls to get_group_load_store_type
7531 and vect_use_strided_gather_scatters_p.
7533 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7534 Alan Hayward <alan.hayward@arm.com>
7535 David Sherwood <david.sherwood@arm.com>
7537 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
7538 optional tree argument.
7539 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
7541 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
7542 but continue to use the current value as a fallback.
7543 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
7544 to compare the updates.
7545 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
7546 (get_load_store_type): Use it when handling a strided access.
7547 (vect_get_strided_load_store_ops): New function.
7548 (vect_get_data_ptr_increment): Likewise.
7549 (vectorizable_load): Handle strided gather loads. Always pass
7550 a step to vect_create_data_ref_ptr and bump_vector_ptr.
7552 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7553 Alan Hayward <alan.hayward@arm.com>
7554 David Sherwood <david.sherwood@arm.com>
7556 * doc/md.texi (gather_load@var{m}): Document.
7557 (mask_gather_load@var{m}): Likewise.
7558 * genopinit.c (main): Add supports_vec_gather_load and
7559 supports_vec_gather_load_cached to target_optabs.
7560 * optabs-tree.c (init_tree_optimization_optabs): Use
7561 ggc_cleared_alloc to allocate target_optabs.
7562 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
7563 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
7565 * internal-fn.h (internal_load_fn_p): Declare.
7566 (internal_gather_scatter_fn_p): Likewise.
7567 (internal_fn_mask_index): Likewise.
7568 (internal_gather_scatter_fn_supported_p): Likewise.
7569 * internal-fn.c (gather_load_direct): New macro.
7570 (expand_gather_load_optab_fn): New function.
7571 (direct_gather_load_optab_supported_p): New macro.
7572 (direct_internal_fn_optab): New function.
7573 (internal_load_fn_p): Likewise.
7574 (internal_gather_scatter_fn_p): Likewise.
7575 (internal_fn_mask_index): Likewise.
7576 (internal_gather_scatter_fn_supported_p): Likewise.
7577 * optabs-query.c (supports_at_least_one_mode_p): New function.
7578 (supports_vec_gather_load_p): Likewise.
7579 * optabs-query.h (supports_vec_gather_load_p): Declare.
7580 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
7581 and memory_type field.
7582 (NUM_PATTERNS): Bump to 15.
7583 * tree-vect-data-refs.c: Include internal-fn.h.
7584 (vect_gather_scatter_fn_p): New function.
7585 (vect_describe_gather_scatter_call): Likewise.
7586 (vect_check_gather_scatter): Try using internal functions for
7587 gather loads. Recognize existing calls to a gather load function.
7588 (vect_analyze_data_refs): Consider using gather loads if
7589 supports_vec_gather_load_p.
7590 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
7591 (vect_get_gather_scatter_offset_type): Likewise.
7592 (vect_convert_mask_for_vectype): Likewise.
7593 (vect_add_conversion_to_patterm): Likewise.
7594 (vect_try_gather_scatter_pattern): Likewise.
7595 (vect_recog_gather_scatter_pattern): New pattern recognizer.
7596 (vect_vect_recog_func_ptrs): Add it.
7597 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7598 internal_fn_mask_index and internal_gather_scatter_fn_p.
7599 (check_load_store_masking): Take the gather_scatter_info as an
7600 argument and handle gather loads.
7601 (vect_get_gather_scatter_ops): New function.
7602 (vectorizable_call): Check internal_load_fn_p.
7603 (vectorizable_load): Likewise. Handle gather load internal
7605 (vectorizable_store): Update call to check_load_store_masking.
7606 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
7607 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
7608 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
7609 (aarch64_gather_scale_operand_d): New predicates.
7610 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
7611 (mask_gather_load<mode>): New insns.
7613 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7614 Alan Hayward <alan.hayward@arm.com>
7615 David Sherwood <david.sherwood@arm.com>
7617 * optabs.def (fold_left_plus_optab): New optab.
7618 * doc/md.texi (fold_left_plus_@var{m}): Document.
7619 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
7620 * internal-fn.c (fold_left_direct): Define.
7621 (expand_fold_left_optab_fn): Likewise.
7622 (direct_fold_left_optab_supported_p): Likewise.
7623 * fold-const-call.c (fold_const_fold_left): New function.
7624 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
7625 * tree-parloops.c (valid_reduction_p): New function.
7626 (gather_scalar_reductions): Use it.
7627 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
7628 (vect_finish_replace_stmt): Declare.
7629 * tree-vect-loop.c (fold_left_reduction_fn): New function.
7630 (needs_fold_left_reduction_p): New function, split out from...
7631 (vect_is_simple_reduction): ...here. Accept reductions that
7632 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
7633 (vect_force_simple_reduction): Also store the reduction type in
7634 the assignment's STMT_VINFO_REDUC_TYPE.
7635 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
7636 (merge_with_identity): New function.
7637 (vect_expand_fold_left): Likewise.
7638 (vectorize_fold_left_reduction): Likewise.
7639 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
7640 scalar phi in place for it. Check for target support and reject
7641 cases that would reassociate the operation. Defer the transform
7642 phase to vectorize_fold_left_reduction.
7643 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
7644 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
7645 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
7647 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7649 * tree-if-conv.c (predicate_mem_writes): Remove redundant
7650 call to ifc_temp_var.
7652 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7653 Alan Hayward <alan.hayward@arm.com>
7654 David Sherwood <david.sherwood@arm.com>
7656 * target.def (legitimize_address_displacement): Take the original
7657 offset as a poly_int.
7658 * targhooks.h (default_legitimize_address_displacement): Update
7660 * targhooks.c (default_legitimize_address_displacement): Likewise.
7661 * doc/tm.texi: Regenerate.
7662 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
7663 as an argument, moving assert of ad->disp == ad->disp_term to...
7664 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
7665 Try calling targetm.legitimize_address_displacement before expanding
7666 the address rather than afterwards, and adjust for the new interface.
7667 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
7668 Match the new hook interface. Handle SVE addresses.
7669 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
7672 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7674 * Makefile.in (OBJS): Add early-remat.o.
7675 * target.def (select_early_remat_modes): New hook.
7676 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
7677 * doc/tm.texi: Regenerate.
7678 * targhooks.h (default_select_early_remat_modes): Declare.
7679 * targhooks.c (default_select_early_remat_modes): New function.
7680 * timevar.def (TV_EARLY_REMAT): New timevar.
7681 * passes.def (pass_early_remat): New pass.
7682 * tree-pass.h (make_pass_early_remat): Declare.
7683 * early-remat.c: New file.
7684 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
7686 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
7688 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7689 Alan Hayward <alan.hayward@arm.com>
7690 David Sherwood <david.sherwood@arm.com>
7692 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
7693 vfm1 with a bound_epilog parameter.
7694 (vect_do_peeling): Update calls accordingly, and move the prologue
7695 call earlier in the function. Treat the base bound_epilog as 0 for
7696 fully-masked loops and retain vf - 1 for other loops. Add 1 to
7697 this base when peeling for gaps.
7698 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
7699 with fully-masked loops.
7700 (vect_estimate_min_profitable_iters): Handle the single peeled
7701 iteration in that case.
7703 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7704 Alan Hayward <alan.hayward@arm.com>
7705 David Sherwood <david.sherwood@arm.com>
7707 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
7708 single-element interleaving even if the size is not a power of 2.
7709 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
7710 accesses for single-element interleaving if the group size is
7713 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7714 Alan Hayward <alan.hayward@arm.com>
7715 David Sherwood <david.sherwood@arm.com>
7717 * doc/md.texi (fold_extract_last_@var{m}): Document.
7718 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
7719 * optabs.def (fold_extract_last_optab): New optab.
7720 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
7721 * internal-fn.c (fold_extract_direct): New macro.
7722 (expand_fold_extract_optab_fn): Likewise.
7723 (direct_fold_extract_optab_supported_p): Likewise.
7724 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
7725 * tree-vect-loop.c (vect_model_reduction_cost): Handle
7726 EXTRACT_LAST_REDUCTION.
7727 (get_initial_def_for_reduction): Do not create an initial vector
7728 for EXTRACT_LAST_REDUCTION reductions.
7729 (vectorizable_reduction): Leave the scalar phi in place for
7730 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
7731 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
7732 epilogue code for EXTRACT_LAST_REDUCTION and defer the
7733 transform phase to vectorizable_condition.
7734 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
7736 (vect_finish_stmt_generation): ...here.
7737 (vect_finish_replace_stmt): New function.
7738 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
7739 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
7741 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
7743 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7744 Alan Hayward <alan.hayward@arm.com>
7745 David Sherwood <david.sherwood@arm.com>
7747 * doc/md.texi (extract_last_@var{m}): Document.
7748 * optabs.def (extract_last_optab): New optab.
7749 * internal-fn.def (EXTRACT_LAST): New internal function.
7750 * internal-fn.c (cond_unary_direct): New macro.
7751 (expand_cond_unary_optab_fn): Likewise.
7752 (direct_cond_unary_optab_supported_p): Likewise.
7753 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
7754 loops using EXTRACT_LAST.
7755 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
7756 (extract_last_<mode>): ...this optab.
7757 (vec_extract<mode><Vel>): Update accordingly.
7759 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7760 Alan Hayward <alan.hayward@arm.com>
7761 David Sherwood <david.sherwood@arm.com>
7763 * target.def (empty_mask_is_expensive): New hook.
7764 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
7765 * doc/tm.texi: Regenerate.
7766 * targhooks.h (default_empty_mask_is_expensive): Declare.
7767 * targhooks.c (default_empty_mask_is_expensive): New function.
7768 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
7769 if the target says that empty masks are expensive.
7770 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
7772 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
7774 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7775 Alan Hayward <alan.hayward@arm.com>
7776 David Sherwood <david.sherwood@arm.com>
7778 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
7779 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
7780 (vect_use_loop_mask_for_alignment_p): New function.
7781 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
7782 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
7783 niters_skip argument. Make sure that the first niters_skip elements
7784 of the first iteration are inactive.
7785 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
7786 Update call to vect_set_loop_masks_directly.
7787 (get_misalign_in_elems): New function, split out from...
7788 (vect_gen_prolog_loop_niters): ...here.
7789 (vect_update_init_of_dr): Take a code argument that specifies whether
7790 the adjustment should be added or subtracted.
7791 (vect_update_init_of_drs): Likewise.
7792 (vect_prepare_for_masked_peels): New function.
7793 (vect_do_peeling): Skip prologue peeling if we're using a mask
7794 instead. Update call to vect_update_inits_of_drs.
7795 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7797 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
7798 alignment. Do not include the number of peeled iterations in
7799 the minimum threshold in that case.
7800 (vectorizable_induction): Adjust the start value down by
7801 LOOP_VINFO_MASK_SKIP_NITERS iterations.
7802 (vect_transform_loop): Call vect_prepare_for_masked_peels.
7803 Take the number of skipped iterations into account when calculating
7805 * tree-vect-stmts.c (vect_gen_while_not): New function.
7807 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7808 Alan Hayward <alan.hayward@arm.com>
7809 David Sherwood <david.sherwood@arm.com>
7811 * doc/sourcebuild.texi (vect_fully_masked): Document.
7812 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
7814 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
7816 (vect_analyze_loop_2): ...here. Don't check the vectorization
7817 factor against the number of loop iterations if the loop is
7820 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7821 Alan Hayward <alan.hayward@arm.com>
7822 David Sherwood <david.sherwood@arm.com>
7824 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
7825 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
7826 (dump_groups): Update accordingly.
7827 (iv_use::mem_type): New member variable.
7828 (address_p): New function.
7829 (record_use): Add a mem_type argument and initialize the new
7831 (record_group_use): Add a mem_type argument. Use address_p.
7832 Remove obsolete null checks of base_object. Update call to record_use.
7833 (find_interesting_uses_op): Update call to record_group_use.
7834 (find_interesting_uses_cond): Likewise.
7835 (find_interesting_uses_address): Likewise.
7836 (get_mem_type_for_internal_fn): New function.
7837 (find_address_like_use): Likewise.
7838 (find_interesting_uses_stmt): Try find_address_like_use before
7839 calling find_interesting_uses_op.
7840 (addr_offset_valid_p): Use the iv mem_type field as the type
7841 of the addressed memory.
7842 (add_autoinc_candidates): Likewise.
7843 (get_address_cost): Likewise.
7844 (split_small_address_groups_p): Use address_p.
7845 (split_address_groups): Likewise.
7846 (add_iv_candidate_for_use): Likewise.
7847 (autoinc_possible_for_pair): Likewise.
7848 (rewrite_groups): Likewise.
7849 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
7850 (determine_group_iv_cost): Update after split of USE_ADDRESS.
7851 (get_alias_ptr_type_for_ptr_address): New function.
7852 (rewrite_use_address): Rewrite address uses in calls that were
7853 identified by find_address_like_use.
7855 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7856 Alan Hayward <alan.hayward@arm.com>
7857 David Sherwood <david.sherwood@arm.com>
7859 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
7861 * gimple-expr.h (is_gimple_addressable: Likewise.
7862 * gimple-expr.c (is_gimple_address): Likewise.
7863 * internal-fn.c (expand_call_mem_ref): New function.
7864 (expand_mask_load_optab_fn): Use it.
7865 (expand_mask_store_optab_fn): Likewise.
7867 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7868 Alan Hayward <alan.hayward@arm.com>
7869 David Sherwood <david.sherwood@arm.com>
7871 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
7872 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
7873 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
7874 (cond_umax@var{mode}): Document.
7875 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
7876 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
7877 (cond_umin_optab, cond_umax_optab): New optabs.
7878 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
7879 (COND_IOR, COND_XOR): New internal functions.
7880 * internal-fn.h (get_conditional_internal_fn): Declare.
7881 * internal-fn.c (cond_binary_direct): New macro.
7882 (expand_cond_binary_optab_fn): Likewise.
7883 (direct_cond_binary_optab_supported_p): Likewise.
7884 (get_conditional_internal_fn): New function.
7885 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
7886 Cope with reduction statements that are vectorized as calls rather
7888 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
7889 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
7890 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
7891 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
7892 (UNSPEC_COND_EOR): New unspecs.
7893 (optab): Add mappings for them.
7894 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
7895 (sve_int_op, sve_fp_op): New int attributes.
7897 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7898 Alan Hayward <alan.hayward@arm.com>
7899 David Sherwood <david.sherwood@arm.com>
7901 * optabs.def (while_ult_optab): New optab.
7902 * doc/md.texi (while_ult@var{m}@var{n}): Document.
7903 * internal-fn.def (WHILE_ULT): New internal function.
7904 * internal-fn.h (direct_internal_fn_supported_p): New override
7905 that takes two types as argument.
7906 * internal-fn.c (while_direct): New macro.
7907 (expand_while_optab_fn): New function.
7908 (convert_optab_supported_p): Likewise.
7909 (direct_while_optab_supported_p): New macro.
7910 * wide-int.h (wi::udiv_ceil): New function.
7911 * tree-vectorizer.h (rgroup_masks): New structure.
7912 (vec_loop_masks): New typedef.
7913 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
7915 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
7916 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
7917 (vect_max_vf): New function.
7918 (slpeel_make_loop_iterate_ntimes): Delete.
7919 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
7920 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
7921 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
7922 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
7923 internal-fn.h, stor-layout.h and optabs-query.h.
7924 (vect_set_loop_mask): New function.
7925 (add_preheader_seq): Likewise.
7926 (add_header_seq): Likewise.
7927 (interleave_supported_p): Likewise.
7928 (vect_maybe_permute_loop_masks): Likewise.
7929 (vect_set_loop_masks_directly): Likewise.
7930 (vect_set_loop_condition_masked): Likewise.
7931 (vect_set_loop_condition_unmasked): New function, split out from
7932 slpeel_make_loop_iterate_ntimes.
7933 (slpeel_make_loop_iterate_ntimes): Rename to..
7934 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
7935 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
7936 (vect_do_peeling): Update call accordingly.
7937 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
7939 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7940 mask_compare_type, can_fully_mask_p and fully_masked_p.
7941 (release_vec_loop_masks): New function.
7942 (_loop_vec_info): Use it to free the loop masks.
7943 (can_produce_all_loop_masks_p): New function.
7944 (vect_get_max_nscalars_per_iter): Likewise.
7945 (vect_verify_full_masking): Likewise.
7946 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
7947 retries, and free the mask rgroups before retrying. Check loop-wide
7948 reasons for disallowing fully-masked loops. Make the final decision
7949 about whether use a fully-masked loop or not.
