1 2018-04-28 Richard Biener <rguenther@suse.de>
3 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
4 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
5 to reflect use. Only add interesting stmts.
7 2018-04-27 Martin Jambor <mjambor@suse.cz>
10 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
11 the jump function allows for passing through aggregate values.
13 2018-04-27 David Malcolm <dmalcolm@redhat.com>
15 * input.h (in_system_header_at): Convert from macro to inline
17 (from_macro_expansion_at): Likewise.
18 (from_macro_definition_at): Likewise.
20 2018-04-27 Jeff Law <law@redhat.com>
22 * config.gcc: Mark tile* targets as deprecated/obsolete.
24 2018-04-27 Richard Biener <rguenther@suse.de>
26 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
29 2018-04-27 Richard Biener <rguenther@suse.de>
31 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
33 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
35 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
36 with Yd constraint. Set "preferred_for_speed" attribute from
37 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
39 (*movdi_internal): Ditto.
40 (movti_interunit splitters): Remove
41 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
42 (movdi_interunit splitters): Ditto.
43 * config/i386/constraints.md (Ye): Remove.
44 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
46 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
49 * config/aarch64/constraints.md (Usg): Limit to 31.
52 2018-04-27 Jakub Jelinek <jakub@redhat.com>
54 PR tree-optimization/85529
55 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
56 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
57 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
58 zero extension or masking of the MSB bit.
59 (optimize_range_tests): Add FIRST_BB argument, pass it through
60 to optimize_range_tests_var_bound.
61 (maybe_optimize_range_tests, reassociate_bb): Adjust
62 optimize_range_tests callers.
64 2018-04-26 Richard Biener <rguenther@suse.de>
65 Jakub Jelinek <jakub@redhat.com>
67 * cgraph.h (symbol_table): Just declare debug method here.
68 * symtab.c (symbol_table::debug): Define.
70 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
72 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
74 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
76 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
77 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
78 (*movdi_internal): Substitute Yi and Yj constraint with x
79 and Ym and Yn constraint with y constraint. Update "isa"
80 attribute and set "preferred_for_speed" attribute from
81 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
82 (*movsi_internal): Ditto.
83 (*movdf_internal): Ditto.
84 (*movsf_internal): Ditto.
85 (*zero_extendsidi2): Ditto.
86 * config/i386/sse.md (vec_set<mode>_0): Ditto.
88 (*vec_extract<ssevecmodelower>_0): Ditto.
89 (*vec_extractv4si_0_zext_sse4): Ditto.
90 (vec_concatv2di): Ditto.
91 (*vec_dup<mode>): Ditto.
92 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
93 * config/i386/constraints.md (Yi): Remove.
98 2018-04-26 Nathan Sidwell <nathan@acm.org>
100 * dumpfile.c (dump_open): New.
101 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
102 (dump_finish): Detect stdio/stderr by value not name.
104 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
106 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
108 2018-04-26 Tom de Vries <tom@codesourcery.com>
111 * config/nvptx/nvptx.c (verify_neutering_jumps)
112 (verify_neutering_labels): New function
113 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
115 2018-04-26 Tom de Vries <tom@codesourcery.com>
118 * config/nvptx/nvptx.c (needs_neutering_p): New function.
119 (nvptx_single): Use needs_neutering_p to skip over insns that do not
122 2018-04-26 Richard Biener <rguenther@suse.de>
123 Tom de Vries <tom@codesourcery.com>
126 * lto-streamer-out.c (output_function): Fixup loops if required to match
127 discovery done in the reader.
129 2018-04-26 Richard Biener <rguenther@suse.de>
131 PR tree-optimization/85116
132 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
133 have a loop exit from the single latch predecessor. Remove
134 case of header with just condition.
135 (ch_base::copy_headers): Exclude infinite loops from any
137 (pass_ch::execute): Record exits.
139 2018-04-26 Richard Biener <rguenther@suse.de>
141 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
142 prologue cost vector and pass it to vect_get_load_cost.
143 (vect_get_peeling_costs_all_drs): Likewise.
144 (vect_peeling_hash_get_lowest_cost): Likewise.
145 (vect_enhance_data_refs_alignment): Likewise.
147 2018-04-26 Richard Biener <rguenther@suse.de>
150 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
151 checking of integer<->pointer conversions.
152 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
153 sign-/zero-extending pointer types.
154 (expand_omp_for_static_chunk): Likewise.
156 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
157 Jean Lee <xiaoyur347@gmail.com>
159 * config/mips/mips.c (mips_asan_shadow_offset): New function.
160 (TARGET_ASAN_SHADOW_OFFSET): Define.
161 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
162 true for -fsanitize=address.
164 2018-04-25 Mark Wielaard <mark@klomp.org>
166 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
169 2018-04-25 Jakub Jelinek <jakub@redhat.com>
171 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
172 than "alu", remove explicit "memory" and "imm_disp" attributes.
173 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
176 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
177 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
180 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
183 * config/i386/i386.c (ix86_expand_builtin): Change memory
184 operand to XI, extend p0 to Pmode.
185 * config/i386/i386.md: Change unspec volatile and operand
186 1 mode to XI, change operand 0 mode to P.
188 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
190 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
191 GET_MODE_MASK before any checking.
192 (nds32_can_use_bset_p): Likewise.
193 (nds32_can_use_btgl_p): Likewise.
195 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
197 * config/nds32/nds32-doubleword.md: New define_split pattern for
198 illegal register number.
200 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
202 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
204 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
206 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
208 2018-04-25 Richard Biener <rguenther@suse.de>
210 * lto-streamer.h (LTO_major_version): Bump to 8.
212 2018-04-25 Jakub Jelinek <jakub@redhat.com>
214 * BASE-VER: Set to 9.0.0.
216 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
218 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
219 in __abskf2 and __powikf2.
221 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
224 * config/aarch64/constraints.md (Usg, Usj): New constraints.
225 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
226 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
227 Use the above on operand 2. Reindent.
228 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
230 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
233 * common/config/i386/i386-common.c (ix86_handle_option): Don't
235 * config/i386/i386.opt (mcet): Removed.
236 * doc/install.texi: Remove -mcet documentation.
237 * doc/invoke.texi: Likewise.
239 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
242 * doc/install.texi: Remove -mcet from bootstrap-cet.
244 2018-04-24 Jakub Jelinek <jakub@redhat.com>
247 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
248 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
252 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
253 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
254 containing a CONST_VECTOR.
256 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
258 * doc/install.texi: Update newlib dependency for nvptx.
260 2018-04-24 Jakub Jelinek <jakub@redhat.com>
263 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
264 instead of INTVAL when shifting x left.
266 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
268 PR tree-optimization/85478
269 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
270 vect_grouped_store_supported for single element vectors.
272 2018-04-24 Richard Biener <rguenther@suse.de>
275 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
276 load cost increase to the case of non-constant step.
278 2018-04-24 Jakub Jelinek <jakub@redhat.com>
281 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
282 destination if any_malformed_asm.
284 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
287 * expr.c (store_field): In the bitfield case, if the value comes from
288 a function call and is returned in registers by means of a PARALLEL,
289 do not change the mode of the temporary unless BLKmode and VOIDmode.
291 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
293 PR rtl-optimization/85423
294 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
295 dependencies to debug insns when the previous insn is non-debug.
297 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
299 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
300 enums into a single definition.
301 (fls): Fix predicates and printing.
304 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
306 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
307 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
308 and short u6 immediate.
309 (check_if_valid_sleep_operand): Remove.
310 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
312 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
314 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
315 flag_always_save_lp condition.
316 * config/nds32/nds32.opt (malways-save-lp): New option.
318 2018-04-22 Shiva Chen <shiva0217@gmail.com>
320 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
321 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
322 * config/nds32/nds32.h
323 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
324 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
326 2018-04-22 Shiva Chen <shiva0217@gmail.com>
328 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
329 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
331 2018-04-22 Shiva Chen <shiva0217@gmail.com>
332 Chung-Ju Wu <jasonwucj@gmail.com>
334 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
336 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
337 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
339 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
341 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
343 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
345 * config/nds32/nds32-protos.h (nds32_data_alignment,
346 nds32_local_alignment): Declare.
347 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
348 nds32_local_alignment): New functions.
349 (TARGET_CONSTANT_ALIGNMENT): Define.
350 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
352 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
354 * config/nds32/nds32.c
355 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
356 (TARGET_MODES_TIEABLE_P): Likewise.
358 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
360 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
363 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
364 Chung-Ju Wu <jasonwucj@gmail.com>
366 * config/nds32/constants.md (unspec_volatile_element): Add enum values
367 for unaligned access.
368 * config/nds32/nds32-intrinsic.c: Implementation of expanding
370 * config/nds32/nds32-intrinsic.md: Likewise.
371 * config/nds32/nds32_intrinsic.h: Likewise.
372 * config/nds32/nds32.h (nds32_builtins): Likewise.
373 * config/nds32/nds32.opt (munaligned-access): New option.
374 * config/nds32/nds32.c (nds32_asm_file_start): Display
375 flag_unaligned_access status.
377 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
379 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
380 -mno-relax is present.
381 * config/riscv/linux.h (LINK_SPEC): Ditto.
383 2018-04-20 Martin Sebor <msebor@redhat.com>
386 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
388 (gimple_fold_builtin_stxcpy_chk): Same.
389 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
391 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
394 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
395 __powikf2 when long double is IEEE 128-bit.
397 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
399 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
400 step to make sure stack always aligned.
402 2018-04-20 Carl Love <cel@us.ibm.com>
405 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
408 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
409 Tom de Vries <tom@codesourcery.com>
412 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
413 Emit insns for calls too.
414 (nvptx_find_par): Always look for worker-level predecessor insn.
415 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
417 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
418 (nvptx_process_pars): Propagate frames for calls.
420 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
423 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
425 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
426 (ix86_handle_option): Don't handle OPT_mibt.
427 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
429 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
431 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
432 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
433 (ix86_target_macros): Define __CET__ with flag_cf_protection
435 * config/i386/i386.c (isa2_opts): Remove -mibt.
436 * config/i386/i386.h (TARGET_IBT): Removed.
437 (TARGET_IBT_P): Likewise.
438 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
439 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
440 * config/i386/i386.opt (mcet): Update help message.
443 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
444 -mcet as an alias for -mshstk.
446 2018-04-20 Richard Biener <rguenther@suse.de>
449 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
450 complexity by forcing a single use of the multiply operand.
452 2018-04-20 Martin Jambor <mjambor@suse.cz>
455 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
456 recursion dependency to only apply to non-clones.
458 2018-04-20 Martin Jambor <mjambor@suse.cz>
461 * ipa-cp.c (create_specialized_node): Check that clones of
462 self-recursive edges exist during IPA-CP.
464 2018-04-19 Toon Moene <toon@moene.org>
466 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
469 2018-04-19 Jakub Jelinek <jakub@redhat.com>
471 PR tree-optimization/85467
472 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
473 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
474 VECTOR_CST element to type.
476 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
479 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
480 * config/i386/i386.md (builtin_setjmp_setup): Removed.
481 (builtin_longjmp): Likewise.
482 (save_stack_nonlocal): New pattern.
483 (restore_stack_nonlocal): Likewise.
485 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
488 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
489 Replace ASM_OUTPUT_LABEL with fprintf.
491 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
494 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
495 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
496 * config/i386/i386-c.c (ix86_target_macros_internal): Also
497 define __IBT__ and __SHSTK__ for -fcf-protection.
498 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
500 (ix86_trampoline_init): Likewise.
501 (x86_output_mi_thunk): Likewise.
502 (ix86_notrack_prefixed_insn_p): Likewise.
503 (ix86_option_override_internal): Don't disallow -fcf-protection.
504 * config/i386/i386.md (rdssp<mode>): Also enable for
506 (incssp<mode>): Likewise.
507 (nop_endbr): Likewise.
508 * config/i386/i386.opt (mcet): Change help message to built-in
512 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
513 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
514 enable CET built-in functions.
516 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
518 * common/config/i386/i386-common.c
519 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
520 OPTION_MASK_ISA_MOVDIRI_UNSET,
521 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
522 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
523 * config.gcc (movdirintrin.h): New header.
524 * config/i386/cpuid.h (bit_MOVDIRI,
525 bit_MOVDIR64B): New bits.
526 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
528 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
529 (VOID, PVOID, PCVOID)): New function types.
530 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
531 __builtin_ia32_directstoreu_u64,
532 __builtin_ia32_movdir64b): New builtins.
533 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
534 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
536 (ix86_valid_target_attribute_inner_p): Ditto.
537 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
538 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
539 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
540 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
541 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
542 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
543 (movdiri<mode>, movdir64b_<mode>): New.
544 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
545 * config/i386/immintrin.h: Include movdirintrin.h.
546 * config/i386/movdirintrin.h: New file.
547 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
549 2018-04-19 Richard Biener <rguenther@suse.de>
552 * cfg.c (clear_bb_flags): When loop state says we have
553 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
555 2018-04-19 Richard Biener <rguenther@suse.de>
557 PR tree-optimization/84737
558 * tree-vect-data-refs.c (vect_copy_ref_info): New function
559 copying restrict info.
560 (vect_setup_realignment): Use it.
561 * tree-vectorizer.h (vect_copy_ref_info): Declare.
562 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
563 the first DR to all generated stores.
564 (vectorizable_load): Likewise for loads.
566 2018-04-19 Jakub Jelinek <jakub@redhat.com>
568 PR tree-optimization/85446
569 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
570 the integral and pointer types to have the same precision.
572 * doc/install.texi: Document --disable-cet being the default and
575 2018-04-18 Martin Liska <mliska@suse.cz>
577 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
580 2018-04-18 Martin Liska <mliska@suse.cz>
583 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
586 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
587 arguments if they are comparable.
589 2018-04-18 Martin Liska <mliska@suse.cz>
592 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
595 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
598 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
601 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
602 ENDBR after calling __morestack.
604 2018-04-18 David Malcolm <dmalcolm@redhat.com>
607 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
608 by using gcc_base_ver to generate a gcc_driver_version, and use
609 it when generating GCC_DRIVER_NAME.
610 * configure: Regenerate.
612 2018-04-18 Jakub Jelinek <jakub@redhat.com>
615 * config.gcc: Obsolete powerpc*-*-*spe*.
617 2018-04-17 Jakub Jelinek <jakub@redhat.com>
620 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
621 (stabstr_D): Change type of unum from unsigned int to
622 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
625 2018-04-17 Jim Wilson <jimw@sifive.com>
628 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
629 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
630 Set arg_pointer_offset after using pretend_args_size.
632 2018-04-17 Jakub Jelinek <jakub@redhat.com>
634 PR rtl-optimization/85431
635 * dse.c (record_store): Ignore zero width stores.
638 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
639 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
640 __builtin_stack_restore rather than after it.
641 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
642 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
643 argument instead of virtual_dynamic_stack_rtx.
645 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
647 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
649 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
650 Add note to error message to explain internal mapping of overloaded
651 built-in function name to non-overloaded built-in function name.
652 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
655 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
658 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
659 where the inputs overlap with the output.
661 2018-04-17 Jakub Jelinek <jakub@redhat.com>
663 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
664 (=v, v) alternative and explicit "memory" attribute.
665 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
666 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
668 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
669 "sselog1" type instead of "sselog".
670 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
671 "sselog". Remove explicit "memory" attribute.
672 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
673 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
675 (vec_extract_hi_v32hi): Merge all alternatives into one, use
676 "sselog1" type instead of "sselog". Remove explicit "memory"
678 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
679 use "sselog1" type instead of "sselog". Remove explicit "memory"
681 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
682 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
684 (vec_extract_hi_v64qi): Merge all alternatives into one, use
685 "sselog1" type instead of "sselog". Remove explicit "memory"
687 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
688 use "sselog1" type instead of "sselog". Remove explicit "memory"
692 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
695 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
698 2018-04-17 Martin Jambor <mjambor@suse.cz>
701 * ipa-cp.c (create_specialized_node): Call
702 expand_all_artificial_thunks if necessary.
704 2018-04-17 Martin Liska <mliska@suse.cz>
707 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
708 in message, remote space in between '_G' and '('.
710 2018-04-17 Jakub Jelinek <jakub@redhat.com>
713 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
714 avx512f_vmcmp<mode>3<round_saeonly_name>,
715 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
716 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
717 avx512f_rndscale<mode><round_saeonly_name>,
718 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
719 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
720 Use %<iptr>2 instead of %2 for -masm=intel.
721 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
722 avx512f_vcvttss2usi<round_saeonly_name>,
723 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
725 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
726 avx512f_vcvttsd2usi<round_saeonly_name>,
727 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
728 Use %q1 instead of %1 for -masm=intel.
729 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
730 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
731 of %3 for -masm=intel.
732 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
734 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
736 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
738 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
739 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
741 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
743 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
744 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
745 %g1 and one with %0 and %1.
746 (avx512er_vmrcp28<mode><round_saeonly_name>,
747 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
749 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
750 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
751 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
752 of %0 and %{%4%} for -masm=intel.
753 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
754 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
755 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
756 order of %0 and %{%5%}%{z%} for -masm=intel.
758 2018-04-17 Jan Hubicka <jh@suse.cz>
761 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
763 2018-04-17 Martin Liska <mliska@suse.cz>
766 * multiple_target.c (create_dispatcher_calls): Set apostrophes
767 for target_clone error message. Make default implementation
768 clone to be a local declaration.
769 (separate_attrs): Add new argument and check for an empty
771 (expand_target_clones): Handle it.
772 (ipa_target_clone): Make redirection just for target_clones
775 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
776 Tom de Vries <tom@codesourcery.com>
779 * omp-expand.c (expand_oacc_for): Add dummy false branch for
780 tiled basic blocks without omp continue statements.
782 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
785 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
786 vec_extract expression as having side effects to make sure it gets
789 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
792 * config/i386/i386.c (get_builtin_code_for_version): Check
795 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
798 * gcc/config.gcc: Support "skylake".
799 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
801 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
802 (processor_target_table): Add "skylake".
803 (ix86_option_override_internal): Add "skylake".
804 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
805 PROCESSOR_CANNONLAKE.
806 (get_builtin_code_for_version): Fix priority for
807 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
808 PROCESSOR_SKYLAKE-AVX512.
809 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
810 (processor_type): Add PROCESSOR_SKYLAKE.
812 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
813 Jason Merrill <jason@redhat.com>
816 * convert.c (convert_to_integer_1): Use direct recursion for
817 enumeral types and types with a precision less than the number
818 of bits in their mode.
820 2018-04-16 Julia Koval <julia.koval@intel.com>
823 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
824 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
826 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
829 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
830 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
831 and -mno-direct-move.
833 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
836 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
837 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
838 Ensure negative shifts result in {0}.
840 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
842 PR rtl-optimization/79916
843 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
844 regs (if any) to define how to gnerate SD moves when LRA is in
847 2018-04-13 Jakub Jelinek <jakub@redhat.com>
849 PR rtl-optimization/85393
850 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
851 * except.c (expand_dw2_landing_pad_for_region): Make static.
852 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
853 a label and unconditional jump to old_bb, rather than
854 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
857 PR rtl-optimization/85376
858 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
859 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
860 instead of a specific value.
862 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
863 Bin Cheng <bin.cheng@arm.com>
865 PR tree-optimization/82965
866 PR tree-optimization/83991
867 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
868 by_profile_only parameter.
869 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
870 information if the loop was predicted to iterate too many times.
871 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
873 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
876 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
879 2018-04-13 Martin Liska <mliska@suse.cz>
880 Jakub Jelinek <jakub@redhat.com>
883 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
884 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
885 * builtins.c (expand_builtin_memory_copy_args): Use
886 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
887 handle dest_addr == pc_rtx.
889 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
892 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
893 asked to not generate direct moves.
894 (fix_trunc<mode>si2_stfiwx): Similar.
895 (fix_trunc<mode>si2_internal): Similar.
897 2018-04-12 Jakub Jelinek <jakub@redhat.com>
900 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
901 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
902 lookup if dest in some wider mode is known to be const0_rtx and
903 if so, record permanent equivalence for it to be ZERO_EXTEND of
904 the narrower mode destination.
906 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
908 * lto-streamer-out.c (output_function): Revert 259346.
909 * omp-expand.c (expand_oacc_for): Likewise.
911 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
913 PR rtl-optimization/85354
914 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
915 * sel-sched.c (sel_global_init): ... here.
917 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
920 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
921 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
922 mode for PE-COFF targets.
923 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
924 (i386_pe_asm_lto_end): Likewise.
925 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
926 (TARGET_ASM_LTO_END): Likewise.
927 * config/i386/winnt.c (saved_debug_info_level): New static variable.
928 (i386_pe_asm_lto_start): New function.
929 (i386_pe_asm_lto_end): Likewise.
931 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
932 Richard Biener <rguenther@suse.de>
935 * lto-streamer-out.c (output_function): Fix CFG loop state before
937 * omp-expand.c (expand_oacc_for): Handle calls to internal
938 functions like regular functions.
940 2018-04-12 Richard Biener <rguenther@suse.de>
943 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
944 for the early LTO debug to properly generate references to it
945 during DIE emission. Do not re-use that for the skeleton for
947 (dwarf2out_early_finish): Likewise.
949 2018-04-12 Jakub Jelinek <jakub@redhat.com>
953 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
954 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
955 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
956 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
957 and output is a reg, avoid creating invalid lowpart subreg, but
958 instead split into a 512-bit move. Don't split if not AVX512VL,
959 input is xmm16+ reg and output is a mem.
960 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
961 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
962 xmm16+ reg and output is a mem.
964 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
966 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
967 also for flag_dwarf2_cfi_asm.
969 2018-04-12 Jakub Jelinek <jakub@redhat.com>
971 PR rtl-optimization/85342
972 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
973 a bool scalar var inside of the loop instead. Don't try to update
974 recog_data.operand after failed apply_change_group.
976 2018-04-12 Tom de Vries <tom@codesourcery.com>
979 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
980 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
981 array with flexible array member as array without given dimension.
982 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
983 argument for undefined param to true.
985 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
988 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
989 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
990 from PowerPC section.
991 * config/rs6000/sysv4.opt (mcall-): Improve help text.
992 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
993 help text that is too long.
994 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
995 help text that is too long.
996 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
997 help text that is too long.
999 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
1001 * config/alpha/alpha.md (stack_probe_internal): Rename
1002 from "probe_stack". Update all callers.
1004 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1006 PR rtl-optimization/84566
1007 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
1008 sched_macro_fuse_insns.
1010 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1013 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
1014 (compute_block_dependences): ... from here.
1016 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1018 PR tree-optimization/85331
1019 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
1020 from int to HOST_WIDE_INT.
1022 2018-04-11 Martin Jambor <mjambor@suse.cz>
1025 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
1026 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
1027 not the same as the source val.
1028 (cgraph_edge_brings_value_p): New parameter.
1029 (gather_edges_for_value): Pass destination value to
1030 cgraph_edge_brings_value_p.
1031 (perhaps_add_new_callers): Likewise.
1032 (get_info_about_necessary_edges): Likewise and exclude values brought
1033 only by self-recursive edges.
1034 (create_specialized_node): Redirect only clones of self-calling edges.
1035 (+self_recursive_pass_through_p): New function.
1036 (find_more_scalar_values_for_callers_subset): Use it.
1037 (find_aggregate_values_for_callers_subset): Likewise.
1038 (known_aggs_to_agg_replacement_list): Removed.
1039 (decide_whether_version_node): Re-calculate known constants for all
1040 remaining context clones.
1042 2018-04-11 Richard Biener <rguenther@suse.de>
1045 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
1046 from early DWARF output.
1047 (dwarf2out_early_finish): Output line info unconditionally into
1048 early DWARF and add reference to it.
1050 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1053 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
1054 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
1055 other than V2DFmode using iptr mode attribute.
1056 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
1058 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1060 PR rtl-optimization/84659
1061 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
1063 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1066 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
1068 (output_loc_list): Pass address of a dummy size variable even in the
1069 locview handling loop.
1070 (index_location_lists): Add comment on why skip_loc_list_entry can't
1073 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
1076 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
1079 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
1082 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1083 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
1084 and -mstring-compare-inline-limit.
1086 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1089 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
1090 for stack clash protection in a register whenever we need it to be in
1093 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1095 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
1096 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
1098 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1101 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
1103 (mlong-double-): Ditto.
1104 * config/rs6000/sysv4.opt (msdata=): Ditto.
1105 (mtls-size=): Ditto.
1107 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
1109 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1110 erroneous entries for
1111 "vector int vec_ldl (int, long int *)", and
1112 "vector unsigned int vec_ldl (int, unsigned long int *)".
1113 Add comments and entries for
1114 "vector bool char vec_ldl (int, bool char *)",
1115 "vector bool short vec_ldl (int, bool short *)",
1116 "vector bool int vec_ldl (int, bool int *)",
1117 "vector bool long long vec_ldl (int, bool long long *)",
1118 "vector pixel vec_ldl (int, pixel *)",
1119 "vector long long vec_ldl (int, long long *)",
1120 "vector unsigned long long vec_ldl (int, unsigned long long *)".
1121 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
1122 type tree bool_long_long_type_node and correct definition of
1123 bool_V2DI_type_node to make reference to this new type tree.
1124 (rs6000_mangle_type): Replace erroneous reference to
1125 bool_long_type_node with bool_long_long_type_node.
1126 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
1127 comments to emphasize sign distinctions for char and int types and
1128 replace RS6000_BTI_bool_long constant with
1129 RS6000_BTI_bool_long_long constant. Also add comment to restrict
1130 use of RS6000_BTI_pixel.
1131 (bool_long_type_node): Remove this macro definition.
1132 (bool_long_long_type_node): New macro definition
1134 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1136 PR rtl-optimization/85300
1137 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
1138 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
1139 simplify_unary_operation fails.
1141 2018-04-10 Martin Liska <mliska@suse.cz>
1143 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
1144 cgraph_edge and ipa_ref.
1146 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1150 * config/i386/sse.md
1151 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
1152 computation of the VEC_MERGE selector from mask.
1153 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
1154 Fix decoding of the VEC_MERGE selector into mask.
1156 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1158 PR tree-optimization/85286
1159 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
1161 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1163 * final.c (final_1): Set insn_last_address as well as
1164 insn_current_address.
1166 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1169 * explow.c (emit_stack_probe): Call validize_mem on memory location
1170 before passing it to gen_probe_stack. Create address operand and
1171 legitimize it for the probe_stack_address case.
1173 2018-04-09 Jan Hubicka <jh@suse.cz>
1176 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
1177 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
1178 * tree.c (free_lang_data_in_type): Fix handling of binfos;
1180 (free_lang_data): Rebuild type inheritance graph.
1182 2018-04-09 Martin Sebor <msebor@redhat.com>
1184 * invoke.texi (-finline-small-functions): Mention other optimization
1186 (-findirect-inlining, -fpartial-inlining): Same.
1187 (-finline-functions-called-once): Same.
1188 (-freorder-blocks-and-partition): Same.
1190 2018-04-09 Jan Hubicka <jh@suse.cz>
1193 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
1194 jumps; choose last target that matches the criteria (i.e.
1195 no partition changes for non-crossing jumps).
1196 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
1197 support for redirecting crossing jumps to non-crossing.
1199 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
1201 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
1202 also for naked functions.
1204 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
1206 * config/arc/arc.md (add_shift): New pattern.
1207 (add_shift2): Likewise.
1208 (sub_shift): Likewise.
1209 (sub_shift_cmp0_noout): Likewise.
1210 (compare_si_ashiftsi): Likewise.
1211 (xbfu_cmp0_noout): New combine pattern.
1212 (xbfu_cmp0"): Likewise.
1213 (movsi_set_cc_insn): Place the predicable variant first.
1214 (commutative_binary_cmp0_noout): Remove clobber.
1215 (commutative_binary_cmp0): New pattern.
1216 (noncommutative_binary_cmp0): Likewise.
1217 (noncommutative_binary_cmp0_noout): Likewise.
1218 (noncommutative_binary_comparison_result_used): Removed.
1219 (rsub_cmp0): New pattern.
1220 (rsub_cmp0_noout): Likewise.
1221 (extzvsi): Changed, keep only meaningful variants.
1222 (SQH, SEZ): New iterators.
1223 (SQH_postfix): New mode attribute.
1224 (SEZ_prefix): New code attribute.
1225 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
1226 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
1227 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
1229 (noncommutative_operator): Check the availability of barrel
1232 2018-04-09 Richard Biener <rguenther@suse.de>
1234 PR tree-optimization/85284
1235 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
1236 Only use the niter constraining form of simple_iv when the exit
1239 2018-04-09 Tom de Vries <tom@codesourcery.com>
1242 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1243 (define_expand "*memory_barrier"): New define_expand.
1244 (define_insn "memory_barrier"): New insn.
1246 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1248 PR rtl-optimization/80463
1249 PR rtl-optimization/83972
1250 PR rtl-optimization/83480
1252 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
1253 correct producer for the insn.
1254 (tidy_control_flow): Fixup seqnos in case of debug insns.
1256 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1258 PR rtl-optimization/83913
1260 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
1261 different sched-times when merging exprs.
1263 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1265 PR rtl-optimization/83962
1267 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
1268 tidy_fallthru_edge and tidy_control_flow.
1270 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1272 PR rtl-optimization/83530
1274 * sel-sched.c (force_next_insn): New global variable.
1275 (remove_insn_for_debug): When force_next_insn is true, also leave only
1276 next insn in the ready list.
1277 (sel_sched_region): When the region wasn't scheduled, make another pass
1278 over it with force_next_insn set to 1.
1280 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1282 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
1284 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1285 for interrupt control.
1286 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1287 functions for interrupt control.
1288 * config/nds32/nds32-intrinsic.md: Likewise.
1289 * config/nds32/nds32_intrinsic.h: Likewise.
1290 * config/nds32/nds32.h (nds32_builtins): Likewise.
1292 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1294 * config/nds32/nds32.c (nds32_init_machine_status,
1295 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
1296 strict_aligned_p field.
1297 (nds32_expand_to_rtl_hook): New function.
1298 (TARGET_EXPAND_TO_RTL_HOOK): Define.
1299 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
1301 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1302 Chung-Ju Wu <jasonwucj@gmail.com>
1304 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
1305 * config/nds32/nds32-n7.md: New file.
1306 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
1307 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
1309 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
1310 * config/nds32/nds32.md (pipeline_model): Add n7.
1311 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
1312 * config/nds32/pipelines.md: Include n7 settings.
1314 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1315 Chung-Ju Wu <jasonwucj@gmail.com>
1317 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
1318 * config/nds32/nds32-e8.md: New file.
1319 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
1320 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
1322 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
1323 * config/nds32/nds32.md (pipeline_model): Add e8.
1324 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
1325 * config/nds32/pipelines.md: Include e8 settings.
1327 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1328 Chung-Ju Wu <jasonwucj@gmail.com>
1330 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
1331 * config/nds32/nds32-n8.md: New file.
1332 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
1333 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
1335 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
1336 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
1337 * config/nds32/nds32.md (pipeline_model): Add n8.
1338 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
1339 * config/nds32/pipelines.md: Include n8 settings.
1341 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1342 Chung-Ju Wu <jasonwucj@gmail.com>
1344 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
1345 * config/nds32/nds32-n9-2r1w.md: New file.
1346 * config/nds32/nds32-n9-3r2w.md: New file.
1347 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
1348 nds32_register_ports): New or modify for cpu n9.
1349 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
1351 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
1352 * config/nds32/nds32-utils.c: New file.
1353 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
1354 TARGET_MUL_SLOW): Define.
1355 * config/nds32/nds32.md (pipeline_model): New attribute.
1356 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
1357 New options that support cpu n9.
1358 * config/nds32/pipelines.md: Include n9 settings.
1359 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
1361 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1363 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
1364 information if necessary.
1365 (output_cond_branch_compare_zero): Likewise.
1366 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
1367 (nds32_target_alignment): Refine for alignment.
1368 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
1369 (FUNCTION_BOUNDARY): Modify.
1370 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
1372 * config/nds32/nds32.opt (malways-align, malign-functions): New.
1374 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1376 * config/nds32/constants.md (unspec_volatile_element): Add values for
1377 TLB operation and data prefetch.
1378 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1379 functions for TLB operation and data prefetch.
1380 * config/nds32/nds32-intrinsic.md: Likewise.
1381 * config/nds32/nds32_intrinsic.h: Likewise.
1382 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
1383 (nds32_print_operand): Likewise.
1384 * config/nds32/nds32.h (nds32_builtins): Likewise.
1386 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
1387 Andrew Pinski <pinsika@gcc.gnu.org>
1390 * match.pd: Use constant_boolean_node of correct type instead of
1391 boolean_true_node or boolean_false_node for simplifying
1392 pointer comparisons to zero.
1394 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1396 PR tree-optimization/80021
1397 * tree.c (verify_type_variant): Make error call in verify_variant_match
1398 translatable and remove final full stop.
1400 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1402 * config/nds32/constants.md (unspec_volatile_element): Add
1403 UNSPEC_VOLATILE_EH_RETURN.
1404 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
1405 nds32_output_stack_pop): Support dwarf exception handling process.
1406 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
1407 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
1408 exception handling process.
1409 (nds32_compute_stack_frame): Likewise.
1410 (nds32_return_addr_rtx): Likewise.
1411 (nds32_initial_elimination_offset): Likewise.
1412 (nds32_expand_prologue): Likewise.
1413 (nds32_expand_epilogue): Likewise.
1414 (nds32_dynamic_chain_address): New function.
1415 * config/nds32/nds32.h (machine_function): Add fields for dwarf
1417 (DYNAMIC_CHAIN_ADDRESS): Define.
1418 (EH_RETURN_DATA_REGNO): Define.
1419 (EH_RETURN_STACKADJ_RTX): Define.
1420 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
1421 patterns for dwarf exception handling.
1423 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1425 * config/nds32/nds32.h: Clean up obsolete macros.
1427 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1429 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1430 Add enum values for particular instructions.
1431 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1432 particular intrinsic functions.
1433 * config/nds32/nds32-intrinsic.md: Likewise.
1434 * config/nds32/nds32_intrinsic.h: Likewise.
1435 * config/nds32/nds32.h (nds32_builtins): Likewise.
1436 * config/nds32/nds32.md (type): Add pbsad and pbsada.
1437 (btst, ave): New patterns for particular instructions.
1439 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1441 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1442 Add enum values for atomic load/store and memory sync.
1443 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
1445 * config/nds32/nds32-intrinsic.md: Likewise.
1446 * config/nds32/nds32_intrinsic.h: Likewise.
1447 * config/nds32/nds32.h (nds32_builtins): Likewise.
1449 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1451 PR tree-optimization/85257
1452 * fold-const.c (native_encode_vector): If not all elts could fit
1453 and off is -1, return 0 rather than offset.
1454 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
1455 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
1456 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
1457 adjust buffer in native_interpret_expr call.
1459 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1461 * config/nds32/constants.md (unspec_volatile_element): Add cache
1462 control enum values.
1463 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
1464 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
1465 * config/nds32/nds32.c (nds32_cctl_names): New.
1466 (nds32_print_operand): Handle cache control register names.
1467 * config/nds32/nds32.h (nds32_builtins): New enum values.
1468 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
1470 * config/nds32/nds32.md (type): Add mmu.
1471 * config/nds32/pipelines.md (simple_insn): Add mmu.
1473 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1475 * config/nds32/nds32.md (type): Remove call.
1476 * config/nds32/pipelines.md (simple_insn): Likewise.
1478 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1480 * config/nds32/constants.md (unspec_volatile_element): Add
1481 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
1482 UNSPEC_VOLATILE_FMFCFG.
1483 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
1484 description for fmfcfg and fmfcsr.
1485 (bdesc_1arg): Add fmtcsr.
1486 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
1487 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
1488 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
1489 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
1490 unspec_fmfcfg): New patterns.
1491 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
1492 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
1493 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
1494 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
1495 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
1496 __nds32__fmfcfg): Define.
1498 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1500 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
1501 intrinsic register names.
1502 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
1503 intrinsic register enum values and macros.
1505 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1507 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
1508 for load/store addressing form.
1509 (nds32_print_operand_address): Likewise.
1511 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
1514 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
1515 based on LABEL_REF. Remove useless assertion.
1516 (pic_address_needs_scratch): Fix formatting.
1517 (sparc_legitimize_pic_address): Minor tweaks.
1518 (sparc_delegitimize_address): Adjust assertion accordingly.
1519 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
1520 into symbolic_operand.
1521 (movsi_high_pic_label_ref): Likewise.
1522 (movsi_lo_sum_pic_label_ref): Likewise.
1523 (movdi_pic_label_ref): Likewise.
1524 (movdi_high_pic_label_ref): Likewise.
1525 (movdi_lo_sum_pic_label_ref): Likewise.
1527 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
1529 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
1530 custom LIB_SPEC setup.
1532 2018-04-06 Ruslan Bukin <br@bsdpad.com>
1533 Kito Cheng <kito.cheng@gmail.com>
1535 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
1536 * config/riscv/freebsd.h: New.
1538 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1540 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
1541 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
1544 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1545 Kito Cheng <kito.cheng@gmail.com>
1547 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
1548 nds32_output_call, nds32_symbol_binds_local_p): New functions.
1549 * config/nds32/nds32-protos.h (nds32_output_call,
1550 nds32_output_return): Declare.
1551 * config/nds32/nds32.md: Refine all the call and return patterns.
1553 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1556 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
1557 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
1559 PR rtl-optimization/84872
1560 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
1561 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
1564 2018-04-06 Tamar Christina <tamar.christina@arm.com>
1566 * expr.c (copy_blkmode_to_reg): Revert 254862.
1567 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
1569 2018-04-06 Richard Biener <rguenther@suse.de>
1572 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
1573 after seeing a component reference with an adjacent field. Treat
1574 refs to arrays at struct end of external decls similar to
1575 refs to unconstrained commons.
1577 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1580 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
1581 look through SAVE_EXPRs with non-side-effects argument. Adjust
1583 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
1586 2018-04-06 Richard Biener <rguenther@suse.de>
1589 * alias.c (find_base_term): New wrapper around find_base_term
1590 unwinding CSELIB_VAL_PTR changes.
1591 (find_base_term): Do not restore CSELIB_VAL_PTR during the
1594 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1596 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
1598 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
1599 constant definitions.
1600 ("nop"): lr 0,0 -> nopr r0
1601 ("nop_lr0", "nop_lr1"): New insn definitions.
1603 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1605 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
1606 NDS32_V3PUSH_AVAILABLE_P macro.
1608 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
1609 Chung-Ju Wu <jasonwucj@gmail.com>
1611 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
1612 (nds32*-*-*): Add float and fpu_config into supported_defaults.
1613 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
1614 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
1615 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
1616 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
1617 * config/nds32/constraints.md: New constraints and checking for hard
1618 float configuration.
1619 * config/nds32/iterators.md: New mode iterator and attribute for hard
1620 float configuration.
1621 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
1623 * config/nds32/nds32-fpu.md: New file.
1624 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
1625 deal with hard float code generation.
1626 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
1628 (abi_type, float_reg_number): New enum type.
1629 * config/nds32/nds32-predicates.c: New predicates for hard float.
1630 * config/nds32/nds32-protos.h: Declare functions for hard float.
1631 * config/nds32/nds32.c: Implementation for hard float configuration.
1632 * config/nds32/nds32.h: Definitions for hard float configuration.
1633 * config/nds32/nds32.md: Include hard float machine description and
1634 modify patterns for hard float configuration.
1635 * config/nds32/nds32.opt: New options for hard float configuration.
1636 * config/nds32/predicates.md: New predicates for hard float
1639 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1641 * common/config/nds32/nds32-common.c
1642 (nds32_option_optimization_table): Enable -mreleax-hint by default.