7950 (vect_estimate_min_profitable_iters): Do not assume that peeling
7951 for the number of iterations will be needed for fully-masked loops.
7952 (vectorizable_reduction): Disable fully-masked loops.
7953 (vectorizable_live_operation): Likewise.
7954 (vect_halve_mask_nunits): New function.
7955 (vect_double_mask_nunits): Likewise.
7956 (vect_record_loop_mask): Likewise.
7957 (vect_get_loop_mask): Likewise.
7958 (vect_transform_loop): Handle the case in which the final loop
7959 iteration might handle a partial vector. Call vect_set_loop_condition
7960 instead of slpeel_make_loop_iterate_ntimes.
7961 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
7962 (check_load_store_masking): New function.
7963 (prepare_load_store_mask): Likewise.
7964 (vectorizable_store): Handle fully-masked loops.
7965 (vectorizable_load): Likewise.
7966 (supportable_widening_operation): Use vect_halve_mask_nunits for
7968 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
7969 (vect_gen_while): New function.
7970 * config/aarch64/aarch64.md (umax<mode>3): New expander.
7971 (aarch64_uqdec<mode>): New insn.
7973 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7974 Alan Hayward <alan.hayward@arm.com>
7975 David Sherwood <david.sherwood@arm.com>
7977 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
7978 (reduc_xor_scal_optab): New optabs.
7979 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
7980 (reduc_xor_scal_@var{m}): Document.
7981 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
7982 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
7984 * fold-const-call.c (fold_const_call): Handle them.
7985 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
7986 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
7987 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
7988 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
7989 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
7990 (UNSPEC_XORV): New unspecs.
7991 (optab): Add entries for them.
7992 (BITWISEV): New int iterator.
7993 (bit_reduc_op): New int attributes.
7995 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7996 Alan Hayward <alan.hayward@arm.com>
7997 David Sherwood <david.sherwood@arm.com>
7999 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
8000 * internal-fn.def (VEC_SHL_INSERT): New internal function.
8001 * optabs.def (vec_shl_insert_optab): New optab.
8002 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
8003 (duplicate_and_interleave): Likewise.
8004 * tree-vect-loop.c: Include internal-fn.h.
8005 (neutral_op_for_slp_reduction): New function, split out from
8006 get_initial_defs_for_reduction.
8007 (get_initial_def_for_reduction): Handle option 2 for variable-length
8008 vectors by loading the neutral value into a vector and then shifting
8009 the initial value into element 0.
8010 (get_initial_defs_for_reduction): Replace the code argument with
8011 the neutral value calculated by neutral_op_for_slp_reduction.
8012 Use gimple_build_vector for constant-length vectors.
8013 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
8014 but the first group_size elements have a neutral value.
8015 Use duplicate_and_interleave otherwise.
8016 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
8017 Update call to get_initial_defs_for_reduction. Handle SLP
8018 reductions for variable-length vectors by creating one vector
8019 result for each scalar result, with the elements associated
8020 with other scalar results stubbed out with the neutral value.
8021 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
8022 Require IFN_VEC_SHL_INSERT for double reductions on
8023 variable-length vectors, or SLP reductions that have
8024 a neutral value. Require can_duplicate_and_interleave_p
8025 support for variable-length unchained SLP reductions if there
8026 is no neutral value, such as for MIN/MAX reductions. Also require
8027 the number of vector elements to be a multiple of the number of
8028 SLP statements when doing variable-length unchained SLP reductions.
8029 Update call to vect_create_epilog_for_reduction.
8030 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
8031 and remove initial values.
8032 (duplicate_and_interleave): Make public.
8033 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
8034 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
8036 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8037 Alan Hayward <alan.hayward@arm.com>
8038 David Sherwood <david.sherwood@arm.com>
8040 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
8041 (can_duplicate_and_interleave_p): New function.
8042 (vect_get_and_check_slp_defs): Take the vector of statements
8043 rather than just the current one. Remove excess parentheses.
8044 Restriction rejectinon of vect_constant_def and vect_external_def
8045 for variable-length vectors to boolean types, or types for which
8046 can_duplicate_and_interleave_p is false.
8047 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
8048 (duplicate_and_interleave): New function.
8049 (vect_get_constant_vectors): Use gimple_build_vector for
8050 constant-length vectors and suitable variable-length constant
8051 vectors. Use duplicate_and_interleave for other variable-length
8052 vectors. Don't defer the update when inserting new statements.
8054 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8055 Alan Hayward <alan.hayward@arm.com>
8056 David Sherwood <david.sherwood@arm.com>
8058 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
8059 min_profitable_iters doesn't go negative.
8061 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8062 Alan Hayward <alan.hayward@arm.com>
8063 David Sherwood <david.sherwood@arm.com>
8065 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
8066 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
8067 * optabs.def (vec_mask_load_lanes_optab): New optab.
8068 (vec_mask_store_lanes_optab): Likewise.
8069 * internal-fn.def (MASK_LOAD_LANES): New internal function.
8070 (MASK_STORE_LANES): Likewise.
8071 * internal-fn.c (mask_load_lanes_direct): New macro.
8072 (mask_store_lanes_direct): Likewise.
8073 (expand_mask_load_optab_fn): Handle masked operations.
8074 (expand_mask_load_lanes_optab_fn): New macro.
8075 (expand_mask_store_optab_fn): Handle masked operations.
8076 (expand_mask_store_lanes_optab_fn): New macro.
8077 (direct_mask_load_lanes_optab_supported_p): Likewise.
8078 (direct_mask_store_lanes_optab_supported_p): Likewise.
8079 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
8081 (vect_load_lanes_supported): Likewise.
8082 * tree-vect-data-refs.c (strip_conversion): New function.
8083 (can_group_stmts_p): Likewise.
8084 (vect_analyze_data_ref_accesses): Use it instead of checking
8085 for a pair of assignments.
8086 (vect_store_lanes_supported): Take a masked_p parameter.
8087 (vect_load_lanes_supported): Likewise.
8088 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
8089 vect_store_lanes_supported and vect_load_lanes_supported.
8090 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8091 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
8092 parameter. Don't allow gaps for masked accesses.
8093 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
8094 and vect_load_lanes_supported.
8095 (get_load_store_type): Take a masked_p parameter and update
8096 call to get_group_load_store_type.
8097 (vectorizable_store): Update call to get_load_store_type.
8098 Handle IFN_MASK_STORE_LANES.
8099 (vectorizable_load): Update call to get_load_store_type.
8100 Handle IFN_MASK_LOAD_LANES.
8102 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8103 Alan Hayward <alan.hayward@arm.com>
8104 David Sherwood <david.sherwood@arm.com>
8106 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
8108 * config/aarch64/aarch64-protos.h
8109 (aarch64_sve_struct_memory_operand_p): Declare.
8110 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
8111 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
8112 (VPRED, vpred): Handle SVE structure modes.
8113 * config/aarch64/constraints.md (Utx): New constraint.
8114 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
8115 (aarch64_sve_struct_nonimmediate_operand): New predicates.
8116 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
8117 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
8118 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
8119 structure modes. Split into pieces after RA.
8120 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
8121 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
8123 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
8124 SVE structure modes.
8125 (aarch64_classify_address): Likewise.
8126 (sizetochar): Move earlier in file.
8127 (aarch64_print_operand): Handle SVE register lists.
8128 (aarch64_array_mode): New function.
8129 (aarch64_sve_struct_memory_operand_p): Likewise.
8130 (TARGET_ARRAY_MODE): Redefine.
8132 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8133 Alan Hayward <alan.hayward@arm.com>
8134 David Sherwood <david.sherwood@arm.com>
8136 * target.def (array_mode): New target hook.
8137 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
8138 * doc/tm.texi: Regenerate.
8139 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
8140 * hooks.c (hook_optmode_mode_uhwi_none): New function.
8141 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
8143 * stor-layout.c (mode_for_array): Likewise. Support polynomial
8146 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8147 Alan Hayward <alan.hayward@arm.com>
8148 David Sherwood <david.sherwood@arm.com>
8150 * fold-const.c (fold_binary_loc): Check the argument types
8151 rather than the result type when testing for a vector operation.
8153 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8155 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
8156 * doc/tm.texi: Regenerate.
8158 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8159 Alan Hayward <alan.hayward@arm.com>
8160 David Sherwood <david.sherwood@arm.com>
8162 * doc/invoke.texi (-msve-vector-bits=): Document new option.
8163 (sve): Document new AArch64 extension.
8164 * doc/md.texi (w): Extend the description of the AArch64
8165 constraint to include SVE vectors.
8166 (Upl, Upa): Document new AArch64 predicate constraints.
8167 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
8169 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
8170 (msve-vector-bits=): New option.
8171 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
8172 SVE when these are disabled.
8173 (sve): New extension.
8174 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
8175 modes. Adjust their number of units based on aarch64_sve_vg.
8176 (MAX_BITSIZE_MODE_ANY_MODE): Define.
8177 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
8178 aarch64_addr_query_type.
8179 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
8180 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
8181 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
8182 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
8183 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
8184 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
8185 (aarch64_simd_imm_zero_p): Delete.
8186 (aarch64_check_zero_based_sve_index_immediate): Declare.
8187 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8188 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8189 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8190 (aarch64_sve_float_mul_immediate_p): Likewise.
8191 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8193 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
8194 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
8195 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
8196 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
8197 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
8198 (aarch64_regmode_natural_size): Likewise.
8199 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
8200 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
8202 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
8203 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
8204 for VG and the SVE predicate registers.
8205 (V_ALIASES): Add a "z"-prefixed alias.
8206 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
8207 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
8208 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
8209 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
8210 (REG_CLASS_NAMES): Add entries for them.
8211 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
8212 and the predicate registers.
8213 (aarch64_sve_vg): Declare.
8214 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
8215 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
8216 (REGMODE_NATURAL_SIZE): Define.
8217 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
8219 * config/aarch64/aarch64.c: Include cfgrtl.h.
8220 (simd_immediate_info): Add a constructor for series vectors,
8221 and an associated step field.
8222 (aarch64_sve_vg): New variable.
8223 (aarch64_dbx_register_number): Handle VG and the predicate registers.
8224 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
8225 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
8226 (VEC_ANY_DATA, VEC_STRUCT): New constants.
8227 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
8228 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
8229 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
8230 (aarch64_get_mask_mode): New functions.
8231 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
8232 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
8233 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
8234 predicate modes and predicate registers. Explicitly restrict
8235 GPRs to modes of 16 bytes or smaller. Only allow FP registers
8236 to store a vector mode if it is recognized by
8237 aarch64_classify_vector_mode.
8238 (aarch64_regmode_natural_size): New function.
8239 (aarch64_hard_regno_caller_save_mode): Return the original mode
8241 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
8242 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
8243 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
8244 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
8246 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
8247 does not overlap dest if the function is frame-related. Handle
8249 (aarch64_split_add_offset): New function.
8250 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
8251 them aarch64_add_offset.
8252 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
8253 and update call to aarch64_sub_sp.
8254 (aarch64_add_cfa_expression): New function.
8255 (aarch64_expand_prologue): Pass extra temporary registers to the
8256 functions above. Handle the case in which we need to emit new
8257 DW_CFA_expressions for registers that were originally saved
8258 relative to the stack pointer, but now have to be expressed
8259 relative to the frame pointer.
8260 (aarch64_output_mi_thunk): Pass extra temporary registers to the
8262 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
8263 IP0 and IP1 values for SVE frames.
8264 (aarch64_expand_vec_series): New function.
8265 (aarch64_expand_sve_widened_duplicate): Likewise.
8266 (aarch64_expand_sve_const_vector): Likewise.
8267 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
8268 Handle SVE constants. Use emit_move_insn to move a force_const_mem
8269 into the register, rather than emitting a SET directly.
8270 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
8271 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
8272 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
8273 (offset_9bit_signed_scaled_p): New functions.
8274 (aarch64_replicate_bitmask_imm): New function.
8275 (aarch64_bitmask_imm): Use it.
8276 (aarch64_cannot_force_const_mem): Reject expressions involving
8277 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
8278 (aarch64_classify_index): Handle SVE indices, by requiring
8279 a plain register index with a scale that matches the element size.
8280 (aarch64_classify_address): Handle SVE addresses. Assert that
8281 the mode of the address is VOIDmode or an integer mode.
8282 Update call to aarch64_classify_symbol.
8283 (aarch64_classify_symbolic_expression): Update call to
8284 aarch64_classify_symbol.
8285 (aarch64_const_vec_all_in_range_p): New function.
8286 (aarch64_print_vector_float_operand): Likewise.
8287 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
8288 "vN" for FP registers with SVE modes. Handle (const ...) vectors
8289 and the FP immediates 1.0 and 0.5.
8290 (aarch64_print_address_internal): Handle SVE addresses.
8291 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
8292 (aarch64_regno_regclass): Handle predicate registers.
8293 (aarch64_secondary_reload): Handle big-endian reloads of SVE
8295 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
8296 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
8297 (aarch64_convert_sve_vector_bits): New function.
8298 (aarch64_override_options): Use it to handle -msve-vector-bits=.
8299 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8301 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
8302 Handle SVE vector and predicate modes. Accept VL-based constants
8303 that need only one temporary register, and VL offsets that require
8304 no temporary registers.
8305 (aarch64_conditional_register_usage): Mark the predicate registers
8306 as fixed if SVE isn't available.
8307 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
8308 Return true for SVE vector and predicate modes.
8309 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
8310 rather than an unsigned int. Handle SVE modes.
8311 (aarch64_preferred_simd_mode): Update call accordingly. Handle
8313 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
8315 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8316 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8317 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8318 (aarch64_sve_float_mul_immediate_p): New functions.
8319 (aarch64_sve_valid_immediate): New function.
8320 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
8321 Explicitly reject structure modes. Check for INDEX constants.
8322 Handle PTRUE and PFALSE constants.
8323 (aarch64_check_zero_based_sve_index_immediate): New function.
8324 (aarch64_simd_imm_zero_p): Delete.
8325 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
8326 vector modes. Accept constants in the range of CNT[BHWD].
8327 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
8328 ask for an Advanced SIMD mode.
8329 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
8330 (aarch64_simd_vector_alignment): Handle SVE predicates.
8331 (aarch64_vectorize_preferred_vector_alignment): New function.
8332 (aarch64_simd_vector_alignment_reachable): Use it instead of
8334 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
8335 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
8337 (MAX_VECT_LEN): Delete.
8338 (expand_vec_perm_d): Add a vec_flags field.
8339 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
8340 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
8341 (aarch64_evpc_ext): Don't apply a big-endian lane correction
8343 (aarch64_evpc_rev): Rename to...
8344 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
8345 (aarch64_evpc_rev_global): New function.
8346 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
8347 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
8349 (aarch64_evpc_sve_tbl): New function.
8350 (aarch64_expand_vec_perm_const_1): Update after rename of
8351 aarch64_evpc_rev. Handle SVE permutes too, trying
8352 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
8353 than aarch64_evpc_tbl.
8354 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
8355 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
8356 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
8357 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
8358 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
8359 (aarch64_expand_sve_vcond): New functions.
8360 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
8361 of aarch64_vector_mode_p.
8362 (aarch64_dwarf_poly_indeterminate_value): New function.
8363 (aarch64_compute_pressure_classes): Likewise.
8364 (aarch64_can_change_mode_class): Likewise.
8365 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
8366 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
8367 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
8368 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
8369 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
8370 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
8371 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
8372 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
8374 (Dn, Dl, Dr): Accept const as well as const_vector.
8375 (Dz): Likewise. Compare against CONST0_RTX.
8376 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
8377 of "vector" where appropriate.
8378 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
8379 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
8380 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
8381 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
8382 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
8383 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
8384 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
8385 (v_int_equiv): Extend to SVE modes.
8386 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
8388 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
8389 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
8390 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
8391 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
8392 (SVE_COND_FP_CMP): New int iterators.
8393 (perm_hilo): Handle the new unpack unspecs.
8394 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
8396 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
8397 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
8398 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
8399 (aarch64_equality_operator, aarch64_constant_vector_operand)
8400 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
8401 (aarch64_sve_nonimmediate_operand): Likewise.
8402 (aarch64_sve_general_operand): Likewise.
8403 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
8404 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
8405 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
8406 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
8407 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
8408 (aarch64_sve_float_arith_immediate): Likewise.
8409 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
8410 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
8411 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
8412 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
8413 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
8414 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
8415 (aarch64_sve_float_arith_operand): Likewise.
8416 (aarch64_sve_float_arith_with_sub_operand): Likewise.
8417 (aarch64_sve_float_mul_operand): Likewise.
8418 (aarch64_sve_vec_perm_operand): Likewise.
8419 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
8420 (aarch64_mov_operand): Accept const_poly_int and const_vector.
8421 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
8422 as well as const_vector.
8423 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
8424 in file. Use CONST0_RTX and CONSTM1_RTX.
8425 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
8426 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
8427 Use aarch64_simd_imm_zero.
8428 * config/aarch64/aarch64-sve.md: New file.
8429 * config/aarch64/aarch64.md: Include it.
8430 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
8431 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
8432 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
8433 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
8434 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
8435 (sve): New attribute.
8436 (enabled): Disable instructions with the sve attribute unless
8438 (movqi, movhi): Pass CONST_POLY_INT operaneds through
8439 aarch64_expand_mov_immediate.
8440 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
8441 CNT[BHSD] immediates.
8442 (movti): Split CONST_POLY_INT moves into two halves.
8443 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
8444 Split additions that need a temporary here if the destination
8445 is the stack pointer.
8446 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
8447 (*add<mode>3_poly_1): New instruction.
8448 (set_clobber_cc): New expander.
8450 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8452 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
8453 parameter and use it instead of GET_MODE_SIZE (innermode). Use
8454 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
8455 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
8456 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
8457 Change innermode from fixed_mode_size to machine_mode.
8458 (simplify_subreg): Update call accordingly. Handle a constant-sized
8459 subreg of a variable-length CONST_VECTOR.
8461 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8462 Alan Hayward <alan.hayward@arm.com>
8463 David Sherwood <david.sherwood@arm.com>
8465 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
8466 (add_offset_to_base): New function, split out from...
8467 (create_mem_ref): ...here. When handling a scale other than 1,
8468 check first whether the address is valid without the offset.
8469 Add it into the base if so, leaving the index and scale as-is.
8471 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8474 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
8475 fold_for_warn before checking if arg2 is INTEGER_CST.
8477 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
8479 * config/rs6000/predicates.md (load_multiple_operation): Delete.
8480 (store_multiple_operation): Delete.
8481 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
8482 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
8483 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
8484 guarded by TARGET_STRING.
8485 (rs6000_output_load_multiple): Delete.
8486 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
8487 OPTION_MASK_STRING / TARGET_STRING handling.
8488 (print_operand) <'N', 'O'>: Add comment that these are unused now.
8489 (const rs6000_opt_masks) <"string">: Change mask to 0.
8490 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
8491 (MASK_STRING): Delete.
8492 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
8494 (load_multiple): Delete.
8501 (store_multiple): Delete.
8508 (movmemsi_8reg): Delete.
8509 (corresponding unnamed define_insn): Delete.
8510 (movmemsi_6reg): Delete.
8511 (corresponding unnamed define_insn): Delete.
8512 (movmemsi_4reg): Delete.
8513 (corresponding unnamed define_insn): Delete.
8514 (movmemsi_2reg): Delete.
8515 (corresponding unnamed define_insn): Delete.
8516 (movmemsi_1reg): Delete.
8517 (corresponding unnamed define_insn): Delete.
8518 * config/rs6000/rs6000.opt (mno-string): New.
8519 (mstring): Replace by deprecation warning stub.
8520 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
8522 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8524 * regrename.c (regrename_do_replace): If replacing the same
8525 reg multiple times, try to reuse last created gen_raw_REG.
8528 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
8529 main to workaround a bug in GDB.
8531 2018-01-12 Tom de Vries <tom@codesourcery.com>
8534 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
8536 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
8538 PR rtl-optimization/80481
8539 * ira-color.c (get_cap_member): New function.
8540 (allocnos_conflict_by_live_ranges_p): Use it.
8541 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
8542 (setup_slot_coalesced_allocno_live_ranges): Ditto.
8544 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
8547 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
8548 (*saddl_se_1): Ditto.
8550 (*ssubl_se_1): Ditto.
8552 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8554 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
8555 rather than wi::to_widest for DR_INITs.
8556 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
8557 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
8558 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
8560 (vect_analyze_group_access_1): Note that here.
8562 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8564 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
8565 polynomial type sizes.
8567 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8569 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
8570 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
8571 (gimple_add_tmp_var): Likewise.
8573 2018-01-12 Martin Liska <mliska@suse.cz>
8575 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
8576 (gimple_alloc_sizes): Likewise.
8577 (dump_gimple_statistics): Use PRIu64 in printf format.
8578 * gimple.h: Change uint64_t to int.
8580 2018-01-12 Martin Liska <mliska@suse.cz>
8582 * tree-core.h: Use uint64_t instead of int.
8583 * tree.c (tree_node_counts): Likewise.
8584 (tree_node_sizes): Likewise.
8585 (dump_tree_statistics): Use PRIu64 in printf format.
8587 2018-01-12 Martin Liska <mliska@suse.cz>
8589 * Makefile.in: As qsort_chk is implemented in vec.c, add
8590 vec.o to linkage of gencfn-macros.
8591 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
8592 passing the info to record_node_allocation_statistics.
8593 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
8595 * ggc-common.c (struct ggc_usage): Add operator== and use
8596 it in operator< and compare function.
8597 * mem-stats.h (struct mem_usage): Likewise.
8598 * vec.c (struct vec_usage): Remove operator< and compare
8599 function. Can be simply inherited.
8601 2018-01-12 Martin Jambor <mjambor@suse.cz>
8604 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
8605 * tree-ssa-math-opts.c: Include domwalk.h.
8606 (convert_mult_to_fma_1): New function.
8607 (fma_transformation_info): New type.
8608 (fma_deferring_state): Likewise.
8609 (cancel_fma_deferring): New function.
8610 (result_of_phi): Likewise.
8611 (last_fma_candidate_feeds_initial_phi): Likewise.
8612 (convert_mult_to_fma): Added deferring logic, split actual
8613 transformation to convert_mult_to_fma_1.
8614 (math_opts_dom_walker): New type.
8615 (math_opts_dom_walker::after_dom_children): New method, body moved
8616 here from pass_optimize_widening_mul::execute, added deferring logic
8618 (pass_optimize_widening_mul::execute): Moved most of code to
8619 math_opts_dom_walker::after_dom_children.
8620 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
8621 * config/i386/i386.c (ix86_option_override_internal): Added
8622 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
8624 2018-01-12 Richard Biener <rguenther@suse.de>
8627 * dwarf2out.c (gen_variable_die): Do not reset old_die for
8628 inline instance vars.
8630 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
8633 * config/rx/rx.c (rx_is_restricted_memory_address):
8636 2018-01-12 Richard Biener <rguenther@suse.de>
8638 PR tree-optimization/80846
8639 * target.def (split_reduction): New target hook.
8640 * targhooks.c (default_split_reduction): New function.
8641 * targhooks.h (default_split_reduction): Declare.
8642 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
8643 target requests first reduce vectors by combining low and high
8645 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
8646 (get_vectype_for_scalar_type_and_size): Export.
8647 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
8648 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
8649 * doc/tm.texi: Regenerate.
8650 * config/i386/i386.c (ix86_split_reduction): Implement
8651 TARGET_VECTORIZE_SPLIT_REDUCTION.
8653 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8656 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
8657 in PIC mode except for TARGET_VXWORKS_RTP.
8658 * config/sparc/sparc.c: Include cfgrtl.h.
8659 (TARGET_INIT_PIC_REG): Define.
8660 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
8661 (sparc_pic_register_p): New predicate.
8662 (sparc_legitimate_address_p): Use it.
8663 (sparc_legitimize_pic_address): Likewise.
8664 (sparc_delegitimize_address): Likewise.
8665 (sparc_mode_dependent_address_p): Likewise.
8666 (gen_load_pcrel_sym): Remove 4th parameter.
8667 (load_got_register): Adjust call to above. Remove obsolete stuff.
8668 (sparc_expand_prologue): Do not call load_got_register here.
8669 (sparc_flat_expand_prologue): Likewise.
8670 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
8671 (sparc_use_pseudo_pic_reg): New function.
8672 (sparc_init_pic_reg): Likewise.
8673 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8674 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
8676 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
8678 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
8679 Add item for branch_cost.
8681 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8683 PR rtl-optimization/83565
8684 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
8685 not extend the result to a larger mode for rotate operations.
8686 (num_sign_bit_copies1): Likewise.
8688 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8691 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
8693 Use values-Xc.o for -pedantic.
8694 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
8696 2018-01-12 Martin Liska <mliska@suse.cz>
8699 * ipa-devirt.c (final_warning_record::grow_type_warnings):
8701 (possible_polymorphic_call_targets): Use it.
8702 (ipa_devirt): Likewise.
8704 2018-01-12 Martin Liska <mliska@suse.cz>
8706 * profile-count.h (enum profile_quality): Use 0 as invalid
8707 enum value of profile_quality.
8709 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
8711 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
8712 -mext-string options.
8714 2018-01-12 Richard Biener <rguenther@suse.de>
8716 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
8717 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
8718 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
8720 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
8722 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
8724 * configure.ac (--with-long-double-format): Add support for the
8725 configuration option to change the default long double format on
8727 * config.gcc (powerpc*-linux*-*): Likewise.
8728 * configure: Regenerate.
8729 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
8730 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
8731 used without modification.
8733 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8735 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
8736 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
8737 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
8738 MISC_BUILTIN_SPEC_BARRIER.
8739 (rs6000_init_builtins): Likewise.
8740 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
8742 (speculation_barrier): New define_insn.
8743 * doc/extend.texi: Document __builtin_speculation_barrier.
8745 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8748 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
8749 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
8750 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
8752 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
8753 integral modes instead of "ss" and "sd".
8754 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
8755 vectors with 32-bit and 64-bit elements.
8756 (vecdupssescalarmodesuffix): New mode attribute.
8757 (vec_dup<mode>): Use it.
8759 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
8762 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
8763 frame if argument is passed on stack.
8765 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8768 * ree.c (combine_reaching_defs): Optimize also
8769 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
8770 reg2=any_extend(exp); reg1=reg2;, formatting fix.
8772 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8775 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
8777 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8780 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
8781 after they are computed.
8783 2018-01-11 Bin Cheng <bin.cheng@arm.com>
8785 PR tree-optimization/83695
8786 * gimple-loop-linterchange.cc
8787 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
8788 reset cached scev information after interchange.
8789 (pass_linterchange::execute): Remove call to scev_reset_htab.
8791 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8793 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
8794 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
8795 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
8796 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
8797 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
8798 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
8799 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
8800 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
8801 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
8802 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
8803 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
8804 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
8805 (V_lane_reg): Likewise.
8806 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
8808 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
8809 (vfmal_lane_low<mode>_intrinsic,
8810 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
8811 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
8812 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
8813 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
8814 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
8815 vfmsl_lane_high<mode>_intrinsic): New define_insns.
8817 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8819 * config/arm/arm-cpus.in (fp16fml): New feature.
8820 (ALL_SIMD): Add fp16fml.
8821 (armv8.2-a): Add fp16fml as an option.
8822 (armv8.3-a): Likewise.
8823 (armv8.4-a): Add fp16fml as part of fp16.
8824 * config/arm/arm.h (TARGET_FP16FML): Define.
8825 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
8827 * config/arm/arm-modes.def (V2HF): Define.
8828 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
8829 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
8830 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
8831 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
8832 vfmsl_low, vfmsl_high): New set of builtins.
8833 * config/arm/iterators.md (PLUSMINUS): New code iterator.
8834 (vfml_op): New code attribute.
8835 (VFMLHALVES): New int iterator.
8836 (VFML, VFMLSEL): New mode attributes.
8837 (V_reg): Define mapping for V2HF.
8838 (V_hi, V_lo): New mode attributes.
8839 (VF_constraint): Likewise.
8840 (vfml_half, vfml_half_selector): New int attributes.
8841 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
8843 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
8844 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
8846 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
8847 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
8848 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
8849 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
8851 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
8852 Document new effective target and option set.
8854 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8856 * config/arm/arm-cpus.in (armv8_4): New feature.
8857 (ARMv8_4a): New fgroup.
8858 (armv8.4-a): New arch.
8859 * config/arm/arm-tables.opt: Regenerate.
8860 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
8861 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
8862 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
8863 Add matching rules for -march=armv8.4-a and extensions.
8864 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
8866 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
8869 * config/rx/rx.md (BW): New mode attribute.
8870 (sync_lock_test_and_setsi): Add mode suffix to insn output.
8872 2018-01-11 Richard Biener <rguenther@suse.de>
8874 PR tree-optimization/83435
8875 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
8876 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
8877 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
8879 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8880 Alan Hayward <alan.hayward@arm.com>
8881 David Sherwood <david.sherwood@arm.com>
8883 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
8885 (aarch64_classify_address): Initialize it. Track polynomial offsets.
8886 (aarch64_print_address_internal): Use it to check for a zero offset.
8888 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8889 Alan Hayward <alan.hayward@arm.com>
8890 David Sherwood <david.sherwood@arm.com>
8892 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
8893 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
8894 Return a poly_int64 rather than a HOST_WIDE_INT.
8895 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
8896 rather than a HOST_WIDE_INT.
8897 * config/aarch64/aarch64.h (aarch64_frame): Protect with
8898 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
8899 hard_fp_offset, frame_size, initial_adjust, callee_offset and
8900 final_offset from HOST_WIDE_INT to poly_int64.
8901 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
8902 to_constant when getting the number of units in an Advanced SIMD
8904 (aarch64_builtin_vectorized_function): Check for a constant number
8906 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
8908 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
8909 attribute instead of GET_MODE_NUNITS.
8910 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
8911 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
8912 GET_MODE_SIZE for fixed-size registers.
8913 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
8914 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
8915 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
8916 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
8917 (aarch64_print_operand, aarch64_print_address_internal)
8918 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
8919 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
8920 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
8921 Handle polynomial GET_MODE_SIZE.
8922 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
8923 wider than SImode without modification.
8924 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
8925 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
8926 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
8927 passing and returning SVE modes.
8928 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
8929 rather than GEN_INT.
8930 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
8931 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
8932 (aarch64_allocate_and_probe_stack_space): Likewise.
8933 (aarch64_layout_frame): Cope with polynomial offsets.
8934 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
8935 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
8937 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
8938 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
8939 poly_int64 rather than a HOST_WIDE_INT.
8940 (aarch64_get_separate_components, aarch64_process_components)
8941 (aarch64_expand_prologue, aarch64_expand_epilogue)
8942 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
8943 (aarch64_anchor_offset): New function, split out from...
8944 (aarch64_legitimize_address): ...here.
8945 (aarch64_builtin_vectorization_cost): Handle polynomial
8946 TYPE_VECTOR_SUBPARTS.
8947 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
8949 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
8950 number of elements from the PARALLEL rather than the mode.
8951 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
8952 rather than GET_MODE_BITSIZE.
8953 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
8954 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
8955 (aarch64_expand_vec_perm_const_1): Handle polynomial
8956 d->perm.length () and d->perm elements.
8957 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
8958 Apply to_constant to d->perm elements.
8959 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
8960 polynomial CONST_VECTOR_NUNITS.
8961 (aarch64_move_pointer): Take amount as a poly_int64 rather
8963 (aarch64_progress_pointer): Avoid temporary variable.
8964 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
8965 the mode attribute instead of GET_MODE.
8967 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8968 Alan Hayward <alan.hayward@arm.com>
8969 David Sherwood <david.sherwood@arm.com>
8971 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
8972 x exists before using it.
8973 (aarch64_add_constant_internal): Rename to...
8974 (aarch64_add_offset_1): ...this. Replace regnum with separate
8975 src and dest rtxes. Handle the case in which they're different,
8976 including when the offset is zero. Replace scratchreg with an rtx.
8977 Use 2 additions if there is no spare register into which we can
8978 move a 16-bit constant.
8979 (aarch64_add_constant): Delete.
8980 (aarch64_add_offset): Replace reg with separate src and dest
8981 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
8982 Use aarch64_add_offset_1.
8983 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
8984 an rtx rather than an int. Take the delta as a poly_int64
8985 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
8986 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
8987 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
8988 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
8989 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
8991 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
8992 aarch64_add_constant.
8994 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8996 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
8997 Use scalar_float_mode.
8999 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9001 * config/aarch64/aarch64-simd.md
9002 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
9003 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
9004 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
9005 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
9006 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
9007 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
9008 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
9009 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
9010 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
9011 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
9013 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9016 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
9017 targ_options->x_arm_arch_string is non NULL.