1644 2018-04-05 Jakub Jelinek <jakub@redhat.com>
1647 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
1648 CONSTRUCTOR_ELT (ctor, ...)->value.
1650 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
1653 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
1655 2018-04-05 Tom de Vries <tom@codesourcery.com>
1658 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
1661 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1662 Kito Cheng <kito.cheng@gmail.com>
1664 * config/nds32/constraints.md (U33): Fine-tune checking condition.
1665 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
1666 * config/nds32/nds32.h (nds32_16bit_address_type): Add
1667 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
1669 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1670 Kito Cheng <kito.cheng@gmail.com>
1672 * config/nds32/constraints.md (Ufe): New memory constraint.
1673 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
1674 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
1675 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
1677 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
1678 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
1680 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1682 * config/nds32/nds32.md: Use optimize_size in the condition for
1683 alu-shift instructions.
1685 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1687 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
1689 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1691 * config/nds32/nds32.md (negsi2): Refine pattern.
1693 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1694 Chung-Ju Wu <jasonwucj@gmail.com>
1696 * config/nds32/iterators.md (shift_rotate): New code iterator.
1697 (shift): New code attribute.
1698 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
1699 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
1700 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
1701 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
1702 bit-wise operations.
1703 (andsi3, *andsi3): Ditto.
1704 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
1705 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
1706 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
1707 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
1708 nds32_ior_operand, nds32_xor_operand): New predicates.
1710 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1712 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
1713 (addsi3, subsi3): ... this.
1715 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1717 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
1719 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1721 * config/nds32/nds32.md: Adjust indention.
1723 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1725 * config/nds32/nds32.md (feature): New attribute.
1727 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1729 * config/nds32/nds32.md (subtype): New attribute.
1731 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
1734 * config/arm/arm-builtins.c (arm_expand_builtin): Change
1735 expansion to perform a bitwise AND of the argument followed by a
1736 boolean negation of the result.
1738 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
1740 PR rtl-optimization/84878
1741 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
1742 the basic block. Assert the use reference is not artificial and that
1743 it has an associated insn.
1745 2018-04-04 Michael Matz <matz@suse.de>
1747 * builtins.c (compute_objsize): Pass correct operand
1748 to array_at_struct_end_p.
1750 2018-04-04 Richard Biener <rguenther@suse.de>
1753 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
1754 from contexts for DINFO_LEVEL_TERSE and below.
1756 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1758 * config/nds32/nds32-doubleword.md (move_<mode>): Require
1759 resiter_operand condition.
1760 * config/nds32/nds32.md (*move<mode>): Ditto.
1762 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1763 Monk Chiang <sh.chiang04@gmail.com>
1765 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
1767 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1769 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
1771 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1772 Kito Cheng <kito.cheng@gmail.com>
1774 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
1775 nds32_cond_code_str, output_cond_branch,
1776 output_cond_branch_compare_zero, nds32_expand_cbranch,
1777 nds32_expand_cstore, nds32_expand_movcc,
1778 nds32_output_cbranchsi4_equality_zero,
1779 nds32_output_cbranchsi4_equality_reg,
1780 nds32_output_cbranchsi4_equality_reg_or_const_int,
1781 nds32_output_cbranchsi4_greater_less_zero: New functions.
1782 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
1783 nds32_expand_cstore, nds32_expand_movcc,
1784 nds32_output_cbranchsi4_equality_zero,
1785 nds32_output_cbranchsi4_equality_reg,
1786 nds32_output_cbranchsi4_equality_reg_or_const_int,
1787 nds32_output_cbranchsi4_greater_less_zero): Declare.
1788 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
1789 nds32_rimm11s_operand): New predicates.
1790 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
1791 * config/nds32/nds32.md: Rewrite all the branch and conditional move
1794 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1796 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
1797 * config/nds32/nds32.md: Ditto.
1798 * config/nds32/pipelines.md: Ditto.
1800 2018-04-04 Richard Biener <rguenther@suse.de>
1802 PR tree-optimization/85168
1803 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
1804 propagating abnormals.
1806 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1808 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
1810 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1811 Kito Cheng <kito.cheng@gmail.com>
1813 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
1814 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
1815 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
1816 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
1817 * config/nds32/nds32.md (sibcall_internal): New.
1818 (sibcall_register): Remove.
1819 (sibcall_immediate): Remove.
1820 (sibcall_value_internal): New.
1821 (sibcall_value_register): Remove.
1822 (sibcall_value_immediate): Remove.
1823 * config/nds32/predicates.md (nds32_general_register_operand): New.
1824 (nds32_call_address_operand): New.
1826 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1828 PR rtl-optimization/85167
1829 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
1830 bb_defs if *split_p, instead preinitialize it to NULL.
1832 PR tree-optimization/85156
1833 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
1834 evaluating the argument multiple times.
1836 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
1838 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
1840 (_mm_cvtpd_ps): Likewise.
1841 (_mm_cvttpd_epi32): Likewise.
1842 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
1843 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
1844 vector, pixel, and bool following altivec.h include.
1846 2018-04-03 Martin Sebor <msebor@redhat.com>
1848 * doc/extend.texi (Common Function Attributes): Clarify.
1849 (const attribute): Likewise.
1850 (pure attribute): Likewise.
1852 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1855 * config/i386/i386.c (ix86_expand_vector_set): Use
1856 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
1858 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
1860 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
1861 instructions when changing rounding bits to preserve precision bits
1862 in the x87 control word.
1864 2018-04-03 Martin Liska <mliska@suse.cz>
1866 PR tree-optimization/82491
1867 * rtl.h (strip_offset_and_add): Replace += suboffset with
1868 poly_uint64 () + suboffset.
1870 2018-03-29 Martin Liska <mliska@suse.cz>
1871 Martin Jambor <mjambor@suse.cz>
1874 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
1875 param_type is not an integral or pointer type.
1877 2018-04-03 Richard Biener <rguenther@suse.de>
1879 * sese.h (recompute_all_dominators): Remove.
1881 2018-04-02 Martin Sebor <msebor@redhat.com>
1883 * doc/invoke.texi (-Wrestrict): Fix typos.
1885 2018-04-02 Jim Wilson <jimw@sifive.com>
1887 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
1888 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
1889 (<optab>di3, <optab>si3_extend): Likewise.
1890 (<optab>si3_mask, <optab>si3_mask_1): New.
1891 (<optab>di3_mask, <optab>di3_mask_1): New.
1892 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
1893 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
1894 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
1896 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
1898 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
1901 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
1903 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
1904 (nds32_canonicalize_comparison): New function.
1906 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1907 Kito Cheng <kito.cheng@gmail.com>
1908 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1910 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
1911 * config/nds32/constants.md (unspec_volatile_element): Add
1912 UNSPEC_VOLATILE_RELAX_GROUP.
1913 * config/nds32/nds32-relax-opt.c: New file.
1914 * config/nds32/nds32-predicates.c
1915 (nds32_symbol_load_store_p): New function.
1916 * config/nds32/nds32-protos.h
1917 (nds32_symbol_load_store_p): Declare function.
1918 (make_pass_nds32_relax_opt): Declare new rtl pass function.
1919 * config/nds32/nds32.c
1920 (nds32_register_pass): New function to register pass.
1921 (nds32_register_passes): New function to register passes.
1922 * config/nds32/nds32.md (relax_group): New pattern.
1923 * config/nds32/nds32.opt (mrelax-hint): New option.
1924 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
1926 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
1928 * config/nds32/t-nds32: Modify files dependency.
1930 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1932 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
1933 (PROFILE_HOOK): Define its implementation.
1935 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1937 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
1938 type and 32-bit size.
1940 2018-04-01 Jakub Jelinek <jakub@redhat.com>
1943 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
1944 (V_128_256): New mode iterator.
1945 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
1946 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
1947 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
1949 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
1950 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
1952 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
1955 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
1956 NaN inputs correctly.
1958 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
1961 * config/rs6000/vsx.md (??r): New mode attribute.
1962 (*vsx_mov<mode>_64bit): Use it.
1963 (*vsx_mov<mode>_32bit): Likewise.
1965 2018-03-30 Martin Sebor <msebor@redhat.com>
1967 PR tree-optimization/84818
1968 * builtins.c (check_access): Use warning_n.
1970 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1973 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
1975 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
1978 2018-03-30 Julia Koval <julia.koval@intel.com>
1981 * x86-tune.def (movx, partial_reg_dependency): Enable for
1984 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
1987 * lra-constraints.c (process_alt_operands): Move setting
1988 this_alternative_matches below.
1990 2018-03-29 Martin Liska <mliska@suse.cz>
1993 * doc/invoke.texi: Document how LTO works with debug info.
1994 Describe auto-load support of binutils. Mention 'x86-64'
1995 as valid option value of -march option.
1997 2018-03-29 Jakub Jelinek <jakub@redhat.com>
1999 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
2002 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
2003 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
2004 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
2007 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2010 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
2011 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
2012 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
2013 for RS6000_BTM_POWERPC64.
2014 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
2015 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
2016 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
2021 2018-03-28 Carl Love <cel@us.ibm.com>
2024 2017-09-27 Carl Love <cel@us.ibm.com>
2026 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
2027 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
2028 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
2031 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2033 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
2034 instead of __vector bool.
2035 (_mm_max_pu8): Likewise.
2036 (_mm_min_pi16): Likewise.
2038 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2041 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
2042 (DIVWEUO): Likewise.
2044 (DIVDEUO): Likewise.
2045 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
2046 DIVWEUO and DIVDEUO.
2047 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
2048 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
2049 (div_extend): Likewise.
2050 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
2052 (__builtin_divweuo): Likewise.
2053 (__builtin_divdeo): Likewise.
2054 (__builtin_divdeuo): Likewise.
2056 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2059 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
2060 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
2062 PR tree-optimization/82004
2063 * gimple-match-head.c (optimize_pow_to_exp): New function.
2064 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
2065 Don't fold to exp if optimize_pow_to_exp is false.
2067 2018-03-28 Martin Liska <mliska@suse.cz>
2070 * calls.c (initialize_argument_information): Fix trailing space.
2071 * common.opt: Fix typo and provide better explanation for
2072 -fsanitize-coverage option.
2073 * config/i386/i386.opt: Fix typo.
2075 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2076 Martin Liska <mliska@suse.cz>
2079 * gimplify.c (asan_poison_variable): Don't do the check for
2080 gimplify_omp_ctxp here.
2081 (gimplify_decl_expr): Do it here.
2082 (gimplify_target_expr): Likewise.
2084 2018-03-28 Martin Liska <mliska@suse.cz>
2087 * config/i386/i386.c (ix86_function_arg_advance): Do not call
2088 chkp_type_bounds_count if MPX is not enabled.
2090 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
2092 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
2094 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
2097 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
2098 function to create the function decl for complex long double
2099 multiply and divide for -mabi=ieeelongdouble.
2100 (init_float128_ieee): Call it.
2102 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
2105 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
2106 -fcf-protection=branch -mibt.
2107 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
2109 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2112 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
2114 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
2117 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
2118 extern array declarations.
2120 2018-03-27 Richard Biener <rguenther@suse.de>
2123 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
2124 explicit single_use checks.
2126 2018-03-27 Richard Biener <rguenther@suse.de>
2128 PR tree-optimization/85082
2129 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
2132 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2134 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
2135 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
2136 Turn on fasynchronous-unwind-tables and funwind-tables.
2138 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
2141 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
2142 (*bmi_blsr_<mode>_ccz): Ditto.
2144 2018-03-26 Tom de Vries <tom@codesourcery.com>
2146 PR tree-optimization/85063
2147 * omp-general.c (offloading_function_p): New function. Factor out
2149 * omp-offload.c (pass_omp_target_link::gate): ... here.
2150 * omp-general.h (offloading_function_p): Declare.
2151 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
2152 with attribute omp declare target for offloading functions.
2154 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
2156 PR tree-optimization/84005
2157 * tree-data-ref.h (get_base_for_alignment): Declare.
2158 * tree-data-ref.c (get_base_for_alignment_1): New function.
2159 (get_base_for_alignment): Likewise.
2160 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
2161 get_base_for_alignment to find a suitable base object, instead
2162 of always using drb->base_address.
2164 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2167 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
2168 known size by default.
2170 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
2173 * lra-constraints.c (process_alt_operands): Don't match BLKmode
2174 and non BLKmode operands.
2176 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2179 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
2180 Clean up attributes.
2182 2018-03-23 Richard Biener <rguenther@suse.de>
2185 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
2186 we are going to emit early debug for LTO.
2188 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2191 * function.c (match_asm_constraints_1): Don't optimize if input
2192 doesn't satisfy general_operand predicate for output's mode.
2195 * alias.c (write_dependence_p): Don't require for x_canonicalized
2196 non-VOIDmode if x has VOIDmode.
2199 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
2200 just don't try to optimize it rather than assert it never happens.
2202 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2204 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
2205 macro expansions for definition of ST_INTERNAL_<mode> and
2206 LD_INTERNAL_<mode> builtins.
2207 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
2209 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
2211 (altivec_expand_st_builtin): Likewise.
2212 (altivec_expand_builtin): Remove calls to deleted functions.
2213 (rs6000_address_for_altivec): Delete this function.
2214 * config/rs6000/vector.md: Remove expands for
2215 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
2217 2018-03-22 Sudakshina Das <sudi.das@arm.com>
2220 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
2221 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
2222 re-computing once computed.
2223 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
2224 (arm_init_machine_status): Initialize
2225 machine->static_chain_stack_bytes.
2227 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2230 * doc/extend.texi: Add four new prototypes for vec_ld.
2231 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
2232 definitions for more logical presentation.
2233 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
2234 entries for V1TI variants of __builtin_altivec_ld builtin.
2235 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
2236 handling of V1TI variant of LVX icode pattern.
2237 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
2238 (rs6000_gimple_fold_builtin): Likewise.
2239 (altivec_init_builtins): Add code to define
2240 __builtin_altivec_lvx_v1ti function.
2242 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2245 * function.c (match_asm_constraints_1): Don't do the optimization
2246 if input isn't a REG, SUBREG, MEM or constant.
2248 2018-03-22 Tom de Vries <tom@codesourcery.com>
2250 PR tree-optimization/84956
2251 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
2252 bb_has_abnormal_pred.
2254 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2257 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
2258 DECL_INITIAL (decl) to decl at the end.
2259 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
2262 2018-03-21 Joseph Myers <joseph@codesourcery.com>
2264 * doc/extend.texi (__builtin_tgmath): Document when complex
2265 integer types are treated as _Complex _Float64.
2267 2018-03-21 Tom de Vries <tom@codesourcery.com>
2269 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
2271 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2273 PR tree-optimization/84960
2274 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
2275 if it is ENTRY block, move them into single succ of ENTRY in that case.
2277 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
2279 PR tree-optimization/84811
2280 * poly-int.h (poly_span_traits): Remove the T3 parameter and
2281 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
2282 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
2283 (known_subrange_p): Update accordingly. Cast each value involved
2284 in the size comparison, rather than casting the result of the
2287 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2289 PR tree-optimization/84982
2290 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
2291 by flipping the least significant bit rather than all bits from
2292 bitpos to bitpos + bitsize - 1.
2294 2018-03-21 Nathan Sidwell <nathan@acm.org>
2296 * doc/extend.texi (Deprecated Features): Remove mention of
2297 long-deleted deprecations.
2299 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2302 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
2303 * configure: Regenerate.
2305 2018-03-21 Tom de Vries <tom@codesourcery.com>
2307 PR tree-optimization/83126
2308 * tree-parloops.c (num_phis): New function.
2309 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
2311 2018-03-21 Nathan Sidwell <nathan@acm.org>
2313 * doc/extend.texi (Deprecated Features): Update deprecated flags,
2314 mention anon-struct/union members and trailing attributes.
2316 2018-03-21 Bin Cheng <bin.cheng@arm.com>
2318 PR tree-optimization/84969
2319 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
2320 builtin memset partitions if they set different rhs values.
2322 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2324 PR rtl-optimization/84989
2325 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
2326 VEC_DUPLICATE with scalar result mode.
2328 2018-03-21 Martin Liska <mliska@suse.cz>
2331 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
2332 not intended return statement.
2334 2018-03-21 Martin Liska <mliska@suse.cz>
2337 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
2338 (chkp_find_bound_slots_1): Limit number of iterations.
2340 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
2343 * Minor grammar fixes for x86 options.
2345 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2348 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
2349 holding REG_CFA_RESTORE notes, instead turn them into a USE.
2351 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
2354 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
2355 (altivec_lvx_<mode>_1op): Likewise.
2356 (altivec_stvx_<mode>_2op): Likewise.
2357 (altivec_stvx_<mode>_1op): Likewise.
2358 (altivec_lvx_<VM2:mode>): New define_expand.
2359 (altivec_stvx_<VM2:mode>): Likewise.
2360 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
2361 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2362 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
2363 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2364 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
2365 (rs6000_gen_lvx): Likewise.
2366 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
2367 (altivec_expand_stv_builtin): Likewise.
2368 (altivec_expand_builtin): Likewise.
2369 * config/rs6000/vector.md: Likewise.
2371 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2374 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
2377 2018-03-20 Richard Biener <rguenther@suse.de>
2380 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
2381 sign-conversions as zero, fall back to standard scalar_stmt
2384 2018-03-20 Martin Liska <mliska@suse.cz>
2387 * predict.c (rebuild_frequencies): Handle case when we have
2388 PROFILE_ABSENT, but flag_guess_branch_prob is false.
2390 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2393 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
2394 flag_section_anchors.
2395 * varasm.c (use_blocks_for_decl_p): Remove hack for
2396 dw2_force_const_mem.
2399 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
2401 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
2402 be created, use lowpart_subreg of operands[0] rather than operands[0]
2404 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
2405 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
2406 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
2407 and n constraint instead of aarch64_shift_imm_di and Usd.
2408 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
2409 (*aarch64_<optab>_reg_minus<mode>3): ... this.
2411 2018-03-20 Sudakshina Das <sudi.das@arm.com>
2414 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
2415 to favor GPR over NEON registers.
2416 (<shift>di3_neon): Likewise.
2418 2018-03-20 Tom de Vries <tom@codesourcery.com>
2421 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
2422 (nvptx_process_pars): Emit bar.sync asap and alap.
2424 2018-03-20 Tom de Vries <tom@codesourcery.com>
2427 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
2428 seen_label if seen_label is already set.
2430 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2433 * config/i386/i386.c (fold_builtin_cpu): For features above 31
2434 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
2435 Use 1U instead of 1. Formatting fixes.
2438 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
2439 instead of TREE_TYPE (s1) for the return value.
2441 2018-03-19 Jakub Jelinek <jakub@redhat.com>
2443 PR tree-optimization/84946
2444 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
2445 bitsize + bitsize in poly_uint64 rather than poly_int64.
2448 * dwarf2asm.c: Include fold-const.c.
2449 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
2450 of decl rather than decl itself.
2452 PR rtl-optimization/84643
2453 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
2455 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
2458 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
2459 calling assemble_variable.
2461 2018-03-19 Sudakshina Das <sudi.das@arm.com>
2464 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
2465 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
2467 2018-03-19 Jim Wilson <jimw@sifive.com>
2470 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
2471 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
2472 (riscv_first_stack_step): Likewise.
2473 (riscv_option_override): Use STACK_BOUNDARY instead of
2475 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
2477 (BIGGEST_ALIGNMENT): Set to 128.
2478 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
2479 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
2482 2018-03-19 Richard Biener <rguenther@suse.de>
2484 PR tree-optimization/84933
2485 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
2486 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
2488 2018-03-19 Richard Biener <rguenther@suse.de>
2490 PR tree-optimization/84859
2491 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
2492 (cond_if_else_store_replacement): Perform sinking operation on
2493 single-store BBs regardless of MAX_STORES_TO_SINK setting.
2494 Generalize what a BB with a single eligible store is.
2496 2018-03-19 Richard Biener <rguenther@suse.de>
2498 PR tree-optimization/84929
2499 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
2500 chrec_is_positive against non-chrec arg.
2502 2018-03-19 Tamar Christina <tamar.christina@arm.com>
2505 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
2507 2018-03-18 Martin Liska <mliska@suse.cz>
2509 PR rtl-optimization/84635
2510 * regrename.c (build_def_use): Use matches_mode only when
2513 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
2515 PR tree-optimization/84913
2516 * tree-vect-loop.c (vectorizable_reduction): Don't try to
2517 vectorize chains of COND_EXPRs.
2519 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2521 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
2523 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2525 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
2527 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2529 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
2531 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2532 Kito Cheng <kito.cheng@gmail.com>
2534 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
2535 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
2536 (nds32_adjust_reg_alloc_order): New function.
2537 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
2539 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
2541 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
2542 nds32_print_operand, nds32_print_operand_address): Use
2543 HOST_WIDE_INT_PRINT_DEC instead.
2545 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2547 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
2549 2018-03-17 Jakub Jelinek <jakub@redhat.com>
2552 * config/i386/i386.c (initial_ix86_tune_features,
2553 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
2555 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
2556 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
2557 rather than 1u << ix86_tune. Formatting fix.
2558 (ix86_option_override_internal): Change ix86_arch_mask from
2559 unsigned int to unsigned HOST_WIDE_INT, initialize to
2560 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
2561 (ix86_function_specific_restore): Likewise.
2563 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2566 * postreload.c (reload_combine_recognize_pattern): Perform
2567 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
2568 truncate_int_for_mode the result for the destination's mode.
2571 * hsa-gen.c (mem_type_for_type): Fix comment typo.
2572 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
2574 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2577 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
2580 * lra-assigns.c (lra_split_hard_reg_for): Don't use
2581 regno_allocno_class_array and sorted_pseudos.
2582 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
2583 insns where regno is used.
2585 2018-03-16 Martin Liska <mliska@suse.cz>
2588 * multiple_target.c (create_dispatcher_calls): Redirect
2589 reference in the symbol table.
2591 2018-03-16 Martin Liska <mliska@suse.cz>
2594 * multiple_target.c (create_dispatcher_calls): Redirect also
2597 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2601 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
2602 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
2603 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
2605 2018-03-16 Julia Koval <julia.koval@intel.com>
2607 * doc/invoke.texi (Skylake Server): Add CLWB.
2608 Cannonlake): Remove CLWB.
2610 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2612 PR tree-optimization/84841
2613 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
2615 (FLOAT_ONE_CONST_TYPE): Define.
2616 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
2617 (sort_by_operand_rank): Put entries with higher constant_type last
2618 rather than first to match comments.
2620 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
2622 * config/nios2/nios2.md (movsi_internal): Fix thinko in
2625 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2628 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
2629 check and use address of target always.
2631 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
2634 * config/i386/i386.c (indirect_thunk_needed): Update comments.
2635 (indirect_thunk_bnd_needed): Likewise.
2636 (indirect_thunks_used): Likewise.
2637 (indirect_thunks_bnd_used): Likewise.
2638 (indirect_return_needed): New.
2639 (indirect_return_bnd_needed): Likewise.
2640 (output_indirect_thunk_function): Add a bool argument for
2642 (output_indirect_thunk_function): Don't generate alias for
2643 function return thunk.
2644 (ix86_code_end): Call output_indirect_thunk_function to generate
2645 function return thunks.
2646 (ix86_output_function_return): Set indirect_return_bnd_needed
2647 and indirect_return_needed instead of indirect_thunk_bnd_needed
2648 and indirect_thunk_needed.
2650 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
2652 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
2653 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
2654 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
2656 2018-03-15 David Malcolm <dmalcolm@redhat.com>
2657 Paul Hua <paul.hua.gm@gmail.com>
2660 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
2662 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
2664 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
2665 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
2668 2018-03-15 Tamar Christina <tamar.christina@arm.com>
2671 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
2672 instead of GET_MODE_SIZE when comparing Units.
2674 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2677 * varasm.c (hash_section): Return an unchangeble hash value
2678 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
2679 Return !aarch64_can_use_per_function_literal_pools_p ().
2681 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2684 * optabs.c (emit_conditional_move): Pass address of cmode's copy
2685 rather than address of cmode as last argument to prepare_cmp_insn.
2687 2018-03-15 Julia Koval <julia.koval@intel.com>
2689 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
2690 F_AVX512VNNI, F_AVX512BITALG): New.
2692 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
2695 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
2696 insn for floating-point loads and stores.
2698 2018-03-14 Carl Love <cel@us.ibm.com>
2700 * config/rs6000/rs6000-c.c: Add macro definitions for
2701 ALTIVEC_BUILTIN_VEC_PERMXOR.
2702 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
2703 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
2704 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
2705 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
2707 * config/doc/extend.texi: Add prototypes for vec_permxor.
2709 2018-03-14 David Malcolm <dmalcolm@redhat.com>
2712 * diagnostic-show-locus.c (class layout_point): Convert m_line
2713 from int to linenum_type.
2714 (line_span::comparator): Use linenum "compare" function when
2715 comparing line numbers.
2716 (test_line_span): New function.
2717 (layout_range::contains_point): Convert param "row" from int to
2719 (layout_range::intersects_line_p): Likewise.
2720 (layout::will_show_line_p): Likewise.
2721 (layout::print_source_line): Likewise.
2722 (layout::should_print_annotation_line_p): Likewise.
2723 (layout::print_annotation_line): Likewise.
2724 (layout::print_leading_fixits): Likewise.
2725 (layout::annotation_line_showed_range_p): Likewise.
2726 (struct line_corrections): Likewise for field m_row.
2727 (line_corrections::line_corrections): Likewise for param "row".
2728 (layout::print_trailing_fixits): Likewise.
2729 (layout::get_state_at_point): Likewise.
2730 (layout::get_x_bound_for_row): Likewise.
2731 (layout::print_line): Likewise.
2732 (diagnostic_show_locus): Likewise for locals "last_line" and
2734 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
2735 * input.c (selftest::test_linenum_comparisons): New function.
2736 (selftest::input_c_tests): Call it.
2737 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
2738 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
2739 * selftest.h (ASSERT_GT): New macro.
2740 (ASSERT_GT_AT): New macro.
2741 (ASSERT_LT): New macro.
2742 (ASSERT_LT_AT): New macro.
2744 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
2746 PR rtl-optimization/84780
2747 * combine.c (distribute_links): Don't make a link based on pc_rtx.
2749 2018-03-14 Martin Liska <mliska@suse.cz>
2751 * tree.c (record_node_allocation_statistics): Use
2752 get_stats_node_kind.
2753 (get_stats_node_kind): New function extracted from
2754 record_node_allocation_statistics.
2755 (free_node): Use get_stats_node_kind.
2757 2018-03-14 Richard Biener <rguenther@suse.de>
2759 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
2760 that the value-set of ANTIC_IN doesn't grow.
2763 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2765 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2766 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2768 2018-03-14 Julia Koval <julia.koval@intel.com>
2770 * config.gcc (icelake-client, icelake-server): New.
2772 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
2773 (initial_ix86_arch_features): Ditto.
2774 (PTA_SKYLAKE): Add SGX.
2775 (PTA_ICELAKE): Remove.
2776 (PTA_ICELAKE_CLIENT): New.
2777 (PTA_ICELAKE_SERVER): New.
2778 (ix86_option_override_internal): Split up icelake on icelake client and
2780 (get_builtin_code_for_version): Ditto.
2781 (fold_builtin_cpu): Ditto.
2782 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
2783 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
2784 * config/i386/i386.h (processor_type): Ditto.
2785 * doc/invoke.texi: Ditto.
2787 2018-03-14 Jakub Jelinek <jakub@redhat.com>
2790 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
2791 INTEGER_CST offset, add it together with bitpos / 8 and
2792 sign extend based on POINTER_SIZE.
2796 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
2799 * config/i386/constraints.md (Yc): New register constraint.
2800 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
2801 Use Yc constraint for alternative 2 of operand 0. Remove
2802 preferred_for_speed attribute.
2804 2018-03-14 Richard Biener <rguenther@suse.de>
2806 PR tree-optimization/84830
2807 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
2808 with the old one to avoid oscillations.
2810 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
2813 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
2815 (assign_by_spills): Return a flag of reload assignment failure.
2816 Do not process the reload assignment failures. Do not spill other
2817 reload pseudos if they has the same reg class. Update n if
2819 (lra_assign): Add a return arg. Set up from the result of
2820 assign_by_spills call.
2821 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2822 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2823 usage_insns if it is not NULL.
2824 (spill_hard_reg_in_range): New function.
2825 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2826 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2827 function prototypes.
2828 (lra_assign): Change prototype.
2829 * lra.c (lra): Add code to deal with fails by splitting hard reg
2832 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
2834 * config/riscv/riscv.opt (mrelax): New option.
2835 * config/riscv/riscv.c (riscv_file_start): Emit ".option
2836 "norelax" when riscv_mrelax is disabled.
2837 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
2839 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2842 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
2843 reassociation for int modes.
2845 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2847 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
2848 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
2850 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
2851 * config/aarch64/aarch64-sve.md
2852 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
2853 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
2854 (*extend<mode><Vwide>2): Rename to...
2855 (aarch64_sve_extend<mode><Vwide>2): ...this.
2856 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
2857 renaming the old pattern to...
2858 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
2860 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
2861 define_expand, renaming the old pattern to...
2862 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
2863 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
2864 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
2865 account when deciding which SVE instruction the optab should use.
2866 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
2868 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2870 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
2871 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
2872 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
2873 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
2874 (tlsdesc_small_<mode>): Turn a define_expand and use
2875 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
2876 (tlsdesc_small_advsimd_<mode>): ...this.
2877 (tlsdesc_small_sve_<mode>): New pattern.
2879 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2881 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
2882 (UNSPEC_UMUL_HIGHPART): New constants.
2883 (MUL_HIGHPART): New int iteraor.
2884 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
2885 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
2887 (*<su>mul<mode>3_highpart): New define_insn.
2889 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2892 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2895 2018-03-13 Martin Liska <mliska@suse.cz>
2898 * (sem_item_optimizer::sem_item_optimizer): Initialize new
2900 (sem_item_optimizer::~sem_item_optimizer): Release it.
2901 (sem_item_optimizer::merge_classes): Register variable aliases.
2902 (sem_item_optimizer::fixup_pt_set): New function.
2903 (sem_item_optimizer::fixup_points_to_sets): Likewise.
2904 * ipa-icf.h: Declare new variables and functions.
2906 2018-03-13 Jakub Jelinek <jakub@redhat.com>
2909 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
2910 integer_pow2p@2 and test integer_pow2p in condition.
2911 (A < 0 ? C : 0): Similarly for @1.
2914 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
2915 characters starting at p contain '\0' character, don't look beyond
2919 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
2920 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
2923 * reg-stack.c (change_stack): Change update_end var from int to
2924 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
2925 also call set_block_for_insn on the newly added insns and rescan.
2928 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
2929 on the last operand.
2932 * tree.c (stabilize_reference_1): Return save_expr (e) for
2933 STATEMENT_LIST even if it doesn't have side-effects.
2935 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
2937 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
2939 2018-03-12 Renlin Li <renlin.li@arm.com>
2941 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
2942 aarch64_output_scalar_simd_mov_immediate.
2944 2018-03-12 Martin Sebor <msebor@redhat.com>
2946 PR tree-optimization/83456
2947 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
2948 for perfectly overlapping calls to memcpy.
2949 (gimple_fold_builtin_memory_chk): Same.
2950 (gimple_fold_builtin_strcpy): Handle no-warning.
2951 (gimple_fold_builtin_stxcpy_chk): Same.
2952 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
2954 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2956 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
2957 parameter. Use it for SFmode.
2958 (rs6000_function_arg_advance_1): Adjust.
2959 (rs6000_function_arg): Adjust.
2960 (rs6000_gimplify_va_arg): Pass false for that new parameter.
2962 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2964 PR rtl-optimization/84169
2965 PR rtl-optimization/84780
2966 * combine.c (can_combine_p): Check for a 2-insn combination whether
2967 the destination register is used between the two insns, too.
2969 2018-03-12 Richard Biener <rguenther@suse.de>
2971 PR tree-optimization/84803
2972 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
2973 for refs DR analysis didn't process.
2975 2018-03-12 Richard Biener <rguenther@suse.de>
2977 PR tree-optimization/84777
2978 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
2979 force-vectorize loops ignore whether we are optimizing for size.
2981 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
2983 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
2984 (TARGET_MD_ASM_ADJUST): Define.
2986 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
2987 Kito Cheng <kito.cheng@gmail.com>
2988 Chung-Ju Wu <jasonwucj@gmail.com>
2990 * config/nds32/nds32.c (nds32_compute_stack_frame,
2991 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
2992 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
2993 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
2994 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
2995 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
2996 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
2997 * config/nds32/nds32.md (prologue, epilogue): Use macro
2998 NDS32_V3PUSH_AVAILABLE_P to do checking.
3000 2018-03-11 Jakub Jelinek <jakub@redhat.com>
3003 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
3004 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
3005 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
3006 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
3007 addition of most attributes on !orig_type_die or the attribute not
3008 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
3010 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3011 Chung-Ju Wu <jasonwucj@gmail.com>
3013 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
3015 * config/nds32/nds32.opt (mvh): New option.
3017 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3018 Chung-Ju Wu <jasonwucj@gmail.com>
3020 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
3022 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
3023 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
3026 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3027 Chung-Ju Wu <jasonwucj@gmail.com>
3029 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
3031 * config/nds32/nds32-multiple.md (strlensi): New pattern.
3032 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
3034 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
3035 Kito Cheng <kito.cheng@gmail.com>
3036 Chung-Ju Wu <jasonwucj@gmail.com>
3038 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
3039 UNSPEC_FFMISM and UNSPEC_FLMISM.
3040 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
3041 for ffb, ffmism and flmism.
3042 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
3043 (unspec_ffmism): Ditto.
3044 (unspec_flmism): Ditto.
3045 (nds32_expand_builtin_impl): Check if string extension is available.
3046 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
3047 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
3049 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
3052 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3055 * lra-assigns.c (assign_by_spills): Return a flag of reload
3056 assignment failure. Do not process the reload assignment
3057 failures. Do not spill other reload pseudos if they has the same
3059 (lra_assign): Add a return arg. Set up from the result of
3060 assign_by_spills call.
3061 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3062 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3063 usage_insns if it is not NULL.
3064 (spill_hard_reg_in_range): New function.
3065 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3066 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3067 function prototypes.
3068 (lra_assign): Change prototype.
3069 * lra.c (lra): Add code to deal with fails by splitting hard reg
3072 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
3075 * config/i386/i386.opt: Replace Enforcment with Enforcement.
3077 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
3080 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
3081 (dw_val_node): Add val_symbolic_view.
3082 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
3083 (symview_upper_bound): New.
3084 (new_line_info_table): Initialize symviews_since_reset.
3085 (dwarf2out_source_line): Count symviews_since_reset and set
3086 symview_upper_bound.
3087 (dw_val_equal_p): Handle symview.
3088 (add_AT_symview): New.
3089 (print_dw_val): Handle symview.
3090 (attr_checksum, attr_checksum_ordered): Likewise.
3091 (same_dw_val_p, size_of_die): Likewise.
3092 (value_format, output_die): Likewise.
3093 (add_high_low_attributes): Use add_AT_symview for entry_view.
3094 (dwarf2out_finish): Reset symview_upper_bound, clear
3097 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
3100 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
3101 Add strict argument and use it.
3102 (rs6000_split_multireg_move): Update for new strict argument.
3103 (mem_operand_gpr): Disallow all non-offsettable addresses.
3104 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
3106 2018-03-09 Jakub Jelinek <jakub@redhat.com>
3109 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
3110 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
3111 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
3114 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
3115 decl, use remap_type if we want to use the type.
3117 2018-03-09 Martin Sebor <msebor@redhat.com>
3119 PR tree-optimization/84526
3120 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3122 (builtin_access::generic_overlap): Be prepared to handle non-array
3125 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
3127 PR rtl-optimization/84682
3128 * lra-constraints.c (process_address_1): Check is_address flag
3129 for address constraints.
3130 (process_alt_operands): Likewise.
3131 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
3132 preprocess_constraints.
3133 * recog.h (preprocess_constraints): Add oploc parameter.
3135 * recog.c (preprocess_constraints): Test address_operand for
3136 CT_ADDRESS constraints.
3138 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3141 * lra-assigns.c (assign_by_spills): Return a flag of reload
3142 assignment failure. Do not process the reload assignment
3143 failures. Do not spill other reload pseudos if they has the same
3145 (lra_assign): Add a return arg. Set up from the result of
3146 assign_by_spills call.
3147 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3148 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3149 usage_insns if it is not NULL.
3150 (spill_hard_reg_in_range): New function.
3151 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3152 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3153 function prototypes.
3154 (lra_assign): Change prototype.
3155 * lra.c (lra): Add code to deal with fails by splitting hard reg
3158 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3161 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
3162 Accept complain bool parameter. Only emit errors if it is true.
3163 (arm_parse_cpu_option_name): Likewise.
3164 (arm_target_thumb_only): Adjust callers of the above.
3165 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
3166 prototype to take a default true bool parameter.
3167 (arm_parse_arch_option_name): Likewise.
3169 2018-03-09 David Malcolm <dmalcolm@redhat.com>
3170 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
3174 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
3175 * configure: Regenerate.
3176 * configure.ac ("linker --version-script option"): New.
3177 ("linker soname option"): New.
3179 2018-03-09 Richard Biener <rguenther@suse.de>
3181 PR tree-optimization/84775
3182 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
3183 immediate uses of predicate stmts and mark them modified.
3186 PR tree-optimization/84178
3187 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3189 (version_loop_for_if_conversion): Delay update_ssa call.
3190 (tree_if_conversion): Delay update_ssa until after predicate
3193 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
3196 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
3197 when the function accesses prior frames.
3199 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3202 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
3203 gen_llsym, otherwise call maybe_gen_llsym.
3206 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
3207 has ',' character inside of it.
3209 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3212 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
3213 as clobbering CC_REGNUM.
3215 2018-03-08 Richard Biener <rguenther@suse.de>
3218 * tree-scalar-evolution.c: Include tree-into-ssa.h.
3219 (follow_copies_to_constant): Do not follow SSA names registered
3222 2018-03-08 Richard Biener <rguenther@suse.de>
3224 PR tree-optimization/84178
3225 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3227 (version_loop_for_if_conversion): Delay update_ssa call.
3228 (tree_if_conversion): Delay update_ssa until after predicate
3231 2018-03-08 David Malcolm <dmalcolm@redhat.com>
3233 PR tree-optimization/84178
3234 * tree-if-conv.c (release_bb_predicate): Remove the
3235 the assertion that the stmts have NULL use_ops.
3236 Discard the statements, asserting that they haven't
3237 yet been added to a BB.
3239 2018-03-08 Richard Biener <rguenther@suse.de>
3241 PR tree-optimization/84746
3242 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
3243 (phi_translate): Pass in destination ANTIC_OUT set.
3244 (phi_translate_1): Likewise. For a simplified result lookup
3245 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
3246 (phi_translate_set): Adjust.
3247 (do_pre_regular_insertion): Likewise.
3248 (do_pre_partial_partial_insertion): Likewise.
3250 2018-03-08 Martin Liska <mliska@suse.cz>
3252 PR gcov-profile/84735
3253 * doc/gcov.texi: Document usage of profile files.
3254 * gcov-io.h: Document changes in the format.
3256 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
3260 * dwarf2out.c (struct dw_line_info_table): Update comments for
3262 (FORCE_RESET_NEXT_VIEW): New.
3263 (FORCE_RESETTING_VIEW_P): New.
3264 (RESETTING_VIEW_P): Check for -1 too.
3265 (ZERO_VIEW_P): Likewise.
3266 (new_line_info_table): Force-reset next view.
3267 (dwarf2out_begin_function): Likewise.
3268 (dwarf2out_source_line): Simplify zero_view_p initialization.
3269 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
3270 view directly. Omit view when omitting .loc at line 0.