9019 2018-01-11 Tamar Christina <tamar.christina@arm.com>
9021 * config/aarch64/aarch64.h
9022 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
9024 2018-01-11 Sudakshina Das <sudi.das@arm.com>
9027 * expmed.c (emit_store_flag_force): Swap if const op0
9028 and change VOIDmode to mode of op0.
9030 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9032 PR rtl-optimization/83761
9033 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
9034 than bytes to mode_for_size.
9036 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9039 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
9040 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
9043 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9046 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
9047 when in layout mode.
9048 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
9049 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
9052 2018-01-10 Michael Collison <michael.collison@arm.com>
9054 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
9055 * config/aarch64/aarch64-option-extension.def: Add
9056 AARCH64_OPT_EXTENSION of 'fp16fml'.
9057 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9058 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
9059 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
9060 * config/aarch64/constraints.md (Ui7): New constraint.
9061 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
9062 (VFMLA_SEL_W): Ditto.
9065 (VFMLA16_LOW): New int iterator.
9066 (VFMLA16_HIGH): Ditto.
9067 (UNSPEC_FMLAL): New unspec.
9068 (UNSPEC_FMLSL): Ditto.
9069 (UNSPEC_FMLAL2): Ditto.
9070 (UNSPEC_FMLSL2): Ditto.
9071 (f16mac): New code attribute.
9072 * config/aarch64/aarch64-simd-builtins.def
9073 (aarch64_fmlal_lowv2sf): Ditto.
9074 (aarch64_fmlsl_lowv2sf): Ditto.
9075 (aarch64_fmlalq_lowv4sf): Ditto.
9076 (aarch64_fmlslq_lowv4sf): Ditto.
9077 (aarch64_fmlal_highv2sf): Ditto.
9078 (aarch64_fmlsl_highv2sf): Ditto.
9079 (aarch64_fmlalq_highv4sf): Ditto.
9080 (aarch64_fmlslq_highv4sf): Ditto.
9081 (aarch64_fmlal_lane_lowv2sf): Ditto.
9082 (aarch64_fmlsl_lane_lowv2sf): Ditto.
9083 (aarch64_fmlal_laneq_lowv2sf): Ditto.
9084 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
9085 (aarch64_fmlalq_lane_lowv4sf): Ditto.
9086 (aarch64_fmlsl_lane_lowv4sf): Ditto.
9087 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
9088 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
9089 (aarch64_fmlal_lane_highv2sf): Ditto.
9090 (aarch64_fmlsl_lane_highv2sf): Ditto.
9091 (aarch64_fmlal_laneq_highv2sf): Ditto.
9092 (aarch64_fmlsl_laneq_highv2sf): Ditto.
9093 (aarch64_fmlalq_lane_highv4sf): Ditto.
9094 (aarch64_fmlsl_lane_highv4sf): Ditto.
9095 (aarch64_fmlalq_laneq_highv4sf): Ditto.
9096 (aarch64_fmlsl_laneq_highv4sf): Ditto.
9097 * config/aarch64/aarch64-simd.md:
9098 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
9099 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9100 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
9101 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9102 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
9103 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
9104 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
9105 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
9106 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
9107 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
9108 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
9109 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
9110 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
9111 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
9112 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
9113 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
9114 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
9115 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
9116 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
9117 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
9118 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
9119 (vfmlsl_low_u32): Ditto.
9120 (vfmlalq_low_u32): Ditto.
9121 (vfmlslq_low_u32): Ditto.
9122 (vfmlal_high_u32): Ditto.
9123 (vfmlsl_high_u32): Ditto.
9124 (vfmlalq_high_u32): Ditto.
9125 (vfmlslq_high_u32): Ditto.
9126 (vfmlal_lane_low_u32): Ditto.
9127 (vfmlsl_lane_low_u32): Ditto.
9128 (vfmlal_laneq_low_u32): Ditto.
9129 (vfmlsl_laneq_low_u32): Ditto.
9130 (vfmlalq_lane_low_u32): Ditto.
9131 (vfmlslq_lane_low_u32): Ditto.
9132 (vfmlalq_laneq_low_u32): Ditto.
9133 (vfmlslq_laneq_low_u32): Ditto.
9134 (vfmlal_lane_high_u32): Ditto.
9135 (vfmlsl_lane_high_u32): Ditto.
9136 (vfmlal_laneq_high_u32): Ditto.
9137 (vfmlsl_laneq_high_u32): Ditto.
9138 (vfmlalq_lane_high_u32): Ditto.
9139 (vfmlslq_lane_high_u32): Ditto.
9140 (vfmlalq_laneq_high_u32): Ditto.
9141 (vfmlslq_laneq_high_u32): Ditto.
9142 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
9143 (AARCH64_FL_FOR_ARCH8_4): New.
9144 (AARCH64_ISA_F16FML): New ISA flag.
9145 (TARGET_F16FML): New feature flag for fp16fml.
9146 (doc/invoke.texi): Document new fp16fml option.
9148 2018-01-10 Michael Collison <michael.collison@arm.com>
9150 * config/aarch64/aarch64-builtins.c:
9151 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
9152 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9153 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
9154 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
9155 (AARCH64_ISA_SHA3): New ISA flag.
9156 (TARGET_SHA3): New feature flag for sha3.
9157 * config/aarch64/iterators.md (sha512_op): New int attribute.
9158 (CRYPTO_SHA512): New int iterator.
9159 (UNSPEC_SHA512H): New unspec.
9160 (UNSPEC_SHA512H2): Ditto.
9161 (UNSPEC_SHA512SU0): Ditto.
9162 (UNSPEC_SHA512SU1): Ditto.
9163 * config/aarch64/aarch64-simd-builtins.def
9164 (aarch64_crypto_sha512hqv2di): New builtin.
9165 (aarch64_crypto_sha512h2qv2di): Ditto.
9166 (aarch64_crypto_sha512su0qv2di): Ditto.
9167 (aarch64_crypto_sha512su1qv2di): Ditto.
9168 (aarch64_eor3qv8hi): Ditto.
9169 (aarch64_rax1qv2di): Ditto.
9170 (aarch64_xarqv2di): Ditto.
9171 (aarch64_bcaxqv8hi): Ditto.
9172 * config/aarch64/aarch64-simd.md:
9173 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
9174 (aarch64_crypto_sha512su0qv2di): Ditto.
9175 (aarch64_crypto_sha512su1qv2di): Ditto.
9176 (aarch64_eor3qv8hi): Ditto.
9177 (aarch64_rax1qv2di): Ditto.
9178 (aarch64_xarqv2di): Ditto.
9179 (aarch64_bcaxqv8hi): Ditto.
9180 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
9181 (vsha512h2q_u64): Ditto.
9182 (vsha512su0q_u64): Ditto.
9183 (vsha512su1q_u64): Ditto.
9184 (veor3q_u16): Ditto.
9185 (vrax1q_u64): Ditto.
9187 (vbcaxq_u16): Ditto.
9188 * config/arm/types.md (crypto_sha512): New type attribute.
9189 (crypto_sha3): Ditto.
9190 (doc/invoke.texi): Document new sha3 option.
9192 2018-01-10 Michael Collison <michael.collison@arm.com>
9194 * config/aarch64/aarch64-builtins.c:
9195 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
9196 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9197 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
9198 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
9199 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
9200 (AARCH64_ISA_SM4): New ISA flag.
9201 (TARGET_SM4): New feature flag for sm4.
9202 * config/aarch64/aarch64-simd-builtins.def
9203 (aarch64_sm3ss1qv4si): Ditto.
9204 (aarch64_sm3tt1aq4si): Ditto.
9205 (aarch64_sm3tt1bq4si): Ditto.
9206 (aarch64_sm3tt2aq4si): Ditto.
9207 (aarch64_sm3tt2bq4si): Ditto.
9208 (aarch64_sm3partw1qv4si): Ditto.
9209 (aarch64_sm3partw2qv4si): Ditto.
9210 (aarch64_sm4eqv4si): Ditto.
9211 (aarch64_sm4ekeyqv4si): Ditto.
9212 * config/aarch64/aarch64-simd.md:
9213 (aarch64_sm3ss1qv4si): Ditto.
9214 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
9215 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
9216 (aarch64_sm4eqv4si): Ditto.
9217 (aarch64_sm4ekeyqv4si): Ditto.
9218 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
9219 (sm3part_op): Ditto.
9220 (CRYPTO_SM3TT): Ditto.
9221 (CRYPTO_SM3PART): Ditto.
9222 (UNSPEC_SM3SS1): New unspec.
9223 (UNSPEC_SM3TT1A): Ditto.
9224 (UNSPEC_SM3TT1B): Ditto.
9225 (UNSPEC_SM3TT2A): Ditto.
9226 (UNSPEC_SM3TT2B): Ditto.
9227 (UNSPEC_SM3PARTW1): Ditto.
9228 (UNSPEC_SM3PARTW2): Ditto.
9229 (UNSPEC_SM4E): Ditto.
9230 (UNSPEC_SM4EKEY): Ditto.
9231 * config/aarch64/constraints.md (Ui2): New constraint.
9232 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
9233 * config/arm/types.md (crypto_sm3): New type attribute.
9234 (crypto_sm4): Ditto.
9235 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
9236 (vsm3tt1aq_u32): Ditto.
9237 (vsm3tt1bq_u32): Ditto.
9238 (vsm3tt2aq_u32): Ditto.
9239 (vsm3tt2bq_u32): Ditto.
9240 (vsm3partw1q_u32): Ditto.
9241 (vsm3partw2q_u32): Ditto.
9242 (vsm4eq_u32): Ditto.
9243 (vsm4ekeyq_u32): Ditto.
9244 (doc/invoke.texi): Document new sm4 option.
9246 2018-01-10 Michael Collison <michael.collison@arm.com>
9248 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
9249 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
9250 (AARCH64_FL_FOR_ARCH8_4): New.
9251 (AARCH64_FL_V8_4): New flag.
9252 (doc/invoke.texi): Document new armv8.4-a option.
9254 2018-01-10 Michael Collison <michael.collison@arm.com>
9256 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9257 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
9258 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
9259 * config/aarch64/aarch64-option-extension.def: Add
9260 AARCH64_OPT_EXTENSION of 'sha2'.
9261 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
9262 (crypto): Disable sha2 and aes if crypto disabled.
9263 (crypto): Enable aes and sha2 if enabled.
9264 (simd): Disable sha2 and aes if simd disabled.
9265 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
9267 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
9268 (TARGET_SHA2): New feature flag for sha2.
9269 (TARGET_AES): New feature flag for aes.
9270 * config/aarch64/aarch64-simd.md:
9271 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
9272 conditional on TARGET_AES.
9273 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
9274 (aarch64_crypto_sha1hsi): Make pattern conditional
9276 (aarch64_crypto_sha1hv4si): Ditto.
9277 (aarch64_be_crypto_sha1hv4si): Ditto.
9278 (aarch64_crypto_sha1su1v4si): Ditto.
9279 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
9280 (aarch64_crypto_sha1su0v4si): Ditto.
9281 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
9282 (aarch64_crypto_sha256su0v4si): Ditto.
9283 (aarch64_crypto_sha256su1v4si): Ditto.
9284 (doc/invoke.texi): Document new aes and sha2 options.
9286 2018-01-10 Martin Sebor <msebor@redhat.com>
9288 PR tree-optimization/83781
9289 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
9292 2018-01-11 Martin Sebor <msebor@gmail.com>
9293 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9295 PR tree-optimization/83501
9296 PR tree-optimization/81703
9298 * tree-ssa-strlen.c (get_string_cst): Rename...
9299 (get_string_len): ...to this. Handle global constants.
9300 (handle_char_store): Adjust.
9302 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
9303 Jim Wilson <jimw@sifive.com>
9305 * config/riscv/riscv-protos.h (riscv_output_return): New.
9306 * config/riscv/riscv.c (struct machine_function): New naked_p field.
9307 (riscv_attribute_table, riscv_output_return),
9308 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
9309 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
9310 (riscv_compute_frame_info): Only compute frame->mask if not a naked
9312 (riscv_expand_prologue): Add early return for naked function.
9313 (riscv_expand_epilogue): Likewise.
9314 (riscv_function_ok_for_sibcall): Return false for naked function.
9315 (riscv_set_current_function): New.
9316 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
9317 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
9318 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
9319 * doc/extend.texi (RISC-V Function Attributes): New.
9321 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
9323 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
9324 check for 128-bit long double before checking TCmode.
9325 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
9326 128-bit long doubles before checking TFmode or TCmode.
9327 (FLOAT128_IBM_P): Likewise.
9329 2018-01-10 Martin Sebor <msebor@redhat.com>
9331 PR tree-optimization/83671
9332 * builtins.c (c_strlen): Unconditionally return zero for the empty
9334 Use -Warray-bounds for warnings.
9335 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
9336 for non-constant array indices with COMPONENT_REF, arrays of
9337 arrays, and pointers to arrays.
9338 (gimple_fold_builtin_strlen): Determine and set length range for
9339 non-constant character arrays.
9341 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
9344 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
9347 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
9349 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
9351 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9354 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
9355 VECTOR_MEM_ALTIVEC_OR_VSX_P.
9356 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
9357 indexed_or_indirect_operand predicate.
9358 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
9359 (*vsx_le_perm_load_v8hi): Likewise.
9360 (*vsx_le_perm_load_v16qi): Likewise.
9361 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
9362 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
9363 (*vsx_le_perm_store_v8hi): Likewise.
9364 (*vsx_le_perm_store_v16qi): Likewise.
9365 (eight unnamed splitters): Likewise.
9367 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9369 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
9370 * config/rs6000/emmintrin.h: Likewise.
9371 * config/rs6000/mmintrin.h: Likewise.
9372 * config/rs6000/xmmintrin.h: Likewise.
9374 2018-01-10 David Malcolm <dmalcolm@redhat.com>
9377 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
9379 * tree.c (tree_nop_conversion): Return true for location wrapper
9381 (maybe_wrap_with_location): New function.
9382 (selftest::check_strip_nops): New function.
9383 (selftest::test_location_wrappers): New function.
9384 (selftest::tree_c_tests): Call it.
9385 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
9386 (maybe_wrap_with_location): New decl.
9387 (EXPR_LOCATION_WRAPPER_P): New macro.
9388 (location_wrapper_p): New inline function.
9389 (tree_strip_any_location_wrapper): New inline function.
9391 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
9394 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
9395 stack_realign_offset for the largest alignment of stack slot
9397 (ix86_find_max_used_stack_alignment): New function.
9398 (ix86_finalize_stack_frame_flags): Use it. Set
9399 max_used_stack_alignment if we don't realign stack.
9400 * config/i386/i386.h (machine_function): Add
9401 max_used_stack_alignment.
9403 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
9405 * config/arm/arm.opt (-mbranch-cost): New option.
9406 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
9409 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
9412 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
9413 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
9415 2018-01-10 Richard Biener <rguenther@suse.de>
9418 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
9419 early out so it also covers the case where we have a non-NULL
9422 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9424 PR tree-optimization/83753
9425 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
9426 for non-strided grouped accesses if the number of elements is 1.
9428 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9431 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
9432 * i386.h (TARGET_USE_GATHER): Define.
9433 * x86-tune.def (X86_TUNE_USE_GATHER): New.
9435 2018-01-10 Martin Liska <mliska@suse.cz>
9438 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
9439 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
9441 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
9442 CLEANUP_NO_PARTITIONING is not set.
9444 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9446 * doc/rtl.texi: Remove documentation of (const ...) wrappers
9447 for vectors, as a partial revert of r254296.
9448 * rtl.h (const_vec_p): Delete.
9449 (const_vec_duplicate_p): Don't test for vector CONSTs.
9450 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
9451 * expmed.c (make_tree): Likewise.
9454 * common.md (E, F): Use CONSTANT_P instead of checking for
9456 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9457 checking for CONST_VECTOR.
9459 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9462 * predict.c (force_edge_cold): Handle in more sane way edges
9465 2018-01-09 Carl Love <cel@us.ibm.com>
9467 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
9469 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
9470 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
9471 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
9472 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
9473 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
9474 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
9475 * config/rs6000/rs6000-protos.h: Add extern defition for
9476 rs6000_generate_float2_double_code.
9477 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
9479 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
9480 (float2_v2df): Add define_expand.
9482 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
9485 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
9486 op_mode in the force_to_mode call.
9488 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9490 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
9491 instead of checking each element individually.
9492 (aarch64_evpc_uzp): Likewise.
9493 (aarch64_evpc_zip): Likewise.