3272 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3274 PR tree-optimization/84740
3275 * tree-switch-conversion.c (process_switch): Call build_constructors
3276 only if info.phi_count is non-zero.
3278 PR tree-optimization/84739
3279 * tree-tailcall.c (find_tail_calls): Check call arguments against
3280 DECL_ARGUMENTS (current_function_decl) rather than
3281 DECL_ARGUMENTS (func) when checking for tail recursion.
3283 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3285 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
3286 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
3287 Volker Reichelt's entry and add entries for people that perform
3288 GCC fuzzy testing and report numerous bugs.
3290 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
3293 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
3294 readonly data in sdata, if that is disabled.
3295 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
3296 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3297 -mreadonly-in-sdata option.
3299 2018-03-07 Martin Sebor <msebor@redhat.com>
3301 PR tree-optimization/84468
3302 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
3303 basic block when looking for nul assignment.
3305 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
3308 * except.h (output_function_exception_table): Adjust prototype.
3309 * except.c (output_function_exception_table): Remove FNNAME parameter
3310 and add SECTION parameter. Ouput one part of the table at a time.
3311 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
3312 the first part of the exception table and emit unwind directives.
3313 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
3314 (i386_pe_seh_cold_init): Likewise.
3315 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
3316 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
3317 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
3318 (ix86_output_call_insn): Emit a nop in one more case for SEH.
3319 * config/i386/winnt.c: Include except.h.
3320 (struct seh_frame_state): Add reg_offset, after_prologue and
3321 in_cold_section fields.
3322 (i386_pe_seh_end_prologue): Set seh->after_prologue.
3323 (i386_pe_seh_cold_init): New function.
3324 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
3325 to seh->in_cold_section.
3326 (seh_emit_push): Record the offset of the push.
3327 (seh_emit_save): Record the offet of the save.
3328 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
3329 Test seh->after_prologue to disregard the epilogue.
3330 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
3331 (i386_pe_end_cold_function): New function.
3333 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3336 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
3337 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
3340 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
3342 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
3343 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
3346 * multiple_target.c: Include tree-inline.h and intl.h.
3347 (expand_target_clones): Diagnose and fail if node->definition and
3348 !tree_versionable_function_p (node->decl).
3350 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
3352 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
3354 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
3355 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
3356 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
3358 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3361 * combine.c (try_combine): Use reg_or_subregno instead of handling
3362 just paradoxical SUBREGs and REGs.
3364 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
3366 * config/arc/arc.c (arc_finalize_pic): Remove function.
3367 (arc_must_save_register): We use single base PIC register, remove
3368 checks to save/restore the PIC register.
3369 (arc_expand_prologue): Likewise.
3370 * config/arc/arc-protos.h (arc_set_default_type_attributes):
3372 (arc_verify_short): Likewise.
3373 (arc_attr_type): Likewise.
3374 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
3375 (walk_stores): Likewise.
3376 (arc_address_cost): Make it static.
3377 (arc_verify_short): Likewise.
3378 (branch_dest): Likewise.
3379 (arc_attr_type): Likewise.
3380 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
3381 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
3382 (arc_final_prescan_insn): Remove inserting the nops due to
3383 hardware hazards. It is done in reorg step.
3384 (insn_length_variant_t): Remove.
3385 (insn_length_parameters_t): Likewise.
3386 (arc_insn_length_parameters): Likewise.
3387 (arc_get_insn_variants): Likewise.
3388 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
3390 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3393 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
3396 PR tree-optimization/84687
3397 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
3399 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
3401 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3403 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
3404 Rename to ppc_speculation_barrier.
3405 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
3406 __builtin_ppc_speculation_barrier.
3408 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3411 * combine.c (combine_simplify_rtx): Don't try to simplify if
3412 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
3415 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
3417 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
3418 to 32 bytes when compiling for POWER9.
3420 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3423 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
3424 regparm >= 3 with no arg reg available also for calls with
3425 flag_force_indirect_call. Pass decl to ix86_function_regparm.
3428 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
3430 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
3432 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
3435 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
3437 2018-03-05 Richard Biener <rguenther@suse.de>
3439 PR tree-optimization/84486
3440 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
3441 When inserting a __builtin_assume_aligned call set the LHS
3442 SSA name alignment info accordingly.
3444 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
3446 PR tree-optimization/84114
3447 * config/aarch64/aarch64.c (aarch64_reassociation_width)
3448 Avoid reassociation of FLOAT_MODE addition.
3450 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
3452 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
3453 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
3454 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
3455 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
3456 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
3457 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
3458 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
3460 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
3461 __builtin_ia32_wbinvd): New builtins.
3462 (SPECIAL_ARGS2): New.
3463 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
3464 (SPECIAL_ARGS2): New.
3465 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
3466 (ix86_valid_target_attribute_inner_p): Ditto.
3467 (ix86_init_mmx_sse_builtins): Add special_args2.
3468 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
3469 TARGET_WBNOINVD_P): New.
3470 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
3471 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
3472 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
3473 * config/i386/immintrin.h (_wbinvd): New intrinsic.
3474 * config/i386/pconfigintrin.h: New file.
3475 * config/i386/wbnoinvdintrin.h: Ditto.
3476 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
3477 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
3479 2018-03-05 Richard Biener <rguenther@suse.de>
3481 PR tree-optimization/84670
3482 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3484 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3485 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3486 (compute_antic_aux): Only assert the number of values in ANTIC_IN
3487 doesn't grow if all successors (recursively) were visited at least
3490 2018-03-05 Richard Biener <rguenther@suse.de>
3492 PR tree-optimization/84650
3493 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
3494 if executed in the loop pipeline.
3496 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
3498 * doc/configfiles.texi (Configuration Files): Move info about
3499 conditionalizing $target-protos.h to...
3500 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
3501 differs from $target-protos.h.
3503 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
3504 Chung-Ju Wu <jasonwucj@gmail.com>
3506 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
3507 * config/nds32/nds32-multiple.md (setmemsi): Define.
3508 * config/nds32/nds32-memory-manipulation.c
3509 (nds32_gen_dup_4_byte_to_word_value): New.
3510 (emit_setmem_word_loop): New.
3511 (emit_setmem_byte_loop): New.
3512 (nds32_expand_setmem_loop): New.
3513 (nds32_expand_setmem_loop_v3m): New.
3514 (nds32_expand_setmem_unroll): New.
3515 (nds32_expand_setmem): New.
3517 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3518 Chung-Ju Wu <jasonwucj@gmail.com>
3520 * config/nds32/nds32-memory-manipulation.c
3521 (nds32_emit_load_store): New.
3522 (nds32_emit_post_inc_load_store): New.
3523 (nds32_emit_mem_move): New.
3524 (nds32_emit_mem_move_block): New.
3525 (nds32_expand_movmemsi_loop_unknown_size): New.
3526 (nds32_expand_movmemsi_loop_known_size): New.
3527 (nds32_expand_movmemsi_loop): New.
3528 (nds32_expand_movmemsi_unroll): New.
3529 (nds32_expand_movmemqi): Rename ...
3530 (nds32_expand_movmemsi): ... to this.
3531 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
3532 (movmemsi): ... to this.
3533 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
3534 (nds32_expand_movmemsi): ... to this.
3536 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3537 Monk Chiang <sh.chiang04@gmail.com>
3538 Chung-Ju Wu <jasonwucj@gmail.com>
3540 * config/nds32/nds32-protos.h
3541 (nds32_expand_load_multiple): New arguments.
3542 (nds32_expand_store_multiple): Ditto.
3543 (nds32_valid_multiple_load_store): Rename ...
3544 (nds32_valid_multiple_load_store_p): ... to this.
3545 * config/nds32/nds32-memory-manipulation.c
3546 (nds32_expand_load_multiple): Refine implementation.
3547 (nds32_expand_store_multiple): Ditto.
3548 * config/nds32/nds32-multiple.md
3549 (load_multiple): Update nds32_expand_load_multiple interface.
3550 (store_multiple): Update nds32_expand_store_multiple interface.
3551 * config/nds32/nds32-predicates.c
3552 (nds32_valid_multiple_load_store): Rename ...
3553 (nds32_valid_multiple_load_store_p): ... to this and refine
3555 * config/nds32/predicates.md
3556 (nds32_load_multiple_and_update_address_operation): New predicate.
3557 (nds32_store_multiple_and_update_address_operation): New predicate.
3559 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3560 Chung-Ju Wu <jasonwucj@gmail.com>
3562 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
3563 (combo): New attribute.
3564 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
3566 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
3568 * config/nds32/nds32.opt: Change -mcmodel= default value.
3570 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
3571 Monk Chiang <sh.chiang04@gmail.com>
3572 Chung-Ju Wu <jasonwucj@gmail.com>
3574 * config/nds32/constants.md (unspec_element): New enum.
3575 * config/nds32/constraints.md (Umw): New constraint.
3576 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
3577 * config/nds32/nds32-intrinsic.md: Likewise.
3578 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
3579 (nds32_valid_smw_lwm_base_p): New.
3580 (nds32_output_smw_single_word): New.
3581 (nds32_output_lmw_single_word): New.
3582 (nds32_expand_unaligned_load): New.
3583 (nds32_expand_unaligned_store): New.
3584 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
3585 (nds32_output_smw_single_word): Declare.
3586 (nds32_output_lmw_single_word): Declare.
3587 (nds32_expand_unaligned_load): Declare.
3588 (nds32_expand_unaligned_store): Declare.
3589 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
3590 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
3591 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
3592 NDS32_BUILTIN_UASTORE_DW.
3593 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
3596 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
3597 Kito Cheng <kito.cheng@gmail.com>
3598 Chung-Ju Wu <jasonwucj@gmail.com>
3600 * config/nds32/nds32-intrinsic.c
3601 (nds32_expand_builtin_null_ftype_reg): Delete.
3602 (nds32_expand_builtin_reg_ftype_imm): Ditto.
3603 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
3604 (nds32_read_argument): New.
3605 (nds32_legitimize_target): Ditto.
3606 (nds32_legitimize_argument): Ditto.
3607 (nds32_check_constant_argument): Ditto.
3608 (nds32_expand_unop_builtin): Ditto.
3609 (nds32_expand_unopimm_builtin): Ditto.
3610 (nds32_expand_binop_builtin): Ditto.
3611 (nds32_builtin_decl_impl): Ditto.
3612 (builtin_description): Ditto.
3613 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
3614 (nds32_init_builtins_impl): Ditto.
3615 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
3616 (nds32_builtin_decl): New.
3617 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
3618 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
3620 2018-03-02 Jeff Law <law@redhat.com>
3622 * reorg.c (stop_search_p): Handle DEBUG_INSN.
3623 (redundant_insn, fill_simple_delay_slots): Likewise.
3624 (fill_slots_from_thread): Likewise.
3625 * resource.c (mark_referenced_resources): Likewise.
3626 (mark_set_resources, find_dead_or_set_registers): Likewise.
3628 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3630 * substring-locations.h (format_warning_va): Formatting fix for
3632 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
3634 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
3635 * substring-locations.c: Include intl.h.
3636 (format_warning_va): Turned into small wrapper around
3637 format_warning_n_va, renamed to ...
3638 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
3639 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
3641 (format_warning_at_substring_n): New function.
3642 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
3643 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
3644 format_warning_at_substring with just a shorter name instead of
3645 const function pointer.
3646 (fmtwarn_n): New function.
3647 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
3648 appropriate, get rid of all the fmtstr temporaries, move conditionals
3649 with G_() wrapped string literals directly into fmtwarn arguments,
3650 cast dir.len to (int), formatting fixes.
3652 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
3654 * doc/invoke.texi: Remove "Cilk Plus" references.
3656 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3657 Richard Biener <rguenther@suse.de>
3660 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
3661 for error or warning attributes if CALL_FROM_THUNK_P is set.
3664 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3667 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
3668 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
3671 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
3672 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
3674 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
3675 instead of -1U in last predictors element's probability member.
3677 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3680 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3681 arguments if they are comparable.
3683 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3685 PR tree-optimization/84634
3686 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
3687 masks and masked_loop_p with a single loop_masks, making sure it's
3688 null for bb vectorization.
3690 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3692 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3693 (vect_analyze_data_ref_access): Use loop->safe_len rather than
3694 loop->force_vectorize to check whether there is no alias.
3696 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3699 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
3701 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
3703 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
3704 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
3705 instead of a loop around prev_real_insn.
3706 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
3710 * config/i386/i386.c (ix86_print_operand): Use conditional
3711 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
3714 2018-03-02 Richard Biener <rguenther@suse.de>
3716 PR tree-optimization/84427
3717 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
3718 (bitmap_set_subtract_values): Rewrite to handle multiple
3721 (prune_clobbered_mems): Likewise.
3722 (phi_translate): Take edge instead of pred/phiblock.
3723 (phi_translate_1): Likewise.
3724 (phi_translate_set): Likewise. Insert all translated
3725 exprs for a value into the set, keeping possibly multiple
3726 expressions per value.
3727 (compute_antic_aux): Adjust for phi_translate changes.
3728 When intersecting union the expressions and prune those
3729 not in the final value set, keeping possibly multiple
3730 expressions per value. Do not use value-insertion
3731 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
3732 all expressions. Add verification that the value-sets
3733 only shrink during iteration.
3734 (compute_partial_antic_aux): Adjust for the phi_translate changes.
3735 (do_pre_regular_insertion): Likewise.
3736 (do_pre_partial_partial_insertion): Likewise.
3738 2018-03-02 Richard Biener <rguenther@suse.de>
3741 * config/darwin.c (saved_debug_info_level): New static global.
3742 (darwin_asm_lto_start): Disable debug info generation for LTO out.
3743 (darwin_asm_lto_end): Restore debug info generation settings.
3745 2018-03-01 Martin Liska <mliska@suse.cz>
3748 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
3751 2018-03-01 Richard Biener <rguenther@suse.de>
3754 * dwarf2out.c (gen_variable_die): Properly handle late VLA
3755 type annotation with LTO when debug was disabled at compile-time.
3757 2018-03-01 Matthew Fortune <mfortune@gmail.com>
3759 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
3761 (mips_final_postscan_insn): Likewise.
3763 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
3765 PR rtl-optimization/84528
3766 * alias.c (init_alias_target): Add commentary.
3767 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
3768 a unique base value if the frame pointer is not eliminated
3769 to the stack pointer.
3771 2018-03-01 Tom de Vries <tom@codesourcery.com>
3773 PR rtl-optimization/83327
3774 * lra-int.h (hard_regs_spilled_into): Declare.
3775 * lra.c (hard_regs_spilled_into): Define.
3776 (init_reg_info): Init hard_regs_spilled_into.
3777 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
3778 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
3779 (process_bb_lives): Handle hard_regs_spilled_into.
3780 (lra_create_live_ranges_1): Before doing liveness propagation, clear
3781 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
3783 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
3785 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
3786 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
3787 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
3788 * config/rs6000/aix72.h: New file.
3790 2018-02-28 Jakub Jelinek <jakub@redhat.com>
3792 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
3793 instead of warning_at with conditional singular and plural messages
3797 * stor-layout.c (update_alignment_for_field): For
3798 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
3799 && !DECL_PACKED (field), do the alignment update, just use
3800 only desired_align instead of MAX (type_align, desired_align)
3802 (place_field): Don't do known_align < desired_align handling
3803 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
3804 is non-NULL, instead do it after rli->prev_field handling and
3805 only if not within a bitfield word. For DECL_PACKED (field)
3806 use type_align of BITS_PER_UNIT.
3808 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
3810 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
3811 superfluous parentheses and trailing spaces.
3813 2018-02-28 Richard Biener <rguenther@suse.de>
3815 PR tree-optimization/84584
3816 * graphite-scop-detection.c (scop_detection::add_scop): Discard
3817 SCoPs with fake exit edge.
3819 2018-02-28 Martin Liska <mliska@suse.cz>
3822 * timevar.c (timer::print): Fix format to properly print 100%
3825 2018-02-28 Richard Biener <rguenther@suse.de>
3828 * genmatch.c (capture_info::walk_match): Do not mark
3829 captured expressions without operands as expr_p given
3830 they act more like predicates and should be subject to
3831 "lost tail" side-effect preserving.
3833 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
3835 PR rtl-optimization/81611
3836 * auto-inc-dec.c (attempt_change): Move dead note from
3837 mem_insn if it's the next use of regno
3838 (find_address): Take address use of reg holding
3839 non-incremented value. Add parm to limit search to the named
3841 (merge_in_block): Attempt to use a mem insn that is the next
3842 use of the original regno.
3844 2018-02-27 Martin Sebor <msebor@redhat.com>
3847 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
3848 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
3850 2018-02-27 Martin Sebor <msebor@redhat.com>
3852 PR translation/84207
3853 * diagnostic-core.h (warning_n, error_n, inform_n): Change
3854 n argument to unsigned HOST_WIDE_INT.
3855 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
3856 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
3857 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
3858 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
3860 2018-02-27 Richard Biener <rguenther@suse.de>
3862 PR tree-optimization/84512
3863 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
3864 Do not use the estimate returned from record_stmt_cost for
3865 the scalar iteration cost but sum properly using add_stmt_cost.
3867 2018-02-27 Richard Biener <rguenther@suse.de>
3869 PR tree-optimization/84466
3870 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3871 Adjust last change to less strictly validate use operands.
3873 2018-02-27 Martin Liska <mliska@suse.cz>
3875 PR gcov-profile/84548
3876 * gcov.c (process_file): Allow partial overlap and consider it
3877 also as group functions.
3878 (output_lines): Properly calculate range of lines for a group.
3880 2018-02-27 Martin Liska <mliska@suse.cz>
3882 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
3883 'ggc' suffixes. Change first column width.
3884 (timer::print): Fix formatting of the column.
3886 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
3888 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
3889 preserve inline entry blocks for the sake of debug inline
3890 entry point markers alone.
3891 (remove_unused_locals): Suggest in comments a better place to
3892 force the preservation of inline entry blocks that are
3893 otherwise unused, but do not preserve them.
3895 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3897 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
3899 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3902 * config/i386/constraints.md (Bs): Replace
3903 ix86_indirect_branch_register with
3904 TARGET_INDIRECT_BRANCH_REGISTER.
3906 * config/i386/i386.md (indirect_jump): Likewise.
3907 (tablejump): Likewise.
3908 (*sibcall_memory): Likewise.
3909 (*sibcall_value_memory): Likewise.
3910 Peepholes of indirect call and jump via memory: Likewise.
3911 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
3912 (*sibcall_value_GOT_32): Likewise.
3913 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3914 (GOT_memory_operand): Likewise.
3915 (call_insn_operand): Likewise.
3916 (sibcall_insn_operand): Likewise.
3917 (GOT32_symbol_operand): Likewise.
3918 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
3920 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3922 PR rtl-optimization/83496
3923 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
3924 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
3925 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
3926 redundant insn, if any.
3927 (relax_delay_slots): Likewise.
3928 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
3930 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
3932 PR tree-optimization/83965
3933 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
3934 that grouped statements are part of a reduction chain. Return
3935 true if the statement is not marked as a reduction itself but
3937 (vect_recog_dot_prod_pattern): Don't check whether the statement
3938 is part of a group here.
3939 (vect_recog_sad_pattern): Likewise.
3940 (vect_recog_widen_sum_pattern): Likewise.
3942 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3945 * final.c (rest_of_clean_state): Also look for calls inside sequences.
3947 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3950 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
3952 (ix86_output_indirect_function_return): New prototype.
3953 (ix86_split_simple_return_pop_internal): Likewise.
3954 * config/i386/i386.c (indirect_return_via_cx): New.
3955 (indirect_return_via_cx_bnd): Likewise.
3956 (indirect_thunk_name): Handle return va CX_REG.
3957 (output_indirect_thunk_function): Create alias for
3958 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
3959 (ix86_output_indirect_jmp): Remove the bool argument.
3960 (ix86_output_indirect_function_return): New function.
3961 (ix86_split_simple_return_pop_internal): Likewise.
3962 * config/i386/i386.md (*indirect_jump): Don't pass false
3963 to ix86_output_indirect_jmp.
3964 (*tablejump_1): Likewise.
3965 (simple_return_pop_internal): Change it to define_insn_and_split.
3966 Call ix86_split_simple_return_pop_internal to split it for
3968 (simple_return_indirect_internal): Call
3969 ix86_output_indirect_function_return instead of
3970 ix86_output_indirect_jmp.
3972 2018-02-26 Jakub Jelinek <jakub@redhat.com>
3975 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
3976 memset and value initialization afterwards.
3978 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
3980 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
3982 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3985 * common/config/aarch64/aarch64-common.c
3986 (aarch_option_optimization_table[]): Switch
3987 off fomit-frame-pointer
3989 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3990 Chung-Ju Wu <jasonwucj@gmail.com>
3992 * config/nds32/nds32-multiple.md (load_multiple): Disallow
3994 (store_multiple): Ditto.
3996 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3998 * config.gcc: Add --with-cpu support for nds32 target.
3999 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
4000 * config/nds32/nds32.opt: Add -mcpu= option.
4002 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
4004 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
4005 isel=yes): Warn for these deprecated options.
4007 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
4009 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
4010 ISA_2_5_MASKS_EMBEDDED.
4012 2018-02-23 Jakub Jelinek <jakub@redhat.com>
4014 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
4015 p->max as pointers rather than using iterative_hash_expr.
4017 2018-02-23 Carl Love <cel@us.ibm.com>
4019 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
4020 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
4022 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
4023 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
4024 P8V_BUILTIN_VEC_VUNSIGNED2.
4026 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
4029 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
4030 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
4032 * lra-constraints.c (curr_insn_transform): Set up
4033 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
4035 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
4036 * lra-eliminations.c (spill_pseudos): Ditto.
4037 (process_insn_for_elimination): Ditto.
4038 * lra-lives.c (reg_early_clobber_p): Use the new macros.
4039 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
4040 LRA_NON_CLOBBERED_ALT.
4042 2018-02-22 Martin Sebor <msebor@redhat.com>
4044 PR tree-optimization/84480
4045 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
4046 to maybe_diag_stxncpy_trunc. Call it.
4047 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
4048 from gimple_fold_builtin_strcpy. Print inlining stack.
4049 (handle_builtin_stxncpy): Print inlining stack.
4050 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
4052 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
4055 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
4056 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
4057 and -fcheck-pointer-bounds are used together.
4058 (indirect_thunk_prefix): New enum.
4059 (indirect_thunk_need_prefix): New function.
4060 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
4061 "_nt" instead of "_bnd" for NOTRACK prefix.
4062 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
4063 (output_indirect_thunk_function): Likewise.
4065 (ix86_code_end): Update output_indirect_thunk_function calls.
4066 (ix86_output_indirect_branch_via_reg): Replace
4067 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
4068 (ix86_output_indirect_branch_via_push): Likewise.
4069 (ix86_output_function_return): Likewise.
4070 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
4071 incompatible with -fcf-protection=branch and
4072 -fcheck-pointer-bounds.
4074 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4077 * config/aarch64/aarch64.c (aarch64_print_address_internal):
4078 Change gcc_assert call to output_operand_lossage.
4080 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4082 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
4084 2018-02-22 DJ Delorie <dj@redhat.com>
4085 Sebastian Perta <sebastian.perta@renesas.com>
4086 Oleg Endo <olegendo@gcc.gnu.org>
4088 * config/rx/rx.c (rx_rtx_costs): New function.
4089 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
4091 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
4093 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
4095 2018-02-22 Martin Liska <mliska@suse.cz>
4098 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
4099 Add "native" as a possible value.
4101 2018-02-22 Martin Liska <mliska@suse.cz>
4104 * config/i386/i386.c (ix86_option_override_internal):
4105 Add "native" as a possible value for -march and -mtune.
4107 2018-02-22 Jakub Jelinek <jakub@redhat.com>
4110 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
4111 to all type variants.
4113 PR tree-optimization/84503
4114 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
4115 width as info->bitpos + info->bitsize - start.
4116 (merged_store_group::merge_overlapping): Simplify width computation.
4117 (check_no_overlap): New function.
4118 (imm_store_chain_info::try_coalesce_bswap): Compute expected
4119 start + width and last_order of the group, fail if check_no_overlap
4121 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
4122 to group if check_no_overlap fails.
4124 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4126 * config/rs6000/altivec.md: Delete contraint arguments to
4127 define_expand, define_split, and define_peephole2, and in
4128 define_insn_and_split if always unused.
4129 * config/rs6000/darwin.md: Ditto.
4130 * config/rs6000/dfp.md: Ditto.
4131 * config/rs6000/rs6000.md: Ditto.
4132 * config/rs6000/sync.md: Ditto.
4133 * config/rs6000/vector.md: Ditto.
4134 * config/rs6000/vsx.md: Ditto.
4136 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4138 * config/rs6000/altivec.md: Write output control strings as braced
4139 blocks instead of double-quoted strings.
4140 * config/rs6000/darwin.md: Ditto.
4141 * config/rs6000/rs6000.md: Ditto.
4142 * config/rs6000/vector.md: Ditto.
4143 * config/rs6000/vsx.md: Ditto.
4145 2018-02-21 Jason Merrill <jason@redhat.com>
4147 PR c++/84314 - ICE with templates and fastcall attribute.
4148 * attribs.c (build_type_attribute_qual_variant): Remove assert.
4150 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4152 * ipa-cp.c (determine_versionability): Fix comment typos.
4154 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4157 * ipa-cp.c (determine_versionability): Do not version functions caling
4160 2018-02-21 Martin Liska <mliska@suse.cz>
4163 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
4164 Add "native" as a possible value.
4165 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
4166 the macro when native cpu detection is available.
4168 2018-02-21 Martin Liska <mliska@suse.cz>
4171 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
4172 Add "native" as a possible value.
4173 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
4174 when native cpu detection is available.
4176 2018-02-21 Jakub Jelinek <jakub@redhat.com>
4177 Martin Sebor <msebor@redhat.com>
4179 PR tree-optimization/84478
4180 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
4182 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
4183 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
4184 support which is conservatively correct, for 2 only stay conservative
4185 for maxlen. Formatting and comment capitalization fixes. Add STRICT
4186 argument to the 2 argument get_range_strlen, adjust 6 arg
4187 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
4189 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
4190 (gimple_fold_builtin_strlen): Pass true as last argument to
4193 2018-02-20 Martin Sebor <msebor@redhat.com>
4196 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
4197 (builtin_memref::set_base_and_offset): Same. Handle inner references.
4198 (builtin_memref::builtin_memref): Factor out parts into
4199 set_base_and_offset and call it.
4201 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4204 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
4205 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
4206 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
4207 search at the associated MODE_INT.
4209 2018-02-20 Jeff Law <law@redhat.com>
4212 PR tree-optimization/81592
4214 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
4215 for range data rather than using global data.
4216 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
4217 range data rather than using global data.
4218 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
4219 pass it to children as needed.
4220 (struct directive::fmtresult): Similarly.
4221 (struct directive::set_width): Similarly.
4222 (struct directive::set_precision): Similarly.
4223 (format_integer, format_directive, parse_directive): Similarly.
4224 (format_none): Accept unnamed vr_values parameter.
4225 (format_percent, format_floating, format_character): Similarly.
4226 (format_string, format_plain): Similarly.
4227 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
4228 the EVRP range analyzer for range data rather than using global data.
4229 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
4230 gimple-ssa-evrp-analyze.h
4231 (class sprintf_dom_walker): Add after_dom_children member function.
4232 Add evrp_range_analyzer member.
4233 (sprintf_dom_walker::before_dom_children): Call into the EVRP
4234 range analyzer as needed.
4235 (sprintf_dom_walker::after_dom_children): New member function.
4236 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
4238 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
4239 (evrp_range_analyzer::pop_to_marker): Likewise.
4241 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4243 PR tree-optimization/84419
4244 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
4245 with the required type if its current type is compatible but
4248 2018-02-20 Jakub Jelinek <jakub@redhat.com>
4251 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
4252 after vectorization.
4254 2018-02-20 Martin Liska <mliska@suse.cz>
4257 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
4258 possible values if we don't have a hint.
4260 2018-02-20 Martin Liska <mliska@suse.cz>
4264 * final.c (shorten_branches): Build align_tab array with one
4266 * opts.c (finish_options): Add alignment option limit check.
4267 (MAX_CODE_ALIGN): Likewise.
4268 (MAX_CODE_ALIGN_VALUE): Likewise.
4269 * doc/invoke.texi: Document maximum allowed option value for
4270 all -falign-* options.
4272 2018-02-19 Jakub Jelinek <jakub@redhat.com>
4275 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
4276 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
4277 * var-tracking.c (emit_note_insn_var_location): Remove all references
4278 to NOTE_INSN_CALL_ARG_LOCATION.
4279 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
4280 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
4281 Use copy_rtx_if_shared.
4282 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
4283 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
4284 (dwarf2out_var_location): Remove handling of
4285 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
4287 * final.c (final_scan_insn): Remove all references to
4288 NOTE_INSN_CALL_ARG_LOCATION.
4289 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
4290 before dumping final insns.
4291 * except.c (emit_note_eh_region_end): Remove all references to
4292 NOTE_INSN_CALL_ARG_LOCATION.
4293 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
4294 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
4295 * config/arc/arc.c (hwloop_optimize): Likewise.
4296 * config/arm/arm.c (create_fix_barrier): Likewise.
4297 * config/s390/s390.c (s390_chunkify_start): Likewise.
4298 * config/sh/sh.c (find_barrier): Likewise.
4299 * config/i386/i386.c (rest_of_insert_endbranch,
4300 ix86_seh_fixup_eh_fallthru): Likewise.
4301 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
4302 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
4303 * config/frv/frv.c (frv_function_prologue): Likewise.
4304 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
4306 (note_outside_basic_block_p): Remove all references to
4307 NOTE_INSN_CALL_ARG_LOCATION.
4308 * gengtype.c (adjust_field_rtx_def): Likewise.
4309 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
4311 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
4312 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
4315 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
4318 PR tree-optimization/84452
4319 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
4320 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
4323 2018-02-19 Martin Liska <mliska@suse.cz>
4326 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
4328 2018-02-19 Martin Liska <mliska@suse.cz>
4329 Richard Sandiford <richard.sandiford@linaro.org>
4331 PR tree-optimization/82491
4332 * gimple-fold.c (get_base_constructor): Make earlier bail out
4335 2018-02-19 Carl Love <cel@us.ibm.com>
4337 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
4338 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
4340 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
4341 P8V_BUILTIN_VEC_NEG.
4343 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
4345 * config/rl78/rl78.md (movdf): New define expand.
4347 2018-02-19 Martin Liska <mliska@suse.cz>
4350 * doc/invoke.texi: Fix typo.
4351 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
4353 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
4355 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
4356 handle rs6000_single_float and rs6000_double_float specially for
4359 2018-02-16 Jeff Law <law@redhat.com>
4361 * config/rx/rx.c (add_pop_cfi_notes): New function.;
4364 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4367 * ipa-inline.c (inline_small_functions): Fix a typo.
4369 2018-02-16 Nathan Sidwell <nathan@acm.org>
4371 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
4373 2018-02-16 Carl Love <cel@us.ibm.com>
4375 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
4376 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
4377 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
4378 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
4379 expansion to P8V_BUILTIN_VEC_FLOAT2.
4381 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
4383 PR rtl-optimization/70023
4384 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
4385 src_regno into account.
4387 2018-02-16 Carl Love <cel@us.ibm.com>
4389 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
4390 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
4391 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
4392 * config/rs6000/rs6000.c: Remove case statements for
4393 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
4394 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
4395 and P9V_BUILTIN_VEC_VINSERT4B.
4396 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
4397 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
4398 * config/rs6000/vsx.md:
4399 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
4402 2018-02-16 Carl Love <cel@us.ibm.com>
4404 * config/rs6000/altivec.h: Add builtin names vec_extract4b
4406 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
4408 * config/rs6000/rs6000-c.c: Add the definitions for
4409 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
4410 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
4411 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
4412 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
4413 definition for insert4b and define insn *insert3b_internal.
4414 * doc/extend.texi: Add documentation for vec_extract4b.
4416 2018-02-16 Nathan Sidwell <nathan@acm.org>
4418 * doc/extend.texi (Backwards Compatibility): Mention friend
4419 injection. Note for-scope is deprecated.
4420 * doc/invoke.texi (-ffriend-injection): Deprecate.
4422 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
4424 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
4425 that moved to I2, also allow destinations that are a paradoxical
4426 subreg (instead of a normal reg).
4428 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
4431 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
4434 2018-02-16 Richard Biener <rguenther@suse.de>
4436 PR tree-optimization/84037
4437 PR tree-optimization/84016
4439 * config/i386/i386.c (ix86_builtin_vectorization_cost):
4440 Adjust vec_construct for the fact we need additional higher latency
4441 128bit inserts for AVX256 and AVX512 vector builds.
4442 (ix86_add_stmt_cost): Scale vector construction cost for
4445 2018-02-16 Richard Biener <rguenther@suse.de>
4447 PR tree-optimization/84417
4448 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
4449 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
4450 (non_rewritable_lvalue_p): Likewise, use poly-ints.
4452 2018-02-16 Martin Liska <mliska@suse.cz>
4455 * internal-fn.def (ASAN_CHECK): Set proper flags.
4456 (ASAN_MARK): Likewise.
4458 2018-02-16 Julia Koval <julia.koval@intel.com>
4460 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
4461 from PTA_CANNONLAKE.
4463 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4466 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
4467 Use ++iter rather than iter++ for std::list iterators.
4468 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
4469 defer deleting them until all nodes in the forest are processed. Do
4470 free even leaf nodes. Change to_process into auto_vec.
4473 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
4474 * vec.h (vec_default_construct): Use memset instead of placement new
4475 if BROKEN_VALUE_INITIALIZATION is defined.
4476 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
4477 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
4480 PR rtl-optimization/83723
4481 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
4482 * lra.c (lra_substitute_pseudo): Likewise. If true, use
4483 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
4485 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
4487 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
4489 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
4491 PR rtl-optimization/81443
4492 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
4493 from inner REGs to paradoxical SUBREGs.
4495 2018-02-16 Richard Biener <rguenther@suse.de>
4497 PR tree-optimization/84399
4498 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4499 For operands we can analyze at their definition make sure we can
4500 analyze them at each use as well.
4502 2018-02-16 Richard Biener <rguenther@suse.de>
4504 PR tree-optimization/84190
4505 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
4506 volatile accesses if the decl isn't volatile.
4508 2018-02-15 Jason Merrill <jason@redhat.com>
4510 PR c++/84314 - ICE with templates and fastcall attribute.
4511 * attribs.c (build_type_attribute_qual_variant): Don't clobber
4512 TYPE_CANONICAL on an existing type.
4514 2018-02-15 Jakub Jelinek <jakub@redhat.com>
4516 PR tree-optimization/84383
4517 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
4518 dstoff nor call operand_equal_p if dstbase is NULL.
4520 PR tree-optimization/84334
4521 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
4522 also a CONSTANT_CLASS_P, punt.
4524 2018-02-14 Jim Wilson <jimw@sifive.com>
4526 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
4527 first SMALL_OPERAND check. New local min_second_step. Move assert
4528 to where locals are set. Add TARGET_RVC support.
4529 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
4531 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
4533 * doc/invoke.texi: Correct -Wformat-overflow code sample.
4535 2018-02-14 Martin Sebor <msebor@redhat.com>
4537 PR tree-optimization/83698
4538 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
4539 arrays constrain the offset range to their bounds.
4540 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
4541 (builtin_access::overlap): Avoid setting the size of overlap if it's
4543 (maybe_diag_overlap): Also consider arrays when deciding what values
4544 of offsets to include in diagnostics.
4546 2018-02-14 Martin Sebor <msebor@redhat.com>
4549 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
4550 that correspond to the kind of a declaration.
4552 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
4555 * config/pa/pa.md: Load address of PIC label using the linkage table
4556 if the label is nonlocal.
4558 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
4560 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
4561 warning message if user requests -maltivec=be.
4562 * doc/invoke.texi: Document deprecation of -maltivec=be.
4564 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
4567 * config/rs6000/rs6000-c.c: Update definitions for
4568 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
4569 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
4571 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4574 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
4575 add _get_ssp intrinsics. Remove argument from
4576 __builtin_ia32_rdssp[d|q].
4577 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
4578 * config/i386/i386-builtin.def: Remove argument from
4579 __builtin_ia32_rdssp[d|q].
4580 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
4581 ix86_expand_special_args_builtin for _rdssp[d|q].
4582 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
4583 Clear register before usage.
4584 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
4585 Add documentation for new _get_ssp and _inc_ssp intrinsics.
4587 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
4589 PR tree-optimization/84357
4590 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
4591 operand 1 of an ARRAY_REF too.
4593 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
4596 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
4597 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4599 (set_of_reg): New struct.
4600 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
4601 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
4602 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4604 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
4605 Split into bitclr, bitset, bitinvert patterns if appropriate.
4606 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
4607 use rx_fuse_in_memory_bitop.
4608 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
4609 to named insn, correct maximum insn length.
4611 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
4614 * machmode.def: Define a complex mode for PARTIAL_INT.
4615 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
4617 * doc/rtl.texi: Document CSPImode.
4618 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
4620 (msp430_hard_regno_nregs_with_padding): Likewise.
4622 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
4625 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
4627 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
4629 PR rtl-optimization/84169
4630 * combine.c (try_combine): New variable split_i2i3. Set it to true if
4631 we generated a parallel as new i3 and we split that to new i2 and i3
4632 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
4633 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
4634 those to i2, not i1. Partially rewrite this scan code.
4636 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4639 * stor-layout.c (place_field): For variable length fields, adjust
4640 offset_align afterwards not just based on the field's alignment,
4641 but also on the size.
4644 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
4645 of exps and logs in the use_exp2 case.
4647 2018-02-13 Jeff Law <law@redhat.com>
4649 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
4652 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
4655 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
4659 * common.opt (gas-loc-support, gas-locview-support): New.
4660 (ginline-points, ginternal-reset-location-views): New.
4661 * doc/invoke.texi: Document them. Use @itemx where intended.
4662 (gvariable-location-views): Adjust.
4663 * target.def (reset_location_view): New.
4664 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
4665 (TARGET_RESET_LOCATION_VIEW): New.
4666 * doc/tm.texi: Rebuilt.
4667 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
4668 (dwarf2out_default_as_locview_support): New.
4669 (output_asm_line_debug_info): Use option variables.
4670 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
4671 (output_loc_list): Likewise.
4672 (add_high_low_attributes): Check option variables.
4673 Don't output entry view attribute in strict mode.
4674 (gen_inlined_subroutine_die): Check option variables.
4675 (dwarf2out_inline_entry): Likewise.
4676 (init_sections_and_labels): Likewise.
4677 (dwarf2out_early_finish): Likewise.
4678 (maybe_reset_location_view): New, from...
4679 (dwarf2out_var_location): ... here. Call it.
4680 * debug.h (dwarf2out_default_as_loc_support): Declare.
4681 (dwarf2out_default_as_locview_support): Declare.
4682 * hooks.c (hook_int_rtx_insn_0): New.
4683 * hooks.h (hook_int_rtx_insn_0): Declare.
4684 * toplev.c (process_options): Take -gas-loc-support and
4685 -gas-locview-support from dwarf2out. Enable
4686 -gvariable-location-views by default only with locview
4687 assembler support. Enable -ginternal-reset-location-views by
4688 default only if the target defines the corresponding hook.
4689 Enable -ginline-points by default if location views are
4690 enabled; force it disabled if statement frontiers are
4692 * tree-inline.c (expand_call_inline): Check option variables.
4693 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
4695 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
4697 PR tree-optimization/84321
4698 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
4699 handling. Also check whether the anti-range contains any values
4700 that satisfy the mask; switch to a VR_RANGE if not.
4702 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
4705 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
4707 2018-02-13 Martin Jambor <mjambor@suse.cz>
4710 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
4711 of call statements, also set location of a load to a temporary.
4713 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
4715 * config/rl78/rl78.c (add_vector_labels): New function.
4716 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
4717 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
4718 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
4719 which checks that no arguments are passed.
4720 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
4721 * doc/extend.texi: Documentation for the new attribute.
4723 2018-02-13 Andreas Schwab <schwab@suse.de>
4725 * config/riscv/linux.h (CPP_SPEC): Define.
4727 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4730 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
4731 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
4732 OPTION_MASK_ISA_AES as first argument to def_builtin_const
4733 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
4734 instead of OPTION_MASK_ISA_PCLMUL as first argument to
4735 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
4736 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
4737 temporarily for AES and PCLMUL builtins.
4739 PR tree-optimization/84339
4740 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
4741 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
4745 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
4746 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
4747 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
4749 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
4751 * omp-simd-clone.h: New file.
4752 * omp-simd-clone.c: Include omp-simd-clone.h.
4753 (expand_simd_clones): No longer static.
4754 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
4755 cgraph.h and omp-simd-clone.h.
4756 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
4757 (vect_recog_widen_shift_pattern): Formatting fix.
4758 (vect_pattern_recog_1): Don't check optab for calls.
4761 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
4762 operands[2] into a REG before using gen_lowpart on it.
4764 2018-02-12 Jeff Law <law@redhat.com>
4767 * config/sh/sh.c (find_barrier): Consider a sibling call
4770 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
4771 successfully back substituting a reg.
4773 2018-02-12 Richard Biener <rguenther@suse.de>
4775 PR tree-optimization/84037
4776 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
4777 parameter, move visited init to caller.
4778 (vect_slp_analyze_operations): Separate cost from validity
4779 check, initialize visited once for all instances.
4780 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
4782 * tree-vect-stmts.c (vect_model_simple_cost): Make early
4784 (vect_model_promotion_demotion_cost): Likewise.
4785 (vectorizable_bswap): Guard cost modeling with !slp_node
4786 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
4788 (vectorizable_call): Likewise.
4789 (vectorizable_conversion): Likewise.
4790 (vectorizable_assignment): Likewise.
4791 (vectorizable_shift): Likewise.
4792 (vectorizable_operation): Likewise.
4793 (vectorizable_store): Likewise.
4794 (vectorizable_load): Likewise.
4795 (vectorizable_condition): Likewise.
4796 (vectorizable_comparison): Likewise.
4798 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
4801 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
4802 (ASAN_MARK): Fix fnspec to account for return value, change pointer
4803 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
4805 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4808 * params.def (inline-min-speedup): Increase from 8 to 15.
4809 (max-inline-insns-auto): Decrease from 40 to 30.
4810 * ipa-split.c (consider_split): Add some buffer for function to
4811 be considered inlining candidate.
4812 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
4815 2018-02-12 Richard Biener <rguenther@suse.de>
4817 PR tree-optimization/84037
4818 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
4819 matched stmts if we cannot swap the non-matched ones.
4821 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4823 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
4824 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
4825 _mm_maskz_scalef_round_ss): New intrinsics.
4826 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
4827 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
4828 __builtin_ia32_scalefss_round): Remove.
4829 (__builtin_ia32_scalefsd_mask_round,
4830 __builtin_ia32_scalefss_mask_round): New intrinsics.
4831 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
4832 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
4833 ((match_operand:VF_128 2 "<round_nimm_predicate>"
4834 "<round_constraint>")): Changed to ...
4835 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
4836 "<round_scalar_constraint>")): ... this.
4837 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
4838 %0, %1, %2<round_op3>}"): Changed to ...
4839 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
4840 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
4841 %2<round_scalar_mask_op3>}"): ... this.
4842 * config/i386/subst.md (round_scalar_nimm_predicate): New.
4844 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4846 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
4847 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
4848 (_mm_maskz_sqrt_round_ss): New intrinsics.
4849 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
4850 (__builtin_ia32_sqrtsd_mask_round)
4851 (__builtin_ia32_sqrtss_mask_round): New builtins.
4852 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
4853 (__builtin_ia32_sqrtss_round): Remove.
4854 (__builtin_ia32_sqrtsd_mask_round)
4855 (__builtin_ia32_sqrtss_mask_round): New builtins.
4856 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
4857 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
4858 ((match_operand:VF_128 1 "vector_operand"
4859 "xBm,<round_constraint>")): Changed to ...
4860 ((match_operand:VF_128 1 "vector_operand"
4861 "xBm,<round_scalar_constraint>")): ... this.
4862 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
4863 %0, %2, %<iptr>1<round_op3>}): Changed to ...
4864 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
4865 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
4866 %<iptr>1<round_scalar_mask_op3>}): ... this.
4867 ((set_attr "prefix" "<round_prefix>")): Changed to ...
4868 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
4870 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
4873 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
4874 Cast vec_cmpeq result to correct type.
4875 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
4876 Cast vec_cmpgt result to correct type.
4878 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
4880 * final.c (final_scan_insn_1): Renamed from...
4881 (final_scan_insn): ... this. New wrapper, to recover
4882 seen from the outermost call in recursive ones.
4883 * config/sparc/sparc.c (output_return): Drop seen from call.
4884 (output_sibcall): Likewise.
4885 * config/visium/visium.c (output_branch): Likewise.
4887 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
4889 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
4892 2018-02-10 Alan Modra <amodra@gmail.com>
4895 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
4896 Specify LR as an input.
4898 2018-02-10 Jakub Jelinek <jakub@redhat.com>
4901 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
4902 remove_member_access_dummy_vars): New functions.
4903 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
4904 lower_omp_1, execute_lower_omp): Use them.
4906 PR rtl-optimization/84308
4907 * shrink-wrap.c (spread_components): Release todo vector.
4909 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
4911 PR rtl-optimization/57193
4912 * ira-color.c (struct allocno_color_data): Add member
4913 conflict_allocno_hard_prefs.
4914 (update_conflict_allocno_hard_prefs): New.
4915 (bucket_allocno_compare_func): Add a preference based on
4916 conflict_allocno_hard_prefs.
4917 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
4918 (color_allocnos): Remove a dead code. Initiate
4919 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
4921 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4924 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
4925 constraint from =wa to wa. Avoid a subreg on the output operand,
4926 instead use a pseudo and subreg it in a move.
4927 (p9_xxbrd_<mode>): Changed to ...
4928 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
4929 (p9_xxbrd_v2df): New expander.
4930 (p9_xxbrw_<mode>): Changed to ...
4931 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
4932 (p9_xxbrw_v4sf): New expander.
4934 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4936 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
4938 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
4941 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
4942 multiply in 32-bit mode.
4943 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
4944 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
4947 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4949 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
4950 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
4951 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
4952 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
4954 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
4957 * dwarf2out.c (is_trivial_indirect_ref): New function.
4958 (dwarf2out_late_global_decl): Do not generate a location
4959 attribute for variables that have a non-trivial DECL_VALUE_EXPR
4960 and that are not defined in the current unit.
4962 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4964 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
4965 instead of a libcall for UNORDERED.
4967 2018-02-09 Tamar Christina <tamar.christina@arm.com>
4970 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
4971 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
4973 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4976 * config/s390/s390.c (s390_set_current_function): Invoke
4977 s390_indirect_branch_settings also if fndecl didn't change.
4979 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4981 * config/rs6000/rs6000.md (blockage): Set length to zero.
4983 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4985 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
4987 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4990 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
4991 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
4995 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
4996 PARALLEL incoming that failed vt_get_decl_and_offset check.
4999 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
5000 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
5002 (get_variable_section): For decls in named .bss* sections pass true as
5003 second argument to bss_initializer_p.
5005 2018-02-09 Marek Polacek <polacek@redhat.com>
5006 Jakub Jelinek <jakub@redhat.com>
5009 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
5010 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
5011 Sync some changes from cxx_fold_indirect_ref.
5013 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5015 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
5017 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
5018 (BLOCK_INLINE_ENTRY_LABEL): New.
5019 (dwarf2out_var_location): Disregard inline entry markers.
5020 (inline_entry_data): New struct.
5021 (inline_entry_data_hasher): New hashtable type.
5022 (inline_entry_data_hasher::hash): New.
5023 (inline_entry_data_hasher::equal): New.
5024 (inline_entry_data_table): New variable.
5025 (add_high_low_attributes): Add DW_AT_entry_pc and
5026 DW_AT_GNU_entry_view attributes if a pending entry is found
5027 in inline_entry_data_table. Add old entry_pc attribute only
5028 if debug nonbinding markers are disabled.
5029 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
5030 markers are enabled.
5031 (block_within_block_p, dwarf2out_inline_entry): New.
5032 (dwarf2out_finish): Check that no entries remained in
5033 inline_entry_data_table.
5034 * final.c (reemit_insn_block_notes): Handle inline entry notes.
5035 (final_scan_insn, notice_source_line): Likewise.
5036 (rest_of_clean_state): Skip inline entry markers.
5037 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
5039 * gimple.c (gimple_build_debug_inline_entry): New.
5040 * gimple.h (enum gimple_debug_subcode): Add
5041 GIMPLE_DEBUG_INLINE_ENTRY.
5042 (gimple_build_debug_inline_entry): Declare.
5043 (gimple_debug_inline_entry_p): New.
5044 (gimple_debug_nonbind_marker_p): Adjust.
5045 * insn-notes.def (INLINE_ENTRY): New.
5046 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
5047 inline entry marker notes.
5048 (print_insn): Likewise.
5049 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
5050 (INSN_DEBUG_MARKER_KIND): Likewise.
5051 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
5052 * tree-inline.c (expand_call_inline): Build and insert
5053 debug_inline_entry stmt.
5054 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
5055 inline entry blocks early, if nonbind markers are enabled.
5056 (dump_scope_block): Dump fragment info.
5057 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
5058 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
5059 (gimple_build_debug_inline_entry): New.
5060 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
5061 Enable/disable inline entry points too.
5062 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
5063 (DEBUG_INSN): Describe inline entry markers.
5065 * common.opt (gvariable-location-views): New.
5066 (gvariable-location-views=incompat5): New.
5067 * config.in: Rebuilt.
5068 * configure: Rebuilt.
5069 * configure.ac: Test assembler for view support.
5070 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
5071 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
5072 * dwarf2out.c (var_loc_view): New typedef.
5073 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
5074 (dwarf2out_locviews_in_attribute): New.
5075 (dwarf2out_locviews_in_loclist): New.
5076 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
5077 (enum dw_line_info_opcode): Add LI_adv_address.
5078 (struct dw_line_info_table): Add view.
5079 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
5080 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
5081 (zero_view_p): New variable.
5082 (ZERO_VIEW_P): New macro.
5083 (output_asm_line_debug_info): New.
5084 (struct var_loc_node): Add view.
5085 (add_AT_view_list, AT_loc_list): New.
5086 (add_var_loc_to_decl): Add view param. Test it against last.
5087 (new_loc_list): Add view params. Record them.
5088 (AT_loc_list_ptr): Handle loc and view lists.
5089 (view_list_to_loc_list_val_node): New.
5090 (print_dw_val): Handle dw_val_class_view_list.
5091 (size_of_die): Likewise.
5092 (value_format): Likewise.
5093 (loc_list_has_views): New.
5094 (gen_llsym): Set vl_symbol too.
5095 (maybe_gen_llsym, skip_loc_list_entry): New.
5096 (dwarf2out_maybe_output_loclist_view_pair): New.
5097 (output_loc_list): Output view list or entries too.
5098 (output_view_list_offset): New.
5099 (output_die): Handle dw_val_class_view_list.
5100 (output_dwarf_version): New.
5101 (output_compilation_unit_header): Use it.
5102 (output_skeleton_debug_sections): Likewise.
5103 (output_rnglists, output_line_info): Likewise.
5104 (output_pubnames, output_aranges): Update version comments.
5105 (output_one_line_info_table): Output view numbers in asm comments.
5106 (dw_loc_list): Determine current endview, pass it to new_loc_list.
5107 Call maybe_gen_llsym.
5108 (loc_list_from_tree_1): Adjust.
5109 (add_AT_location_description): Create view list attribute if
5110 needed, check it's absent otherwise.
5111 (convert_cfa_to_fb_loc_list): Adjust.
5112 (maybe_emit_file): Call output_asm_line_debug_info for test.
5113 (dwarf2out_var_location): Reset views as needed. Precompute
5114 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
5115 attribute. Set view.
5116 (new_line_info_table): Reset next view.
5117 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
5118 (dwarf2out_source_line): Likewise. Output view resets and labels to
5119 the assembler, or select appropriate line info opcodes.
5120 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
5121 (optimize_string_length): Catch it. Adjust.
5122 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
5123 dw_val_class_view_list, and remove it if no longer needed.
5124 (hash_loc_list): Hash view numbers.
5125 (loc_list_hasher::equal): Compare them.
5126 (optimize_location_lists): Check whether a view list symbol is
5127 needed, and whether the locview attribute is present, and
5128 whether they match. Remove the locview attribute if no longer
5130 (index_location_lists): Call skip_loc_list_entry for test.
5131 (dwarf2out_finish): Call output_asm_line_debug_info for test.
5132 Use output_dwarf_version.
5133 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
5134 (struct dw_val_node): Add val_view_list.
5135 * final.c (SEEN_NEXT_VIEW): New.
5136 (set_next_view_needed): New.
5137 (clear_next_view_needed): New.
5138 (maybe_output_next_view): New.
5139 (final_start_function): Rename to...
5140 (final_start_function_1): ... this. Take pointer to FIRST,
5141 add SEEN parameter. Emit param bindings in the initial view.
5142 (final_start_function): Reintroduce SEEN-less interface.
5143 (final): Rename to...
5144 (final_1): ... this. Take SEEN parameter. Output final pending
5145 next view at the end.
5146 (final): Reintroduce seen-less interface.
5147 (final_scan_insn): Output pending next view before switching
5148 sections or ending a block. Mark the next view as needed when
5149 outputting variable locations. Notify debug backend of section
5150 changes, and of location view changes.
5151 (rest_of_handle_final): Adjust.
5152 * toplev.c (process_options): Autodetect value for debug variable
5153 location views option. Warn on incompat5 without -gdwarf-5.
5154 * doc/invoke.texi (gvariable-location-views): New.
5155 (gvariable-location-views=incompat5): New.
5156 (gno-variable-location-views): New.
5158 2018-02-08 David Malcolm <dmalcolm@redhat.com>
5160 PR tree-optimization/84136
5161 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
5162 that the result of find_edge is non-NULL.
5164 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5167 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
5168 storing integer register in SImode. Fix cost of 256 and 512
5169 byte aligned SSE register store.
5171 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5173 * config/i386/i386.c (ix86_multiplication_cost): Fix
5174 multiplication cost for TARGET_AVX512DQ.
5176 2018-02-08 Marek Polacek <polacek@redhat.com>
5178 PR tree-optimization/84238
5179 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
5182 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5184 PR tree-optimization/84265
5185 * tree-vect-stmts.c (vectorizable_store): Don't treat
5186 VMAT_CONTIGUOUS accesses as grouped.
5187 (vectorizable_load): Likewise.
5189 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5191 PR tree-optimization/81635
5192 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
5193 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
5194 (test_round_for_mask): New functions.
5195 (wide_int_cc_tests): Call test_round_for_mask.
5196 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
5197 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
5198 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
5199 range returned by get_range_info.
5201 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5204 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
5205 * symtab.c: Include builtins.h
5206 (symtab_node::output_to_lto_symbol_table_p): Move here
5207 from lto-streamer-out.c:output_symbol_p.
5208 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
5209 (output_symbol_p): Move all logic to symtab.c
5210 (produce_symtab): Update.
5212 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5214 * config/s390/s390-opts.h (enum indirect_branch): Define.
5215 * config/s390/s390-protos.h (s390_return_addr_from_memory)
5216 (s390_indirect_branch_via_thunk)
5217 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
5218 (enum s390_indirect_branch_type): Define.
5219 * config/s390/s390.c (struct s390_frame_layout, struct
5220 machine_function): Remove.
5221 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
5222 (indirect_branch_table_label_no, indirect_branch_table_name):
5224 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
5225 (enum s390_indirect_branch_option): Define.
5226 (s390_return_addr_from_memory): New function.
5227 (s390_handle_string_attribute): New function.
5228 (s390_attribute_table): Add new attribute handler.
5229 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
5230 (s390_indirect_branch_via_thunk): New function.
5231 (s390_indirect_branch_via_inline_thunk): New function.
5232 (s390_function_ok_for_sibcall): When jumping via thunk disallow
5233 sibling call optimization for non z10 compiles.
5234 (s390_emit_call): Force indirect branch target to be a single
5235 register. Add r1 clobber for non-z10 compiles.
5236 (s390_emit_epilogue): Emit return jump via return_use expander.
5237 (s390_reorg): Handle JUMP_INSNs as execute targets.
5238 (s390_option_override_internal): Perform validity checks for the
5239 new command line options.
5240 (s390_indirect_branch_attrvalue): New function.
5241 (s390_indirect_branch_settings): New function.
5242 (s390_set_current_function): Invoke s390_indirect_branch_settings.
5243 (s390_output_indirect_thunk_function): New function.
5244 (s390_code_end): Implement target hook.
5245 (s390_case_values_threshold): Implement target hook.
5246 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
5248 * config/s390/s390.h (struct s390_frame_layout)
5249 (struct machine_function): Move here from s390.c.
5250 (TARGET_INDIRECT_BRANCH_NOBP_RET)
5251 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
5252 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
5253 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
5254 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
5255 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
5256 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
5257 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
5258 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
5259 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
5260 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
5261 (mnemonic attribute): Add values which aren't recognized
5263 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
5264 pattern for branch conversion. Fix mnemonic attribute.
5265 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
5266 indirect branch via thunk if requested.
5267 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
5268 ("*indirect_jump"): Disable for branch conversion using out of
5270 ("indirect_jump_via_thunk<mode>_z10")
5271 ("indirect_jump_via_thunk<mode>")
5272 ("indirect_jump_via_inlinethunk<mode>_z10")
5273 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
5274 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
5275 ("casesi_jump_via_inlinethunk<mode>_z10")
5276 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
5277 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
5278 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
5279 ("*indirect2_jump"): Disable for branch conversion.
5280 ("casesi_jump"): Turn into expander and expand patterns for branch
5282 ("return_use"): New expander.
5283 ("*return"): Emit return via thunk and rename it to ...
5284 ("*return<mode>"): ... this one.
5285 * config/s390/s390.opt: Add new options and and enum for the
5288 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5290 * lra-constraints.c (match_reload): Unconditionally use
5291 gen_lowpart_SUBREG, rather than selecting between that
5292 and equivalent gen_rtx_SUBREG code.
5294 2018-02-08 Richard Biener <rguenther@suse.de>
5296 PR tree-optimization/84233
5297 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
5298 changed flag instead of boguously re-using phi_inserted.
5300 2018-02-08 Martin Jambor <mjambor@suse.cz>
5302 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
5303 static local variables.
5305 2018-02-08 Richard Biener <rguenther@suse.de>
5307 PR tree-optimization/84278
5308 * tree-vect-stmts.c (vectorizable_store): When looking for
5309 smaller vector types to perform grouped strided loads/stores
5310 make sure the mode is supported by the target.
5311 (vectorizable_load): Likewise.
5313 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5315 * config/aarch64/aarch64.c (aarch64_components_for_bb):
5316 Increase LDP/STP opportunities by adding adjacent callee-saves.
5318 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5320 PR rtl-optimization/84068
5321 PR rtl-optimization/83459
5322 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
5324 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
5326 PR tree-optimization/84224
5327 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
5328 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
5331 2018-02-07 Iain Sandoe <iain@codesourcery.com>
5334 * config/rs6000/altivec.md (*restore_world): Remove LR use.
5335 * config/rs6000/predicates.md (restore_world_operation): Adjust op
5336 count, remove one USE.
5338 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
5340 * doc/install.texi (Configuration): Document the
5341 --with-long-double-format={ibm,ieee} PowerPC configuration
5345 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
5346 Convert from define_expand to be define_insn_and_split. Rework
5347 float/double/_Float128 conversions to QI/HI/SImode to work with
5348 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
5349 conversions to QI/HImode types did a store and then a load to
5350 truncate the value. For conversions to VSX registers, don't split
5351 the insn, instead emit the code directly. Use the code iterator
5352 any_fix to combine signed and unsigned conversions.
5353 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
5354 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
5355 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5356 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
5357 (fix_<mode>di2_hw): Likewise.
5358 (fixuns_<mode>di2_hw): Likewise.
5359 (fix_<mode>si2_hw): Likewise.
5360 (fixuns_<mode>si2_hw): Likewise.
5361 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
5362 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5363 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
5364 fix<uns>_trunc<SFDF:mode>si2_p8.
5365 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
5367 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
5368 (fix<uns>_<mode>_mem): Likewise.
5369 (fctiw<u>z_<mode>_mem): Likewise.
5370 (fix<uns>_<mode>_mem): Likewise.
5371 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
5372 the register allocator from doing a direct move to the GPRs to do
5373 a store, and instead use the ISA 3.0 store byte/half-word from
5374 vector register instruction. For IEEE 128-bit floating point,
5375 also optimize stores of 32-bit ints.
5376 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
5378 2018-02-07 Alan Hayward <alan.hayward@arm.com>
5380 * genextract.c (push_pathstr_operand): New function to support
5382 (walk_rtx): Call push_pathstr_operand.
5383 (print_path): Support [a-zA-Z].
5385 2018-02-07 Richard Biener <rguenther@suse.de>
5387 PR tree-optimization/84037
5388 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
5389 (cse_and_gimplify_to_preheader): Declare.
5390 (vect_get_place_in_interleaving_chain): Likewise.
5391 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5393 (_loop_vec_info::~_loop_vec_info): Delete it.
5394 (cse_and_gimplify_to_preheader): New function.
5395 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
5396 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
5397 (vectorizable_load): Likewise. For grouped stores always base
5398 the IV on the first element.
5399 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
5400 condition before gimplifying.
5402 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5404 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
5405 *DIV_EXPR and *MOD_EXPR.
5407 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
5410 * config/i386/i386.c (ix86_option_override_internal): Mask out
5411 the CF_SET bit when checking -fcf-protection.
5413 2018-02-07 Tom de Vries <tom@codesourcery.com>
5416 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
5419 2018-02-07 Richard Biener <rguenther@suse.de>
5421 PR tree-optimization/84204
5422 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
5425 PR tree-optimization/84205
5426 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
5427 special-case isl_ast_op_zdiv_r.
5429 PR tree-optimization/84223
5430 * graphite-scop-detection.c (gather_bbs::before_dom_children):
5431 Only add conditions from within the region.
5432 (gather_bbs::after_dom_children): Adjust.
5434 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
5437 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
5438 * config/avr/avr.md: Only post-reload split REG-REG moves if
5439 either register is GENERAL_REG_P.
5441 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5443 PR tree-optimization/84235
5444 * tree-ssa-scopedtables.c
5445 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
5446 if the subtraction is performed in floating point type where NaNs are
5447 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
5448 build 1. Formatting fix.
5450 2018-02-06 Jakub Jelinek <jakub@redhat.com>
5453 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
5454 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
5455 and skip it regardless of bb boundaries. Use CALL_P macro,
5456 don't test INSN_P (insn) together with CALL_P or JUMP_P check
5457 unnecessarily, formatting fix.
5459 2018-02-06 Michael Collison <michael.collison@arm.com>
5461 * config/arm/thumb2.md:
5462 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
5463 (*thumb_mov_notscc): Ditto.
5465 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
5468 * config/rs6000/rs6000.md (su code attribute): Use "u" for
5469 unsigned_fix, not "s".
5471 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5473 * configure.ac (gcc_fn_eh_frame_ro): New function.
5474 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
5475 correct .eh_frame permissions.
5476 * configure: Regenerate.
5478 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
5480 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
5483 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5485 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5486 Display warning message for -mno-speculate-indirect-jumps.
5488 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
5490 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
5492 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
5494 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
5496 PR tree-optimization/84225
5497 * tree-eh.c (find_trapping_overflow): Only call
5498 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
5500 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5503 * config/i386/i386.c: Reimplement the check of possible options
5504 -mibt/-mshstk conbination. Change error messages.
5505 * doc/invoke.texi: Fix a typo: remove extra '='.
5507 2018-02-06 Marek Polacek <polacek@redhat.com>
5509 PR tree-optimization/84228
5510 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
5512 2018-02-06 Tamar Christina <tamar.christina@arm.com>
5515 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
5516 emitted arch directives.
5517 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
5518 __ARM_FEATURE_COPROC before changing architectures.
5520 2018-02-06 Richard Biener <rguenther@suse.de>
5522 * config/i386/i386.c (print_reg): Fix typo.
5523 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
5525 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
5527 * configure: Regenerate.
5529 2018-02-05 Martin Sebor <msebor@redhat.com>
5531 PR tree-optimization/83369
5532 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
5535 2018-02-05 Martin Liska <mliska@suse.cz>
5537 * doc/invoke.texi: Cherry-pick upstream r323995.
5539 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
5541 * ira.c (ira_init_register_move_cost): Adjust comment.
5543 2018-02-05 Martin Liska <mliska@suse.cz>
5545 PR gcov-profile/84137
5546 * doc/gcov.texi: Fix typo in documentation.
5548 2018-02-05 Martin Liska <mliska@suse.cz>
5550 PR gcov-profile/83879
5551 * doc/gcov.texi: Document necessity of --dynamic-list-data when
5552 using dlopen functionality.
5554 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
5556 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
5557 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
5558 _mm_maskz_range_ss, _mm_mask_range_round_ss,
5559 _mm_maskz_range_round_ss): New intrinsics.
5560 (__builtin_ia32_rangesd128_round)
5561 (__builtin_ia32_rangess128_round): Remove.
5562 (__builtin_ia32_rangesd128_mask_round,
5563 __builtin_ia32_rangess128_mask_round): New builtins.
5564 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
5565 __builtin_ia32_rangess128_round): Remove.
5566 (__builtin_ia32_rangesd128_mask_round,
5567 __builtin_ia32_rangess128_mask_round): New builtins.
5568 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
5569 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
5570 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
5571 "<round_saeonly_constraint>")): Changed to ...
5572 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
5573 "<round_saeonly_scalar_constraint>")): ... this.
5574 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
5575 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
5576 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
5577 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
5578 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
5580 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
5582 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
5584 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
5585 Remove all values except native, 8540 and 8548.
5587 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
5589 * config/i386/i386.c (ix86_output_function_return): Pass
5590 INVALID_REGNUM, instead of -1, as invalid register number to
5591 indirect_thunk_name and output_indirect_thunk.
5593 2018-02-02 Julia Koval <julia.koval@intel.com>
5595 * config.gcc: Add -march=icelake.
5596 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
5597 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
5598 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
5599 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
5600 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
5601 (processor_target_table): Add icelake.
5602 (ix86_option_override_internal): Handle new PTAs.
5603 (get_builtin_code_for_version): Handle icelake.
5604 (M_INTEL_COREI7_ICELAKE): New.
5605 (fold_builtin_cpu): Handle icelake.
5606 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
5607 * doc/invoke.texi: Add -march=icelake.
5609 2018-02-02 Julia Koval <julia.koval@intel.com>
5611 * config/i386/i386.c (ix86_option_override_internal): Change flags type
5612 to wide_int_bitmask.
5613 * wide-int-bitmask.h: New.
5615 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5618 * config/i386/i386.md: Replace Pmode with word_mode in
5619 builtin_setjmp_setup and builtin_longjmp to support x32.
5621 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
5625 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
5627 (rs6000_supported_cpu_names): New static variable.
5628 (linux_cpu_translation_table): Likewise.
5629 (elf_platform) <cpu>: Define new static variable and use it.
5630 Translate kernel AT_PLATFORM name to canonical name if needed.
5631 Error if platform name is unknown.
5633 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
5636 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
5638 2018-02-01 Jeff Law <law@redhat.com>
5641 * config/i386/i386.c (release_scratch_register_on_entry): Add new
5642 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
5643 the scratch if RELEASE_VIA_POP is false.
5644 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
5645 If we have to save a temporary register, decrement SIZE appropriately.
5646 Pass new arguments to release_scratch_register_on_entry.
5647 (ix86_adjust_stack_and_probe): Likewise.
5648 (ix86_emit_probe_stack_range): Pass new arguments to
5649 release_scratch_register_on_entry.
5651 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
5653 PR rtl-optimization/84157
5654 * combine.c (change_zero_ext): Use REG_P predicate in
5655 front of HARD_REGISTER_P predicate.
5657 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
5659 * config/avr/avr.c (avr_option_override): Move disabling of
5660 -fdelete-null-pointer-checks to...
5661 * common/config/avr/avr-common.c (avr_option_optimization_table):
5664 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5666 PR tree-optimization/81635
5667 * tree-data-ref.c (split_constant_offset_1): For types that
5668 wrap on overflow, try to use range info to prove that wrapping
5671 2018-02-01 Renlin Li <renlin.li@arm.com>
5674 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
5676 (aarch64_register_move_cost): Likewise.
5677 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
5679 (REG_CLASS_NAMES): Likewise.
5680 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
5681 TAILCALL_ADDR_REGS. Remove IP registers.
5682 * config/aarch64/aarch64.md (Ucs): Update register constraint.
5684 2018-02-01 Richard Biener <rguenther@suse.de>
5686 * domwalk.h (dom_walker::dom_walker): Add additional constructor
5687 for specifying RPO order and allow NULL for that.
5688 * domwalk.c (dom_walker::dom_walker): Likewise.
5689 (dom_walker::walk): Handle NULL RPO order.
5690 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
5692 (rewrite_update_dom_walker): Likewise.
5693 (mark_def_dom_walker): Likewise.
5695 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5697 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
5698 (aarch64_maybe_expand_sve_subreg_move): Declare.
5699 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
5700 * config/aarch64/predicates.md (aarch64_any_register_operand): New
5702 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
5703 that are semantically a reverse operation.
5704 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
5705 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
5706 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
5708 (aarch64_can_change_mode_class): For big-endian, forbid changes
5709 between two SVE modes if they have different element sizes.
5711 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5713 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
5714 the TImode handling for big-endian targets.
5716 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5718 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
5719 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
5721 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
5722 Remove BSWAP handing for big-endian targets and use the form of
5723 LD1RQ appropariate for the mode.
5725 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5727 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
5728 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
5731 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5734 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
5735 check for operands that need to go through aarch64_sve_reload_be.
5737 2018-02-01 Jakub Jelinek <jakub@redhat.com>
5739 PR tree-optimization/81661
5740 PR tree-optimization/84117
5741 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
5742 * tree-eh.c: Include gimplify.h.
5743 (find_trapping_overflow, replace_trapping_overflow,
5744 rewrite_to_non_trapping_overflow): New functions.
5745 * tree-vect-loop.c: Include tree-eh.h.
5746 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
5747 * tree-data-ref.c: Include tree-eh.h.
5748 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
5750 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
5752 PR rtl-optimization/84123
5753 * combine.c (change_zero_ext): Check if hard register satisfies
5754 can_change_dest_mode before calling gen_lowpart_SUBREG.
5756 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
5759 * ira.c (ira_init_register_move_cost): Remove assert.
5761 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5763 PR rtl-optimization/84071
5764 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
5765 * doc/tm.texi: Regenerate.
5767 2018-01-31 Richard Biener <rguenther@suse.de>
5769 PR tree-optimization/84132
5770 * tree-data-ref.c (analyze_miv_subscript): Properly
5771 check whether evolution_function_is_affine_multivariate_p
5772 before calling gcd_of_steps_may_divide_p.
5774 2018-01-31 Julia Koval <julia.koval@intel.com>
5777 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
5778 * config/i386/i386.md (rdpid_rex64) New.
5779 (rdpid): Make 32bit only.
5781 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
5784 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
5785 an IDENTIFIER_NODE for FUNCTION_TYPE's.
5787 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5790 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5792 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5794 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5796 PR rtl-optimization/84071
5797 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
5798 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
5800 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5802 * config/arc/arc.c (arc_handle_aux_attribute): New function.
5803 (arc_attribute_table): Add 'aux' attribute.
5804 (arc_in_small_data_p): Consider aux like variables.
5805 (arc_is_aux_reg_p): New function.
5806 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
5807 (arc_get_aux_arg): New function.
5808 (prepare_move_operands): Handle aux-register access.
5809 (arc_handle_aux_attribute): New function.
5810 * doc/extend.texi (ARC Variable attributes): Add subsection.
5812 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5814 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
5815 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
5816 (arc_attribute_table): Add 'uncached' attribute.
5817 (arc_print_operand): Print '.di' flag for uncached memory
5819 (arc_in_small_data_p): Do not consider for small data the uncached
5821 (arc_is_uncached_mem_p): New function.
5822 * config/arc/predicates.md (compact_store_memory_operand): Check
5823 for uncached memory accesses.
5824 (nonvol_nonimm_operand): Likewise.
5825 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
5827 2018-01-31 Jakub Jelinek <jakub@redhat.com>
5830 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
5831 falign-loops=): Add Optimization flag.
5833 2018-01-30 Jeff Law <law@redhat.com>
5836 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
5837 INT_REGISTERS_SAVED. Check it prior to calling
5838 get_scratch_register_on_entry.
5839 (ix86_adjust_stack_and_probe): Similarly.
5840 (ix86_emit_probe_stack_range): Similarly.
5841 (ix86_expand_prologue): Corresponding changes.
5843 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5846 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
5847 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
5849 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
5852 * lra-constraints.c (curr_insn_transform): Process AND in the
5855 2018-01-30 Jakub Jelinek <jakub@redhat.com>
5857 PR rtl-optimization/83986
5858 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
5859 dependence against last_pending_memory_flush in addition to
5862 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
5864 PR tree-optimization/81611
5865 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
5868 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5871 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
5874 2018-01-30 Richard Biener <rguenther@suse.de>
5875 Jakub Jelinek <jakub@redhat.com>
5877 PR tree-optimization/84111
5878 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
5879 inner loops added during recursion, as they don't have up-to-date
5882 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5885 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
5886 (can_inline_edge_by_limits_p): ... here.
5887 (can_early_inline_edge_p, check_callers,
5888 update_caller_keys, update_callee_keys, recursive_inlining,
5889 add_new_edges_to_heap, speculation_useful_p,
5890 inline_small_functions,
5891 inline_small_functions, flatten_function,
5892 inline_to_all_callers_1): Update.
5894 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5896 * profile-count.c (profile_count::combine_with_ipa_count): Handle
5899 2018-01-30 Richard Biener <rguenther@suse.de>
5901 PR tree-optimization/83008
5902 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
5903 invariant and constant vector uses in stmts when they need
5906 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5909 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
5910 * configure: Regenerate.
5912 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5914 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
5916 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
5917 Use gen_rtx_REG rather than gen_lowpart.
5919 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5921 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
5922 rather than 0 when creating partial subregs.
5924 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5926 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
5929 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
5932 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
5933 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
5934 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
5935 flags. This restores the settings used before the 2017-07-24.
5936 Turning off pre increment/decrement/modify allows IVOPTS to
5937 optimize DF/SF loops where the index is an int.
5939 2018-01-29 Richard Biener <rguenther@suse.de>
5940 Kelvin Nilsen <kelvin@gcc.gnu.org>
5943 * tree-vect-stmts.c (vectorizable_call): Don't call
5944 targetm.vectorize_builtin_md_vectorized_function if callee is
5947 2018-01-22 Carl Love <cel@us.ibm.com>
5949 * doc/extend.tex: Fix typo in second arg in
5950 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
5952 2018-01-29 Richard Biener <rguenther@suse.de>
5954 PR tree-optimization/84086
5955 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
5956 (flush_ssaname_freelist): When SSA names were released reset
5957 the SCEV hash table.
5959 2018-01-29 Richard Biener <rguenther@suse.de>
5961 PR tree-optimization/84057
5962 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
5963 removed paths when removing edges.
5965 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
5967 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
5968 -mfunction-return=@var{choice}.
5970 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
5973 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
5975 (layout::print_source_line): Likewise.
5976 (test_get_line_width_without_trailing_whitespace): Add test cases.
5978 2018-01-27 Jakub Jelinek <jakub@redhat.com>
5981 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
5984 2018-01-26 Jim Wilson <jimw@sifive.com>
5986 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
5988 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
5991 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5993 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
5994 and CMP + SUB-immediate -> SUBS.
5996 2018-01-26 Martin Sebor <msebor@redhat.com>
5998 PR tree-optimization/83896
5999 * tree-ssa-strlen.c (get_string_len): Rename...
6000 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
6001 Avoid assuming length is constant.
6002 (handle_char_store): Use HOST_WIDE_INT for string length.
6004 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
6007 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
6008 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
6010 2018-01-26 Richard Biener <rguenther@suse.de>
6012 PR rtl-optimization/84003
6013 * dse.c (record_store): Only record redundant stores when
6014 the earlier store aliases at least all accesses the later one does.
6016 2018-01-26 Jakub Jelinek <jakub@redhat.com>
6018 PR rtl-optimization/83985
6019 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
6020 REG_CFA_RESTORE insns.
6021 (delete_unmarked_insns): Don't ignore separate shrink wrapping
6022 REG_CFA_RESTORE insns here.
6025 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
6026 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
6028 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6030 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
6031 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
6032 (arc_init): Likewise.
6033 (arc_override_options): Likewise.
6034 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
6036 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
6038 * config/arc/arc.h (TARGET_DBNZ): Define.
6039 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
6040 properly set the tune attribute.
6041 (dbnz): Use TARGET_DBNZ guard.
6042 * config/arc/arc.opt (mtune): Add core3 option.
6044 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6046 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
6047 recognize new pic like addresses.
6048 (arc_delegitimize_address): Clean up.
6050 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6052 * config/arc/arc-arches.def: Option mrf16 valid for all
6054 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
6055 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
6056 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
6057 * config/arc/arc-tables.opt: Regenerate.
6058 * config/arc/arc.c (arc_conditional_register_usage): Handle
6059 reduced register file case.
6060 (arc_file_start): Set must have build attributes.
6061 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
6063 * config/arc/arc.opt (mrf16): Add new option.
6064 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
6065 * config/arc/genmultilib.awk: Handle new mrf16 option.
6066 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
6067 * config/arc/t-multilib: Regenerate.
6068 * doc/invoke.texi (ARC Options): Document mrf16 option.
6070 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6072 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
6073 * config/arc/arc.c (arc_handle_secure_attribute): New function.
6074 (arc_attribute_table): Add 'secure_call' attribute.
6075 (arc_print_operand): Print secure call operand.
6076 (arc_function_ok_for_sibcall): Don't optimize tail calls when
6078 (arc_is_secure_call_p): New function. * config/arc/arc.md
6079 (call_i): Add support for sjli instruction.
6080 (call_value_i): Likewise.
6081 * config/arc/constraints.md (Csc): New constraint.
6083 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6084 John Eric Martin <John.Martin@emmicro-us.com>
6086 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
6087 * config/arc/arc.c (_arc_jli_section): New struct.
6088 (arc_jli_section): New type.
6089 (rc_jli_sections): New static variable.
6090 (arc_handle_jli_attribute): New function.
6091 (arc_attribute_table): Add jli_always and jli_fixed attribute.
6092 (arc_file_end): New function.
6093 (TARGET_ASM_FILE_END): Define.
6094 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
6095 (arc_add_jli_section): New function.
6096 (jli_call_scan): Likewise.
6097 (arc_reorg): Call jli_call_scan.
6098 (arc_output_addsi): Remove 'S' from printing asm operand.
6099 (arc_is_jli_call_p): New function.
6100 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
6102 (movhi_insn): Likewise.
6103 (movsi_insn): Likewise.