9494 (aarch64_evpc_ext): Likewise.
9495 (aarch64_evpc_rev): Likewise.
9496 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
9497 instead of checking each element individually. Return true without
9499 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
9500 whether all selected elements come from the same input, instead of
9501 checking each element individually. Remove calls to gen_rtx_REG,
9502 start_sequence and end_sequence and instead assert that no rtl is
9505 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9507 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
9508 order of HIGH and CONST checks.
9510 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9512 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
9513 if the destination isn't an SSA_NAME.
9515 2018-01-09 Richard Biener <rguenther@suse.de>
9517 PR tree-optimization/83668
9518 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
9520 (canonicalize_loop_form): ... here, renamed from ...
9521 (canonicalize_loop_closed_ssa_form): ... this and amended to
9522 swap successor edges for loop exit blocks to make us use
9523 the RPO order we need for initial schedule generation.
9525 2018-01-09 Joseph Myers <joseph@codesourcery.com>
9527 PR tree-optimization/64811
9528 * match.pd: When optimizing comparisons with Inf, avoid
9529 introducing or losing exceptions from comparisons with NaN.
9531 2018-01-09 Martin Liska <mliska@suse.cz>
9534 * asan.c (shadow_mem_size): Add gcc_assert.
9536 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
9538 Don't save registers in main().
9541 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
9542 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
9543 * config/avr/avr.c (avr_set_current_function): Don't error if
9544 naked, OS_task or OS_main are specified at the same time.
9545 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
9547 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
9549 * common/config/avr/avr-common.c (avr_option_optimization_table):
9550 Switch on -mmain-is-OS_task for optimizing compilations.
9552 2018-01-09 Richard Biener <rguenther@suse.de>
9554 PR tree-optimization/83572
9555 * graphite.c: Include cfganal.h.
9556 (graphite_transform_loops): Connect infinite loops to exit
9557 and remove fake edges at the end.
9559 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9561 * ipa-inline.c (edge_badness): Revert accidental checkin.
9563 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9566 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
9567 symbols; not inline clones.
9569 2018-01-09 Jakub Jelinek <jakub@redhat.com>
9572 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
9573 hard registers. Formatting fixes.
9575 PR preprocessor/83722
9576 * gcc.c (try_generate_repro): Pass
9577 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
9578 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
9581 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
9582 Kito Cheng <kito.cheng@gmail.com>
9584 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
9585 (riscv_leaf_function_p): Delete.
9586 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
9588 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9590 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
9592 (do_ifelse): New function.
9593 (do_isel): New function.
9594 (do_sub3): New function.
9595 (do_add3): New function.
9596 (do_load_mask_compare): New function.
9597 (do_overlap_load_compare): New function.
9598 (expand_compare_loop): New function.
9599 (expand_block_compare): Call expand_compare_loop() when appropriate.
9600 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
9602 (-mblock-compare-inline-loop-limit): New option.
9604 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9607 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
9608 Reverse order of second and third operands in first alternative.
9609 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
9610 of first and second elements in UNSPEC_VPERMR vector.
9611 (altivec_expand_vec_perm_le): Likewise.
9613 2018-01-08 Jeff Law <law@redhat.com>
9615 PR rtl-optimizatin/81308
9616 * tree-switch-conversion.c (cfg_altered): New file scoped static.
9617 (process_switch): If group_case_labels makes a change, then set
9619 (pass_convert_switch::execute): If a switch is converted, then
9620 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
9622 PR rtl-optimization/81308
9623 * recog.c (split_all_insns): Conditionally cleanup the CFG after
9626 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
9628 PR target/83663 - Revert r255946
9629 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
9630 generation for cases where splatting a value is not useful.
9631 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
9632 across a vec_duplicate and a paradoxical subreg forming a vector
9633 mode to a vec_concat.
9635 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9637 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
9638 -march=armv8.3-a variants.
9639 * config/arm/t-multilib: Likewise.
9640 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
9642 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9644 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
9646 (cceq_ior_compare_complement): Give it a name so I can use it, and
9647 change boolean_or_operator predicate to boolean_operator so it can
9648 be used to generate a crand.
9649 (eqne): New code iterator.
9650 (bd/bd_neg): New code_attrs.
9651 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
9652 a single define_insn.
9653 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
9654 decrement (bdnzt/bdnzf/bdzt/bdzf).
9655 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
9656 with the new names of the branch decrement patterns, and added the
9657 names of the branch decrement conditional patterns.
9659 2018-01-08 Richard Biener <rguenther@suse.de>
9661 PR tree-optimization/83563
9662 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
9665 2018-01-08 Richard Biener <rguenther@suse.de>
9668 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
9670 2018-01-08 Richard Biener <rguenther@suse.de>
9672 PR tree-optimization/83685
9673 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
9674 references to abnormals.
9676 2018-01-08 Richard Biener <rguenther@suse.de>
9679 * dwarf2out.c (output_indirect_strings): Handle empty
9680 skeleton_debug_str_hash.
9681 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
9683 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9685 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
9686 (emit_store_direct): Likewise.
9687 (arc_trampoline_adjust_address): Likewise.
9688 (arc_asm_trampoline_template): New function.
9689 (arc_initialize_trampoline): Use asm_trampoline_template.
9690 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
9691 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
9692 * config/arc/arc.md (flush_icache): Delete pattern.
9694 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9696 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
9697 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
9700 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9703 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
9704 by not USED_FOR_TARGET.
9705 (make_pass_resolve_sw_modes): Likewise.
9707 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9709 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
9712 2018-01-08 Richard Biener <rguenther@suse.de>
9715 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
9717 2018-01-08 Richard Biener <rguenther@suse.de>
9720 * match.pd ((t * 2) / 2) -> t): Add missing :c.
9722 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
9725 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
9726 basic blocks with a small number of successors.
9727 (convert_control_dep_chain_into_preds): Improve handling of
9729 (dump_predicates): Split apart into...
9730 (dump_pred_chain): ...here...
9731 (dump_pred_info): ...and here.
9732 (can_one_predicate_be_invalidated_p): Add debugging printfs.
9733 (can_chain_union_be_invalidated_p): Improve check for invalidation
9735 (uninit_uses_cannot_happen): Avoid unnecessary if
9736 convert_control_dep_chain_into_preds yielded nothing.
9738 2018-01-06 Martin Sebor <msebor@redhat.com>
9740 PR tree-optimization/83640
9741 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
9742 subtracting negative offset from size.
9743 (builtin_access::overlap): Adjust offset bounds of the access to fall
9744 within the size of the object if possible.
9746 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
9748 PR rtl-optimization/83699
9749 * expmed.c (extract_bit_field_1): Restrict the vector usage of
9750 extract_bit_field_as_subreg to cases in which the extracted
9751 value is also a vector.
9753 * lra-constraints.c (process_alt_operands): Test for the equivalence
9754 substitutions when detecting a possible reload cycle.
9756 2018-01-06 Jakub Jelinek <jakub@redhat.com>
9759 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
9760 by default if flag_selective_schedling{,2}. Formatting fixes.
9762 PR rtl-optimization/83682
9763 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
9764 if it has non-VECTOR_MODE element mode.
9765 (vec_duplicate_p): Likewise.
9768 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
9769 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
9771 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9774 * config/i386/i386-builtin.def
9775 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
9776 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
9777 Require also OPTION_MASK_ISA_AVX512F in addition to
9778 OPTION_MASK_ISA_GFNI.
9779 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
9780 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
9781 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
9782 to OPTION_MASK_ISA_GFNI.
9783 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
9784 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
9785 OPTION_MASK_ISA_AVX512BW.
9786 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
9787 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
9788 addition to OPTION_MASK_ISA_GFNI.
9789 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
9790 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
9791 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
9792 to OPTION_MASK_ISA_GFNI.
9793 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
9794 a requirement for all ISAs rather than any of them with a few
9796 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
9798 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
9799 bitmasks to be enabled with 3 exceptions, instead of requiring any
9800 enabled ISA with lots of exceptions.
9801 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
9802 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
9803 Change avx512bw in isa attribute to avx512f.
9804 * config/i386/sgxintrin.h: Add license boilerplate.
9805 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
9806 to __AVX512F__ and __AVX512VL to __AVX512VL__.
9807 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
9808 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
9810 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
9811 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
9812 temporarily sse2 rather than sse if not enabled already.
9815 * config/i386/sse.md (VI248_VLBW): Rename to ...
9816 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
9817 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
9818 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
9819 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
9820 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
9821 mode iterator instead of VI248_VLBW.
9823 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
9825 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
9826 (record_modified): Skip clobbers; add debug output.
9827 (param_change_prob): Use sreal frequencies.
9829 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9831 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
9832 punt for user-aligned variables.
9834 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9836 * tree-chrec.c (chrec_contains_symbols): Return true for
9839 2018-01-05 Sudakshina Das <sudi.das@arm.com>
9842 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
9843 of (x|y) == x for BICS pattern.
9845 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9847 PR tree-optimization/83605
9848 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
9849 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
9852 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
9854 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
9855 * config/epiphany/rtems.h: New file.
9857 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9858 Uros Bizjak <ubizjak@gmail.com>
9861 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
9862 QIreg_operand instead of register_operand predicate.
9863 * config/i386/i386.c (ix86_rop_should_change_byte_p,
9864 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
9865 comments instead of -fmitigate[-_]rop.
9867 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9870 * cgraphunit.c (symbol_table::compile): Switch to text_section
9871 before calling assembly_start debug hook.
9872 * run-rtl-passes.c (run_rtl_passes): Likewise.
9875 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9877 * tree-vrp.c (extract_range_from_binary_expr_1): Check
9878 range_int_cst_p rather than !symbolic_range_p before calling
9879 extract_range_from_multiplicative_op_1.
9881 2018-01-04 Jeff Law <law@redhat.com>
9883 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
9884 redundant test in assertion.
9886 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9888 * doc/rtl.texi: Document machine_mode wrapper classes.
9890 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9892 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
9895 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9897 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
9898 the VEC_PERM_EXPR fold to fail.
9900 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9903 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
9904 to switched_sections.
9906 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9909 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
9912 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
9915 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
9916 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
9918 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9921 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
9922 is BLKmode and bitpos not zero or mode change is needed.
9924 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9927 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
9930 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
9933 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
9934 instead of MULT rtx. Update all corresponding splitters.
9936 (*ssub<modesuffix>): Ditto.
9938 (*cmp_sadd_di): Update split patterns.
9939 (*cmp_sadd_si): Ditto.
9940 (*cmp_sadd_sidi): Ditto.
9941 (*cmp_ssub_di): Ditto.
9942 (*cmp_ssub_si): Ditto.
9943 (*cmp_ssub_sidi): Ditto.
9944 * config/alpha/predicates.md (const23_operand): New predicate.
9945 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
9946 Look for ASHIFT, not MULT inner operand.
9947 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
9949 2018-01-04 Martin Liska <mliska@suse.cz>
9951 PR gcov-profile/83669
9952 * gcov.c (output_intermediate_file): Add version to intermediate
9954 * doc/gcov.texi: Document new field 'version' in intermediate
9955 file format. Fix location of '-k' option of gcov command.
9957 2018-01-04 Martin Liska <mliska@suse.cz>
9960 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
9962 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9964 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
9966 2018-01-03 Martin Sebor <msebor@redhat.com>
9968 PR tree-optimization/83655
9969 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
9970 checking calls with invalid arguments.
9972 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9974 * tree-vect-stmts.c (vect_get_store_rhs): New function.
9975 (vectorizable_mask_load_store): Delete.
9976 (vectorizable_call): Return false for masked loads and stores.
9977 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
9978 instead of gimple_assign_rhs1.
9979 (vectorizable_load): Handle IFN_MASK_LOAD.
9980 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
9982 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9984 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
9986 (vectorizable_mask_load_store): ...here.
9987 (vectorizable_load): ...and here.
9989 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9991 * tree-vect-stmts.c (vect_build_all_ones_mask)
9992 (vect_build_zero_merge_argument): New functions, split out from...
9993 (vectorizable_load): ...here.
9995 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9997 * tree-vect-stmts.c (vect_check_store_rhs): New function,
9999 (vectorizable_mask_load_store): ...here.
10000 (vectorizable_store): ...and here.
10002 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10004 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
10006 (vectorizable_mask_load_store): ...here.
10008 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10010 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
10011 (vect_model_store_cost): Take a vec_load_store_type instead of a
10013 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
10014 (vect_model_store_cost): Take a vec_load_store_type instead of a
10016 (vectorizable_mask_load_store): Update accordingly.
10017 (vectorizable_store): Likewise.
10018 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
10020 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10022 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
10023 IFN_MASK_LOAD calls here rather than...
10024 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
10026 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10027 Alan Hayward <alan.hayward@arm.com>
10028 David Sherwood <david.sherwood@arm.com>
10030 * expmed.c (extract_bit_field_1): For vector extracts,
10031 fall back to extract_bit_field_as_subreg if vec_extract
10034 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10035 Alan Hayward <alan.hayward@arm.com>
10036 David Sherwood <david.sherwood@arm.com>
10038 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
10039 they are variable or constant sized.
10040 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
10041 slots for constant-sized data.
10043 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10044 Alan Hayward <alan.hayward@arm.com>
10045 David Sherwood <david.sherwood@arm.com>
10047 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
10048 handling COND_EXPRs with boolean comparisons, try to find a better
10049 basis for the mask type than the boolean itself.
10051 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10053 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
10054 is calculated and how it can be overridden.
10055 * genmodes.c (max_bitsize_mode_any_mode): New variable.
10056 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
10058 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
10061 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10062 Alan Hayward <alan.hayward@arm.com>
10063 David Sherwood <david.sherwood@arm.com>
10065 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
10066 Remove the mode argument.
10067 (aarch64_simd_valid_immediate): Remove the mode and inverse
10069 * config/aarch64/iterators.md (bitsize): New iterator.
10070 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
10071 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
10072 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
10073 aarch64_simd_valid_immediate.
10074 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
10075 (aarch64_reg_or_bic_imm): Likewise.
10076 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
10077 with an insn_type enum and msl with a modifier_type enum.
10078 Replace element_width with a scalar_mode. Change the shift
10079 to unsigned int. Add constructors for scalar_float_mode and
10080 scalar_int_mode elements.
10081 (aarch64_vect_float_const_representable_p): Delete.
10082 (aarch64_can_const_movi_rtx_p)
10083 (aarch64_simd_scalar_immediate_valid_for_move)
10084 (aarch64_simd_make_constant): Update call to
10085 aarch64_simd_valid_immediate.
10086 (aarch64_advsimd_valid_immediate_hs): New function.
10087 (aarch64_advsimd_valid_immediate): Likewise.
10088 (aarch64_simd_valid_immediate): Remove mode and inverse
10089 arguments. Rewrite to use the above. Use const_vec_duplicate_p
10090 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
10091 and aarch64_float_const_representable_p on the result.
10092 (aarch64_output_simd_mov_immediate): Remove mode argument.
10093 Update call to aarch64_simd_valid_immediate and use of
10094 simd_immediate_info.
10095 (aarch64_output_scalar_simd_mov_immediate): Update call
10098 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10099 Alan Hayward <alan.hayward@arm.com>
10100 David Sherwood <david.sherwood@arm.com>
10102 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
10103 (mode_nunits): Likewise CONST_MODE_NUNITS.
10104 * machmode.def (ADJUST_NUNITS): Document.
10105 * genmodes.c (mode_data::need_nunits_adj): New field.
10106 (blank_mode): Update accordingly.
10107 (adj_nunits): New variable.
10108 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
10110 (emit_mode_size_inline): Set need_bytesize_adj for all modes
10111 listed in adj_nunits.
10112 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
10113 listed in adj_nunits. Don't emit case statements for such modes.
10114 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
10115 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
10116 nothing if adj_nunits is nonnull.
10117 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
10118 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
10119 (emit_mode_fbit): Update use of print_maybe_const_decl.
10120 (emit_move_size): Likewise. Treat the array as non-const
10122 (emit_mode_adjustments): Handle adj_nunits.
10124 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10126 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
10127 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
10128 (VECTOR_MODES): Use it.
10129 (make_vector_modes): Take the prefix as an argument.
10131 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10132 Alan Hayward <alan.hayward@arm.com>
10133 David Sherwood <david.sherwood@arm.com>
10135 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
10136 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
10137 for MODE_VECTOR_BOOL.
10138 * machmode.def (VECTOR_BOOL_MODE): Document.
10139 * genmodes.c (VECTOR_BOOL_MODE): New macro.