6104 (movsi_set_cc_insn): Likewise.
6105 (loadqi_update): Likewise.
6106 (load_zeroextendqisi_update): Likewise.
6107 (load_signextendqisi_update): Likewise.
6108 (loadhi_update): Likewise.
6109 (load_zeroextendhisi_update): Likewise.
6110 (load_signextendhisi_update): Likewise.
6111 (loadsi_update): Likewise.
6112 (loadsf_update): Likewise.
6113 (movsicc_insn): Likewise.
6114 (bset_insn): Likewise.
6115 (bxor_insn): Likewise.
6116 (bclr_insn): Likewise.
6117 (bmsk_insn): Likewise.
6118 (bicsi3_insn): Likewise.
6119 (cmpsi_cc_c_insn): Likewise.
6120 (movsi_ne): Likewise.
6121 (movsi_cond_exec): Likewise.
6122 (clrsbsi2): Likewise.
6133 (call_i): Remove 'S' asm letter, add jli instruction.
6134 (call_value_i): Likewise.
6135 * config/arc/arc.op (mjli-always): New option.
6136 * config/arc/constraints.md (Cji): New constraint.
6137 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
6139 (subsf3_fpx): Likewise.
6140 (mulsf3_fpx): Likewise.
6141 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
6143 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
6145 * doc/invoke.texi (ARC): Document mjli-always option.
6147 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
6149 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
6150 avoid addition with 0 and use incw and decw where possible.
6152 2018-01-26 Richard Biener <rguenther@suse.de>
6154 PR tree-optimization/81082
6155 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
6156 association if it requires casting to unsigned.
6157 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
6158 from fold_plusminus_mult_expr to catch important cases late when
6159 range info is available.
6161 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6163 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
6164 * configure.ac (hidden_linkonce): New test.
6165 * configure: Regenerate.
6166 * config.in: Regenerate.
6168 2018-01-26 Julia Koval <julia.koval@intel.com>
6170 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
6171 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
6172 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
6173 _mm_mask_bitshuffle_epi64_mask): Fix type.
6174 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
6175 USI_FTYPE_V4DI_V4DI_USI): Remove.
6176 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
6177 __builtin_ia32_vpshufbitqmb256_mask,
6178 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
6179 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
6180 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
6182 2018-01-26 Alan Modra <amodra@gmail.com>
6185 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
6186 UNSPEC_VBPERMQ. Sort other unspecs.
6188 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
6190 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
6192 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
6195 * predict.c (drop_profile): Do not push/pop cfun; update also
6197 (handle_missing_profiles): Fix logic looking for zero profiles.
6199 2018-01-25 Jakub Jelinek <jakub@redhat.com>
6202 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
6203 on functions with #pragma omp declare simd or functions with simd
6205 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
6206 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
6207 Remove trailing \n from warning_at calls.
6209 2018-01-25 Tom de Vries <tom@codesourcery.com>
6212 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6213 for neutered workers.
6215 2018-01-24 Joseph Myers <joseph@codesourcery.com>
6218 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
6219 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
6221 2018-01-24 Jeff Law <law@redhat.com>
6224 * i386.c (get_probe_interval): Move to earlier point.
6225 (ix86_compute_frame_layout): If -fstack-clash-protection and
6226 the frame is larger than the probe interval, then use pushes
6227 to save registers rather than reg->mem moves.
6228 (ix86_expand_prologue): Remove conditional for int_registers_saved
6231 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
6234 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
6235 min/max for never referenced object.
6237 2018-01-24 Jakub Jelinek <jakub@redhat.com>
6240 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
6242 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
6243 attributes from DECL_ATTRIBUTES (decl) without affecting
6244 DECL_ATTRIBUTES (current_function_decl).
6245 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
6246 functions with non-NULL DECL_ABSTRACT_ORIGIN.
6248 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
6250 PR tree-optimization/83979
6251 * fold-const.c (fold_comparison): Use constant_boolean_node
6252 instead of boolean_{true,false}_node.
6254 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
6256 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
6259 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6261 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6262 Simplify the clause that sets the length attribute.
6263 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6264 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
6265 clause that sets the length attribute.
6266 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6268 2018-01-24 Tom de Vries <tom@codesourcery.com>
6271 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
6272 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
6273 Add strict parameter.
6274 (prevent_branch_around_nothing): Insert dummy insn between branch to
6275 label and label with no ptx insn inbetween.
6276 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
6278 2018-01-24 Tom de Vries <tom@codesourcery.com>
6281 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6282 for neutered threads in warp.
6283 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
6285 2018-01-24 Richard Biener <rguenther@suse.de>
6287 PR tree-optimization/83176
6288 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
6291 2018-01-24 Richard Biener <rguenther@suse.de>
6293 PR tree-optimization/82819
6294 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
6295 code generating pluses that are no-ops in the target precision.
6297 2018-01-24 Richard Biener <rguenther@suse.de>
6300 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
6302 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6304 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
6305 to merge probabilities.
6306 * predict.c (probably_never_executed): Also mark as cold functions
6307 with global 0 profile and guessed local profile.
6308 * profile-count.c (profile_probability::combine_with_count): New
6310 * profile-count.h (profile_probability::operator*,
6311 profile_probability::operator*=, profile_probability::operator/,
6312 profile_probability::operator/=): Reduce precision to adjusted
6313 and set value to guessed on contradictory divisions.
6314 (profile_probability::combine_with_freq): Remove.
6315 (profile_probability::combine_wiht_count): Declare.
6316 (profile_count::force_nonzero):: Set to adjusted.
6317 (profile_count::probability_in):: Set quality to adjusted.
6318 * tree-ssa-tail-merge.c (replace_block_by): Use
6321 2018-01-23 Andrew Waterman <andrew@sifive.com>
6322 Jim Wilson <jimw@sifive.com>
6324 * config/riscv/riscv.c (riscv_stack_boundary): New.
6325 (riscv_option_override): Set riscv_stack_boundary. Handle
6326 riscv_preferred_stack_boundary_arg.
6327 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
6328 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
6329 (STACK_BOUNDARY): Set to riscv_stack_boundary.
6330 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
6331 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
6332 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
6334 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
6337 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
6338 of struct ix86_frame.
6339 (ix86_expand_epilogue): Likewise. Add a local variable for
6340 the reg_save_offset field in struct ix86_frame.
6342 2018-01-23 Bin Cheng <bin.cheng@arm.com>
6344 PR tree-optimization/82604
6345 * tree-loop-distribution.c (enum partition_kind): New enum item
6346 PKIND_PARTIAL_MEMSET.
6347 (partition_builtin_p): Support above new enum item.
6348 (generate_code_for_partition): Ditto.
6349 (compute_access_range): Differentiate cases that equality can be
6350 proven at all loops, the innermost loops or no loops.
6351 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
6352 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
6353 (finalize_partitions, distribute_loop): Don't fuse partition of
6354 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
6355 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
6358 2018-01-23 Martin Liska <mliska@suse.cz>
6360 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
6361 order to ignore the predictor.
6362 (PRED_POLYMORPHIC_CALL): Likewise.
6363 (PRED_RECURSIVE_CALL): Likewise.
6365 2018-01-23 Martin Liska <mliska@suse.cz>
6367 * tree-profile.c (tree_profiling): Print function header to
6368 aware reader which function we are working on.
6369 * value-prof.c (gimple_find_values_to_profile): Do not print
6370 not interesting value histograms.
6372 2018-01-23 Martin Liska <mliska@suse.cz>
6374 * profile-count.h (enum profile_quality): Add
6375 profile_uninitialized as the first value. Do not number values
6376 as they are zero based.
6377 (profile_count::verify): Update sanity check.
6378 (profile_probability::verify): Likewise.
6380 2018-01-23 Nathan Sidwell <nathan@acm.org>
6382 * doc/invoke.texi (ffor-scope): Deprecate.
6384 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6386 PR tree-optimization/83510
6387 * domwalk.c (set_all_edges_as_executable): New function.
6388 (dom_walker::dom_walker): Convert bool param
6389 "skip_unreachable_blocks" to enum reachability. Move setup of
6390 edge flags to set_all_edges_as_executable and only do it when
6391 reachability is REACHABLE_BLOCKS.
6392 * domwalk.h (enum dom_walker::reachability): New enum.
6393 (dom_walker::dom_walker): Convert bool param
6394 "skip_unreachable_blocks" to enum reachability.
6395 (set_all_edges_as_executable): New decl.
6396 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
6397 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
6399 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
6400 but converting true to REACHABLE_BLOCKS.
6401 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
6403 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
6404 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
6405 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
6407 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
6408 if check_all_array_refs will be called.
6410 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6412 * tree.c (selftest::test_location_wrappers): Add more test
6415 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6417 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
6418 (selftest::test_bit_in_range): Likewise.
6420 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6423 * doc/sourcebuild.texi (vect_float): Say that the selector
6424 only describes the situation when -funsafe-math-optimizations is on.
6425 (vect_float_strict): Document.
6427 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6429 PR tree-optimization/83965
6430 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
6431 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
6432 instead of checking only for a reduction.
6433 (vect_recog_widen_sum_pattern): Likewise.
6435 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6437 * predict.c (probably_never_executed): Only use precise profile info.
6438 (compute_function_frequency): Skip after inlining hack since we now
6439 have quality checking.
6441 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6443 * profile-count.h (profile_probability::very_unlikely,
6444 profile_probability::unlikely, profile_probability::even): Set
6445 precision to guessed.
6447 2018-01-23 Richard Biener <rguenther@suse.de>
6449 PR tree-optimization/83963
6450 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
6451 Properly terminate dominator walk when crossing the exit edge not
6452 when visiting its source block.
6454 2018-01-23 Jakub Jelinek <jakub@redhat.com>
6457 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
6458 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
6460 2018-01-22 Jakub Jelinek <jakub@redhat.com>
6462 PR tree-optimization/83957
6463 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
6464 semicolon after for body surrounded by braces.
6466 PR tree-optimization/83081
6467 * profile-count.h (profile_probability::split): New method.
6468 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
6469 Use profile_probability::split.
6470 (do_compare_rtx_and_jump): Fix adjustment of probabilities
6471 when splitting a single conditional jump into 2.
6473 2018-01-22 David Malcolm <dmalcolm@redhat.com>
6475 PR tree-optimization/69452
6476 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
6479 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6481 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
6482 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
6483 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
6485 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6487 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
6488 * config/rl78/rl78.md (movdi): New define_expand.
6489 * config/rl78/rl78.c (rl78_split_movdi): New function.
6491 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
6494 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
6496 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
6497 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
6498 128-bit to produce an UNSPEC move to get the double word with the
6499 signbit and then a shift directly to do signbit.
6500 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
6501 implementation with a new version that just does either a direct
6502 move or a regular move. Move memory interface to separate insns.
6503 Move insns so they are next to the expander.
6504 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
6505 with signbit move. Split big and little endian case.
6506 (signbit<mode>2_dm_mem_le): Likewise.
6507 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
6508 (signbit<mode>2_dm2): Likewise.
6510 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6512 * config/rl78/rl78.md (anddi3): New define_expand.
6514 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6516 * config/rl78/rl78.md (umindi3): New define_expand.
6518 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6520 * config/rl78/rl78.md (smindi3): New define_expand.
6522 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6524 * config/rl78/rl78.md (smaxdi3): New define_expand.
6526 2018-01-22 Carl Love <cel@us.ibm.com>
6528 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
6529 LVX_V1TI): Add macro expansion.
6530 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
6531 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
6532 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
6533 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
6534 Change check to determine if the instruction is a byte reversing
6535 entry. Fix typo in comment.
6536 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
6537 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
6538 Add def_builtin calls for new builtins.
6539 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
6540 Add define_insn expansion.
6542 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6544 * config/rl78/rl78.md (umaxdi3): New define_expand.
6546 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6548 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
6549 for non-QImode registers.
6551 2018-01-22 Richard Biener <rguenther@suse.de>
6553 PR tree-optimization/83963
6554 * graphite-scop-detection.c (scop_detection::get_sese): Delay
6555 including the loop exit block.
6556 (scop_detection::merge_sese): Likewise.
6557 (scop_detection::add_scop): Do it here instead.
6559 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6561 * doc/sourcebuild.texi (arm_softfloat): Document.
6563 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
6566 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
6567 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
6568 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
6570 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6571 David Edelsohn <dje.gcc@gmail.com>
6574 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6575 Change "crset eq" to "crset 2".
6576 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6577 (*call_indirect_aix<mode>_nospec): Likewise.
6578 (*call_value_indirect_aix<mode>_nospec): Likewise.
6579 (*call_indirect_elfv2<mode>_nospec): Likewise.
6580 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
6581 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
6582 change assembly output from . to $.
6583 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6584 (indirect_jump<mode>_nospec): Change assembly output from . to $.
6585 (*tablejump<mode>_internal1_nospec): Likewise.
6587 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
6590 * config/sh/sh_optimize_sett_clrt.cc:
6591 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
6593 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
6595 PR tree-optimization/83940
6596 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
6597 offset_dt to vect_constant_def rather than vect_unknown_def_type.
6598 (vect_check_load_store_mask): Add a mask_dt_out parameter and
6599 use it to pass back the definition type.
6600 (vect_check_store_rhs): Likewise rhs_dt_out.
6601 (vect_build_gather_load_calls): Add a mask_dt argument and use
6602 it instead of a call to vect_is_simple_use.
6603 (vectorizable_store): Update calls to vect_check_load_store_mask
6604 and vect_check_store_rhs. Use the dt returned by the latter instead
6605 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
6606 instead of calls to vect_is_simple_use. Pass the scalar rather
6607 than the vector operand to vect_is_simple_use when handling
6608 second and subsequent copies of an rhs value.
6609 (vectorizable_load): Update calls to vect_check_load_store_mask
6610 and vect_build_gather_load_calls. Use the cached mask_dt and
6611 gs_info.offset_dt instead of calls to vect_is_simple_use.
6613 2018-01-20 Jakub Jelinek <jakub@redhat.com>
6616 * tree-emutls.c: Include gimplify.h.
6617 (lower_emutls_2): New function.
6618 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
6619 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
6620 it before further processing.
6623 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
6624 UINTVAL (trueop1) instead of INTVAL (op1).
6626 2018-01-19 Jakub Jelinek <jakub@redhat.com>
6630 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
6631 INCOMING_FRAME_SP_OFFSET if not defined.
6632 (scan_trace): Add ENTRY argument. If true and
6633 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
6634 emit a note to adjust the CFA offset.
6635 (create_cfi_notes): Adjust scan_trace callers.
6636 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
6637 INCOMING_FRAME_SP_OFFSET in the CIE.
6638 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
6639 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
6641 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
6642 * doc/tm.texi: Regenerated.
6644 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6646 PR rtl-optimization/83147
6647 * lra-constraints.c (remove_inheritance_pseudos): Use
6648 lra_substitute_pseudo_within_insn.
6650 2018-01-19 Tom de Vries <tom@codesourcery.com>
6651 Cesar Philippidis <cesar@codesourcery.com>
6654 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
6656 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
6659 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
6660 spaces for function labels.
6662 2018-01-19 Martin Liska <mliska@suse.cz>
6664 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
6665 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
6666 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
6667 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
6668 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
6669 (PRED_CONST_RETURN): Change from 69 to 65.
6670 (PRED_NULL_RETURN): Change from 91 to 71.
6671 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
6672 (PRED_LOOP_GUARD): Change from 66 to 73.
6674 2018-01-19 Martin Liska <mliska@suse.cz>
6676 * predict.c (predict_insn_def): Add new assert.
6677 (struct branch_predictor): Change type to signed integer.
6678 (test_prediction_value_range): Amend test to cover
6680 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
6681 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
6682 (PRED_LOOP_ITERATIONS_MAX): Likewise.
6683 (PRED_LOOP_IV_COMPARE): Likewise.
6684 * predict.h (PROB_UNINITIALIZED): Define new constant.
6686 2018-01-19 Martin Liska <mliska@suse.cz>
6688 * predict.c (dump_prediction): Add new format for
6689 analyze_brprob.py script which is enabled with -details
6691 * profile-count.h (precise_p): New function.
6693 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6695 PR tree-optimization/83922
6696 * tree-vect-loop.c (vect_verify_full_masking): Return false if
6697 there are no statements that need masking.
6698 (vect_active_double_reduction_p): New function.
6699 (vect_analyze_loop_operations): Use it when handling phis that
6700 are not in the loop header.
6702 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6704 PR tree-optimization/83914
6705 * tree-vect-loop.c (vectorizable_induction): Don't convert
6706 init_expr or apply the peeling adjustment for inductions
6707 that are nested within the vectorized loop.
6709 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6711 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
6714 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6718 * function.h (gimplify_parameters): Add gimple_seq * argument.
6719 * function.c: Include gimple.h and options.h.
6720 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
6721 for the added local temporaries if needed.
6722 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
6723 if there are any parameter cleanups, wrap whole body into a
6724 try/finally with the cleanups.
6726 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
6729 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
6730 Use GET_MODE_CLASS for scalar floating point.
6732 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6736 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
6737 Fix call of call_cgraph_insertion_hooks.
6739 2018-01-18 Martin Sebor <msebor@redhat.com>
6741 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
6743 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6746 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
6749 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
6752 * common.opt: (-ffile-prefix-map): New option.
6753 * opts.c (common_handle_option): Defer it.
6754 * opts-global.c (handle_common_deferred_options): Handle it.
6755 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
6756 * file-prefix-map.h: New file.
6757 (remap_debug_filename, add_debug_prefix_map): ...here.
6758 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
6759 * final.c (debug_prefix_map, add_debug_prefix_map
6760 remap_debug_filename): Move to...
6761 * file-prefix-map.c: New file.
6762 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
6763 generalize, get rid of alloca(), use strrchr() instead of strchr().
6764 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
6765 Implement in terms of add_prefix_map().
6766 (remap_macro_filename, remap_debug_filename): Implement in term of
6768 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
6769 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
6770 * dbxout.c: Include file-prefix-map.h.
6771 * varasm.c: Likewise.
6772 * vmsdbgout.c: Likewise.
6773 * xcoffout.c: Likewise.
6774 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
6775 * doc/cppopts.texi (-fmacro-prefix-map): Document.
6776 * doc/invoke.texi (-ffile-prefix-map): Document.
6777 (-fdebug-prefix-map): Update description.
6779 2018-01-18 Martin Liska <mliska@suse.cz>
6781 * config/i386/i386.c (indirect_thunk_name): Document that also
6783 (output_indirect_thunk): Document why both instructions
6784 (pause and lfence) are generated.
6786 2018-01-18 Richard Biener <rguenther@suse.de>
6788 PR tree-optimization/83887
6789 * graphite-scop-detection.c
6790 (scop_detection::get_nearest_dom_with_single_entry): Remove.
6791 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
6792 (scop_detection::merge_sese): Re-implement with a flood-fill
6793 algorithm that properly finds a SESE region if it exists.
6795 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6798 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
6799 pointer_diff optimizations use view_convert instead of convert.
6801 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6803 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6804 Generate different code for -mno-speculate-indirect-jumps.
6805 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6806 (*call_indirect_aix<mode>): Disable for
6807 -mno-speculate-indirect-jumps.
6808 (*call_indirect_aix<mode>_nospec): New define_insn.
6809 (*call_value_indirect_aix<mode>): Disable for
6810 -mno-speculate-indirect-jumps.
6811 (*call_value_indirect_aix<mode>_nospec): New define_insn.
6812 (*sibcall_nonlocal_sysv<mode>): Generate different code for
6813 -mno-speculate-indirect-jumps.
6814 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6816 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6818 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
6819 long double type, set the flags for noting the default long double
6820 type, even if we don't pass or return a long double type.
6822 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
6825 * ipa-inline.c (flatten_function): Do not overwrite final inlining
6828 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
6830 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
6831 support for merge[hl].
6832 (fold_mergehl_helper): New helper function.
6833 (tree-vector-builder.h): New #include for tree_vector_builder usage.
6834 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
6835 (altivec_vmrglw_direct): Add xxmrglw insn.
6837 2018-01-17 Andrew Waterman <andrew@sifive.com>
6839 * config/riscv/riscv.c (riscv_conditional_register_usage): If
6840 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
6842 2018-01-17 David Malcolm <dmalcolm@redhat.com>
6845 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
6846 call the lto_location_cache before reading the
6847 DECL_SOURCE_LOCATION of the types.
6849 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
6850 Richard Sandiford <richard.sandiford@linaro.org>
6852 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
6853 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
6854 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
6855 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
6856 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
6858 * config/aarch64/constraints.md (aarch64_movti_operand):
6860 * config/aarch64/predicates.md (Uti): Add new constraint.
6862 2018-01-17 Carl Love <cel@us.ibm.com>
6864 * config/rs6000/vsx.md (define_expand xl_len_r,
6865 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
6866 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
6868 (define_expand, define_insn): Move the shift left from the
6869 define_insn to the define_expand for lxvl and stxvl instructions.
6870 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
6871 and XL_LEN_R definitions to PURE.
6873 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
6875 * config/i386/i386.c (indirect_thunk_name): Declare regno
6876 as unsigned int. Compare regno with INVALID_REGNUM.
6877 (output_indirect_thunk): Ditto.
6878 (output_indirect_thunk_function): Ditto.
6879 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
6880 in the call to output_indirect_thunk_function.
6882 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
6885 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
6886 rather than the size of inner_type to determine the stack slot size
6887 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
6889 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
6892 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
6895 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
6897 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
6898 endian Linux systems to optionally enable multilibs for selecting
6899 the long double type if the user configured an explicit type.
6900 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
6901 have no long double multilibs if not defined.
6902 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
6903 warn if the user used -mabi={ieee,ibm}longdouble and we built
6904 multilibs for long double.
6905 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
6906 appropriate multilib option.
6907 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
6909 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
6910 for building long double multilibs.
6911 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
6913 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
6915 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
6918 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
6920 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
6923 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
6926 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
6929 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6931 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
6932 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
6934 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
6936 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
6937 different rtl trees depending on TARGET_64BIT.
6938 (rs6000_gen_lvx): Likewise.
6940 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6942 * config/visium/visium.md (nop): Tweak comment.
6943 (hazard_nop): Likewise.
6945 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6947 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
6948 -mspeculate-indirect-jumps.
6949 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
6950 for -mno-speculate-indirect-jumps.
6951 (*call_indirect_elfv2<mode>_nospec): New define_insn.
6952 (*call_value_indirect_elfv2<mode>): Disable for
6953 -mno-speculate-indirect-jumps.
6954 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
6955 (indirect_jump): Emit different RTL for
6956 -mno-speculate-indirect-jumps.
6957 (*indirect_jump<mode>): Disable for
6958 -mno-speculate-indirect-jumps.
6959 (*indirect_jump<mode>_nospec): New define_insn.
6960 (tablejump): Emit different RTL for
6961 -mno-speculate-indirect-jumps.
6962 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
6963 (tablejumpsi_nospec): New define_expand.
6964 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
6965 (tablejumpdi_nospec): New define_expand.
6966 (*tablejump<mode>_internal1): Disable for
6967 -mno-speculate-indirect-jumps.
6968 (*tablejump<mode>_internal1_nospec): New define_insn.
6969 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
6972 2018-01-16 Artyom Skrobov tyomitch@gmail.com
6974 * caller-save.c (insert_save): Drop unnecessary parameter. All
6977 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6978 Richard Biener <rguenth@suse.de>
6981 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
6982 return early, inline manually is_gimple_sizepos. Make sure if we
6983 call gimplify_expr we don't end up with a gimple constant.
6984 * tree.c (variably_modified_type_p): Don't return true for
6985 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
6986 * gimplify.h (is_gimple_sizepos): Remove.
6988 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6990 PR tree-optimization/83857
6991 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
6992 vectorizable_live_operation for pure SLP statements.
6993 (vectorizable_live_operation): Handle PHIs.
6995 2018-01-16 Richard Biener <rguenther@suse.de>
6997 PR tree-optimization/83867
6998 * tree-vect-stmts.c (vect_transform_stmt): Precompute
6999 nested_in_vect_loop_p since the scalar stmt may get invalidated.
7001 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7004 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
7005 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
7006 If off is not INTEGER_CST, issue a may not be aligned warning
7007 rather than isn't aligned. Use isn%'t rather than isn't.
7008 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
7010 <case MULT_EXPR>: Improve the case when bottom and one of the
7011 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
7012 operand, in that case check if the other operand is multiple of
7013 bottom divided by the INTEGER_CST operand.
7015 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7018 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
7019 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
7020 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
7021 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
7022 * config/pa/pa.c (pa_function_arg_advance): Likewise.
7023 (pa_function_arg, pa_arg_partial_bytes): Likewise.
7024 (pa_function_arg_size): New function.
7026 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7028 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
7029 in a separate statement.
7031 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7033 PR tree-optimization/83847
7034 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
7035 group gathers and scatters.
7037 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7039 PR rtl-optimization/86620
7040 * params.def (max-sched-ready-insns): Bump minimum value to 1.
7042 PR rtl-optimization/83213
7043 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
7044 to last if both are JUMP_INSNs.
7046 PR tree-optimization/83843
7047 * gimple-ssa-store-merging.c
7048 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
7049 store_immediate_info for bswap/nop orig_stores.
7051 2018-01-15 Andrew Waterman <andrew@sifive.com>
7053 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
7055 <UDIV>: Increase cost if !TARGET_DIV.
7057 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
7059 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
7060 (define_attr "cr_logical_3op"): New.
7061 (cceq_ior_compare): Adjust.
7062 (cceq_ior_compare_complement): Adjust.
7063 (*cceq_rev_compare): Adjust.
7064 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
7065 (is_cracked_insn): Adjust.
7066 (insn_must_be_first_in_group): Adjust.
7067 * config/rs6000/40x.md: Adjust.
7068 * config/rs6000/440.md: Adjust.
7069 * config/rs6000/476.md: Adjust.
7070 * config/rs6000/601.md: Adjust.
7071 * config/rs6000/603.md: Adjust.
7072 * config/rs6000/6xx.md: Adjust.
7073 * config/rs6000/7450.md: Adjust.
7074 * config/rs6000/7xx.md: Adjust.
7075 * config/rs6000/8540.md: Adjust.
7076 * config/rs6000/cell.md: Adjust.
7077 * config/rs6000/e300c2c3.md: Adjust.
7078 * config/rs6000/e500mc.md: Adjust.
7079 * config/rs6000/e500mc64.md: Adjust.
7080 * config/rs6000/e5500.md: Adjust.
7081 * config/rs6000/e6500.md: Adjust.
7082 * config/rs6000/mpc.md: Adjust.
7083 * config/rs6000/power4.md: Adjust.
7084 * config/rs6000/power5.md: Adjust.
7085 * config/rs6000/power6.md: Adjust.
7086 * config/rs6000/power7.md: Adjust.
7087 * config/rs6000/power8.md: Adjust.
7088 * config/rs6000/power9.md: Adjust.
7089 * config/rs6000/rs64.md: Adjust.
7090 * config/rs6000/titan.md: Adjust.
7092 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7094 * config/i386/predicates.md (indirect_branch_operand): Rewrite
7095 ix86_indirect_branch_register logic.
7097 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7099 * config/i386/constraints.md (Bs): Update
7100 ix86_indirect_branch_register check. Don't check
7101 ix86_indirect_branch_register with GOT_memory_operand.
7103 * config/i386/predicates.md (GOT_memory_operand): Don't check
7104 ix86_indirect_branch_register here.
7105 (GOT32_symbol_operand): Likewise.
7107 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7109 * config/i386/predicates.md (constant_call_address_operand):
7110 Rewrite ix86_indirect_branch_register logic.
7111 (sibcall_insn_operand): Likewise.
7113 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7115 * config/i386/constraints.md (Bs): Replace
7116 ix86_indirect_branch_thunk_register with
7117 ix86_indirect_branch_register.
7119 * config/i386/i386.md (indirect_jump): Likewise.
7120 (tablejump): Likewise.
7121 (*sibcall_memory): Likewise.
7122 (*sibcall_value_memory): Likewise.
7123 Peepholes of indirect call and jump via memory: Likewise.
7124 * config/i386/i386.opt: Likewise.
7125 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7126 (GOT_memory_operand): Likewise.
7127 (call_insn_operand): Likewise.
7128 (sibcall_insn_operand): Likewise.
7129 (GOT32_symbol_operand): Likewise.
7131 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7134 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
7135 type rather than type addr's type points to.
7136 (expand_omp_atomic_mutex): Likewise.
7137 (expand_omp_atomic): Likewise.
7139 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7142 * config/i386/i386.c (output_indirect_thunk_function): Use
7143 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
7144 for __x86_return_thunk.
7146 2018-01-15 Richard Biener <rguenther@suse.de>
7149 * expmed.c (extract_bit_field_1): Fix typo.
7151 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7154 * config/arm/iterators.md (VF): New mode iterator.
7155 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
7156 Remove integer-related logic from pattern.
7157 (neon_vabd<mode>_3): Likewise.
7159 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7162 * common.opt (fstrict-overflow): No longer an alias.
7163 (fwrapv-pointer): New option.
7164 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
7165 also for pointer types based on flag_wrapv_pointer.
7166 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
7167 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
7168 opts->x_flag_wrapv got set.
7169 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
7170 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
7171 POINTER_TYPE_OVERFLOW_UNDEFINED.
7172 * match.pd: Likewise in address comparison pattern.
7173 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
7175 2018-01-15 Richard Biener <rguenther@suse.de>
7178 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
7179 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
7180 Reset type names to their identifier if their TYPE_DECL doesn't
7181 have linkage (and thus is used for ODR and devirt).
7182 (save_debug_info_for_decl): Remove.
7183 (save_debug_info_for_type): Likewise.
7184 (add_tree_to_fld_list): Adjust.
7185 * tree-pretty-print.c (dump_generic_node): Make dumping of
7186 type names more robust.
7188 2018-01-15 Richard Biener <rguenther@suse.de>
7190 * BASE-VER: Bump to 8.0.1.
7192 2018-01-14 Martin Sebor <msebor@redhat.com>
7195 * builtins.c (check_access): Avoid warning when the no-warning bit
7198 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
7200 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
7201 * ira-color (allocno_hard_regs_compare): Likewise.
7203 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
7206 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
7207 Use .pushsection/.popsection.
7209 2018-01-14 Martin Sebor <msebor@redhat.com>
7212 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
7214 2018-01-14 Jakub Jelinek <jakub@redhat.com>
7216 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
7217 entry from extra_headers.
7218 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
7219 extra_headers, make the list bitwise identical to the i?86-*-* one.
7221 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7223 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
7224 -mcmodel=large with -mindirect-branch=thunk,
7225 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
7226 -mfunction-return=thunk-extern.
7227 * doc/invoke.texi: Document -mcmodel=large is incompatible with
7228 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
7229 -mfunction-return=thunk and -mfunction-return=thunk-extern.
7231 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7233 * config/i386/i386.c (print_reg): Print the name of the full
7234 integer register without '%'.
7235 (ix86_print_operand): Handle 'V'.
7236 * doc/extend.texi: Document 'V' modifier.
7238 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7240 * config/i386/constraints.md (Bs): Disallow memory operand for
7241 -mindirect-branch-register.
7243 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7244 (GOT_memory_operand): Likewise.
7245 (call_insn_operand): Likewise.
7246 (sibcall_insn_operand): Likewise.
7247 (GOT32_symbol_operand): Likewise.
7248 * config/i386/i386.md (indirect_jump): Call convert_memory_address
7249 for -mindirect-branch-register.
7250 (tablejump): Likewise.
7251 (*sibcall_memory): Likewise.
7252 (*sibcall_value_memory): Likewise.
7253 Disallow peepholes of indirect call and jump via memory for
7254 -mindirect-branch-register.
7255 (*call_pop): Replace m with Bw.
7256 (*call_value_pop): Likewise.
7257 (*sibcall_pop_memory): Replace m with Bs.
7258 * config/i386/i386.opt (mindirect-branch-register): New option.
7259 * doc/invoke.texi: Document -mindirect-branch-register option.
7261 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7263 * config/i386/i386-protos.h (ix86_output_function_return): New.
7264 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
7265 set function_return_type.
7266 (indirect_thunk_name): Add ret_p to indicate thunk for function
7268 (output_indirect_thunk_function): Pass false to
7269 indirect_thunk_name.
7270 (ix86_output_indirect_branch_via_reg): Likewise.
7271 (ix86_output_indirect_branch_via_push): Likewise.
7272 (output_indirect_thunk_function): Create alias for function
7273 return thunk if regno < 0.
7274 (ix86_output_function_return): New function.
7275 (ix86_handle_fndecl_attribute): Handle function_return.
7276 (ix86_attribute_table): Add function_return.
7277 * config/i386/i386.h (machine_function): Add
7278 function_return_type.
7279 * config/i386/i386.md (simple_return_internal): Use
7280 ix86_output_function_return.
7281 (simple_return_internal_long): Likewise.
7282 * config/i386/i386.opt (mfunction-return=): New option.
7283 (indirect_branch): Mention -mfunction-return=.
7284 * doc/extend.texi: Document function_return function attribute.
7285 * doc/invoke.texi: Document -mfunction-return= option.
7287 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7289 * config/i386/i386-opts.h (indirect_branch): New.
7290 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
7291 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
7292 with local indirect jump when converting indirect call and jump.
7293 (ix86_set_indirect_branch_type): New.
7294 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
7295 (indirectlabelno): New.
7296 (indirect_thunk_needed): Likewise.
7297 (indirect_thunk_bnd_needed): Likewise.
7298 (indirect_thunks_used): Likewise.
7299 (indirect_thunks_bnd_used): Likewise.
7300 (INDIRECT_LABEL): Likewise.
7301 (indirect_thunk_name): Likewise.
7302 (output_indirect_thunk): Likewise.
7303 (output_indirect_thunk_function): Likewise.
7304 (ix86_output_indirect_branch_via_reg): Likewise.
7305 (ix86_output_indirect_branch_via_push): Likewise.
7306 (ix86_output_indirect_branch): Likewise.
7307 (ix86_output_indirect_jmp): Likewise.
7308 (ix86_code_end): Call output_indirect_thunk_function if needed.
7309 (ix86_output_call_insn): Call ix86_output_indirect_branch if
7311 (ix86_handle_fndecl_attribute): Handle indirect_branch.
7312 (ix86_attribute_table): Add indirect_branch.
7313 * config/i386/i386.h (machine_function): Add indirect_branch_type
7314 and has_local_indirect_jump.
7315 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
7317 (tablejump): Likewise.
7318 (*indirect_jump): Use ix86_output_indirect_jmp.
7319 (*tablejump_1): Likewise.
7320 (simple_return_indirect_internal): Likewise.
7321 * config/i386/i386.opt (mindirect-branch=): New option.
7322 (indirect_branch): New.
7325 (thunk-inline): Likewise.
7326 (thunk-extern): Likewise.
7327 * doc/extend.texi: Document indirect_branch function attribute.
7328 * doc/invoke.texi: Document -mindirect-branch= option.
7330 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
7333 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
7335 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
7337 * ipa-inline.c (want_inline_small_function_p): Return false if
7338 inlining has already failed with CIF_FINAL_ERROR.
7339 (update_caller_keys): Call want_inline_small_function_p before
7341 (update_callee_keys): Likewise.
7343 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7345 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
7347 (rs6000_quadword_masked_address_p): Likewise.
7348 (quad_aligned_load_p): Likewise.
7349 (quad_aligned_store_p): Likewise.
7350 (const_load_sequence_p): Add comment to describe the outer-most loop.
7351 (mimic_memory_attributes_and_flags): New function.
7352 (rs6000_gen_stvx): Likewise.
7353 (replace_swapped_aligned_store): Likewise.
7354 (rs6000_gen_lvx): Likewise.
7355 (replace_swapped_aligned_load): Likewise.
7356 (replace_swapped_load_constant): Capitalize argument name in
7357 comment describing this function.
7358 (rs6000_analyze_swaps): Add a third pass to search for vector loads
7359 and stores that access quad-word aligned addresses and replace
7360 with stvx or lvx instructions when appropriate.
7361 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
7362 New function prototype.
7363 (rs6000_quadword_masked_address_p): Likewise.
7364 (rs6000_gen_lvx): Likewise.
7365 (rs6000_gen_stvx): Likewise.
7366 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
7367 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
7368 when memory address is aligned.
7369 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
7370 this split to select lvx instruction when memory address is aligned.
7371 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
7372 instruction when memory address is aligned.
7373 (*vsx_le_perm_load_v16qi): Likewise.
7374 (four unnamed splitters): Modify to select the stvx instruction
7375 when memory is aligned.
7377 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
7379 * predict.c (determine_unlikely_bbs): Handle correctly BBs
7380 which appears in the queue multiple times.
7382 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7383 Alan Hayward <alan.hayward@arm.com>
7384 David Sherwood <david.sherwood@arm.com>
7386 * tree-vectorizer.h (vec_lower_bound): New structure.
7387 (_loop_vec_info): Add check_nonzero and lower_bounds.
7388 (LOOP_VINFO_CHECK_NONZERO): New macro.
7389 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
7390 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
7391 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
7392 fields. Make seg_len the distance travelled, not including the
7394 (dr_direction_indicator): Declare.
7395 (dr_zero_step_indicator): Likewise.
7396 (dr_known_forward_stride_p): Likewise.
7397 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
7399 (runtime_alias_check_p): Allow runtime alias checks with
7401 (operator ==): Compare access_size and align.
7402 (prune_runtime_alias_test_list): Rework for new distinction between
7403 the access_size and seg_len.
7404 (create_intersect_range_checks_index): Likewise. Cope with polynomial
7406 (get_segment_min_max): New function.
7407 (create_intersect_range_checks): Use it.
7408 (dr_step_indicator): New function.
7409 (dr_direction_indicator): Likewise.
7410 (dr_zero_step_indicator): Likewise.
7411 (dr_known_forward_stride_p): Likewise.
7412 * tree-loop-distribution.c (data_ref_segment_size): Return
7413 DR_STEP * (niters - 1).
7414 (compute_alias_check_pairs): Update call to the dr_with_seg_len
7416 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
7417 (vect_preserves_scalar_order_p): New function, split out from...
7418 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
7419 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
7420 (vect_vfa_access_size): New function.
7421 (vect_vfa_align): Likewise.
7422 (vect_compile_time_alias): Take access_size_a and access_b arguments.
7423 (dump_lower_bound): New function.
7424 (vect_check_lower_bound): Likewise.
7425 (vect_small_gap_p): Likewise.
7426 (vectorizable_with_step_bound_p): Likewise.
7427 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
7428 depencies if the vectorization factor is 1. Convert the checks
7429 for nonzero steps into checks on the bounds of DR_STEP. Try using
7430 a bunds check for variable steps if the minimum required step is
7431 relatively small. Update calls to the dr_with_seg_len
7432 constructor and to vect_compile_time_alias.
7433 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
7435 (vect_loop_versioning): Call it.
7436 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
7438 (vect_estimate_min_profitable_iters): Account for any bounds checks.
7440 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7441 Alan Hayward <alan.hayward@arm.com>
7442 David Sherwood <david.sherwood@arm.com>
7444 * doc/sourcebuild.texi (vect_scatter_store): Document.
7445 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
7447 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
7449 * genopinit.c (main): Add supports_vec_scatter_store and
7450 supports_vec_scatter_store_cached to target_optabs.
7451 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
7452 IFN_MASK_SCATTER_STORE.
7453 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
7455 * internal-fn.h (internal_store_fn_p): Declare.
7456 (internal_fn_stored_value_index): Likewise.
7457 * internal-fn.c (scatter_store_direct): New macro.
7458 (expand_scatter_store_optab_fn): New function.
7459 (direct_scatter_store_optab_supported_p): New macro.
7460 (internal_store_fn_p): New function.
7461 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
7462 IFN_MASK_SCATTER_STORE.
7463 (internal_fn_mask_index): Likewise.
7464 (internal_fn_stored_value_index): New function.
7465 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
7467 * optabs-query.h (supports_vec_scatter_store_p): Declare.
7468 * optabs-query.c (supports_vec_scatter_store_p): New function.
7469 * tree-vectorizer.h (vect_get_store_rhs): Declare.
7470 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
7471 true for scatter stores.
7472 (vect_gather_scatter_fn_p): Handle scatter stores too.
7473 (vect_check_gather_scatter): Consider using scatter stores if
7474 supports_vec_scatter_store_p.
7475 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
7477 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7478 internal_fn_stored_value_index.
7479 (check_load_store_masking): Handle scatter stores too.
7480 (vect_get_store_rhs): Make public.
7481 (vectorizable_call): Use internal_store_fn_p.
7482 (vectorizable_store): Handle scatter store internal functions.
7483 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
7484 when deciding whether the end of the group has been reached.
7485 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
7486 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
7487 (mask_scatter_store<mode>): New insns.
7489 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7490 Alan Hayward <alan.hayward@arm.com>
7491 David Sherwood <david.sherwood@arm.com>
7493 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
7494 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
7495 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
7497 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
7498 Use vect_truncate_gather_scatter_offset if we can't treat the
7499 operation as a normal gather load or scatter store.
7500 (get_group_load_store_type): Take the gather_scatter_info
7501 as argument. Try using a gather load or scatter store for
7502 single-element groups.
7503 (get_load_store_type): Update calls to get_group_load_store_type
7504 and vect_use_strided_gather_scatters_p.
7506 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7507 Alan Hayward <alan.hayward@arm.com>
7508 David Sherwood <david.sherwood@arm.com>
7510 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
7511 optional tree argument.
7512 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
7514 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
7515 but continue to use the current value as a fallback.
7516 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
7517 to compare the updates.
7518 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
7519 (get_load_store_type): Use it when handling a strided access.
7520 (vect_get_strided_load_store_ops): New function.
7521 (vect_get_data_ptr_increment): Likewise.
7522 (vectorizable_load): Handle strided gather loads. Always pass
7523 a step to vect_create_data_ref_ptr and bump_vector_ptr.
7525 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7526 Alan Hayward <alan.hayward@arm.com>
7527 David Sherwood <david.sherwood@arm.com>
7529 * doc/md.texi (gather_load@var{m}): Document.
7530 (mask_gather_load@var{m}): Likewise.
7531 * genopinit.c (main): Add supports_vec_gather_load and
7532 supports_vec_gather_load_cached to target_optabs.
7533 * optabs-tree.c (init_tree_optimization_optabs): Use
7534 ggc_cleared_alloc to allocate target_optabs.
7535 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
7536 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
7538 * internal-fn.h (internal_load_fn_p): Declare.
7539 (internal_gather_scatter_fn_p): Likewise.
7540 (internal_fn_mask_index): Likewise.
7541 (internal_gather_scatter_fn_supported_p): Likewise.
7542 * internal-fn.c (gather_load_direct): New macro.
7543 (expand_gather_load_optab_fn): New function.
7544 (direct_gather_load_optab_supported_p): New macro.
7545 (direct_internal_fn_optab): New function.
7546 (internal_load_fn_p): Likewise.
7547 (internal_gather_scatter_fn_p): Likewise.
7548 (internal_fn_mask_index): Likewise.
7549 (internal_gather_scatter_fn_supported_p): Likewise.
7550 * optabs-query.c (supports_at_least_one_mode_p): New function.
7551 (supports_vec_gather_load_p): Likewise.
7552 * optabs-query.h (supports_vec_gather_load_p): Declare.
7553 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
7554 and memory_type field.
7555 (NUM_PATTERNS): Bump to 15.
7556 * tree-vect-data-refs.c: Include internal-fn.h.
7557 (vect_gather_scatter_fn_p): New function.
7558 (vect_describe_gather_scatter_call): Likewise.
7559 (vect_check_gather_scatter): Try using internal functions for
7560 gather loads. Recognize existing calls to a gather load function.
7561 (vect_analyze_data_refs): Consider using gather loads if
7562 supports_vec_gather_load_p.
7563 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
7564 (vect_get_gather_scatter_offset_type): Likewise.
7565 (vect_convert_mask_for_vectype): Likewise.
7566 (vect_add_conversion_to_patterm): Likewise.
7567 (vect_try_gather_scatter_pattern): Likewise.
7568 (vect_recog_gather_scatter_pattern): New pattern recognizer.
7569 (vect_vect_recog_func_ptrs): Add it.
7570 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7571 internal_fn_mask_index and internal_gather_scatter_fn_p.
7572 (check_load_store_masking): Take the gather_scatter_info as an
7573 argument and handle gather loads.
7574 (vect_get_gather_scatter_ops): New function.
7575 (vectorizable_call): Check internal_load_fn_p.
7576 (vectorizable_load): Likewise. Handle gather load internal
7578 (vectorizable_store): Update call to check_load_store_masking.
7579 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
7580 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
7581 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
7582 (aarch64_gather_scale_operand_d): New predicates.
7583 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
7584 (mask_gather_load<mode>): New insns.
7586 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7587 Alan Hayward <alan.hayward@arm.com>
7588 David Sherwood <david.sherwood@arm.com>
7590 * optabs.def (fold_left_plus_optab): New optab.
7591 * doc/md.texi (fold_left_plus_@var{m}): Document.
7592 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
7593 * internal-fn.c (fold_left_direct): Define.
7594 (expand_fold_left_optab_fn): Likewise.
7595 (direct_fold_left_optab_supported_p): Likewise.
7596 * fold-const-call.c (fold_const_fold_left): New function.
7597 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
7598 * tree-parloops.c (valid_reduction_p): New function.
7599 (gather_scalar_reductions): Use it.
7600 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
7601 (vect_finish_replace_stmt): Declare.
7602 * tree-vect-loop.c (fold_left_reduction_fn): New function.
7603 (needs_fold_left_reduction_p): New function, split out from...
7604 (vect_is_simple_reduction): ...here. Accept reductions that
7605 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
7606 (vect_force_simple_reduction): Also store the reduction type in
7607 the assignment's STMT_VINFO_REDUC_TYPE.
7608 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
7609 (merge_with_identity): New function.
7610 (vect_expand_fold_left): Likewise.
7611 (vectorize_fold_left_reduction): Likewise.
7612 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
7613 scalar phi in place for it. Check for target support and reject
7614 cases that would reassociate the operation. Defer the transform
7615 phase to vectorize_fold_left_reduction.
7616 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
7617 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
7618 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
7620 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7622 * tree-if-conv.c (predicate_mem_writes): Remove redundant
7623 call to ifc_temp_var.
7625 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7626 Alan Hayward <alan.hayward@arm.com>
7627 David Sherwood <david.sherwood@arm.com>
7629 * target.def (legitimize_address_displacement): Take the original
7630 offset as a poly_int.
7631 * targhooks.h (default_legitimize_address_displacement): Update
7633 * targhooks.c (default_legitimize_address_displacement): Likewise.
7634 * doc/tm.texi: Regenerate.
7635 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
7636 as an argument, moving assert of ad->disp == ad->disp_term to...
7637 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
7638 Try calling targetm.legitimize_address_displacement before expanding
7639 the address rather than afterwards, and adjust for the new interface.
7640 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
7641 Match the new hook interface. Handle SVE addresses.
7642 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
7645 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7647 * Makefile.in (OBJS): Add early-remat.o.
7648 * target.def (select_early_remat_modes): New hook.
7649 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
7650 * doc/tm.texi: Regenerate.
7651 * targhooks.h (default_select_early_remat_modes): Declare.
7652 * targhooks.c (default_select_early_remat_modes): New function.
7653 * timevar.def (TV_EARLY_REMAT): New timevar.
7654 * passes.def (pass_early_remat): New pass.
7655 * tree-pass.h (make_pass_early_remat): Declare.
7656 * early-remat.c: New file.
7657 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
7659 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
7661 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7662 Alan Hayward <alan.hayward@arm.com>
7663 David Sherwood <david.sherwood@arm.com>
7665 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
7666 vfm1 with a bound_epilog parameter.
7667 (vect_do_peeling): Update calls accordingly, and move the prologue
7668 call earlier in the function. Treat the base bound_epilog as 0 for
7669 fully-masked loops and retain vf - 1 for other loops. Add 1 to
7670 this base when peeling for gaps.
7671 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
7672 with fully-masked loops.
7673 (vect_estimate_min_profitable_iters): Handle the single peeled
7674 iteration in that case.
7676 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7677 Alan Hayward <alan.hayward@arm.com>
7678 David Sherwood <david.sherwood@arm.com>
7680 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
7681 single-element interleaving even if the size is not a power of 2.
7682 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
7683 accesses for single-element interleaving if the group size is
7686 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7687 Alan Hayward <alan.hayward@arm.com>
7688 David Sherwood <david.sherwood@arm.com>
7690 * doc/md.texi (fold_extract_last_@var{m}): Document.
7691 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
7692 * optabs.def (fold_extract_last_optab): New optab.
7693 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
7694 * internal-fn.c (fold_extract_direct): New macro.
7695 (expand_fold_extract_optab_fn): Likewise.
7696 (direct_fold_extract_optab_supported_p): Likewise.
7697 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
7698 * tree-vect-loop.c (vect_model_reduction_cost): Handle
7699 EXTRACT_LAST_REDUCTION.
7700 (get_initial_def_for_reduction): Do not create an initial vector
7701 for EXTRACT_LAST_REDUCTION reductions.
7702 (vectorizable_reduction): Leave the scalar phi in place for
7703 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
7704 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
7705 epilogue code for EXTRACT_LAST_REDUCTION and defer the
7706 transform phase to vectorizable_condition.
7707 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
7709 (vect_finish_stmt_generation): ...here.
7710 (vect_finish_replace_stmt): New function.
7711 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
7712 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
7714 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
7716 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7717 Alan Hayward <alan.hayward@arm.com>
7718 David Sherwood <david.sherwood@arm.com>
7720 * doc/md.texi (extract_last_@var{m}): Document.
7721 * optabs.def (extract_last_optab): New optab.
7722 * internal-fn.def (EXTRACT_LAST): New internal function.
7723 * internal-fn.c (cond_unary_direct): New macro.
7724 (expand_cond_unary_optab_fn): Likewise.
7725 (direct_cond_unary_optab_supported_p): Likewise.
7726 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
7727 loops using EXTRACT_LAST.
7728 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
7729 (extract_last_<mode>): ...this optab.
7730 (vec_extract<mode><Vel>): Update accordingly.
7732 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7733 Alan Hayward <alan.hayward@arm.com>
7734 David Sherwood <david.sherwood@arm.com>
7736 * target.def (empty_mask_is_expensive): New hook.
7737 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
7738 * doc/tm.texi: Regenerate.
7739 * targhooks.h (default_empty_mask_is_expensive): Declare.
7740 * targhooks.c (default_empty_mask_is_expensive): New function.
7741 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
7742 if the target says that empty masks are expensive.
7743 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
7745 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
7747 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7748 Alan Hayward <alan.hayward@arm.com>
7749 David Sherwood <david.sherwood@arm.com>
7751 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
7752 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
7753 (vect_use_loop_mask_for_alignment_p): New function.
7754 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
7755 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
7756 niters_skip argument. Make sure that the first niters_skip elements
7757 of the first iteration are inactive.
7758 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
7759 Update call to vect_set_loop_masks_directly.
7760 (get_misalign_in_elems): New function, split out from...
7761 (vect_gen_prolog_loop_niters): ...here.
7762 (vect_update_init_of_dr): Take a code argument that specifies whether
7763 the adjustment should be added or subtracted.
7764 (vect_update_init_of_drs): Likewise.
7765 (vect_prepare_for_masked_peels): New function.
7766 (vect_do_peeling): Skip prologue peeling if we're using a mask
7767 instead. Update call to vect_update_inits_of_drs.
7768 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7770 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
7771 alignment. Do not include the number of peeled iterations in
7772 the minimum threshold in that case.
7773 (vectorizable_induction): Adjust the start value down by
7774 LOOP_VINFO_MASK_SKIP_NITERS iterations.
7775 (vect_transform_loop): Call vect_prepare_for_masked_peels.
7776 Take the number of skipped iterations into account when calculating
7778 * tree-vect-stmts.c (vect_gen_while_not): New function.
7780 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7781 Alan Hayward <alan.hayward@arm.com>
7782 David Sherwood <david.sherwood@arm.com>
7784 * doc/sourcebuild.texi (vect_fully_masked): Document.
7785 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
7787 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
7789 (vect_analyze_loop_2): ...here. Don't check the vectorization
7790 factor against the number of loop iterations if the loop is
7793 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7794 Alan Hayward <alan.hayward@arm.com>
7795 David Sherwood <david.sherwood@arm.com>
7797 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
7798 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
7799 (dump_groups): Update accordingly.
7800 (iv_use::mem_type): New member variable.
7801 (address_p): New function.
7802 (record_use): Add a mem_type argument and initialize the new
7804 (record_group_use): Add a mem_type argument. Use address_p.
7805 Remove obsolete null checks of base_object. Update call to record_use.
7806 (find_interesting_uses_op): Update call to record_group_use.
7807 (find_interesting_uses_cond): Likewise.
7808 (find_interesting_uses_address): Likewise.
7809 (get_mem_type_for_internal_fn): New function.
7810 (find_address_like_use): Likewise.
7811 (find_interesting_uses_stmt): Try find_address_like_use before
7812 calling find_interesting_uses_op.
7813 (addr_offset_valid_p): Use the iv mem_type field as the type
7814 of the addressed memory.
7815 (add_autoinc_candidates): Likewise.
7816 (get_address_cost): Likewise.
7817 (split_small_address_groups_p): Use address_p.
7818 (split_address_groups): Likewise.
7819 (add_iv_candidate_for_use): Likewise.
7820 (autoinc_possible_for_pair): Likewise.
7821 (rewrite_groups): Likewise.
7822 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
7823 (determine_group_iv_cost): Update after split of USE_ADDRESS.
7824 (get_alias_ptr_type_for_ptr_address): New function.
7825 (rewrite_use_address): Rewrite address uses in calls that were
7826 identified by find_address_like_use.
7828 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7829 Alan Hayward <alan.hayward@arm.com>
7830 David Sherwood <david.sherwood@arm.com>
7832 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
7834 * gimple-expr.h (is_gimple_addressable: Likewise.
7835 * gimple-expr.c (is_gimple_address): Likewise.
7836 * internal-fn.c (expand_call_mem_ref): New function.
7837 (expand_mask_load_optab_fn): Use it.
7838 (expand_mask_store_optab_fn): Likewise.
7840 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7841 Alan Hayward <alan.hayward@arm.com>
7842 David Sherwood <david.sherwood@arm.com>
7844 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
7845 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
7846 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
7847 (cond_umax@var{mode}): Document.
7848 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
7849 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
7850 (cond_umin_optab, cond_umax_optab): New optabs.
7851 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
7852 (COND_IOR, COND_XOR): New internal functions.
7853 * internal-fn.h (get_conditional_internal_fn): Declare.
7854 * internal-fn.c (cond_binary_direct): New macro.
7855 (expand_cond_binary_optab_fn): Likewise.
7856 (direct_cond_binary_optab_supported_p): Likewise.
7857 (get_conditional_internal_fn): New function.
7858 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
7859 Cope with reduction statements that are vectorized as calls rather
7861 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
7862 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
7863 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
7864 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
7865 (UNSPEC_COND_EOR): New unspecs.
7866 (optab): Add mappings for them.
7867 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
7868 (sve_int_op, sve_fp_op): New int attributes.
7870 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7871 Alan Hayward <alan.hayward@arm.com>
7872 David Sherwood <david.sherwood@arm.com>
7874 * optabs.def (while_ult_optab): New optab.
7875 * doc/md.texi (while_ult@var{m}@var{n}): Document.
7876 * internal-fn.def (WHILE_ULT): New internal function.
7877 * internal-fn.h (direct_internal_fn_supported_p): New override
7878 that takes two types as argument.
7879 * internal-fn.c (while_direct): New macro.
7880 (expand_while_optab_fn): New function.
7881 (convert_optab_supported_p): Likewise.
7882 (direct_while_optab_supported_p): New macro.
7883 * wide-int.h (wi::udiv_ceil): New function.
7884 * tree-vectorizer.h (rgroup_masks): New structure.
7885 (vec_loop_masks): New typedef.
7886 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
7888 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
7889 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
7890 (vect_max_vf): New function.
7891 (slpeel_make_loop_iterate_ntimes): Delete.
7892 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
7893 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
7894 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
7895 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
7896 internal-fn.h, stor-layout.h and optabs-query.h.
7897 (vect_set_loop_mask): New function.
7898 (add_preheader_seq): Likewise.
7899 (add_header_seq): Likewise.
7900 (interleave_supported_p): Likewise.
7901 (vect_maybe_permute_loop_masks): Likewise.
7902 (vect_set_loop_masks_directly): Likewise.
7903 (vect_set_loop_condition_masked): Likewise.
7904 (vect_set_loop_condition_unmasked): New function, split out from
7905 slpeel_make_loop_iterate_ntimes.
7906 (slpeel_make_loop_iterate_ntimes): Rename to..
7907 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
7908 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
7909 (vect_do_peeling): Update call accordingly.
7910 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
7912 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7913 mask_compare_type, can_fully_mask_p and fully_masked_p.
7914 (release_vec_loop_masks): New function.
7915 (_loop_vec_info): Use it to free the loop masks.
7916 (can_produce_all_loop_masks_p): New function.
7917 (vect_get_max_nscalars_per_iter): Likewise.
7918 (vect_verify_full_masking): Likewise.
7919 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
7920 retries, and free the mask rgroups before retrying. Check loop-wide
7921 reasons for disallowing fully-masked loops. Make the final decision
7922 about whether use a fully-masked loop or not.
7923 (vect_estimate_min_profitable_iters): Do not assume that peeling
7924 for the number of iterations will be needed for fully-masked loops.
7925 (vectorizable_reduction): Disable fully-masked loops.
7926 (vectorizable_live_operation): Likewise.
7927 (vect_halve_mask_nunits): New function.
7928 (vect_double_mask_nunits): Likewise.
7929 (vect_record_loop_mask): Likewise.
7930 (vect_get_loop_mask): Likewise.
7931 (vect_transform_loop): Handle the case in which the final loop
7932 iteration might handle a partial vector. Call vect_set_loop_condition
7933 instead of slpeel_make_loop_iterate_ntimes.
7934 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
7935 (check_load_store_masking): New function.
7936 (prepare_load_store_mask): Likewise.
7937 (vectorizable_store): Handle fully-masked loops.
7938 (vectorizable_load): Likewise.
7939 (supportable_widening_operation): Use vect_halve_mask_nunits for
7941 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
7942 (vect_gen_while): New function.
7943 * config/aarch64/aarch64.md (umax<mode>3): New expander.
7944 (aarch64_uqdec<mode>): New insn.
7946 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7947 Alan Hayward <alan.hayward@arm.com>
7948 David Sherwood <david.sherwood@arm.com>
7950 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
7951 (reduc_xor_scal_optab): New optabs.
7952 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
7953 (reduc_xor_scal_@var{m}): Document.
7954 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
7955 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
7957 * fold-const-call.c (fold_const_call): Handle them.
7958 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
7959 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
7960 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
7961 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
7962 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
7963 (UNSPEC_XORV): New unspecs.
7964 (optab): Add entries for them.
7965 (BITWISEV): New int iterator.
7966 (bit_reduc_op): New int attributes.
7968 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7969 Alan Hayward <alan.hayward@arm.com>
7970 David Sherwood <david.sherwood@arm.com>
7972 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
7973 * internal-fn.def (VEC_SHL_INSERT): New internal function.
7974 * optabs.def (vec_shl_insert_optab): New optab.
7975 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
7976 (duplicate_and_interleave): Likewise.
7977 * tree-vect-loop.c: Include internal-fn.h.
7978 (neutral_op_for_slp_reduction): New function, split out from
7979 get_initial_defs_for_reduction.
7980 (get_initial_def_for_reduction): Handle option 2 for variable-length
7981 vectors by loading the neutral value into a vector and then shifting
7982 the initial value into element 0.
7983 (get_initial_defs_for_reduction): Replace the code argument with
7984 the neutral value calculated by neutral_op_for_slp_reduction.
7985 Use gimple_build_vector for constant-length vectors.
7986 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
7987 but the first group_size elements have a neutral value.
7988 Use duplicate_and_interleave otherwise.
7989 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
7990 Update call to get_initial_defs_for_reduction. Handle SLP
7991 reductions for variable-length vectors by creating one vector
7992 result for each scalar result, with the elements associated
7993 with other scalar results stubbed out with the neutral value.
7994 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
7995 Require IFN_VEC_SHL_INSERT for double reductions on
7996 variable-length vectors, or SLP reductions that have
7997 a neutral value. Require can_duplicate_and_interleave_p
7998 support for variable-length unchained SLP reductions if there
7999 is no neutral value, such as for MIN/MAX reductions. Also require
8000 the number of vector elements to be a multiple of the number of
8001 SLP statements when doing variable-length unchained SLP reductions.
8002 Update call to vect_create_epilog_for_reduction.
8003 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
8004 and remove initial values.
8005 (duplicate_and_interleave): Make public.
8006 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
8007 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
8009 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8010 Alan Hayward <alan.hayward@arm.com>
8011 David Sherwood <david.sherwood@arm.com>
8013 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
8014 (can_duplicate_and_interleave_p): New function.
8015 (vect_get_and_check_slp_defs): Take the vector of statements
8016 rather than just the current one. Remove excess parentheses.
8017 Restriction rejectinon of vect_constant_def and vect_external_def
8018 for variable-length vectors to boolean types, or types for which
8019 can_duplicate_and_interleave_p is false.
8020 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
8021 (duplicate_and_interleave): New function.
8022 (vect_get_constant_vectors): Use gimple_build_vector for
8023 constant-length vectors and suitable variable-length constant
8024 vectors. Use duplicate_and_interleave for other variable-length
8025 vectors. Don't defer the update when inserting new statements.
8027 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8028 Alan Hayward <alan.hayward@arm.com>
8029 David Sherwood <david.sherwood@arm.com>
8031 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
8032 min_profitable_iters doesn't go negative.
8034 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8035 Alan Hayward <alan.hayward@arm.com>
8036 David Sherwood <david.sherwood@arm.com>
8038 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
8039 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
8040 * optabs.def (vec_mask_load_lanes_optab): New optab.
8041 (vec_mask_store_lanes_optab): Likewise.
8042 * internal-fn.def (MASK_LOAD_LANES): New internal function.
8043 (MASK_STORE_LANES): Likewise.
8044 * internal-fn.c (mask_load_lanes_direct): New macro.
8045 (mask_store_lanes_direct): Likewise.
8046 (expand_mask_load_optab_fn): Handle masked operations.
8047 (expand_mask_load_lanes_optab_fn): New macro.
8048 (expand_mask_store_optab_fn): Handle masked operations.
8049 (expand_mask_store_lanes_optab_fn): New macro.
8050 (direct_mask_load_lanes_optab_supported_p): Likewise.
8051 (direct_mask_store_lanes_optab_supported_p): Likewise.
8052 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
8054 (vect_load_lanes_supported): Likewise.
8055 * tree-vect-data-refs.c (strip_conversion): New function.
8056 (can_group_stmts_p): Likewise.
8057 (vect_analyze_data_ref_accesses): Use it instead of checking
8058 for a pair of assignments.
8059 (vect_store_lanes_supported): Take a masked_p parameter.
8060 (vect_load_lanes_supported): Likewise.
8061 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
8062 vect_store_lanes_supported and vect_load_lanes_supported.
8063 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8064 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
8065 parameter. Don't allow gaps for masked accesses.
8066 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
8067 and vect_load_lanes_supported.
8068 (get_load_store_type): Take a masked_p parameter and update
8069 call to get_group_load_store_type.
8070 (vectorizable_store): Update call to get_load_store_type.
8071 Handle IFN_MASK_STORE_LANES.
8072 (vectorizable_load): Update call to get_load_store_type.
8073 Handle IFN_MASK_LOAD_LANES.
8075 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8076 Alan Hayward <alan.hayward@arm.com>
8077 David Sherwood <david.sherwood@arm.com>
8079 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
8081 * config/aarch64/aarch64-protos.h
8082 (aarch64_sve_struct_memory_operand_p): Declare.
8083 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
8084 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
8085 (VPRED, vpred): Handle SVE structure modes.
8086 * config/aarch64/constraints.md (Utx): New constraint.
8087 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
8088 (aarch64_sve_struct_nonimmediate_operand): New predicates.
8089 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
8090 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
8091 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
8092 structure modes. Split into pieces after RA.
8093 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
8094 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
8096 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
8097 SVE structure modes.
8098 (aarch64_classify_address): Likewise.
8099 (sizetochar): Move earlier in file.
8100 (aarch64_print_operand): Handle SVE register lists.
8101 (aarch64_array_mode): New function.
8102 (aarch64_sve_struct_memory_operand_p): Likewise.
8103 (TARGET_ARRAY_MODE): Redefine.
8105 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8106 Alan Hayward <alan.hayward@arm.com>
8107 David Sherwood <david.sherwood@arm.com>
8109 * target.def (array_mode): New target hook.
8110 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
8111 * doc/tm.texi: Regenerate.
8112 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
8113 * hooks.c (hook_optmode_mode_uhwi_none): New function.
8114 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
8116 * stor-layout.c (mode_for_array): Likewise. Support polynomial
8119 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8120 Alan Hayward <alan.hayward@arm.com>
8121 David Sherwood <david.sherwood@arm.com>
8123 * fold-const.c (fold_binary_loc): Check the argument types
8124 rather than the result type when testing for a vector operation.
8126 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8128 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
8129 * doc/tm.texi: Regenerate.
8131 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8132 Alan Hayward <alan.hayward@arm.com>
8133 David Sherwood <david.sherwood@arm.com>
8135 * doc/invoke.texi (-msve-vector-bits=): Document new option.
8136 (sve): Document new AArch64 extension.
8137 * doc/md.texi (w): Extend the description of the AArch64
8138 constraint to include SVE vectors.
8139 (Upl, Upa): Document new AArch64 predicate constraints.
8140 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
8142 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
8143 (msve-vector-bits=): New option.
8144 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
8145 SVE when these are disabled.
8146 (sve): New extension.
8147 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
8148 modes. Adjust their number of units based on aarch64_sve_vg.
8149 (MAX_BITSIZE_MODE_ANY_MODE): Define.
8150 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
8151 aarch64_addr_query_type.
8152 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
8153 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
8154 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
8155 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
8156 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
8157 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
8158 (aarch64_simd_imm_zero_p): Delete.
8159 (aarch64_check_zero_based_sve_index_immediate): Declare.
8160 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8161 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8162 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8163 (aarch64_sve_float_mul_immediate_p): Likewise.
8164 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8166 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
8167 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
8168 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
8169 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
8170 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
8171 (aarch64_regmode_natural_size): Likewise.
8172 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
8173 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
8175 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
8176 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
8177 for VG and the SVE predicate registers.
8178 (V_ALIASES): Add a "z"-prefixed alias.
8179 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
8180 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
8181 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
8182 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
8183 (REG_CLASS_NAMES): Add entries for them.
8184 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
8185 and the predicate registers.
8186 (aarch64_sve_vg): Declare.
8187 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
8188 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
8189 (REGMODE_NATURAL_SIZE): Define.
8190 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
8192 * config/aarch64/aarch64.c: Include cfgrtl.h.
8193 (simd_immediate_info): Add a constructor for series vectors,
8194 and an associated step field.
8195 (aarch64_sve_vg): New variable.
8196 (aarch64_dbx_register_number): Handle VG and the predicate registers.
8197 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
8198 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
8199 (VEC_ANY_DATA, VEC_STRUCT): New constants.
8200 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
8201 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
8202 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
8203 (aarch64_get_mask_mode): New functions.
8204 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
8205 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
8206 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
8207 predicate modes and predicate registers. Explicitly restrict
8208 GPRs to modes of 16 bytes or smaller. Only allow FP registers
8209 to store a vector mode if it is recognized by
8210 aarch64_classify_vector_mode.
8211 (aarch64_regmode_natural_size): New function.
8212 (aarch64_hard_regno_caller_save_mode): Return the original mode
8214 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
8215 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
8216 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
8217 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
8219 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
8220 does not overlap dest if the function is frame-related. Handle
8222 (aarch64_split_add_offset): New function.
8223 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
8224 them aarch64_add_offset.
8225 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
8226 and update call to aarch64_sub_sp.
8227 (aarch64_add_cfa_expression): New function.
8228 (aarch64_expand_prologue): Pass extra temporary registers to the
8229 functions above. Handle the case in which we need to emit new
8230 DW_CFA_expressions for registers that were originally saved
8231 relative to the stack pointer, but now have to be expressed
8232 relative to the frame pointer.
8233 (aarch64_output_mi_thunk): Pass extra temporary registers to the
8235 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
8236 IP0 and IP1 values for SVE frames.
8237 (aarch64_expand_vec_series): New function.
8238 (aarch64_expand_sve_widened_duplicate): Likewise.
8239 (aarch64_expand_sve_const_vector): Likewise.
8240 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
8241 Handle SVE constants. Use emit_move_insn to move a force_const_mem
8242 into the register, rather than emitting a SET directly.
8243 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
8244 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
8245 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
8246 (offset_9bit_signed_scaled_p): New functions.
8247 (aarch64_replicate_bitmask_imm): New function.
8248 (aarch64_bitmask_imm): Use it.
8249 (aarch64_cannot_force_const_mem): Reject expressions involving
8250 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
8251 (aarch64_classify_index): Handle SVE indices, by requiring
8252 a plain register index with a scale that matches the element size.
8253 (aarch64_classify_address): Handle SVE addresses. Assert that
8254 the mode of the address is VOIDmode or an integer mode.
8255 Update call to aarch64_classify_symbol.
8256 (aarch64_classify_symbolic_expression): Update call to
8257 aarch64_classify_symbol.
8258 (aarch64_const_vec_all_in_range_p): New function.
8259 (aarch64_print_vector_float_operand): Likewise.
8260 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
8261 "vN" for FP registers with SVE modes. Handle (const ...) vectors
8262 and the FP immediates 1.0 and 0.5.
8263 (aarch64_print_address_internal): Handle SVE addresses.
8264 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
8265 (aarch64_regno_regclass): Handle predicate registers.
8266 (aarch64_secondary_reload): Handle big-endian reloads of SVE
8268 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
8269 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
8270 (aarch64_convert_sve_vector_bits): New function.
8271 (aarch64_override_options): Use it to handle -msve-vector-bits=.
8272 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8274 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
8275 Handle SVE vector and predicate modes. Accept VL-based constants
8276 that need only one temporary register, and VL offsets that require
8277 no temporary registers.
8278 (aarch64_conditional_register_usage): Mark the predicate registers
8279 as fixed if SVE isn't available.
8280 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
8281 Return true for SVE vector and predicate modes.
8282 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
8283 rather than an unsigned int. Handle SVE modes.
8284 (aarch64_preferred_simd_mode): Update call accordingly. Handle
8286 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
8288 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8289 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8290 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8291 (aarch64_sve_float_mul_immediate_p): New functions.
8292 (aarch64_sve_valid_immediate): New function.
8293 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
8294 Explicitly reject structure modes. Check for INDEX constants.
8295 Handle PTRUE and PFALSE constants.
8296 (aarch64_check_zero_based_sve_index_immediate): New function.
8297 (aarch64_simd_imm_zero_p): Delete.
8298 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
8299 vector modes. Accept constants in the range of CNT[BHWD].
8300 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
8301 ask for an Advanced SIMD mode.
8302 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
8303 (aarch64_simd_vector_alignment): Handle SVE predicates.
8304 (aarch64_vectorize_preferred_vector_alignment): New function.
8305 (aarch64_simd_vector_alignment_reachable): Use it instead of
8307 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
8308 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
8310 (MAX_VECT_LEN): Delete.
8311 (expand_vec_perm_d): Add a vec_flags field.
8312 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
8313 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
8314 (aarch64_evpc_ext): Don't apply a big-endian lane correction
8316 (aarch64_evpc_rev): Rename to...
8317 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
8318 (aarch64_evpc_rev_global): New function.
8319 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
8320 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
8322 (aarch64_evpc_sve_tbl): New function.
8323 (aarch64_expand_vec_perm_const_1): Update after rename of
8324 aarch64_evpc_rev. Handle SVE permutes too, trying
8325 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
8326 than aarch64_evpc_tbl.
8327 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
8328 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
8329 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
8330 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
8331 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
8332 (aarch64_expand_sve_vcond): New functions.
8333 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
8334 of aarch64_vector_mode_p.
8335 (aarch64_dwarf_poly_indeterminate_value): New function.
8336 (aarch64_compute_pressure_classes): Likewise.
8337 (aarch64_can_change_mode_class): Likewise.
8338 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
8339 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
8340 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
8341 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
8342 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
8343 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
8344 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
8345 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
8347 (Dn, Dl, Dr): Accept const as well as const_vector.
8348 (Dz): Likewise. Compare against CONST0_RTX.
8349 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
8350 of "vector" where appropriate.
8351 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
8352 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
8353 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
8354 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
8355 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
8356 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
8357 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
8358 (v_int_equiv): Extend to SVE modes.
8359 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
8361 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
8362 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
8363 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
8364 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
8365 (SVE_COND_FP_CMP): New int iterators.
8366 (perm_hilo): Handle the new unpack unspecs.
8367 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
8369 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
8370 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
8371 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
8372 (aarch64_equality_operator, aarch64_constant_vector_operand)
8373 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
8374 (aarch64_sve_nonimmediate_operand): Likewise.
8375 (aarch64_sve_general_operand): Likewise.
8376 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
8377 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
8378 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
8379 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
8380 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
8381 (aarch64_sve_float_arith_immediate): Likewise.
8382 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
8383 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
8384 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
8385 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
8386 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
8387 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
8388 (aarch64_sve_float_arith_operand): Likewise.
8389 (aarch64_sve_float_arith_with_sub_operand): Likewise.
8390 (aarch64_sve_float_mul_operand): Likewise.
8391 (aarch64_sve_vec_perm_operand): Likewise.
8392 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
8393 (aarch64_mov_operand): Accept const_poly_int and const_vector.
8394 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
8395 as well as const_vector.
8396 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
8397 in file. Use CONST0_RTX and CONSTM1_RTX.
8398 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
8399 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
8400 Use aarch64_simd_imm_zero.
8401 * config/aarch64/aarch64-sve.md: New file.
8402 * config/aarch64/aarch64.md: Include it.
8403 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
8404 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
8405 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
8406 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
8407 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
8408 (sve): New attribute.
8409 (enabled): Disable instructions with the sve attribute unless
8411 (movqi, movhi): Pass CONST_POLY_INT operaneds through
8412 aarch64_expand_mov_immediate.
8413 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
8414 CNT[BHSD] immediates.
8415 (movti): Split CONST_POLY_INT moves into two halves.
8416 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
8417 Split additions that need a temporary here if the destination
8418 is the stack pointer.
8419 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
8420 (*add<mode>3_poly_1): New instruction.
8421 (set_clobber_cc): New expander.
8423 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8425 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
8426 parameter and use it instead of GET_MODE_SIZE (innermode). Use
8427 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
8428 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
8429 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
8430 Change innermode from fixed_mode_size to machine_mode.
8431 (simplify_subreg): Update call accordingly. Handle a constant-sized
8432 subreg of a variable-length CONST_VECTOR.
8434 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8435 Alan Hayward <alan.hayward@arm.com>
8436 David Sherwood <david.sherwood@arm.com>
8438 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
8439 (add_offset_to_base): New function, split out from...
8440 (create_mem_ref): ...here. When handling a scale other than 1,
8441 check first whether the address is valid without the offset.
8442 Add it into the base if so, leaving the index and scale as-is.
8444 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8447 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
8448 fold_for_warn before checking if arg2 is INTEGER_CST.
8450 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
8452 * config/rs6000/predicates.md (load_multiple_operation): Delete.
8453 (store_multiple_operation): Delete.
8454 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
8455 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
8456 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
8457 guarded by TARGET_STRING.
8458 (rs6000_output_load_multiple): Delete.
8459 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
8460 OPTION_MASK_STRING / TARGET_STRING handling.
8461 (print_operand) <'N', 'O'>: Add comment that these are unused now.
8462 (const rs6000_opt_masks) <"string">: Change mask to 0.
8463 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
8464 (MASK_STRING): Delete.
8465 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
8467 (load_multiple): Delete.
8474 (store_multiple): Delete.
8481 (movmemsi_8reg): Delete.
8482 (corresponding unnamed define_insn): Delete.
8483 (movmemsi_6reg): Delete.
8484 (corresponding unnamed define_insn): Delete.
8485 (movmemsi_4reg): Delete.
8486 (corresponding unnamed define_insn): Delete.
8487 (movmemsi_2reg): Delete.
8488 (corresponding unnamed define_insn): Delete.
8489 (movmemsi_1reg): Delete.
8490 (corresponding unnamed define_insn): Delete.
8491 * config/rs6000/rs6000.opt (mno-string): New.
8492 (mstring): Replace by deprecation warning stub.
8493 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
8495 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8497 * regrename.c (regrename_do_replace): If replacing the same
8498 reg multiple times, try to reuse last created gen_raw_REG.
8501 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
8502 main to workaround a bug in GDB.
8504 2018-01-12 Tom de Vries <tom@codesourcery.com>
8507 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
8509 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
8511 PR rtl-optimization/80481
8512 * ira-color.c (get_cap_member): New function.
8513 (allocnos_conflict_by_live_ranges_p): Use it.
8514 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
8515 (setup_slot_coalesced_allocno_live_ranges): Ditto.
8517 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
8520 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
8521 (*saddl_se_1): Ditto.
8523 (*ssubl_se_1): Ditto.
8525 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8527 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
8528 rather than wi::to_widest for DR_INITs.
8529 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
8530 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
8531 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
8533 (vect_analyze_group_access_1): Note that here.
8535 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8537 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
8538 polynomial type sizes.
8540 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8542 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
8543 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
8544 (gimple_add_tmp_var): Likewise.
8546 2018-01-12 Martin Liska <mliska@suse.cz>
8548 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
8549 (gimple_alloc_sizes): Likewise.
8550 (dump_gimple_statistics): Use PRIu64 in printf format.
8551 * gimple.h: Change uint64_t to int.
8553 2018-01-12 Martin Liska <mliska@suse.cz>
8555 * tree-core.h: Use uint64_t instead of int.
8556 * tree.c (tree_node_counts): Likewise.
8557 (tree_node_sizes): Likewise.
8558 (dump_tree_statistics): Use PRIu64 in printf format.
8560 2018-01-12 Martin Liska <mliska@suse.cz>
8562 * Makefile.in: As qsort_chk is implemented in vec.c, add
8563 vec.o to linkage of gencfn-macros.
8564 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
8565 passing the info to record_node_allocation_statistics.
8566 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
8568 * ggc-common.c (struct ggc_usage): Add operator== and use
8569 it in operator< and compare function.
8570 * mem-stats.h (struct mem_usage): Likewise.
8571 * vec.c (struct vec_usage): Remove operator< and compare
8572 function. Can be simply inherited.
8574 2018-01-12 Martin Jambor <mjambor@suse.cz>
8577 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
8578 * tree-ssa-math-opts.c: Include domwalk.h.
8579 (convert_mult_to_fma_1): New function.
8580 (fma_transformation_info): New type.
8581 (fma_deferring_state): Likewise.
8582 (cancel_fma_deferring): New function.
8583 (result_of_phi): Likewise.