10140 (make_vector_bool_mode): New function.
10141 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
10143 * lto-streamer-in.c (lto_input_mode_table): Likewise.
10144 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
10146 * stor-layout.c (int_mode_for_mode): Likewise.
10147 * tree.c (build_vector_type_for_mode): Likewise.
10148 * varasm.c (output_constant_pool_2): Likewise.
10149 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
10150 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
10151 for MODE_VECTOR_BOOL.
10152 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
10153 of mode class checks.
10154 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
10155 instead of a list of mode class checks.
10156 (expand_vector_scalar_condition): Likewise.
10157 (type_for_widest_vector_mode): Handle BImode as an inner mode.
10159 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10160 Alan Hayward <alan.hayward@arm.com>
10161 David Sherwood <david.sherwood@arm.com>
10163 * machmode.h (mode_size): Change from unsigned short to
10165 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
10166 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10167 or if measurement_type is not polynomial.
10168 (fixed_size_mode::includes_p): Check for constant-sized modes.
10169 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
10170 return a poly_uint16 rather than an unsigned short.
10171 (emit_mode_size): Change the type of mode_size from unsigned short
10172 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
10173 (emit_mode_adjustments): Cope with polynomial vector sizes.
10174 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10176 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10178 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
10179 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
10180 * caller-save.c (setup_save_areas): Likewise.
10181 (replace_reg_with_saved_mem): Likewise.
10182 * calls.c (emit_library_call_value_1): Likewise.
10183 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
10184 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
10185 (gen_lowpart_for_combine): Likewise.
10186 * convert.c (convert_to_integer_1): Likewise.
10187 * cse.c (equiv_constant, cse_insn): Likewise.
10188 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
10189 (cselib_subst_to_values): Likewise.
10190 * dce.c (word_dce_process_block): Likewise.
10191 * df-problems.c (df_word_lr_mark_ref): Likewise.
10192 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
10193 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
10194 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
10195 (rtl_for_decl_location): Likewise.
10196 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
10197 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
10198 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
10199 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
10200 (expand_expr_real_1): Likewise.
10201 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
10202 (pad_below): Likewise.
10203 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10204 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
10205 * ira.c (get_subreg_tracking_sizes): Likewise.
10206 * ira-build.c (ira_create_allocno_objects): Likewise.
10207 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
10208 (ira_sort_regnos_for_alter_reg): Likewise.
10209 * ira-costs.c (record_operand_costs): Likewise.
10210 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
10211 (resolve_simple_move): Likewise.
10212 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
10213 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
10214 (lra_constraints): Likewise.
10215 (CONST_POOL_OK_P): Reject variable-sized modes.
10216 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
10217 (add_pseudo_to_slot, lra_spill): Likewise.
10218 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10219 * optabs-query.c (get_best_extraction_insn): Likewise.
10220 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10221 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
10222 (expand_mult_highpart, valid_multiword_target_p): Likewise.
10223 * recog.c (offsettable_address_addr_space_p): Likewise.
10224 * regcprop.c (maybe_mode_change): Likewise.
10225 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
10226 * regrename.c (build_def_use): Likewise.
10227 * regstat.c (dump_reg_info): Likewise.
10228 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
10229 (find_reloads, find_reloads_subreg_address): Likewise.
10230 * reload1.c (eliminate_regs_1): Likewise.
10231 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
10232 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
10233 (simplify_binary_operation_1, simplify_subreg): Likewise.
10234 * targhooks.c (default_function_arg_padding): Likewise.
10235 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
10236 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
10237 (verify_gimple_assign_ternary): Likewise.
10238 * tree-inline.c (estimate_move_cost): Likewise.
10239 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10240 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
10241 (get_address_cost_ainc): Likewise.
10242 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
10243 (vect_supportable_dr_alignment): Likewise.
10244 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10245 (vectorizable_reduction): Likewise.
10246 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
10247 (vectorizable_operation, vectorizable_load): Likewise.
10248 * tree.c (build_same_sized_truth_vector_type): Likewise.
10249 * valtrack.c (cleanup_auto_inc_dec): Likewise.
10250 * var-tracking.c (emit_note_insn_var_location): Likewise.
10251 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
10252 (ADDR_VEC_ALIGN): Likewise.
10254 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10255 Alan Hayward <alan.hayward@arm.com>
10256 David Sherwood <david.sherwood@arm.com>
10258 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
10260 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10261 or if measurement_type is polynomial.
10262 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
10263 * combine.c (make_extraction): Likewise.
10264 * dse.c (find_shift_sequence): Likewise.
10265 * dwarf2out.c (mem_loc_descriptor): Likewise.
10266 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
10267 (extract_bit_field, extract_low_bits): Likewise.
10268 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
10269 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
10270 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
10271 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
10272 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10273 * reload.c (find_reloads): Likewise.
10274 * reload1.c (alter_reg): Likewise.
10275 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
10276 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
10277 * tree-if-conv.c (predicate_mem_writes): Likewise.
10278 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
10279 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
10280 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
10281 * valtrack.c (dead_debug_insert_temp): Likewise.
10282 * varasm.c (mergeable_constant_section): Likewise.
10283 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
10285 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10286 Alan Hayward <alan.hayward@arm.com>
10287 David Sherwood <david.sherwood@arm.com>
10289 * expr.c (expand_assignment): Cope with polynomial mode sizes
10290 when assigning to a CONCAT.
10292 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10293 Alan Hayward <alan.hayward@arm.com>
10294 David Sherwood <david.sherwood@arm.com>
10296 * machmode.h (mode_precision): Change from unsigned short to
10298 (mode_to_precision): Return a poly_uint16 rather than an unsigned
10300 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
10301 or if measurement_type is not polynomial.
10302 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
10303 in which the mode is already known to be a scalar_int_mode.
10304 * genmodes.c (emit_mode_precision): Change the type of mode_precision
10305 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
10307 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10308 for GET_MODE_PRECISION.
10309 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10310 for GET_MODE_PRECISION.
10311 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
10313 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
10314 (expand_field_assignment, make_extraction): Likewise.
10315 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
10316 (get_last_value): Likewise.
10317 * convert.c (convert_to_integer_1): Likewise.
10318 * cse.c (cse_insn): Likewise.
10319 * expr.c (expand_expr_real_1): Likewise.
10320 * lra-constraints.c (simplify_operand_subreg): Likewise.
10321 * optabs-query.c (can_atomic_load_p): Likewise.
10322 * optabs.c (expand_atomic_load): Likewise.
10323 (expand_atomic_store): Likewise.
10324 * ree.c (combine_reaching_defs): Likewise.
10325 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
10326 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
10327 * tree.h (type_has_mode_precision_p): Likewise.
10328 * ubsan.c (instrument_si_overflow): Likewise.
10330 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10331 Alan Hayward <alan.hayward@arm.com>
10332 David Sherwood <david.sherwood@arm.com>
10334 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
10335 polynomial numbers of units.
10336 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
10337 (valid_vector_subparts_p): New function.
10338 (build_vector_type): Remove temporary shim and take the number
10339 of units as a poly_uint64 rather than an int.
10340 (build_opaque_vector_type): Take the number of units as a
10341 poly_uint64 rather than an int.
10342 * tree.c (build_vector_from_ctor): Handle polynomial
10343 TYPE_VECTOR_SUBPARTS.
10344 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
10345 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
10346 (build_vector_from_val): If the number of units is variable,
10347 use build_vec_duplicate_cst for constant operands and
10348 VEC_DUPLICATE_EXPR otherwise.
10349 (make_vector_type): Remove temporary is_constant ().
10350 (build_vector_type, build_opaque_vector_type): Take the number of
10351 units as a poly_uint64 rather than an int.
10352 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
10354 * cfgexpand.c (expand_debug_expr): Likewise.
10355 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
10356 (store_constructor, expand_expr_real_1): Likewise.
10357 (const_scalar_mask_from_tree): Likewise.
10358 * fold-const-call.c (fold_const_reduction): Likewise.
10359 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
10360 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
10361 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
10362 (fold_relational_const): Likewise.
10363 (native_interpret_vector): Likewise. Change the size from an
10364 int to an unsigned int.
10365 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
10366 TYPE_VECTOR_SUBPARTS.
10367 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
10368 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
10369 duplicating a non-constant operand into a variable-length vector.
10370 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
10371 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
10372 * ipa-icf.c (sem_variable::equals): Likewise.
10373 * match.pd: Likewise.
10374 * omp-simd-clone.c (simd_clone_subparts): Likewise.
10375 * print-tree.c (print_node): Likewise.
10376 * stor-layout.c (layout_type): Likewise.
10377 * targhooks.c (default_builtin_vectorization_cost): Likewise.
10378 * tree-cfg.c (verify_gimple_comparison): Likewise.
10379 (verify_gimple_assign_binary): Likewise.
10380 (verify_gimple_assign_ternary): Likewise.
10381 (verify_gimple_assign_single): Likewise.
10382 * tree-pretty-print.c (dump_generic_node): Likewise.
10383 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10384 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
10385 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
10386 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
10387 (vect_shift_permute_load_chain): Likewise.
10388 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
10389 (expand_vector_condition, optimize_vector_constructor): Likewise.
10390 (lower_vec_perm, get_compute_type): Likewise.
10391 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10392 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
10393 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
10394 (vect_recog_mask_conversion_pattern): Likewise.
10395 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
10396 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
10397 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10398 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
10399 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
10400 (vectorizable_shift, vectorizable_operation, vectorizable_store)
10401 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
10402 (supportable_widening_operation): Likewise.
10403 (supportable_narrowing_operation): Likewise.
10404 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
10406 * varasm.c (output_constant): Likewise.
10408 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10409 Alan Hayward <alan.hayward@arm.com>
10410 David Sherwood <david.sherwood@arm.com>
10412 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
10413 so that both the length == 3 and length != 3 cases set up their
10414 own permute vectors. Add comments explaining why we know the
10415 number of elements is constant.
10416 (vect_permute_load_chain): Likewise.
10418 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10419 Alan Hayward <alan.hayward@arm.com>
10420 David Sherwood <david.sherwood@arm.com>
10422 * machmode.h (mode_nunits): Change from unsigned char to
10424 (ONLY_FIXED_SIZE_MODES): New macro.
10425 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
10426 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
10427 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
10429 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
10430 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
10431 or if measurement_type is not polynomial.
10432 * genmodes.c (ZERO_COEFFS): New macro.
10433 (emit_mode_nunits_inline): Make mode_nunits_inline return a
10435 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
10436 Use ZERO_COEFFS when emitting initializers.
10437 * data-streamer.h (bp_pack_poly_value): New function.
10438 (bp_unpack_poly_value): Likewise.
10439 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10440 for GET_MODE_NUNITS.
10441 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10442 for GET_MODE_NUNITS.
10443 * tree.c (make_vector_type): Remove temporary shim and make
10444 the real function take the number of units as a poly_uint64
10445 rather than an int.
10446 (build_vector_type_for_mode): Handle polynomial nunits.
10447 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
10448 * emit-rtl.c (const_vec_series_p_1): Likewise.
10449 (gen_rtx_CONST_VECTOR): Likewise.
10450 * fold-const.c (test_vec_duplicate_folding): Likewise.
10451 * genrecog.c (validate_pattern): Likewise.
10452 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
10453 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10454 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
10455 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
10456 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
10457 * rtlanal.c (subreg_get_info): Likewise.
10458 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10459 (vect_grouped_load_supported): Likewise.
10460 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
10461 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
10462 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
10463 (simplify_const_unary_operation, simplify_binary_operation_1)
10464 (simplify_const_binary_operation, simplify_ternary_operation)
10465 (test_vector_ops_duplicate, test_vector_ops): Likewise.
10466 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
10467 instead of CONST_VECTOR_NUNITS.
10468 * varasm.c (output_constant_pool_2): Likewise.
10469 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
10470 explicit-encoded elements in the XVEC for variable-length vectors.
10472 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10474 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
10476 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10477 Alan Hayward <alan.hayward@arm.com>
10478 David Sherwood <david.sherwood@arm.com>
10480 * coretypes.h (fixed_size_mode): Declare.
10481 (fixed_size_mode_pod): New typedef.
10482 * builtins.h (target_builtins::x_apply_args_mode)
10483 (target_builtins::x_apply_result_mode): Change type to
10484 fixed_size_mode_pod.
10485 * builtins.c (apply_args_size, apply_result_size, result_vector)
10486 (expand_builtin_apply_args_1, expand_builtin_apply)
10487 (expand_builtin_return): Update accordingly.
10489 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10491 * cse.c (hash_rtx_cb): Hash only the encoded elements.
10492 * cselib.c (cselib_hash_rtx): Likewise.
10493 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
10494 CONST_VECTOR encoding.
10496 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10497 Jeff Law <law@redhat.com>
10500 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
10501 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
10502 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
10503 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
10506 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
10507 explicitly probe *sp in a noreturn function if there were any callee
10508 register saves or frame pointer is needed.
10510 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10513 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
10514 BLKmode for ternary, binary or unary expressions.
10517 * var-tracking.c (delete_vta_debug_insn): New inline function.
10518 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
10519 insns from get_insns () to NULL instead of each bb separately.
10520 Use delete_vta_debug_insn. No longer static.
10521 (vt_debug_insns_local, variable_tracking_main_1): Adjust
10522 delete_vta_debug_insns callers.
10523 * rtl.h (delete_vta_debug_insns): Declare.
10524 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
10525 instead of variable_tracking_main.
10527 2018-01-03 Martin Sebor <msebor@redhat.com>
10529 PR tree-optimization/83603
10530 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
10531 arguments past the endof the argument list in functions declared
10532 without a prototype.
10533 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
10534 Avoid checking when arguments are null.
10536 2018-01-03 Martin Sebor <msebor@redhat.com>
10539 * doc/extend.texi (attribute const): Fix a typo.
10540 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
10541 issuing -Wsuggest-attribute for void functions.
10543 2018-01-03 Martin Sebor <msebor@redhat.com>
10545 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
10546 offset_int::from instead of wide_int::to_shwi.
10547 (maybe_diag_overlap): Remove assertion.
10548 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
10549 * gimple-ssa-sprintf.c (format_directive): Same.
10550 (parse_directive): Same.
10551 (sprintf_dom_walker::compute_format_length): Same.
10552 (try_substitute_return_value): Same.
10554 2018-01-03 Jeff Law <law@redhat.com>
10556 PR middle-end/83654
10557 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
10558 non-constant residual for zero at runtime and avoid probing in
10559 that case. Reorganize code for trailing problem to mirror handling
10562 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10564 PR tree-optimization/83501
10565 * tree-ssa-strlen.c (get_string_cst): New.
10566 (handle_char_store): Call get_string_cst.
10568 2018-01-03 Martin Liska <mliska@suse.cz>
10570 PR tree-optimization/83593
10571 * tree-ssa-strlen.c: Include tree-cfg.h.
10572 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
10573 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
10574 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
10576 (strlen_dom_walker::before_dom_children): Call
10577 gimple_purge_dead_eh_edges. Dump tranformation with details
10579 (strlen_dom_walker::before_dom_children): Update call by adding
10580 new argument cleanup_eh.
10581 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
10583 2018-01-03 Martin Liska <mliska@suse.cz>
10586 * cif-code.def (VARIADIC_THUNK): New enum value.
10587 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
10590 2018-01-03 Jan Beulich <jbeulich@suse.com>
10592 * sse.md (mov<mode>_internal): Tighten condition for when to use
10593 vmovdqu<ssescalarsize> for TI and OI modes.
10595 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10597 Update copyright years.
10599 2018-01-03 Martin Liska <mliska@suse.cz>
10602 * ipa-visibility.c (function_and_variable_visibility): Skip
10603 functions with noipa attribure.
10605 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10607 * gcc.c (process_command): Update copyright notice dates.
10608 * gcov-dump.c (print_version): Ditto.
10609 * gcov.c (print_version): Ditto.
10610 * gcov-tool.c (print_version): Ditto.
10611 * gengtype.c (create_file): Ditto.
10612 * doc/cpp.texi: Bump @copying's copyright year.
10613 * doc/cppinternals.texi: Ditto.
10614 * doc/gcc.texi: Ditto.
10615 * doc/gccint.texi: Ditto.
10616 * doc/gcov.texi: Ditto.
10617 * doc/install.texi: Ditto.
10618 * doc/invoke.texi: Ditto.
10620 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10622 * vector-builder.h (vector_builder::m_full_nelts): Change from
10623 unsigned int to poly_uint64.