8584 (last_fma_candidate_feeds_initial_phi): Likewise.
8585 (convert_mult_to_fma): Added deferring logic, split actual
8586 transformation to convert_mult_to_fma_1.
8587 (math_opts_dom_walker): New type.
8588 (math_opts_dom_walker::after_dom_children): New method, body moved
8589 here from pass_optimize_widening_mul::execute, added deferring logic
8591 (pass_optimize_widening_mul::execute): Moved most of code to
8592 math_opts_dom_walker::after_dom_children.
8593 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
8594 * config/i386/i386.c (ix86_option_override_internal): Added
8595 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
8597 2018-01-12 Richard Biener <rguenther@suse.de>
8600 * dwarf2out.c (gen_variable_die): Do not reset old_die for
8601 inline instance vars.
8603 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
8606 * config/rx/rx.c (rx_is_restricted_memory_address):
8609 2018-01-12 Richard Biener <rguenther@suse.de>
8611 PR tree-optimization/80846
8612 * target.def (split_reduction): New target hook.
8613 * targhooks.c (default_split_reduction): New function.
8614 * targhooks.h (default_split_reduction): Declare.
8615 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
8616 target requests first reduce vectors by combining low and high
8618 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
8619 (get_vectype_for_scalar_type_and_size): Export.
8620 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
8621 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
8622 * doc/tm.texi: Regenerate.
8623 * config/i386/i386.c (ix86_split_reduction): Implement
8624 TARGET_VECTORIZE_SPLIT_REDUCTION.
8626 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8629 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
8630 in PIC mode except for TARGET_VXWORKS_RTP.
8631 * config/sparc/sparc.c: Include cfgrtl.h.
8632 (TARGET_INIT_PIC_REG): Define.
8633 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
8634 (sparc_pic_register_p): New predicate.
8635 (sparc_legitimate_address_p): Use it.
8636 (sparc_legitimize_pic_address): Likewise.
8637 (sparc_delegitimize_address): Likewise.
8638 (sparc_mode_dependent_address_p): Likewise.
8639 (gen_load_pcrel_sym): Remove 4th parameter.
8640 (load_got_register): Adjust call to above. Remove obsolete stuff.
8641 (sparc_expand_prologue): Do not call load_got_register here.
8642 (sparc_flat_expand_prologue): Likewise.
8643 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
8644 (sparc_use_pseudo_pic_reg): New function.
8645 (sparc_init_pic_reg): Likewise.
8646 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8647 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
8649 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
8651 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
8652 Add item for branch_cost.
8654 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8656 PR rtl-optimization/83565
8657 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
8658 not extend the result to a larger mode for rotate operations.
8659 (num_sign_bit_copies1): Likewise.
8661 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8664 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
8666 Use values-Xc.o for -pedantic.
8667 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
8669 2018-01-12 Martin Liska <mliska@suse.cz>
8672 * ipa-devirt.c (final_warning_record::grow_type_warnings):
8674 (possible_polymorphic_call_targets): Use it.
8675 (ipa_devirt): Likewise.
8677 2018-01-12 Martin Liska <mliska@suse.cz>
8679 * profile-count.h (enum profile_quality): Use 0 as invalid
8680 enum value of profile_quality.
8682 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
8684 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
8685 -mext-string options.
8687 2018-01-12 Richard Biener <rguenther@suse.de>
8689 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
8690 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
8691 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
8693 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
8695 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
8697 * configure.ac (--with-long-double-format): Add support for the
8698 configuration option to change the default long double format on
8700 * config.gcc (powerpc*-linux*-*): Likewise.
8701 * configure: Regenerate.
8702 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
8703 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
8704 used without modification.
8706 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8708 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
8709 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
8710 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
8711 MISC_BUILTIN_SPEC_BARRIER.
8712 (rs6000_init_builtins): Likewise.
8713 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
8715 (speculation_barrier): New define_insn.
8716 * doc/extend.texi: Document __builtin_speculation_barrier.
8718 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8721 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
8722 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
8723 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
8725 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
8726 integral modes instead of "ss" and "sd".
8727 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
8728 vectors with 32-bit and 64-bit elements.
8729 (vecdupssescalarmodesuffix): New mode attribute.
8730 (vec_dup<mode>): Use it.
8732 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
8735 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
8736 frame if argument is passed on stack.
8738 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8741 * ree.c (combine_reaching_defs): Optimize also
8742 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
8743 reg2=any_extend(exp); reg1=reg2;, formatting fix.
8745 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8748 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
8750 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8753 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
8754 after they are computed.
8756 2018-01-11 Bin Cheng <bin.cheng@arm.com>
8758 PR tree-optimization/83695
8759 * gimple-loop-linterchange.cc
8760 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
8761 reset cached scev information after interchange.
8762 (pass_linterchange::execute): Remove call to scev_reset_htab.
8764 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8766 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
8767 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
8768 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
8769 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
8770 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
8771 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
8772 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
8773 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
8774 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
8775 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
8776 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
8777 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
8778 (V_lane_reg): Likewise.
8779 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
8781 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
8782 (vfmal_lane_low<mode>_intrinsic,
8783 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
8784 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
8785 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
8786 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
8787 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
8788 vfmsl_lane_high<mode>_intrinsic): New define_insns.
8790 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8792 * config/arm/arm-cpus.in (fp16fml): New feature.
8793 (ALL_SIMD): Add fp16fml.
8794 (armv8.2-a): Add fp16fml as an option.
8795 (armv8.3-a): Likewise.
8796 (armv8.4-a): Add fp16fml as part of fp16.
8797 * config/arm/arm.h (TARGET_FP16FML): Define.
8798 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
8800 * config/arm/arm-modes.def (V2HF): Define.
8801 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
8802 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
8803 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
8804 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
8805 vfmsl_low, vfmsl_high): New set of builtins.
8806 * config/arm/iterators.md (PLUSMINUS): New code iterator.
8807 (vfml_op): New code attribute.
8808 (VFMLHALVES): New int iterator.
8809 (VFML, VFMLSEL): New mode attributes.
8810 (V_reg): Define mapping for V2HF.
8811 (V_hi, V_lo): New mode attributes.
8812 (VF_constraint): Likewise.
8813 (vfml_half, vfml_half_selector): New int attributes.
8814 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
8816 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
8817 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
8819 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
8820 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
8821 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
8822 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
8824 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
8825 Document new effective target and option set.
8827 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8829 * config/arm/arm-cpus.in (armv8_4): New feature.
8830 (ARMv8_4a): New fgroup.
8831 (armv8.4-a): New arch.
8832 * config/arm/arm-tables.opt: Regenerate.
8833 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
8834 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
8835 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
8836 Add matching rules for -march=armv8.4-a and extensions.
8837 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
8839 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
8842 * config/rx/rx.md (BW): New mode attribute.
8843 (sync_lock_test_and_setsi): Add mode suffix to insn output.
8845 2018-01-11 Richard Biener <rguenther@suse.de>
8847 PR tree-optimization/83435
8848 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
8849 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
8850 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
8852 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8853 Alan Hayward <alan.hayward@arm.com>
8854 David Sherwood <david.sherwood@arm.com>
8856 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
8858 (aarch64_classify_address): Initialize it. Track polynomial offsets.
8859 (aarch64_print_address_internal): Use it to check for a zero offset.
8861 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8862 Alan Hayward <alan.hayward@arm.com>
8863 David Sherwood <david.sherwood@arm.com>
8865 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
8866 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
8867 Return a poly_int64 rather than a HOST_WIDE_INT.
8868 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
8869 rather than a HOST_WIDE_INT.
8870 * config/aarch64/aarch64.h (aarch64_frame): Protect with
8871 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
8872 hard_fp_offset, frame_size, initial_adjust, callee_offset and
8873 final_offset from HOST_WIDE_INT to poly_int64.
8874 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
8875 to_constant when getting the number of units in an Advanced SIMD
8877 (aarch64_builtin_vectorized_function): Check for a constant number
8879 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
8881 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
8882 attribute instead of GET_MODE_NUNITS.
8883 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
8884 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
8885 GET_MODE_SIZE for fixed-size registers.
8886 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
8887 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
8888 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
8889 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
8890 (aarch64_print_operand, aarch64_print_address_internal)
8891 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
8892 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
8893 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
8894 Handle polynomial GET_MODE_SIZE.
8895 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
8896 wider than SImode without modification.
8897 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
8898 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
8899 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
8900 passing and returning SVE modes.
8901 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
8902 rather than GEN_INT.
8903 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
8904 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
8905 (aarch64_allocate_and_probe_stack_space): Likewise.
8906 (aarch64_layout_frame): Cope with polynomial offsets.
8907 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
8908 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
8910 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
8911 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
8912 poly_int64 rather than a HOST_WIDE_INT.
8913 (aarch64_get_separate_components, aarch64_process_components)
8914 (aarch64_expand_prologue, aarch64_expand_epilogue)
8915 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
8916 (aarch64_anchor_offset): New function, split out from...
8917 (aarch64_legitimize_address): ...here.
8918 (aarch64_builtin_vectorization_cost): Handle polynomial
8919 TYPE_VECTOR_SUBPARTS.
8920 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
8922 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
8923 number of elements from the PARALLEL rather than the mode.
8924 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
8925 rather than GET_MODE_BITSIZE.
8926 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
8927 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
8928 (aarch64_expand_vec_perm_const_1): Handle polynomial
8929 d->perm.length () and d->perm elements.
8930 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
8931 Apply to_constant to d->perm elements.
8932 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
8933 polynomial CONST_VECTOR_NUNITS.
8934 (aarch64_move_pointer): Take amount as a poly_int64 rather
8936 (aarch64_progress_pointer): Avoid temporary variable.
8937 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
8938 the mode attribute instead of GET_MODE.
8940 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8941 Alan Hayward <alan.hayward@arm.com>
8942 David Sherwood <david.sherwood@arm.com>
8944 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
8945 x exists before using it.
8946 (aarch64_add_constant_internal): Rename to...
8947 (aarch64_add_offset_1): ...this. Replace regnum with separate
8948 src and dest rtxes. Handle the case in which they're different,
8949 including when the offset is zero. Replace scratchreg with an rtx.
8950 Use 2 additions if there is no spare register into which we can
8951 move a 16-bit constant.
8952 (aarch64_add_constant): Delete.
8953 (aarch64_add_offset): Replace reg with separate src and dest
8954 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
8955 Use aarch64_add_offset_1.
8956 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
8957 an rtx rather than an int. Take the delta as a poly_int64
8958 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
8959 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
8960 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
8961 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
8962 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
8964 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
8965 aarch64_add_constant.
8967 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8969 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
8970 Use scalar_float_mode.
8972 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8974 * config/aarch64/aarch64-simd.md
8975 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
8976 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
8977 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
8978 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
8979 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
8980 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
8981 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
8982 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
8983 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
8984 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
8986 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8989 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
8990 targ_options->x_arm_arch_string is non NULL.
8992 2018-01-11 Tamar Christina <tamar.christina@arm.com>
8994 * config/aarch64/aarch64.h
8995 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
8997 2018-01-11 Sudakshina Das <sudi.das@arm.com>
9000 * expmed.c (emit_store_flag_force): Swap if const op0
9001 and change VOIDmode to mode of op0.
9003 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9005 PR rtl-optimization/83761
9006 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
9007 than bytes to mode_for_size.
9009 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9012 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
9013 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
9016 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9019 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
9020 when in layout mode.
9021 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
9022 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
9025 2018-01-10 Michael Collison <michael.collison@arm.com>
9027 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
9028 * config/aarch64/aarch64-option-extension.def: Add
9029 AARCH64_OPT_EXTENSION of 'fp16fml'.
9030 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9031 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
9032 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
9033 * config/aarch64/constraints.md (Ui7): New constraint.
9034 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
9035 (VFMLA_SEL_W): Ditto.
9038 (VFMLA16_LOW): New int iterator.
9039 (VFMLA16_HIGH): Ditto.
9040 (UNSPEC_FMLAL): New unspec.
9041 (UNSPEC_FMLSL): Ditto.
9042 (UNSPEC_FMLAL2): Ditto.
9043 (UNSPEC_FMLSL2): Ditto.
9044 (f16mac): New code attribute.
9045 * config/aarch64/aarch64-simd-builtins.def
9046 (aarch64_fmlal_lowv2sf): Ditto.
9047 (aarch64_fmlsl_lowv2sf): Ditto.
9048 (aarch64_fmlalq_lowv4sf): Ditto.
9049 (aarch64_fmlslq_lowv4sf): Ditto.
9050 (aarch64_fmlal_highv2sf): Ditto.
9051 (aarch64_fmlsl_highv2sf): Ditto.
9052 (aarch64_fmlalq_highv4sf): Ditto.
9053 (aarch64_fmlslq_highv4sf): Ditto.
9054 (aarch64_fmlal_lane_lowv2sf): Ditto.
9055 (aarch64_fmlsl_lane_lowv2sf): Ditto.
9056 (aarch64_fmlal_laneq_lowv2sf): Ditto.
9057 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
9058 (aarch64_fmlalq_lane_lowv4sf): Ditto.
9059 (aarch64_fmlsl_lane_lowv4sf): Ditto.
9060 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
9061 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
9062 (aarch64_fmlal_lane_highv2sf): Ditto.
9063 (aarch64_fmlsl_lane_highv2sf): Ditto.
9064 (aarch64_fmlal_laneq_highv2sf): Ditto.
9065 (aarch64_fmlsl_laneq_highv2sf): Ditto.
9066 (aarch64_fmlalq_lane_highv4sf): Ditto.
9067 (aarch64_fmlsl_lane_highv4sf): Ditto.
9068 (aarch64_fmlalq_laneq_highv4sf): Ditto.
9069 (aarch64_fmlsl_laneq_highv4sf): Ditto.
9070 * config/aarch64/aarch64-simd.md:
9071 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
9072 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9073 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
9074 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9075 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
9076 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
9077 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
9078 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
9079 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
9080 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
9081 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
9082 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
9083 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
9084 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
9085 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
9086 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
9087 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
9088 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
9089 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
9090 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
9091 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
9092 (vfmlsl_low_u32): Ditto.
9093 (vfmlalq_low_u32): Ditto.
9094 (vfmlslq_low_u32): Ditto.
9095 (vfmlal_high_u32): Ditto.
9096 (vfmlsl_high_u32): Ditto.
9097 (vfmlalq_high_u32): Ditto.
9098 (vfmlslq_high_u32): Ditto.
9099 (vfmlal_lane_low_u32): Ditto.
9100 (vfmlsl_lane_low_u32): Ditto.
9101 (vfmlal_laneq_low_u32): Ditto.
9102 (vfmlsl_laneq_low_u32): Ditto.
9103 (vfmlalq_lane_low_u32): Ditto.
9104 (vfmlslq_lane_low_u32): Ditto.
9105 (vfmlalq_laneq_low_u32): Ditto.
9106 (vfmlslq_laneq_low_u32): Ditto.
9107 (vfmlal_lane_high_u32): Ditto.
9108 (vfmlsl_lane_high_u32): Ditto.
9109 (vfmlal_laneq_high_u32): Ditto.
9110 (vfmlsl_laneq_high_u32): Ditto.
9111 (vfmlalq_lane_high_u32): Ditto.
9112 (vfmlslq_lane_high_u32): Ditto.
9113 (vfmlalq_laneq_high_u32): Ditto.
9114 (vfmlslq_laneq_high_u32): Ditto.
9115 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
9116 (AARCH64_FL_FOR_ARCH8_4): New.
9117 (AARCH64_ISA_F16FML): New ISA flag.
9118 (TARGET_F16FML): New feature flag for fp16fml.
9119 (doc/invoke.texi): Document new fp16fml option.
9121 2018-01-10 Michael Collison <michael.collison@arm.com>
9123 * config/aarch64/aarch64-builtins.c:
9124 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
9125 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9126 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
9127 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
9128 (AARCH64_ISA_SHA3): New ISA flag.
9129 (TARGET_SHA3): New feature flag for sha3.
9130 * config/aarch64/iterators.md (sha512_op): New int attribute.
9131 (CRYPTO_SHA512): New int iterator.
9132 (UNSPEC_SHA512H): New unspec.
9133 (UNSPEC_SHA512H2): Ditto.
9134 (UNSPEC_SHA512SU0): Ditto.
9135 (UNSPEC_SHA512SU1): Ditto.
9136 * config/aarch64/aarch64-simd-builtins.def
9137 (aarch64_crypto_sha512hqv2di): New builtin.
9138 (aarch64_crypto_sha512h2qv2di): Ditto.
9139 (aarch64_crypto_sha512su0qv2di): Ditto.
9140 (aarch64_crypto_sha512su1qv2di): Ditto.
9141 (aarch64_eor3qv8hi): Ditto.
9142 (aarch64_rax1qv2di): Ditto.
9143 (aarch64_xarqv2di): Ditto.
9144 (aarch64_bcaxqv8hi): Ditto.
9145 * config/aarch64/aarch64-simd.md:
9146 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
9147 (aarch64_crypto_sha512su0qv2di): Ditto.
9148 (aarch64_crypto_sha512su1qv2di): Ditto.
9149 (aarch64_eor3qv8hi): Ditto.
9150 (aarch64_rax1qv2di): Ditto.
9151 (aarch64_xarqv2di): Ditto.
9152 (aarch64_bcaxqv8hi): Ditto.
9153 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
9154 (vsha512h2q_u64): Ditto.
9155 (vsha512su0q_u64): Ditto.
9156 (vsha512su1q_u64): Ditto.
9157 (veor3q_u16): Ditto.
9158 (vrax1q_u64): Ditto.
9160 (vbcaxq_u16): Ditto.
9161 * config/arm/types.md (crypto_sha512): New type attribute.
9162 (crypto_sha3): Ditto.
9163 (doc/invoke.texi): Document new sha3 option.
9165 2018-01-10 Michael Collison <michael.collison@arm.com>
9167 * config/aarch64/aarch64-builtins.c:
9168 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
9169 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9170 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
9171 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
9172 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
9173 (AARCH64_ISA_SM4): New ISA flag.
9174 (TARGET_SM4): New feature flag for sm4.
9175 * config/aarch64/aarch64-simd-builtins.def
9176 (aarch64_sm3ss1qv4si): Ditto.
9177 (aarch64_sm3tt1aq4si): Ditto.
9178 (aarch64_sm3tt1bq4si): Ditto.
9179 (aarch64_sm3tt2aq4si): Ditto.
9180 (aarch64_sm3tt2bq4si): Ditto.
9181 (aarch64_sm3partw1qv4si): Ditto.
9182 (aarch64_sm3partw2qv4si): Ditto.
9183 (aarch64_sm4eqv4si): Ditto.
9184 (aarch64_sm4ekeyqv4si): Ditto.
9185 * config/aarch64/aarch64-simd.md:
9186 (aarch64_sm3ss1qv4si): Ditto.
9187 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
9188 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
9189 (aarch64_sm4eqv4si): Ditto.
9190 (aarch64_sm4ekeyqv4si): Ditto.
9191 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
9192 (sm3part_op): Ditto.
9193 (CRYPTO_SM3TT): Ditto.
9194 (CRYPTO_SM3PART): Ditto.
9195 (UNSPEC_SM3SS1): New unspec.
9196 (UNSPEC_SM3TT1A): Ditto.
9197 (UNSPEC_SM3TT1B): Ditto.
9198 (UNSPEC_SM3TT2A): Ditto.
9199 (UNSPEC_SM3TT2B): Ditto.
9200 (UNSPEC_SM3PARTW1): Ditto.
9201 (UNSPEC_SM3PARTW2): Ditto.
9202 (UNSPEC_SM4E): Ditto.
9203 (UNSPEC_SM4EKEY): Ditto.
9204 * config/aarch64/constraints.md (Ui2): New constraint.
9205 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
9206 * config/arm/types.md (crypto_sm3): New type attribute.
9207 (crypto_sm4): Ditto.
9208 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
9209 (vsm3tt1aq_u32): Ditto.
9210 (vsm3tt1bq_u32): Ditto.
9211 (vsm3tt2aq_u32): Ditto.
9212 (vsm3tt2bq_u32): Ditto.
9213 (vsm3partw1q_u32): Ditto.
9214 (vsm3partw2q_u32): Ditto.
9215 (vsm4eq_u32): Ditto.
9216 (vsm4ekeyq_u32): Ditto.
9217 (doc/invoke.texi): Document new sm4 option.
9219 2018-01-10 Michael Collison <michael.collison@arm.com>
9221 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
9222 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
9223 (AARCH64_FL_FOR_ARCH8_4): New.
9224 (AARCH64_FL_V8_4): New flag.
9225 (doc/invoke.texi): Document new armv8.4-a option.
9227 2018-01-10 Michael Collison <michael.collison@arm.com>
9229 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9230 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
9231 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
9232 * config/aarch64/aarch64-option-extension.def: Add
9233 AARCH64_OPT_EXTENSION of 'sha2'.
9234 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
9235 (crypto): Disable sha2 and aes if crypto disabled.
9236 (crypto): Enable aes and sha2 if enabled.
9237 (simd): Disable sha2 and aes if simd disabled.
9238 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
9240 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
9241 (TARGET_SHA2): New feature flag for sha2.
9242 (TARGET_AES): New feature flag for aes.
9243 * config/aarch64/aarch64-simd.md:
9244 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
9245 conditional on TARGET_AES.
9246 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
9247 (aarch64_crypto_sha1hsi): Make pattern conditional
9249 (aarch64_crypto_sha1hv4si): Ditto.
9250 (aarch64_be_crypto_sha1hv4si): Ditto.
9251 (aarch64_crypto_sha1su1v4si): Ditto.
9252 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
9253 (aarch64_crypto_sha1su0v4si): Ditto.
9254 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
9255 (aarch64_crypto_sha256su0v4si): Ditto.
9256 (aarch64_crypto_sha256su1v4si): Ditto.
9257 (doc/invoke.texi): Document new aes and sha2 options.
9259 2018-01-10 Martin Sebor <msebor@redhat.com>
9261 PR tree-optimization/83781
9262 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
9265 2018-01-11 Martin Sebor <msebor@gmail.com>
9266 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9268 PR tree-optimization/83501
9269 PR tree-optimization/81703
9271 * tree-ssa-strlen.c (get_string_cst): Rename...
9272 (get_string_len): ...to this. Handle global constants.
9273 (handle_char_store): Adjust.
9275 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
9276 Jim Wilson <jimw@sifive.com>
9278 * config/riscv/riscv-protos.h (riscv_output_return): New.
9279 * config/riscv/riscv.c (struct machine_function): New naked_p field.
9280 (riscv_attribute_table, riscv_output_return),
9281 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
9282 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
9283 (riscv_compute_frame_info): Only compute frame->mask if not a naked
9285 (riscv_expand_prologue): Add early return for naked function.
9286 (riscv_expand_epilogue): Likewise.
9287 (riscv_function_ok_for_sibcall): Return false for naked function.
9288 (riscv_set_current_function): New.
9289 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
9290 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
9291 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
9292 * doc/extend.texi (RISC-V Function Attributes): New.
9294 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
9296 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
9297 check for 128-bit long double before checking TCmode.
9298 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
9299 128-bit long doubles before checking TFmode or TCmode.
9300 (FLOAT128_IBM_P): Likewise.
9302 2018-01-10 Martin Sebor <msebor@redhat.com>
9304 PR tree-optimization/83671
9305 * builtins.c (c_strlen): Unconditionally return zero for the empty
9307 Use -Warray-bounds for warnings.
9308 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
9309 for non-constant array indices with COMPONENT_REF, arrays of
9310 arrays, and pointers to arrays.
9311 (gimple_fold_builtin_strlen): Determine and set length range for
9312 non-constant character arrays.
9314 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
9317 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
9320 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
9322 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
9324 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9327 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
9328 VECTOR_MEM_ALTIVEC_OR_VSX_P.
9329 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
9330 indexed_or_indirect_operand predicate.
9331 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
9332 (*vsx_le_perm_load_v8hi): Likewise.
9333 (*vsx_le_perm_load_v16qi): Likewise.
9334 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
9335 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
9336 (*vsx_le_perm_store_v8hi): Likewise.
9337 (*vsx_le_perm_store_v16qi): Likewise.
9338 (eight unnamed splitters): Likewise.
9340 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9342 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
9343 * config/rs6000/emmintrin.h: Likewise.
9344 * config/rs6000/mmintrin.h: Likewise.
9345 * config/rs6000/xmmintrin.h: Likewise.
9347 2018-01-10 David Malcolm <dmalcolm@redhat.com>
9350 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
9352 * tree.c (tree_nop_conversion): Return true for location wrapper
9354 (maybe_wrap_with_location): New function.
9355 (selftest::check_strip_nops): New function.
9356 (selftest::test_location_wrappers): New function.
9357 (selftest::tree_c_tests): Call it.
9358 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
9359 (maybe_wrap_with_location): New decl.
9360 (EXPR_LOCATION_WRAPPER_P): New macro.
9361 (location_wrapper_p): New inline function.
9362 (tree_strip_any_location_wrapper): New inline function.
9364 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
9367 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
9368 stack_realign_offset for the largest alignment of stack slot
9370 (ix86_find_max_used_stack_alignment): New function.
9371 (ix86_finalize_stack_frame_flags): Use it. Set
9372 max_used_stack_alignment if we don't realign stack.
9373 * config/i386/i386.h (machine_function): Add
9374 max_used_stack_alignment.
9376 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
9378 * config/arm/arm.opt (-mbranch-cost): New option.
9379 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
9382 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
9385 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
9386 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
9388 2018-01-10 Richard Biener <rguenther@suse.de>
9391 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
9392 early out so it also covers the case where we have a non-NULL
9395 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9397 PR tree-optimization/83753
9398 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
9399 for non-strided grouped accesses if the number of elements is 1.
9401 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9404 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
9405 * i386.h (TARGET_USE_GATHER): Define.
9406 * x86-tune.def (X86_TUNE_USE_GATHER): New.
9408 2018-01-10 Martin Liska <mliska@suse.cz>
9411 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
9412 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
9414 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
9415 CLEANUP_NO_PARTITIONING is not set.
9417 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9419 * doc/rtl.texi: Remove documentation of (const ...) wrappers
9420 for vectors, as a partial revert of r254296.
9421 * rtl.h (const_vec_p): Delete.
9422 (const_vec_duplicate_p): Don't test for vector CONSTs.
9423 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
9424 * expmed.c (make_tree): Likewise.
9427 * common.md (E, F): Use CONSTANT_P instead of checking for
9429 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9430 checking for CONST_VECTOR.
9432 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9435 * predict.c (force_edge_cold): Handle in more sane way edges
9438 2018-01-09 Carl Love <cel@us.ibm.com>
9440 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
9442 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
9443 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
9444 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
9445 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
9446 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
9447 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
9448 * config/rs6000/rs6000-protos.h: Add extern defition for
9449 rs6000_generate_float2_double_code.
9450 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
9452 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
9453 (float2_v2df): Add define_expand.
9455 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
9458 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
9459 op_mode in the force_to_mode call.
9461 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9463 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
9464 instead of checking each element individually.
9465 (aarch64_evpc_uzp): Likewise.
9466 (aarch64_evpc_zip): Likewise.
9467 (aarch64_evpc_ext): Likewise.
9468 (aarch64_evpc_rev): Likewise.
9469 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
9470 instead of checking each element individually. Return true without
9472 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
9473 whether all selected elements come from the same input, instead of
9474 checking each element individually. Remove calls to gen_rtx_REG,
9475 start_sequence and end_sequence and instead assert that no rtl is
9478 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9480 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
9481 order of HIGH and CONST checks.
9483 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9485 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
9486 if the destination isn't an SSA_NAME.
9488 2018-01-09 Richard Biener <rguenther@suse.de>
9490 PR tree-optimization/83668
9491 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
9493 (canonicalize_loop_form): ... here, renamed from ...
9494 (canonicalize_loop_closed_ssa_form): ... this and amended to
9495 swap successor edges for loop exit blocks to make us use
9496 the RPO order we need for initial schedule generation.
9498 2018-01-09 Joseph Myers <joseph@codesourcery.com>
9500 PR tree-optimization/64811
9501 * match.pd: When optimizing comparisons with Inf, avoid
9502 introducing or losing exceptions from comparisons with NaN.
9504 2018-01-09 Martin Liska <mliska@suse.cz>
9507 * asan.c (shadow_mem_size): Add gcc_assert.
9509 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
9511 Don't save registers in main().
9514 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
9515 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
9516 * config/avr/avr.c (avr_set_current_function): Don't error if
9517 naked, OS_task or OS_main are specified at the same time.
9518 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
9520 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
9522 * common/config/avr/avr-common.c (avr_option_optimization_table):
9523 Switch on -mmain-is-OS_task for optimizing compilations.
9525 2018-01-09 Richard Biener <rguenther@suse.de>
9527 PR tree-optimization/83572
9528 * graphite.c: Include cfganal.h.
9529 (graphite_transform_loops): Connect infinite loops to exit
9530 and remove fake edges at the end.
9532 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9534 * ipa-inline.c (edge_badness): Revert accidental checkin.
9536 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9539 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
9540 symbols; not inline clones.
9542 2018-01-09 Jakub Jelinek <jakub@redhat.com>
9545 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
9546 hard registers. Formatting fixes.
9548 PR preprocessor/83722
9549 * gcc.c (try_generate_repro): Pass
9550 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
9551 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
9554 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
9555 Kito Cheng <kito.cheng@gmail.com>
9557 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
9558 (riscv_leaf_function_p): Delete.
9559 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
9561 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9563 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
9565 (do_ifelse): New function.
9566 (do_isel): New function.
9567 (do_sub3): New function.
9568 (do_add3): New function.
9569 (do_load_mask_compare): New function.
9570 (do_overlap_load_compare): New function.
9571 (expand_compare_loop): New function.
9572 (expand_block_compare): Call expand_compare_loop() when appropriate.
9573 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
9575 (-mblock-compare-inline-loop-limit): New option.
9577 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9580 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
9581 Reverse order of second and third operands in first alternative.
9582 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
9583 of first and second elements in UNSPEC_VPERMR vector.
9584 (altivec_expand_vec_perm_le): Likewise.
9586 2018-01-08 Jeff Law <law@redhat.com>
9588 PR rtl-optimizatin/81308
9589 * tree-switch-conversion.c (cfg_altered): New file scoped static.
9590 (process_switch): If group_case_labels makes a change, then set
9592 (pass_convert_switch::execute): If a switch is converted, then
9593 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
9595 PR rtl-optimization/81308
9596 * recog.c (split_all_insns): Conditionally cleanup the CFG after
9599 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
9601 PR target/83663 - Revert r255946
9602 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
9603 generation for cases where splatting a value is not useful.
9604 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
9605 across a vec_duplicate and a paradoxical subreg forming a vector
9606 mode to a vec_concat.
9608 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9610 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
9611 -march=armv8.3-a variants.
9612 * config/arm/t-multilib: Likewise.
9613 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
9615 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9617 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
9619 (cceq_ior_compare_complement): Give it a name so I can use it, and
9620 change boolean_or_operator predicate to boolean_operator so it can
9621 be used to generate a crand.
9622 (eqne): New code iterator.
9623 (bd/bd_neg): New code_attrs.
9624 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
9625 a single define_insn.
9626 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
9627 decrement (bdnzt/bdnzf/bdzt/bdzf).
9628 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
9629 with the new names of the branch decrement patterns, and added the
9630 names of the branch decrement conditional patterns.
9632 2018-01-08 Richard Biener <rguenther@suse.de>
9634 PR tree-optimization/83563
9635 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
9638 2018-01-08 Richard Biener <rguenther@suse.de>
9641 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
9643 2018-01-08 Richard Biener <rguenther@suse.de>
9645 PR tree-optimization/83685
9646 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
9647 references to abnormals.
9649 2018-01-08 Richard Biener <rguenther@suse.de>
9652 * dwarf2out.c (output_indirect_strings): Handle empty
9653 skeleton_debug_str_hash.
9654 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
9656 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9658 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
9659 (emit_store_direct): Likewise.
9660 (arc_trampoline_adjust_address): Likewise.
9661 (arc_asm_trampoline_template): New function.
9662 (arc_initialize_trampoline): Use asm_trampoline_template.
9663 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
9664 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
9665 * config/arc/arc.md (flush_icache): Delete pattern.
9667 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9669 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
9670 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
9673 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9676 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
9677 by not USED_FOR_TARGET.
9678 (make_pass_resolve_sw_modes): Likewise.
9680 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9682 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
9685 2018-01-08 Richard Biener <rguenther@suse.de>
9688 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
9690 2018-01-08 Richard Biener <rguenther@suse.de>
9693 * match.pd ((t * 2) / 2) -> t): Add missing :c.
9695 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
9698 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
9699 basic blocks with a small number of successors.
9700 (convert_control_dep_chain_into_preds): Improve handling of
9702 (dump_predicates): Split apart into...
9703 (dump_pred_chain): ...here...
9704 (dump_pred_info): ...and here.
9705 (can_one_predicate_be_invalidated_p): Add debugging printfs.
9706 (can_chain_union_be_invalidated_p): Improve check for invalidation
9708 (uninit_uses_cannot_happen): Avoid unnecessary if
9709 convert_control_dep_chain_into_preds yielded nothing.
9711 2018-01-06 Martin Sebor <msebor@redhat.com>
9713 PR tree-optimization/83640
9714 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
9715 subtracting negative offset from size.
9716 (builtin_access::overlap): Adjust offset bounds of the access to fall
9717 within the size of the object if possible.
9719 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
9721 PR rtl-optimization/83699
9722 * expmed.c (extract_bit_field_1): Restrict the vector usage of
9723 extract_bit_field_as_subreg to cases in which the extracted
9724 value is also a vector.
9726 * lra-constraints.c (process_alt_operands): Test for the equivalence
9727 substitutions when detecting a possible reload cycle.
9729 2018-01-06 Jakub Jelinek <jakub@redhat.com>
9732 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
9733 by default if flag_selective_schedling{,2}. Formatting fixes.
9735 PR rtl-optimization/83682
9736 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
9737 if it has non-VECTOR_MODE element mode.
9738 (vec_duplicate_p): Likewise.
9741 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
9742 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
9744 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9747 * config/i386/i386-builtin.def
9748 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
9749 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
9750 Require also OPTION_MASK_ISA_AVX512F in addition to
9751 OPTION_MASK_ISA_GFNI.
9752 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
9753 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
9754 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
9755 to OPTION_MASK_ISA_GFNI.
9756 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
9757 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
9758 OPTION_MASK_ISA_AVX512BW.
9759 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
9760 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
9761 addition to OPTION_MASK_ISA_GFNI.
9762 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
9763 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
9764 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
9765 to OPTION_MASK_ISA_GFNI.
9766 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
9767 a requirement for all ISAs rather than any of them with a few
9769 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
9771 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
9772 bitmasks to be enabled with 3 exceptions, instead of requiring any
9773 enabled ISA with lots of exceptions.
9774 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
9775 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
9776 Change avx512bw in isa attribute to avx512f.
9777 * config/i386/sgxintrin.h: Add license boilerplate.
9778 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
9779 to __AVX512F__ and __AVX512VL to __AVX512VL__.
9780 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
9781 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
9783 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
9784 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
9785 temporarily sse2 rather than sse if not enabled already.
9788 * config/i386/sse.md (VI248_VLBW): Rename to ...
9789 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
9790 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
9791 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
9792 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
9793 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
9794 mode iterator instead of VI248_VLBW.
9796 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
9798 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
9799 (record_modified): Skip clobbers; add debug output.
9800 (param_change_prob): Use sreal frequencies.
9802 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9804 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
9805 punt for user-aligned variables.
9807 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9809 * tree-chrec.c (chrec_contains_symbols): Return true for
9812 2018-01-05 Sudakshina Das <sudi.das@arm.com>
9815 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
9816 of (x|y) == x for BICS pattern.
9818 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9820 PR tree-optimization/83605
9821 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
9822 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
9825 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
9827 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
9828 * config/epiphany/rtems.h: New file.
9830 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9831 Uros Bizjak <ubizjak@gmail.com>
9834 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
9835 QIreg_operand instead of register_operand predicate.
9836 * config/i386/i386.c (ix86_rop_should_change_byte_p,
9837 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
9838 comments instead of -fmitigate[-_]rop.
9840 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9843 * cgraphunit.c (symbol_table::compile): Switch to text_section
9844 before calling assembly_start debug hook.
9845 * run-rtl-passes.c (run_rtl_passes): Likewise.
9848 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9850 * tree-vrp.c (extract_range_from_binary_expr_1): Check
9851 range_int_cst_p rather than !symbolic_range_p before calling
9852 extract_range_from_multiplicative_op_1.
9854 2018-01-04 Jeff Law <law@redhat.com>
9856 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
9857 redundant test in assertion.
9859 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9861 * doc/rtl.texi: Document machine_mode wrapper classes.
9863 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9865 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
9868 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9870 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
9871 the VEC_PERM_EXPR fold to fail.
9873 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9876 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
9877 to switched_sections.
9879 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9882 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
9885 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
9888 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
9889 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
9891 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9894 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
9895 is BLKmode and bitpos not zero or mode change is needed.
9897 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9900 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
9903 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
9906 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
9907 instead of MULT rtx. Update all corresponding splitters.
9909 (*ssub<modesuffix>): Ditto.
9911 (*cmp_sadd_di): Update split patterns.
9912 (*cmp_sadd_si): Ditto.
9913 (*cmp_sadd_sidi): Ditto.
9914 (*cmp_ssub_di): Ditto.
9915 (*cmp_ssub_si): Ditto.
9916 (*cmp_ssub_sidi): Ditto.
9917 * config/alpha/predicates.md (const23_operand): New predicate.
9918 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
9919 Look for ASHIFT, not MULT inner operand.
9920 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
9922 2018-01-04 Martin Liska <mliska@suse.cz>
9924 PR gcov-profile/83669
9925 * gcov.c (output_intermediate_file): Add version to intermediate
9927 * doc/gcov.texi: Document new field 'version' in intermediate
9928 file format. Fix location of '-k' option of gcov command.
9930 2018-01-04 Martin Liska <mliska@suse.cz>
9933 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
9935 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9937 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
9939 2018-01-03 Martin Sebor <msebor@redhat.com>
9941 PR tree-optimization/83655
9942 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
9943 checking calls with invalid arguments.
9945 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9947 * tree-vect-stmts.c (vect_get_store_rhs): New function.
9948 (vectorizable_mask_load_store): Delete.
9949 (vectorizable_call): Return false for masked loads and stores.
9950 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
9951 instead of gimple_assign_rhs1.
9952 (vectorizable_load): Handle IFN_MASK_LOAD.
9953 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
9955 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9957 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
9959 (vectorizable_mask_load_store): ...here.
9960 (vectorizable_load): ...and here.
9962 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9964 * tree-vect-stmts.c (vect_build_all_ones_mask)
9965 (vect_build_zero_merge_argument): New functions, split out from...
9966 (vectorizable_load): ...here.
9968 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9970 * tree-vect-stmts.c (vect_check_store_rhs): New function,
9972 (vectorizable_mask_load_store): ...here.
9973 (vectorizable_store): ...and here.
9975 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9977 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
9979 (vectorizable_mask_load_store): ...here.
9981 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9983 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
9984 (vect_model_store_cost): Take a vec_load_store_type instead of a
9986 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
9987 (vect_model_store_cost): Take a vec_load_store_type instead of a
9989 (vectorizable_mask_load_store): Update accordingly.
9990 (vectorizable_store): Likewise.
9991 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
9993 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9995 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
9996 IFN_MASK_LOAD calls here rather than...
9997 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
9999 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10000 Alan Hayward <alan.hayward@arm.com>
10001 David Sherwood <david.sherwood@arm.com>
10003 * expmed.c (extract_bit_field_1): For vector extracts,
10004 fall back to extract_bit_field_as_subreg if vec_extract
10007 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10008 Alan Hayward <alan.hayward@arm.com>
10009 David Sherwood <david.sherwood@arm.com>
10011 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
10012 they are variable or constant sized.