10624 (vector_builder::full_nelts): Update prototype accordingly.
10625 (vector_builder::new_vector): Likewise.
10626 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
10627 (vector_builder::operator ==): Likewise.
10628 (vector_builder::finalize): Likewise.
10629 * int-vector-builder.h (int_vector_builder::int_vector_builder):
10630 Take the number of elements as a poly_uint64 rather than an
10632 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
10633 from unsigned int to poly_uint64.
10634 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
10635 (vec_perm_indices::new_vector): Likewise.
10636 (vec_perm_indices::length): Likewise.
10637 (vec_perm_indices::nelts_per_input): Likewise.
10638 (vec_perm_indices::input_nelts): Likewise.
10639 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
10640 number of elements per input as a poly_uint64 rather than an
10641 unsigned int. Use the original encoding for variable-length
10642 vectors, rather than clamping each individual element.
10643 For the second and subsequent elements in each pattern,
10644 clamp the step and base before clamping their sum.
10645 (vec_perm_indices::series_p): Handle polynomial element counts.
10646 (vec_perm_indices::all_in_range_p): Likewise.
10647 (vec_perm_indices_to_tree): Likewise.
10648 (vec_perm_indices_to_rtx): Likewise.
10649 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
10650 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
10651 (tree_vector_builder::new_binary_operation): Handle polynomial
10652 element counts. Return false if we need to know the number
10653 of elements at compile time.
10654 * fold-const.c (fold_vec_perm): Punt if the number of elements
10655 isn't known at compile time.
10657 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10659 * vec-perm-indices.h (vec_perm_builder): Change element type
10660 from HOST_WIDE_INT to poly_int64.
10661 (vec_perm_indices::element_type): Update accordingly.
10662 (vec_perm_indices::clamp): Handle polynomial element_types.
10663 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10664 (vec_perm_indices::all_in_range_p): Likewise.
10665 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
10667 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
10668 polynomial vec_perm_indices element types.
10669 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
10670 * fold-const.c (fold_vec_perm): Likewise.
10671 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
10672 * tree-vect-generic.c (lower_vec_perm): Likewise.
10673 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10674 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
10675 element type to HOST_WIDE_INT.
10677 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10678 Alan Hayward <alan.hayward@arm.com>
10679 David Sherwood <david.sherwood@arm.com>
10681 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
10682 rather than an int. Use plus_constant.
10683 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
10684 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
10686 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10687 Alan Hayward <alan.hayward@arm.com>
10688 David Sherwood <david.sherwood@arm.com>
10690 * calls.c (emit_call_1, expand_call): Change struct_value_size from
10691 a HOST_WIDE_INT to a poly_int64.
10693 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10694 Alan Hayward <alan.hayward@arm.com>
10695 David Sherwood <david.sherwood@arm.com>
10697 * calls.c (load_register_parameters): Cope with polynomial
10698 mode sizes. Require a constant size for BLKmode parameters
10699 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
10700 forces a parameter to be padded at the lsb end in order to
10701 fill a complete number of words, require the parameter size
10702 to be ordered wrt UNITS_PER_WORD.
10704 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10705 Alan Hayward <alan.hayward@arm.com>
10706 David Sherwood <david.sherwood@arm.com>
10708 * reload1.c (spill_stack_slot_width): Change element type
10709 from unsigned int to poly_uint64_pod.
10710 (alter_reg): Treat mode sizes as polynomial.
10712 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10713 Alan Hayward <alan.hayward@arm.com>
10714 David Sherwood <david.sherwood@arm.com>
10716 * reload.c (complex_word_subreg_p): New function.
10717 (reload_inner_reg_of_subreg, push_reload): Use it.
10719 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10720 Alan Hayward <alan.hayward@arm.com>
10721 David Sherwood <david.sherwood@arm.com>
10723 * lra-constraints.c (process_alt_operands): Reject matched
10724 operands whose sizes aren't ordered.
10725 (match_reload): Refer to this check here.
10727 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10728 Alan Hayward <alan.hayward@arm.com>
10729 David Sherwood <david.sherwood@arm.com>
10731 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
10732 that the mode size is in the set {1, 2, 4, 8, 16}.
10734 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10735 Alan Hayward <alan.hayward@arm.com>
10736 David Sherwood <david.sherwood@arm.com>
10738 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
10739 Use plus_constant instead of gen_rtx_PLUS.
10741 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10742 Alan Hayward <alan.hayward@arm.com>
10743 David Sherwood <david.sherwood@arm.com>
10745 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
10746 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
10747 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
10748 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
10749 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
10750 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
10751 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
10752 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
10753 * config/i386/i386.c (ix86_push_rounding): ...this new function.
10754 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
10756 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
10757 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
10758 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
10759 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
10760 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
10761 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
10762 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
10763 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
10764 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
10765 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
10767 * expr.c (emit_move_resolve_push): Treat the input and result
10768 of PUSH_ROUNDING as a poly_int64.
10769 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
10770 (emit_push_insn): Likewise.
10771 * lra-eliminations.c (mark_not_eliminable): Likewise.
10772 * recog.c (push_operand): Likewise.
10773 * reload1.c (elimination_effects): Likewise.
10774 * rtlanal.c (nonzero_bits1): Likewise.
10775 * calls.c (store_one_arg): Likewise. Require the padding to be
10776 known at compile time.
10778 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10779 Alan Hayward <alan.hayward@arm.com>
10780 David Sherwood <david.sherwood@arm.com>
10782 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
10783 Use plus_constant instead of gen_rtx_PLUS.
10785 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10786 Alan Hayward <alan.hayward@arm.com>
10787 David Sherwood <david.sherwood@arm.com>
10789 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
10790 rather than an int.
10792 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10793 Alan Hayward <alan.hayward@arm.com>
10794 David Sherwood <david.sherwood@arm.com>
10796 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
10797 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
10798 via stack temporaries. Treat the mode size as polynomial too.
10800 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10801 Alan Hayward <alan.hayward@arm.com>
10802 David Sherwood <david.sherwood@arm.com>
10804 * expr.c (expand_expr_real_2): When handling conversions involving
10805 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
10806 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
10807 as a poly_uint64 too.
10809 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10810 Alan Hayward <alan.hayward@arm.com>
10811 David Sherwood <david.sherwood@arm.com>
10813 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
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 * combine.c (can_change_dest_mode): Handle polynomial
10820 REGMODE_NATURAL_SIZE.
10821 * expmed.c (store_bit_field_1): Likewise.
10822 * expr.c (store_constructor): Likewise.
10823 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
10824 and polynomial REGMODE_NATURAL_SIZE.
10825 (gen_lowpart_common): Likewise.
10826 * reginfo.c (record_subregs_of_mode): Likewise.
10827 * rtlanal.c (read_modify_subreg_p): Likewise.
10829 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10830 Alan Hayward <alan.hayward@arm.com>
10831 David Sherwood <david.sherwood@arm.com>
10833 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
10834 numbers of elements.
10836 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10837 Alan Hayward <alan.hayward@arm.com>
10838 David Sherwood <david.sherwood@arm.com>
10840 * match.pd: Cope with polynomial numbers of vector 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 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
10847 in a POINTER_PLUS_EXPR.
10849 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10850 Alan Hayward <alan.hayward@arm.com>
10851 David Sherwood <david.sherwood@arm.com>
10853 * omp-simd-clone.c (simd_clone_subparts): New function.
10854 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
10855 (ipa_simd_modify_function_body): Likewise.
10857 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10858 Alan Hayward <alan.hayward@arm.com>
10859 David Sherwood <david.sherwood@arm.com>
10861 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
10862 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
10863 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
10864 (expand_vector_condition, vector_element): Likewise.
10865 (subparts_gt): New function.
10866 (get_compute_type): Use subparts_gt.
10867 (count_type_subparts): Delete.
10868 (expand_vector_operations_1): Use subparts_gt instead of
10869 count_type_subparts.
10871 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10872 Alan Hayward <alan.hayward@arm.com>
10873 David Sherwood <david.sherwood@arm.com>
10875 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
10876 (vect_compile_time_alias): ...this new function. Do the calculation
10877 on poly_ints rather than trees.
10878 (vect_prune_runtime_alias_test_list): Update call accordingly.
10880 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10881 Alan Hayward <alan.hayward@arm.com>
10882 David Sherwood <david.sherwood@arm.com>
10884 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
10886 (vect_schedule_slp_instance): Likewise.
10888 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10889 Alan Hayward <alan.hayward@arm.com>
10890 David Sherwood <david.sherwood@arm.com>
10892 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
10893 constant and extern definitions for variable-length vectors.
10894 (vect_get_constant_vectors): Note that the number of units
10895 is known to be constant.
10897 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10898 Alan Hayward <alan.hayward@arm.com>
10899 David Sherwood <david.sherwood@arm.com>
10901 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
10902 of units as polynomial. Choose between WIDE and NARROW based
10905 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10906 Alan Hayward <alan.hayward@arm.com>
10907 David Sherwood <david.sherwood@arm.com>
10909 * tree-vect-stmts.c (simd_clone_subparts): New function.
10910 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
10912 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10913 Alan Hayward <alan.hayward@arm.com>
10914 David Sherwood <david.sherwood@arm.com>
10916 * tree-vect-stmts.c (vectorizable_call): Treat the number of
10917 vectors as polynomial. Use build_index_vector for
10918 IFN_GOMP_SIMD_LANE.
10920 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10921 Alan Hayward <alan.hayward@arm.com>
10922 David Sherwood <david.sherwood@arm.com>
10924 * tree-vect-stmts.c (get_load_store_type): Treat the number of
10925 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
10926 for variable-length vectors.
10927 (vectorizable_mask_load_store): Treat the number of units as
10928 polynomial, asserting that it is constant if the condition has
10929 already been enforced.
10930 (vectorizable_store, vectorizable_load): Likewise.
10932 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10933 Alan Hayward <alan.hayward@arm.com>
10934 David Sherwood <david.sherwood@arm.com>
10936 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
10937 of units as polynomial. Punt if we can't tell at compile time
10938 which vector contains the final result.
10940 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10941 Alan Hayward <alan.hayward@arm.com>
10942 David Sherwood <david.sherwood@arm.com>
10944 * tree-vect-loop.c (vectorizable_induction): Treat the number
10945 of units as polynomial. Punt on SLP inductions. Use an integer
10946 VEC_SERIES_EXPR for variable-length integer reductions. Use a
10947 cast of such a series for variable-length floating-point
10950 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10951 Alan Hayward <alan.hayward@arm.com>
10952 David Sherwood <david.sherwood@arm.com>
10954 * tree.h (build_index_vector): Declare.
10955 * tree.c (build_index_vector): New function.
10956 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
10957 of units as polynomial, forcibly converting it to a constant if
10958 vectorizable_reduction has already enforced the condition.
10959 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
10960 to create a {1,2,3,...} vector.
10961 (vectorizable_reduction): Treat the number of units as polynomial.
10962 Choose vectype_in based on the largest scalar element size rather
10963 than the smallest number of units. Enforce the restrictions
10966 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10967 Alan Hayward <alan.hayward@arm.com>
10968 David Sherwood <david.sherwood@arm.com>
10970 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
10971 number of units as polynomial.
10973 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10974 Alan Hayward <alan.hayward@arm.com>
10975 David Sherwood <david.sherwood@arm.com>
10977 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
10978 * target.def (autovectorize_vector_sizes): Return the vector sizes
10979 by pointer, using vector_sizes rather than a bitmask.
10980 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
10981 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
10982 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
10984 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
10985 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
10986 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
10987 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
10988 * omp-general.c (omp_max_vf): Likewise.
10989 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10990 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
10991 * tree-vect-loop.c (vect_analyze_loop): Likewise.
10992 * tree-vect-slp.c (vect_slp_bb): Likewise.
10993 * doc/tm.texi: Regenerate.
10994 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
10996 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
10997 the vector size as a poly_uint64 rather than an unsigned int.
10998 (current_vector_size): Change from an unsigned int to a poly_uint64.
10999 (get_vectype_for_scalar_type): Update accordingly.
11000 * tree.h (build_truth_vector_type): Take the size and number of
11001 units as a poly_uint64 rather than an unsigned int.
11002 (build_vector_type): Add a temporary overload that takes
11003 the number of units as a poly_uint64 rather than an unsigned int.
11004 * tree.c (make_vector_type): Likewise.
11005 (build_truth_vector_type): Take the number of units as a poly_uint64
11006 rather than an unsigned int.
11008 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11009 Alan Hayward <alan.hayward@arm.com>
11010 David Sherwood <david.sherwood@arm.com>
11012 * target.def (get_mask_mode): Take the number of units and length
11013 as poly_uint64s rather than unsigned ints.
11014 * targhooks.h (default_get_mask_mode): Update accordingly.
11015 * targhooks.c (default_get_mask_mode): Likewise.
11016 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
11017 * doc/tm.texi: Regenerate.
11019 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11020 Alan Hayward <alan.hayward@arm.com>
11021 David Sherwood <david.sherwood@arm.com>
11023 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
11024 * omp-general.c (omp_max_vf): Likewise.
11025 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
11026 (expand_omp_simd): Handle polynomial safelen.
11027 * omp-low.c (omplow_simd_context): Add a default constructor.
11028 (omplow_simd_context::max_vf): Change from int to poly_uint64.
11029 (lower_rec_simd_input_clauses): Update accordingly.
11030 (lower_rec_input_clauses): Likewise.
11032 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11033 Alan Hayward <alan.hayward@arm.com>
11034 David Sherwood <david.sherwood@arm.com>
11036 * tree-vectorizer.h (vect_nunits_for_cost): New function.
11037 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
11038 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
11039 (vect_analyze_slp_cost): Likewise.
11040 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
11041 (vect_model_load_cost): Likewise.
11043 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11044 Alan Hayward <alan.hayward@arm.com>
11045 David Sherwood <david.sherwood@arm.com>
11047 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
11048 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
11049 from an unsigned int * to a poly_uint64_pod *.
11050 (calculate_unrolling_factor): New function.
11051 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
11053 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11054 Alan Hayward <alan.hayward@arm.com>
11055 David Sherwood <david.sherwood@arm.com>
11057 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
11058 from an unsigned int to a poly_uint64.
11059 (_loop_vec_info::slp_unrolling_factor): Likewise.
11060 (_loop_vec_info::vectorization_factor): Change from an int
11062 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
11063 (vect_get_num_vectors): New function.
11064 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
11065 (vect_get_num_copies): Use vect_get_num_vectors.
11066 (vect_analyze_data_ref_dependences): Change max_vf from an int *
11067 to an unsigned int *.
11068 (vect_analyze_data_refs): Change min_vf from an int * to a
11070 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11071 than an unsigned HOST_WIDE_INT.
11072 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
11073 (vect_analyze_data_ref_dependence): Change max_vf from an int *
11074 to an unsigned int *.
11075 (vect_analyze_data_ref_dependences): Likewise.
11076 (vect_compute_data_ref_alignment): Handle polynomial vf.
11077 (vect_enhance_data_refs_alignment): Likewise.
11078 (vect_prune_runtime_alias_test_list): Likewise.
11079 (vect_shift_permute_load_chain): Likewise.
11080 (vect_supportable_dr_alignment): Likewise.
11081 (dependence_distance_ge_vf): Take the vectorization factor as a
11082 poly_uint64 rather than an unsigned HOST_WIDE_INT.
11083 (vect_analyze_data_refs): Change min_vf from an int * to a
11085 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
11086 vfm1 as a poly_uint64 rather than an int. Make the same change
11087 for the returned bound_scalar.
11088 (vect_gen_vector_loop_niters): Handle polynomial vf.
11089 (vect_do_peeling): Likewise. Update call to
11090 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
11091 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
11093 * tree-vect-loop.c (vect_determine_vectorization_factor)
11094 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
11095 (vect_get_known_peeling_cost): Likewise.
11096 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
11097 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
11098 (vect_transform_loop): Likewise. Use the lowest possible VF when
11099 updating the upper bounds of the loop.
11100 (vect_min_worthwhile_factor): Make static. Return an unsigned int
11101 rather than an int.
11102 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
11103 polynomial unroll factors.
11104 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
11105 (vect_make_slp_decision): Likewise.
11106 (vect_supported_load_permutation_p): Likewise, and polynomial
11108 (vect_analyze_slp_cost): Handle polynomial vf.
11109 (vect_slp_analyze_node_operations): Likewise.
11110 (vect_slp_analyze_bb_1): Likewise.
11111 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11112 than an unsigned HOST_WIDE_INT.
11113 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
11114 (vectorizable_load): Handle polynomial vf.
11115 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
11117 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
11119 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11120 Alan Hayward <alan.hayward@arm.com>
11121 David Sherwood <david.sherwood@arm.com>
11123 * match.pd: Handle bit operations involving three constants
11124 and try to fold one pair.
11126 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11128 * tree-vect-loop-manip.c: Include gimple-fold.h.
11129 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
11130 niters_maybe_zero parameters. Handle other cases besides a step of 1.
11131 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
11132 Add a path that uses a step of VF instead of 1, but disable it
11134 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
11135 and niters_no_overflow parameters. Update calls to
11136 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
11137 Create a new SSA name if the latter choses to use a ste other
11138 than zero, and return it via niters_vector_mult_vf_var.
11139 * tree-vect-loop.c (vect_transform_loop): Update calls to
11140 vect_do_peeling, vect_gen_vector_loop_niters and
11141 slpeel_make_loop_iterate_ntimes.
11142 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
11143 (vect_gen_vector_loop_niters): Update declarations after above changes.
11145 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
11147 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
11148 128-bit round to integer instructions.
11149 (ceil<mode>2): Likewise.
11150 (btrunc<mode>2): Likewise.
11151 (round<mode>2): Likewise.
11153 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11155 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
11156 unaligned VSX load/store on P8/P9.
11157 (expand_block_clear): Allow the use of unaligned VSX
11158 load/store on P8/P9.
11160 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11162 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
11164 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
11165 swap associated with both a load and a store.
11167 2018-01-02 Andrew Waterman <andrew@sifive.com>
11169 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
11170 * config/riscv/riscv.md (clear_cache): Use it.
11172 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
11174 * web.c: Remove out-of-date comment.
11176 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11178 * expr.c (fixup_args_size_notes): Check that any existing
11179 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
11180 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
11181 (emit_single_push_insn): ...here.
11183 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11185 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
11186 (const_vector_encoded_nelts): New function.
11187 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
11188 (const_vector_int_elt, const_vector_elt): Declare.
11189 * emit-rtl.c (const_vector_int_elt_1): New function.
11190 (const_vector_elt): Likewise.
11191 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
11192 of CONST_VECTOR_ELT.
11194 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11196 * expr.c: Include rtx-vector-builder.h.
11197 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
11198 directly on the tree encoding.
11199 (const_vector_from_tree): Likewise.
11200 * optabs.c: Include rtx-vector-builder.h.
11201 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
11202 sequence of "u" values.
11203 * vec-perm-indices.c: Include rtx-vector-builder.h.
11204 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
11205 directly on the vec_perm_indices encoding.
11207 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11209 * doc/rtl.texi (const_vector): Describe new encoding scheme.
11210 * Makefile.in (OBJS): Add rtx-vector-builder.o.
11211 * rtx-vector-builder.h: New file.
11212 * rtx-vector-builder.c: Likewise.
11213 * rtl.h (rtx_def::u2): Add a const_vector field.
11214 (CONST_VECTOR_NPATTERNS): New macro.
11215 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
11216 (CONST_VECTOR_DUPLICATE_P): Likewise.
11217 (CONST_VECTOR_STEPPED_P): Likewise.
11218 (CONST_VECTOR_ENCODED_ELT): Likewise.
11219 (const_vec_duplicate_p): Check for a duplicated vector encoding.
11220 (unwrap_const_vec_duplicate): Likewise.
11221 (const_vec_series_p): Check for a non-duplicated vector encoding.
11222 Say that the function only returns true for integer vectors.
11223 * emit-rtl.c: Include rtx-vector-builder.h.
11224 (gen_const_vec_duplicate_1): Delete.
11225 (gen_const_vector): Call gen_const_vec_duplicate instead of
11226 gen_const_vec_duplicate_1.
11227 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
11228 (gen_const_vec_duplicate): Use rtx_vector_builder.
11229 (gen_const_vec_series): Likewise.
11230 (gen_rtx_CONST_VECTOR): Likewise.
11231 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
11232 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11233 Build a new vector rather than modifying a CONST_VECTOR in-place.
11234 (handle_special_swappables): Update call accordingly.
11235 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
11236 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11237 Build a new vector rather than modifying a CONST_VECTOR in-place.
11238 (handle_special_swappables): Update call accordingly.
11240 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11242 * simplify-rtx.c (simplify_const_binary_operation): Use
11243 CONST_VECTOR_ELT instead of XVECEXP.
11245 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11247 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
11248 the selector elements to be different from the data elements
11249 if the selector is a VECTOR_CST.
11250 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
11251 ssizetype for the selector.
11253 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11255 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
11256 before testing each element individually.
11257 * tree-vect-generic.c (lower_vec_perm): Likewise.
11259 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11261 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
11262 * selftest-run-tests.c (selftest::run_tests): Call it.
11263 * vector-builder.h (vector_builder::operator ==): New function.
11264 (vector_builder::operator !=): Likewise.
11265 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
11266 (vec_perm_indices::all_from_input_p): New function.
11267 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11268 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
11269 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
11270 instead of reading the VECTOR_CST directly. Detect whether both
11271 vector inputs are the same before constructing the vec_perm_indices,
11272 and update the number of inputs argument accordingly. Use the
11273 utility functions added above. Only construct sel2 if we need to.
11275 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11277 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
11278 the broadcast of the low byte.
11279 (expand_mult_highpart): Use an explicit encoding for the permutes.
11280 * optabs-query.c (can_mult_highpart_p): Likewise.
11281 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
11282 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11283 (vectorizable_bswap): Likewise.
11284 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
11285 explicit encoding for the power-of-2 permutes.
11286 (vect_permute_store_chain): Likewise.
11287 (vect_grouped_load_supported): Likewise.
11288 (vect_permute_load_chain): Likewise.
11290 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11292 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
11293 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
11294 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
11295 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11296 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
11297 (vect_gen_perm_mask_any): Likewise.
11299 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11301 * int-vector-builder.h: New file.
11302 * vec-perm-indices.h: Include int-vector-builder.h.
11303 (vec_perm_indices): Redefine as an int_vector_builder.
11304 (auto_vec_perm_indices): Delete.
11305 (vec_perm_builder): Redefine as a stand-alone class.
11306 (vec_perm_indices::vec_perm_indices): New function.
11307 (vec_perm_indices::clamp): Likewise.
11308 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
11309 (vec_perm_indices::new_vector): New function.
11310 (vec_perm_indices::new_expanded_vector): Update for new
11311 vec_perm_indices class.
11312 (vec_perm_indices::rotate_inputs): New function.
11313 (vec_perm_indices::all_in_range_p): Operate directly on the
11314 encoded form, without computing elided elements.
11315 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
11316 encoding. Update for new vec_perm_indices class.
11317 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
11318 the given vec_perm_builder.
11319 (expand_vec_perm_var): Update vec_perm_builder constructor.
11320 (expand_mult_highpart): Use vec_perm_builder instead of
11321 auto_vec_perm_indices.
11322 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
11323 vec_perm_indices instead of auto_vec_perm_indices. Use a single
11324 or double series encoding as appropriate.
11325 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
11326 vec_perm_indices instead of auto_vec_perm_indices.
11327 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11328 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11329 (vect_permute_store_chain): Likewise.
11330 (vect_grouped_load_supported): Likewise.
11331 (vect_permute_load_chain): Likewise.
11332 (vect_shift_permute_load_chain): Likewise.
11333 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11334 (vect_transform_slp_perm_load): Likewise.
11335 (vect_schedule_slp_instance): Likewise.
11336 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11337 (vectorizable_mask_load_store): Likewise.
11338 (vectorizable_bswap): Likewise.
11339 (vectorizable_store): Likewise.
11340 (vectorizable_load): Likewise.
11341 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
11342 vec_perm_indices instead of auto_vec_perm_indices. Use
11343 tree_to_vec_perm_builder to read the vector from a tree.
11344 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
11345 vec_perm_builder instead of a vec_perm_indices.
11346 (have_whole_vector_shift): Use vec_perm_builder and
11347 vec_perm_indices instead of auto_vec_perm_indices. Leave the
11348 truncation to calc_vec_perm_mask_for_shift.
11349 (vect_create_epilog_for_reduction): Likewise.
11350 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
11351 from auto_vec_perm_indices to vec_perm_indices.
11352 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11353 instead of changing individual elements.
11354 (aarch64_vectorize_vec_perm_const): Use new_vector to install
11355 the vector in d.perm.
11356 * config/arm/arm.c (expand_vec_perm_d::perm): Change
11357 from auto_vec_perm_indices to vec_perm_indices.
11358 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11359 instead of changing individual elements.
11360 (arm_vectorize_vec_perm_const): Use new_vector to install
11361 the vector in d.perm.
11362 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
11363 Update vec_perm_builder constructor.
11364 (rs6000_expand_interleave): Likewise.
11365 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
11366 (rs6000_expand_interleave): Likewise.
11368 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11370 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
11371 to qimode could truncate the indices.
11372 * optabs.c (expand_vec_perm_var): Likewise.
11374 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11376 * Makefile.in (OBJS): Add vec-perm-indices.o.
11377 * vec-perm-indices.h: New file.
11378 * vec-perm-indices.c: Likewise.
11379 * target.h (vec_perm_indices): Replace with a forward class
11381 (auto_vec_perm_indices): Move to vec-perm-indices.h.
11382 * optabs.h: Include vec-perm-indices.h.
11383 (expand_vec_perm): Delete.
11384 (selector_fits_mode_p, expand_vec_perm_var): Declare.
11385 (expand_vec_perm_const): Declare.
11386 * target.def (vec_perm_const_ok): Replace with...
11387 (vec_perm_const): ...this new hook.
11388 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
11389 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
11390 * doc/tm.texi: Regenerate.
11391 * optabs.def (vec_perm_const): Delete.
11392 * doc/md.texi (vec_perm_const): Likewise.
11393 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
11394 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
11395 expand_vec_perm for constant permutation vectors. Assert that
11396 the mode of variable permutation vectors is the integer equivalent
11397 of the mode that is being permuted.
11398 * optabs-query.h (selector_fits_mode_p): Declare.
11399 * optabs-query.c: Include vec-perm-indices.h.
11400 (selector_fits_mode_p): New function.
11401 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
11402 is defined, instead of checking whether the vec_perm_const_optab
11403 exists. Use targetm.vectorize.vec_perm_const instead of
11404 targetm.vectorize.vec_perm_const_ok. Check whether the indices
11405 fit in the vector mode before using a variable permute.
11406 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
11407 vec_perm_indices instead of an rtx.
11408 (expand_vec_perm): Replace with...
11409 (expand_vec_perm_const): ...this new function. Take the selector
11410 as a vec_perm_indices rather than an rtx. Also take the mode of
11411 the selector. Update call to shift_amt_for_vec_perm_mask.
11412 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
11413 Use vec_perm_indices::new_expanded_vector to expand the original
11414 selector into bytes. Check whether the indices fit in the vector
11415 mode before using a variable permute.
11416 (expand_vec_perm_var): Make global.
11417 (expand_mult_highpart): Use expand_vec_perm_const.
11418 * fold-const.c: Includes vec-perm-indices.h.
11419 * tree-ssa-forwprop.c: Likewise.
11420 * tree-vect-data-refs.c: Likewise.
11421 * tree-vect-generic.c: Likewise.
11422 * tree-vect-loop.c: Likewise.
11423 * tree-vect-slp.c: Likewise.
11424 * tree-vect-stmts.c: Likewise.
11425 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
11427 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
11428 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
11429 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
11430 (aarch64_vectorize_vec_perm_const): ...this new function.
11431 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11432 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11433 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
11434 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
11435 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11436 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11437 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
11439 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
11440 check for NEON modes.
11441 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
11442 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
11443 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
11444 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
11446 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
11447 the old VEC_PERM_CONST conditions.
11448 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
11449 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
11450 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
11451 (ia64_vectorize_vec_perm_const_ok): Merge into...
11452 (ia64_vectorize_vec_perm_const): ...this new function.
11453 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
11454 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
11455 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
11456 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
11457 * config/mips/mips.c (mips_expand_vec_perm_const)
11458 (mips_vectorize_vec_perm_const_ok): Merge into...
11459 (mips_vectorize_vec_perm_const): ...this new function.
11460 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
11461 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
11462 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
11463 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
11464 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
11465 (rs6000_expand_vec_perm_const): Delete.
11466 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
11468 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11469 (altivec_expand_vec_perm_const_le): Take each operand individually.
11470 Operate on constant selectors rather than rtxes.
11471 (altivec_expand_vec_perm_const): Likewise. Update call to
11472 altivec_expand_vec_perm_const_le.
11473 (rs6000_expand_vec_perm_const): Delete.
11474 (rs6000_vectorize_vec_perm_const_ok): Delete.
11475 (rs6000_vectorize_vec_perm_const): New function.
11476 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11477 an element count and rtx array.
11478 (rs6000_expand_extract_even): Update call accordingly.
11479 (rs6000_expand_interleave): Likewise.
11480 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
11481 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
11482 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
11483 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
11484 (rs6000_expand_vec_perm_const): Delete.
11485 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11486 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11487 (altivec_expand_vec_perm_const_le): Take each operand individually.
11488 Operate on constant selectors rather than rtxes.
11489 (altivec_expand_vec_perm_const): Likewise. Update call to
11490 altivec_expand_vec_perm_const_le.
11491 (rs6000_expand_vec_perm_const): Delete.
11492 (rs6000_vectorize_vec_perm_const_ok): Delete.
11493 (rs6000_vectorize_vec_perm_const): New function. Remove stray
11494 reference to the SPE evmerge intructions.
11495 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11496 an element count and rtx array.
11497 (rs6000_expand_extract_even): Update call accordingly.
11498 (rs6000_expand_interleave): Likewise.
11499 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
11500 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
11502 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11504 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11506 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
11507 vector mode and that that mode matches the mode of the data
11509 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
11510 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
11511 directly using expand_vec_perm_1 when forcing selectors into
11513 (expand_vec_perm_var): New function, split out from expand_vec_perm.
11515 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11517 * optabs-query.h (can_vec_perm_p): Delete.
11518 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
11519 * optabs-query.c (can_vec_perm_p): Split into...
11520 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
11521 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
11522 particular selector is valid.
11523 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11524 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11525 (vect_grouped_load_supported): Likewise.
11526 (vect_shift_permute_load_chain): Likewise.
11527 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11528 (vect_transform_slp_perm_load): Likewise.
11529 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11530 (vectorizable_bswap): Likewise.
11531 (vect_gen_perm_mask_checked): Likewise.
11532 * fold-const.c (fold_ternary_loc): Likewise. Don't take
11533 implementations of variable permutation vectors into account
11534 when deciding which selector to use.
11535 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
11536 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
11537 with a false third argument.
11538 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
11539 to test whether the constant selector is valid and can_vec_perm_var_p
11540 to test whether a variable selector is valid.
11542 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11544 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
11545 * optabs-query.c (can_vec_perm_p): Likewise.
11546 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
11547 instead of vec_perm_indices.
11548 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
11549 (vect_gen_perm_mask_checked): Likewise,
11550 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
11551 (vect_gen_perm_mask_checked): Likewise,
11553 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11555 * optabs-query.h (qimode_for_vec_perm): Declare.
11556 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
11557 (qimode_for_vec_perm): ...this new function.
11558 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
11560 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11562 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
11563 does not have a conditional at the top.
11565 2018-01-02 Richard Biener <rguenther@suse.de>
11567 * ipa-inline.c (big_speedup_p): Fix expression.
11569 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11572 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
11575 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11579 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
11580 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
11581 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
11582 cond_taken_branch_cost 3->4.
11584 2018-01-01 Jakub Jelinek <jakub@redhat.com>
11586 PR tree-optimization/83581
11587 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
11588 TODO_cleanup_cfg if any changes have been made.
11590 PR middle-end/83608
11591 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
11592 convert_modes if target mode has the right side, but different mode
11595 PR middle-end/83609
11596 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
11597 last argument when extracting from CONCAT. If either from_real or
11598 from_imag is NULL, use expansion through memory. If result is not
11599 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
11600 the parts directly to inner mode, if even that fails, use expansion
11603 PR middle-end/83623
11604 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
11605 check for bswap in mode rather than HImode and use that in expand_unop
11608 Copyright (C) 2018 Free Software Foundation, Inc.
11610 Copying and distribution of this file, with or without modification,
11611 are permitted in any medium without royalty provided the copyright
11612 notice and this notice are preserved.