10013 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
10014 slots for constant-sized data.
10016 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10017 Alan Hayward <alan.hayward@arm.com>
10018 David Sherwood <david.sherwood@arm.com>
10020 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
10021 handling COND_EXPRs with boolean comparisons, try to find a better
10022 basis for the mask type than the boolean itself.
10024 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10026 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
10027 is calculated and how it can be overridden.
10028 * genmodes.c (max_bitsize_mode_any_mode): New variable.
10029 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
10031 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
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 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
10039 Remove the mode argument.
10040 (aarch64_simd_valid_immediate): Remove the mode and inverse
10042 * config/aarch64/iterators.md (bitsize): New iterator.
10043 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
10044 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
10045 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
10046 aarch64_simd_valid_immediate.
10047 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
10048 (aarch64_reg_or_bic_imm): Likewise.
10049 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
10050 with an insn_type enum and msl with a modifier_type enum.
10051 Replace element_width with a scalar_mode. Change the shift
10052 to unsigned int. Add constructors for scalar_float_mode and
10053 scalar_int_mode elements.
10054 (aarch64_vect_float_const_representable_p): Delete.
10055 (aarch64_can_const_movi_rtx_p)
10056 (aarch64_simd_scalar_immediate_valid_for_move)
10057 (aarch64_simd_make_constant): Update call to
10058 aarch64_simd_valid_immediate.
10059 (aarch64_advsimd_valid_immediate_hs): New function.
10060 (aarch64_advsimd_valid_immediate): Likewise.
10061 (aarch64_simd_valid_immediate): Remove mode and inverse
10062 arguments. Rewrite to use the above. Use const_vec_duplicate_p
10063 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
10064 and aarch64_float_const_representable_p on the result.
10065 (aarch64_output_simd_mov_immediate): Remove mode argument.
10066 Update call to aarch64_simd_valid_immediate and use of
10067 simd_immediate_info.
10068 (aarch64_output_scalar_simd_mov_immediate): Update call
10071 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10072 Alan Hayward <alan.hayward@arm.com>
10073 David Sherwood <david.sherwood@arm.com>
10075 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
10076 (mode_nunits): Likewise CONST_MODE_NUNITS.
10077 * machmode.def (ADJUST_NUNITS): Document.
10078 * genmodes.c (mode_data::need_nunits_adj): New field.
10079 (blank_mode): Update accordingly.
10080 (adj_nunits): New variable.
10081 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
10083 (emit_mode_size_inline): Set need_bytesize_adj for all modes
10084 listed in adj_nunits.
10085 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
10086 listed in adj_nunits. Don't emit case statements for such modes.
10087 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
10088 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
10089 nothing if adj_nunits is nonnull.
10090 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
10091 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
10092 (emit_mode_fbit): Update use of print_maybe_const_decl.
10093 (emit_move_size): Likewise. Treat the array as non-const
10095 (emit_mode_adjustments): Handle adj_nunits.
10097 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10099 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
10100 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
10101 (VECTOR_MODES): Use it.
10102 (make_vector_modes): Take the prefix as an argument.
10104 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10105 Alan Hayward <alan.hayward@arm.com>
10106 David Sherwood <david.sherwood@arm.com>
10108 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
10109 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
10110 for MODE_VECTOR_BOOL.
10111 * machmode.def (VECTOR_BOOL_MODE): Document.
10112 * genmodes.c (VECTOR_BOOL_MODE): New macro.
10113 (make_vector_bool_mode): New function.
10114 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
10116 * lto-streamer-in.c (lto_input_mode_table): Likewise.
10117 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
10119 * stor-layout.c (int_mode_for_mode): Likewise.
10120 * tree.c (build_vector_type_for_mode): Likewise.
10121 * varasm.c (output_constant_pool_2): Likewise.
10122 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
10123 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
10124 for MODE_VECTOR_BOOL.
10125 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
10126 of mode class checks.
10127 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
10128 instead of a list of mode class checks.
10129 (expand_vector_scalar_condition): Likewise.
10130 (type_for_widest_vector_mode): Handle BImode as an inner mode.
10132 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10133 Alan Hayward <alan.hayward@arm.com>
10134 David Sherwood <david.sherwood@arm.com>
10136 * machmode.h (mode_size): Change from unsigned short to
10138 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
10139 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10140 or if measurement_type is not polynomial.
10141 (fixed_size_mode::includes_p): Check for constant-sized modes.
10142 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
10143 return a poly_uint16 rather than an unsigned short.
10144 (emit_mode_size): Change the type of mode_size from unsigned short
10145 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
10146 (emit_mode_adjustments): Cope with polynomial vector sizes.
10147 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10149 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10151 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
10152 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
10153 * caller-save.c (setup_save_areas): Likewise.
10154 (replace_reg_with_saved_mem): Likewise.
10155 * calls.c (emit_library_call_value_1): Likewise.
10156 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
10157 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
10158 (gen_lowpart_for_combine): Likewise.
10159 * convert.c (convert_to_integer_1): Likewise.
10160 * cse.c (equiv_constant, cse_insn): Likewise.
10161 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
10162 (cselib_subst_to_values): Likewise.
10163 * dce.c (word_dce_process_block): Likewise.
10164 * df-problems.c (df_word_lr_mark_ref): Likewise.
10165 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
10166 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
10167 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
10168 (rtl_for_decl_location): Likewise.
10169 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
10170 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
10171 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
10172 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
10173 (expand_expr_real_1): Likewise.
10174 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
10175 (pad_below): Likewise.
10176 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10177 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
10178 * ira.c (get_subreg_tracking_sizes): Likewise.
10179 * ira-build.c (ira_create_allocno_objects): Likewise.
10180 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
10181 (ira_sort_regnos_for_alter_reg): Likewise.
10182 * ira-costs.c (record_operand_costs): Likewise.
10183 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
10184 (resolve_simple_move): Likewise.
10185 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
10186 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
10187 (lra_constraints): Likewise.
10188 (CONST_POOL_OK_P): Reject variable-sized modes.
10189 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
10190 (add_pseudo_to_slot, lra_spill): Likewise.
10191 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10192 * optabs-query.c (get_best_extraction_insn): Likewise.
10193 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10194 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
10195 (expand_mult_highpart, valid_multiword_target_p): Likewise.
10196 * recog.c (offsettable_address_addr_space_p): Likewise.
10197 * regcprop.c (maybe_mode_change): Likewise.
10198 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
10199 * regrename.c (build_def_use): Likewise.
10200 * regstat.c (dump_reg_info): Likewise.
10201 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
10202 (find_reloads, find_reloads_subreg_address): Likewise.
10203 * reload1.c (eliminate_regs_1): Likewise.
10204 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
10205 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
10206 (simplify_binary_operation_1, simplify_subreg): Likewise.
10207 * targhooks.c (default_function_arg_padding): Likewise.
10208 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
10209 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
10210 (verify_gimple_assign_ternary): Likewise.
10211 * tree-inline.c (estimate_move_cost): Likewise.
10212 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10213 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
10214 (get_address_cost_ainc): Likewise.
10215 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
10216 (vect_supportable_dr_alignment): Likewise.
10217 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10218 (vectorizable_reduction): Likewise.
10219 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
10220 (vectorizable_operation, vectorizable_load): Likewise.
10221 * tree.c (build_same_sized_truth_vector_type): Likewise.
10222 * valtrack.c (cleanup_auto_inc_dec): Likewise.
10223 * var-tracking.c (emit_note_insn_var_location): Likewise.
10224 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
10225 (ADDR_VEC_ALIGN): Likewise.
10227 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10228 Alan Hayward <alan.hayward@arm.com>
10229 David Sherwood <david.sherwood@arm.com>
10231 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
10233 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10234 or if measurement_type is polynomial.
10235 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
10236 * combine.c (make_extraction): Likewise.
10237 * dse.c (find_shift_sequence): Likewise.
10238 * dwarf2out.c (mem_loc_descriptor): Likewise.
10239 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
10240 (extract_bit_field, extract_low_bits): Likewise.
10241 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
10242 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
10243 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
10244 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
10245 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10246 * reload.c (find_reloads): Likewise.
10247 * reload1.c (alter_reg): Likewise.
10248 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
10249 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
10250 * tree-if-conv.c (predicate_mem_writes): Likewise.
10251 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
10252 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
10253 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
10254 * valtrack.c (dead_debug_insert_temp): Likewise.
10255 * varasm.c (mergeable_constant_section): Likewise.
10256 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
10258 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10259 Alan Hayward <alan.hayward@arm.com>
10260 David Sherwood <david.sherwood@arm.com>
10262 * expr.c (expand_assignment): Cope with polynomial mode sizes
10263 when assigning to a CONCAT.
10265 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10266 Alan Hayward <alan.hayward@arm.com>
10267 David Sherwood <david.sherwood@arm.com>
10269 * machmode.h (mode_precision): Change from unsigned short to
10271 (mode_to_precision): Return a poly_uint16 rather than an unsigned
10273 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
10274 or if measurement_type is not polynomial.
10275 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
10276 in which the mode is already known to be a scalar_int_mode.
10277 * genmodes.c (emit_mode_precision): Change the type of mode_precision
10278 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
10280 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10281 for GET_MODE_PRECISION.
10282 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10283 for GET_MODE_PRECISION.
10284 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
10286 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
10287 (expand_field_assignment, make_extraction): Likewise.
10288 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
10289 (get_last_value): Likewise.
10290 * convert.c (convert_to_integer_1): Likewise.
10291 * cse.c (cse_insn): Likewise.
10292 * expr.c (expand_expr_real_1): Likewise.
10293 * lra-constraints.c (simplify_operand_subreg): Likewise.
10294 * optabs-query.c (can_atomic_load_p): Likewise.
10295 * optabs.c (expand_atomic_load): Likewise.
10296 (expand_atomic_store): Likewise.
10297 * ree.c (combine_reaching_defs): Likewise.
10298 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
10299 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
10300 * tree.h (type_has_mode_precision_p): Likewise.
10301 * ubsan.c (instrument_si_overflow): Likewise.
10303 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10304 Alan Hayward <alan.hayward@arm.com>
10305 David Sherwood <david.sherwood@arm.com>
10307 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
10308 polynomial numbers of units.
10309 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
10310 (valid_vector_subparts_p): New function.
10311 (build_vector_type): Remove temporary shim and take the number
10312 of units as a poly_uint64 rather than an int.
10313 (build_opaque_vector_type): Take the number of units as a
10314 poly_uint64 rather than an int.
10315 * tree.c (build_vector_from_ctor): Handle polynomial
10316 TYPE_VECTOR_SUBPARTS.
10317 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
10318 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
10319 (build_vector_from_val): If the number of units is variable,
10320 use build_vec_duplicate_cst for constant operands and
10321 VEC_DUPLICATE_EXPR otherwise.
10322 (make_vector_type): Remove temporary is_constant ().
10323 (build_vector_type, build_opaque_vector_type): Take the number of
10324 units as a poly_uint64 rather than an int.
10325 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
10327 * cfgexpand.c (expand_debug_expr): Likewise.
10328 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
10329 (store_constructor, expand_expr_real_1): Likewise.
10330 (const_scalar_mask_from_tree): Likewise.
10331 * fold-const-call.c (fold_const_reduction): Likewise.
10332 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
10333 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
10334 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
10335 (fold_relational_const): Likewise.
10336 (native_interpret_vector): Likewise. Change the size from an
10337 int to an unsigned int.
10338 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
10339 TYPE_VECTOR_SUBPARTS.
10340 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
10341 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
10342 duplicating a non-constant operand into a variable-length vector.
10343 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
10344 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
10345 * ipa-icf.c (sem_variable::equals): Likewise.
10346 * match.pd: Likewise.
10347 * omp-simd-clone.c (simd_clone_subparts): Likewise.
10348 * print-tree.c (print_node): Likewise.
10349 * stor-layout.c (layout_type): Likewise.
10350 * targhooks.c (default_builtin_vectorization_cost): Likewise.
10351 * tree-cfg.c (verify_gimple_comparison): Likewise.
10352 (verify_gimple_assign_binary): Likewise.
10353 (verify_gimple_assign_ternary): Likewise.
10354 (verify_gimple_assign_single): Likewise.
10355 * tree-pretty-print.c (dump_generic_node): Likewise.
10356 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10357 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
10358 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
10359 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
10360 (vect_shift_permute_load_chain): Likewise.
10361 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
10362 (expand_vector_condition, optimize_vector_constructor): Likewise.
10363 (lower_vec_perm, get_compute_type): Likewise.
10364 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10365 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
10366 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
10367 (vect_recog_mask_conversion_pattern): Likewise.
10368 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
10369 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
10370 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10371 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
10372 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
10373 (vectorizable_shift, vectorizable_operation, vectorizable_store)
10374 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
10375 (supportable_widening_operation): Likewise.
10376 (supportable_narrowing_operation): Likewise.
10377 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
10379 * varasm.c (output_constant): Likewise.
10381 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10382 Alan Hayward <alan.hayward@arm.com>
10383 David Sherwood <david.sherwood@arm.com>
10385 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
10386 so that both the length == 3 and length != 3 cases set up their
10387 own permute vectors. Add comments explaining why we know the
10388 number of elements is constant.
10389 (vect_permute_load_chain): Likewise.
10391 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10392 Alan Hayward <alan.hayward@arm.com>
10393 David Sherwood <david.sherwood@arm.com>
10395 * machmode.h (mode_nunits): Change from unsigned char to
10397 (ONLY_FIXED_SIZE_MODES): New macro.
10398 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
10399 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
10400 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
10402 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
10403 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
10404 or if measurement_type is not polynomial.
10405 * genmodes.c (ZERO_COEFFS): New macro.
10406 (emit_mode_nunits_inline): Make mode_nunits_inline return a
10408 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
10409 Use ZERO_COEFFS when emitting initializers.
10410 * data-streamer.h (bp_pack_poly_value): New function.
10411 (bp_unpack_poly_value): Likewise.
10412 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10413 for GET_MODE_NUNITS.
10414 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10415 for GET_MODE_NUNITS.
10416 * tree.c (make_vector_type): Remove temporary shim and make
10417 the real function take the number of units as a poly_uint64
10418 rather than an int.
10419 (build_vector_type_for_mode): Handle polynomial nunits.
10420 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
10421 * emit-rtl.c (const_vec_series_p_1): Likewise.
10422 (gen_rtx_CONST_VECTOR): Likewise.
10423 * fold-const.c (test_vec_duplicate_folding): Likewise.
10424 * genrecog.c (validate_pattern): Likewise.
10425 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
10426 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10427 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
10428 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
10429 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
10430 * rtlanal.c (subreg_get_info): Likewise.
10431 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10432 (vect_grouped_load_supported): Likewise.
10433 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
10434 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
10435 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
10436 (simplify_const_unary_operation, simplify_binary_operation_1)
10437 (simplify_const_binary_operation, simplify_ternary_operation)
10438 (test_vector_ops_duplicate, test_vector_ops): Likewise.
10439 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
10440 instead of CONST_VECTOR_NUNITS.
10441 * varasm.c (output_constant_pool_2): Likewise.
10442 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
10443 explicit-encoded elements in the XVEC for variable-length vectors.
10445 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10447 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
10449 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10450 Alan Hayward <alan.hayward@arm.com>
10451 David Sherwood <david.sherwood@arm.com>
10453 * coretypes.h (fixed_size_mode): Declare.
10454 (fixed_size_mode_pod): New typedef.
10455 * builtins.h (target_builtins::x_apply_args_mode)
10456 (target_builtins::x_apply_result_mode): Change type to
10457 fixed_size_mode_pod.
10458 * builtins.c (apply_args_size, apply_result_size, result_vector)
10459 (expand_builtin_apply_args_1, expand_builtin_apply)
10460 (expand_builtin_return): Update accordingly.
10462 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10464 * cse.c (hash_rtx_cb): Hash only the encoded elements.
10465 * cselib.c (cselib_hash_rtx): Likewise.
10466 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
10467 CONST_VECTOR encoding.
10469 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10470 Jeff Law <law@redhat.com>
10473 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
10474 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
10475 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
10476 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
10479 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
10480 explicitly probe *sp in a noreturn function if there were any callee
10481 register saves or frame pointer is needed.
10483 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10486 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
10487 BLKmode for ternary, binary or unary expressions.
10490 * var-tracking.c (delete_vta_debug_insn): New inline function.
10491 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
10492 insns from get_insns () to NULL instead of each bb separately.
10493 Use delete_vta_debug_insn. No longer static.
10494 (vt_debug_insns_local, variable_tracking_main_1): Adjust
10495 delete_vta_debug_insns callers.
10496 * rtl.h (delete_vta_debug_insns): Declare.
10497 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
10498 instead of variable_tracking_main.
10500 2018-01-03 Martin Sebor <msebor@redhat.com>
10502 PR tree-optimization/83603
10503 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
10504 arguments past the endof the argument list in functions declared
10505 without a prototype.
10506 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
10507 Avoid checking when arguments are null.
10509 2018-01-03 Martin Sebor <msebor@redhat.com>
10512 * doc/extend.texi (attribute const): Fix a typo.
10513 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
10514 issuing -Wsuggest-attribute for void functions.
10516 2018-01-03 Martin Sebor <msebor@redhat.com>
10518 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
10519 offset_int::from instead of wide_int::to_shwi.
10520 (maybe_diag_overlap): Remove assertion.
10521 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
10522 * gimple-ssa-sprintf.c (format_directive): Same.
10523 (parse_directive): Same.
10524 (sprintf_dom_walker::compute_format_length): Same.
10525 (try_substitute_return_value): Same.
10527 2018-01-03 Jeff Law <law@redhat.com>
10529 PR middle-end/83654
10530 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
10531 non-constant residual for zero at runtime and avoid probing in
10532 that case. Reorganize code for trailing problem to mirror handling
10535 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10537 PR tree-optimization/83501
10538 * tree-ssa-strlen.c (get_string_cst): New.
10539 (handle_char_store): Call get_string_cst.
10541 2018-01-03 Martin Liska <mliska@suse.cz>
10543 PR tree-optimization/83593
10544 * tree-ssa-strlen.c: Include tree-cfg.h.
10545 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
10546 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
10547 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
10549 (strlen_dom_walker::before_dom_children): Call
10550 gimple_purge_dead_eh_edges. Dump tranformation with details
10552 (strlen_dom_walker::before_dom_children): Update call by adding
10553 new argument cleanup_eh.
10554 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
10556 2018-01-03 Martin Liska <mliska@suse.cz>
10559 * cif-code.def (VARIADIC_THUNK): New enum value.
10560 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
10563 2018-01-03 Jan Beulich <jbeulich@suse.com>
10565 * sse.md (mov<mode>_internal): Tighten condition for when to use
10566 vmovdqu<ssescalarsize> for TI and OI modes.
10568 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10570 Update copyright years.
10572 2018-01-03 Martin Liska <mliska@suse.cz>
10575 * ipa-visibility.c (function_and_variable_visibility): Skip
10576 functions with noipa attribure.
10578 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10580 * gcc.c (process_command): Update copyright notice dates.
10581 * gcov-dump.c (print_version): Ditto.
10582 * gcov.c (print_version): Ditto.
10583 * gcov-tool.c (print_version): Ditto.
10584 * gengtype.c (create_file): Ditto.
10585 * doc/cpp.texi: Bump @copying's copyright year.
10586 * doc/cppinternals.texi: Ditto.
10587 * doc/gcc.texi: Ditto.
10588 * doc/gccint.texi: Ditto.
10589 * doc/gcov.texi: Ditto.
10590 * doc/install.texi: Ditto.
10591 * doc/invoke.texi: Ditto.
10593 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10595 * vector-builder.h (vector_builder::m_full_nelts): Change from
10596 unsigned int to poly_uint64.
10597 (vector_builder::full_nelts): Update prototype accordingly.
10598 (vector_builder::new_vector): Likewise.
10599 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
10600 (vector_builder::operator ==): Likewise.
10601 (vector_builder::finalize): Likewise.
10602 * int-vector-builder.h (int_vector_builder::int_vector_builder):
10603 Take the number of elements as a poly_uint64 rather than an
10605 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
10606 from unsigned int to poly_uint64.
10607 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
10608 (vec_perm_indices::new_vector): Likewise.
10609 (vec_perm_indices::length): Likewise.
10610 (vec_perm_indices::nelts_per_input): Likewise.
10611 (vec_perm_indices::input_nelts): Likewise.
10612 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
10613 number of elements per input as a poly_uint64 rather than an
10614 unsigned int. Use the original encoding for variable-length
10615 vectors, rather than clamping each individual element.
10616 For the second and subsequent elements in each pattern,
10617 clamp the step and base before clamping their sum.
10618 (vec_perm_indices::series_p): Handle polynomial element counts.
10619 (vec_perm_indices::all_in_range_p): Likewise.
10620 (vec_perm_indices_to_tree): Likewise.
10621 (vec_perm_indices_to_rtx): Likewise.
10622 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
10623 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
10624 (tree_vector_builder::new_binary_operation): Handle polynomial
10625 element counts. Return false if we need to know the number
10626 of elements at compile time.
10627 * fold-const.c (fold_vec_perm): Punt if the number of elements
10628 isn't known at compile time.
10630 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10632 * vec-perm-indices.h (vec_perm_builder): Change element type
10633 from HOST_WIDE_INT to poly_int64.
10634 (vec_perm_indices::element_type): Update accordingly.
10635 (vec_perm_indices::clamp): Handle polynomial element_types.
10636 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10637 (vec_perm_indices::all_in_range_p): Likewise.
10638 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
10640 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
10641 polynomial vec_perm_indices element types.
10642 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
10643 * fold-const.c (fold_vec_perm): Likewise.
10644 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
10645 * tree-vect-generic.c (lower_vec_perm): Likewise.
10646 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10647 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
10648 element type to HOST_WIDE_INT.
10650 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10651 Alan Hayward <alan.hayward@arm.com>
10652 David Sherwood <david.sherwood@arm.com>
10654 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
10655 rather than an int. Use plus_constant.
10656 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
10657 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
10659 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10660 Alan Hayward <alan.hayward@arm.com>
10661 David Sherwood <david.sherwood@arm.com>
10663 * calls.c (emit_call_1, expand_call): Change struct_value_size from
10664 a HOST_WIDE_INT to a poly_int64.
10666 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10667 Alan Hayward <alan.hayward@arm.com>
10668 David Sherwood <david.sherwood@arm.com>
10670 * calls.c (load_register_parameters): Cope with polynomial
10671 mode sizes. Require a constant size for BLKmode parameters
10672 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
10673 forces a parameter to be padded at the lsb end in order to
10674 fill a complete number of words, require the parameter size
10675 to be ordered wrt UNITS_PER_WORD.
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 * reload1.c (spill_stack_slot_width): Change element type
10682 from unsigned int to poly_uint64_pod.
10683 (alter_reg): Treat mode sizes as polynomial.
10685 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10686 Alan Hayward <alan.hayward@arm.com>
10687 David Sherwood <david.sherwood@arm.com>
10689 * reload.c (complex_word_subreg_p): New function.
10690 (reload_inner_reg_of_subreg, push_reload): Use it.
10692 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10693 Alan Hayward <alan.hayward@arm.com>
10694 David Sherwood <david.sherwood@arm.com>
10696 * lra-constraints.c (process_alt_operands): Reject matched
10697 operands whose sizes aren't ordered.
10698 (match_reload): Refer to this check here.
10700 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10701 Alan Hayward <alan.hayward@arm.com>
10702 David Sherwood <david.sherwood@arm.com>
10704 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
10705 that the mode size is in the set {1, 2, 4, 8, 16}.
10707 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10708 Alan Hayward <alan.hayward@arm.com>
10709 David Sherwood <david.sherwood@arm.com>
10711 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
10712 Use plus_constant instead of gen_rtx_PLUS.
10714 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10715 Alan Hayward <alan.hayward@arm.com>
10716 David Sherwood <david.sherwood@arm.com>
10718 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
10719 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
10720 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
10721 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
10722 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
10723 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
10724 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
10725 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
10726 * config/i386/i386.c (ix86_push_rounding): ...this new function.
10727 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
10729 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
10730 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
10731 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
10732 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
10733 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
10734 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
10735 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
10736 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
10737 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
10738 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
10740 * expr.c (emit_move_resolve_push): Treat the input and result
10741 of PUSH_ROUNDING as a poly_int64.
10742 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
10743 (emit_push_insn): Likewise.
10744 * lra-eliminations.c (mark_not_eliminable): Likewise.
10745 * recog.c (push_operand): Likewise.
10746 * reload1.c (elimination_effects): Likewise.
10747 * rtlanal.c (nonzero_bits1): Likewise.
10748 * calls.c (store_one_arg): Likewise. Require the padding to be
10749 known at compile time.
10751 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10752 Alan Hayward <alan.hayward@arm.com>
10753 David Sherwood <david.sherwood@arm.com>
10755 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
10756 Use plus_constant instead of gen_rtx_PLUS.
10758 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10759 Alan Hayward <alan.hayward@arm.com>
10760 David Sherwood <david.sherwood@arm.com>
10762 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
10763 rather than an int.
10765 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10766 Alan Hayward <alan.hayward@arm.com>
10767 David Sherwood <david.sherwood@arm.com>
10769 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
10770 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
10771 via stack temporaries. Treat the mode size as polynomial too.
10773 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10774 Alan Hayward <alan.hayward@arm.com>
10775 David Sherwood <david.sherwood@arm.com>
10777 * expr.c (expand_expr_real_2): When handling conversions involving
10778 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
10779 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
10780 as a poly_uint64 too.
10782 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10783 Alan Hayward <alan.hayward@arm.com>
10784 David Sherwood <david.sherwood@arm.com>
10786 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
10788 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10789 Alan Hayward <alan.hayward@arm.com>
10790 David Sherwood <david.sherwood@arm.com>
10792 * combine.c (can_change_dest_mode): Handle polynomial
10793 REGMODE_NATURAL_SIZE.
10794 * expmed.c (store_bit_field_1): Likewise.
10795 * expr.c (store_constructor): Likewise.
10796 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
10797 and polynomial REGMODE_NATURAL_SIZE.
10798 (gen_lowpart_common): Likewise.
10799 * reginfo.c (record_subregs_of_mode): Likewise.
10800 * rtlanal.c (read_modify_subreg_p): Likewise.
10802 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10803 Alan Hayward <alan.hayward@arm.com>
10804 David Sherwood <david.sherwood@arm.com>
10806 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
10807 numbers of elements.
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 * match.pd: Cope with polynomial numbers of vector elements.
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 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
10820 in a POINTER_PLUS_EXPR.
10822 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10823 Alan Hayward <alan.hayward@arm.com>
10824 David Sherwood <david.sherwood@arm.com>
10826 * omp-simd-clone.c (simd_clone_subparts): New function.
10827 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
10828 (ipa_simd_modify_function_body): Likewise.
10830 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10831 Alan Hayward <alan.hayward@arm.com>
10832 David Sherwood <david.sherwood@arm.com>
10834 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
10835 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
10836 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
10837 (expand_vector_condition, vector_element): Likewise.
10838 (subparts_gt): New function.
10839 (get_compute_type): Use subparts_gt.
10840 (count_type_subparts): Delete.
10841 (expand_vector_operations_1): Use subparts_gt instead of
10842 count_type_subparts.
10844 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10845 Alan Hayward <alan.hayward@arm.com>
10846 David Sherwood <david.sherwood@arm.com>
10848 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
10849 (vect_compile_time_alias): ...this new function. Do the calculation
10850 on poly_ints rather than trees.
10851 (vect_prune_runtime_alias_test_list): Update call accordingly.
10853 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10854 Alan Hayward <alan.hayward@arm.com>
10855 David Sherwood <david.sherwood@arm.com>
10857 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
10859 (vect_schedule_slp_instance): Likewise.
10861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10862 Alan Hayward <alan.hayward@arm.com>
10863 David Sherwood <david.sherwood@arm.com>
10865 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
10866 constant and extern definitions for variable-length vectors.
10867 (vect_get_constant_vectors): Note that the number of units
10868 is known to be constant.
10870 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10871 Alan Hayward <alan.hayward@arm.com>
10872 David Sherwood <david.sherwood@arm.com>
10874 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
10875 of units as polynomial. Choose between WIDE and NARROW based
10878 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10879 Alan Hayward <alan.hayward@arm.com>
10880 David Sherwood <david.sherwood@arm.com>
10882 * tree-vect-stmts.c (simd_clone_subparts): New function.
10883 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
10885 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10886 Alan Hayward <alan.hayward@arm.com>
10887 David Sherwood <david.sherwood@arm.com>
10889 * tree-vect-stmts.c (vectorizable_call): Treat the number of
10890 vectors as polynomial. Use build_index_vector for
10891 IFN_GOMP_SIMD_LANE.
10893 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10894 Alan Hayward <alan.hayward@arm.com>
10895 David Sherwood <david.sherwood@arm.com>
10897 * tree-vect-stmts.c (get_load_store_type): Treat the number of
10898 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
10899 for variable-length vectors.
10900 (vectorizable_mask_load_store): Treat the number of units as
10901 polynomial, asserting that it is constant if the condition has
10902 already been enforced.
10903 (vectorizable_store, vectorizable_load): Likewise.
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-loop.c (vectorizable_live_operation): Treat the number
10910 of units as polynomial. Punt if we can't tell at compile time
10911 which vector contains the final result.
10913 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10914 Alan Hayward <alan.hayward@arm.com>
10915 David Sherwood <david.sherwood@arm.com>
10917 * tree-vect-loop.c (vectorizable_induction): Treat the number
10918 of units as polynomial. Punt on SLP inductions. Use an integer
10919 VEC_SERIES_EXPR for variable-length integer reductions. Use a
10920 cast of such a series for variable-length floating-point
10923 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10924 Alan Hayward <alan.hayward@arm.com>
10925 David Sherwood <david.sherwood@arm.com>
10927 * tree.h (build_index_vector): Declare.
10928 * tree.c (build_index_vector): New function.
10929 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
10930 of units as polynomial, forcibly converting it to a constant if
10931 vectorizable_reduction has already enforced the condition.
10932 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
10933 to create a {1,2,3,...} vector.
10934 (vectorizable_reduction): Treat the number of units as polynomial.
10935 Choose vectype_in based on the largest scalar element size rather
10936 than the smallest number of units. Enforce the restrictions
10939 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10940 Alan Hayward <alan.hayward@arm.com>
10941 David Sherwood <david.sherwood@arm.com>
10943 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
10944 number of units as polynomial.
10946 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10947 Alan Hayward <alan.hayward@arm.com>
10948 David Sherwood <david.sherwood@arm.com>
10950 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
10951 * target.def (autovectorize_vector_sizes): Return the vector sizes
10952 by pointer, using vector_sizes rather than a bitmask.
10953 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
10954 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
10955 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
10957 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
10958 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
10959 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
10960 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
10961 * omp-general.c (omp_max_vf): Likewise.
10962 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10963 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
10964 * tree-vect-loop.c (vect_analyze_loop): Likewise.
10965 * tree-vect-slp.c (vect_slp_bb): Likewise.
10966 * doc/tm.texi: Regenerate.
10967 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
10969 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
10970 the vector size as a poly_uint64 rather than an unsigned int.
10971 (current_vector_size): Change from an unsigned int to a poly_uint64.
10972 (get_vectype_for_scalar_type): Update accordingly.
10973 * tree.h (build_truth_vector_type): Take the size and number of
10974 units as a poly_uint64 rather than an unsigned int.
10975 (build_vector_type): Add a temporary overload that takes
10976 the number of units as a poly_uint64 rather than an unsigned int.
10977 * tree.c (make_vector_type): Likewise.
10978 (build_truth_vector_type): Take the number of units as a poly_uint64
10979 rather than an unsigned int.
10981 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10982 Alan Hayward <alan.hayward@arm.com>
10983 David Sherwood <david.sherwood@arm.com>
10985 * target.def (get_mask_mode): Take the number of units and length
10986 as poly_uint64s rather than unsigned ints.
10987 * targhooks.h (default_get_mask_mode): Update accordingly.
10988 * targhooks.c (default_get_mask_mode): Likewise.
10989 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
10990 * doc/tm.texi: Regenerate.
10992 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10993 Alan Hayward <alan.hayward@arm.com>
10994 David Sherwood <david.sherwood@arm.com>
10996 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
10997 * omp-general.c (omp_max_vf): Likewise.
10998 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
10999 (expand_omp_simd): Handle polynomial safelen.
11000 * omp-low.c (omplow_simd_context): Add a default constructor.
11001 (omplow_simd_context::max_vf): Change from int to poly_uint64.
11002 (lower_rec_simd_input_clauses): Update accordingly.
11003 (lower_rec_input_clauses): Likewise.
11005 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11006 Alan Hayward <alan.hayward@arm.com>
11007 David Sherwood <david.sherwood@arm.com>
11009 * tree-vectorizer.h (vect_nunits_for_cost): New function.
11010 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
11011 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
11012 (vect_analyze_slp_cost): Likewise.
11013 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
11014 (vect_model_load_cost): Likewise.
11016 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11017 Alan Hayward <alan.hayward@arm.com>
11018 David Sherwood <david.sherwood@arm.com>
11020 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
11021 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
11022 from an unsigned int * to a poly_uint64_pod *.
11023 (calculate_unrolling_factor): New function.
11024 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
11026 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11027 Alan Hayward <alan.hayward@arm.com>
11028 David Sherwood <david.sherwood@arm.com>
11030 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
11031 from an unsigned int to a poly_uint64.
11032 (_loop_vec_info::slp_unrolling_factor): Likewise.
11033 (_loop_vec_info::vectorization_factor): Change from an int
11035 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
11036 (vect_get_num_vectors): New function.
11037 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
11038 (vect_get_num_copies): Use vect_get_num_vectors.
11039 (vect_analyze_data_ref_dependences): Change max_vf from an int *
11040 to an unsigned int *.
11041 (vect_analyze_data_refs): Change min_vf from an int * to a
11043 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11044 than an unsigned HOST_WIDE_INT.
11045 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
11046 (vect_analyze_data_ref_dependence): Change max_vf from an int *
11047 to an unsigned int *.
11048 (vect_analyze_data_ref_dependences): Likewise.
11049 (vect_compute_data_ref_alignment): Handle polynomial vf.
11050 (vect_enhance_data_refs_alignment): Likewise.
11051 (vect_prune_runtime_alias_test_list): Likewise.
11052 (vect_shift_permute_load_chain): Likewise.
11053 (vect_supportable_dr_alignment): Likewise.
11054 (dependence_distance_ge_vf): Take the vectorization factor as a
11055 poly_uint64 rather than an unsigned HOST_WIDE_INT.
11056 (vect_analyze_data_refs): Change min_vf from an int * to a
11058 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
11059 vfm1 as a poly_uint64 rather than an int. Make the same change
11060 for the returned bound_scalar.
11061 (vect_gen_vector_loop_niters): Handle polynomial vf.
11062 (vect_do_peeling): Likewise. Update call to
11063 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
11064 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
11066 * tree-vect-loop.c (vect_determine_vectorization_factor)
11067 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
11068 (vect_get_known_peeling_cost): Likewise.
11069 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
11070 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
11071 (vect_transform_loop): Likewise. Use the lowest possible VF when
11072 updating the upper bounds of the loop.
11073 (vect_min_worthwhile_factor): Make static. Return an unsigned int
11074 rather than an int.
11075 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
11076 polynomial unroll factors.
11077 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
11078 (vect_make_slp_decision): Likewise.
11079 (vect_supported_load_permutation_p): Likewise, and polynomial
11081 (vect_analyze_slp_cost): Handle polynomial vf.
11082 (vect_slp_analyze_node_operations): Likewise.
11083 (vect_slp_analyze_bb_1): Likewise.
11084 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11085 than an unsigned HOST_WIDE_INT.
11086 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
11087 (vectorizable_load): Handle polynomial vf.
11088 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
11090 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
11092 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11093 Alan Hayward <alan.hayward@arm.com>
11094 David Sherwood <david.sherwood@arm.com>
11096 * match.pd: Handle bit operations involving three constants
11097 and try to fold one pair.
11099 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11101 * tree-vect-loop-manip.c: Include gimple-fold.h.
11102 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
11103 niters_maybe_zero parameters. Handle other cases besides a step of 1.
11104 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
11105 Add a path that uses a step of VF instead of 1, but disable it
11107 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
11108 and niters_no_overflow parameters. Update calls to
11109 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
11110 Create a new SSA name if the latter choses to use a ste other
11111 than zero, and return it via niters_vector_mult_vf_var.
11112 * tree-vect-loop.c (vect_transform_loop): Update calls to
11113 vect_do_peeling, vect_gen_vector_loop_niters and
11114 slpeel_make_loop_iterate_ntimes.
11115 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
11116 (vect_gen_vector_loop_niters): Update declarations after above changes.
11118 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
11120 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
11121 128-bit round to integer instructions.
11122 (ceil<mode>2): Likewise.
11123 (btrunc<mode>2): Likewise.
11124 (round<mode>2): Likewise.
11126 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11128 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
11129 unaligned VSX load/store on P8/P9.
11130 (expand_block_clear): Allow the use of unaligned VSX
11131 load/store on P8/P9.
11133 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11135 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
11137 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
11138 swap associated with both a load and a store.
11140 2018-01-02 Andrew Waterman <andrew@sifive.com>
11142 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
11143 * config/riscv/riscv.md (clear_cache): Use it.
11145 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
11147 * web.c: Remove out-of-date comment.
11149 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11151 * expr.c (fixup_args_size_notes): Check that any existing
11152 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
11153 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
11154 (emit_single_push_insn): ...here.
11156 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11158 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
11159 (const_vector_encoded_nelts): New function.
11160 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
11161 (const_vector_int_elt, const_vector_elt): Declare.
11162 * emit-rtl.c (const_vector_int_elt_1): New function.
11163 (const_vector_elt): Likewise.
11164 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
11165 of CONST_VECTOR_ELT.
11167 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11169 * expr.c: Include rtx-vector-builder.h.
11170 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
11171 directly on the tree encoding.
11172 (const_vector_from_tree): Likewise.
11173 * optabs.c: Include rtx-vector-builder.h.
11174 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
11175 sequence of "u" values.
11176 * vec-perm-indices.c: Include rtx-vector-builder.h.
11177 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
11178 directly on the vec_perm_indices encoding.
11180 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11182 * doc/rtl.texi (const_vector): Describe new encoding scheme.
11183 * Makefile.in (OBJS): Add rtx-vector-builder.o.
11184 * rtx-vector-builder.h: New file.
11185 * rtx-vector-builder.c: Likewise.
11186 * rtl.h (rtx_def::u2): Add a const_vector field.
11187 (CONST_VECTOR_NPATTERNS): New macro.
11188 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
11189 (CONST_VECTOR_DUPLICATE_P): Likewise.
11190 (CONST_VECTOR_STEPPED_P): Likewise.
11191 (CONST_VECTOR_ENCODED_ELT): Likewise.
11192 (const_vec_duplicate_p): Check for a duplicated vector encoding.
11193 (unwrap_const_vec_duplicate): Likewise.
11194 (const_vec_series_p): Check for a non-duplicated vector encoding.
11195 Say that the function only returns true for integer vectors.
11196 * emit-rtl.c: Include rtx-vector-builder.h.
11197 (gen_const_vec_duplicate_1): Delete.
11198 (gen_const_vector): Call gen_const_vec_duplicate instead of
11199 gen_const_vec_duplicate_1.
11200 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
11201 (gen_const_vec_duplicate): Use rtx_vector_builder.
11202 (gen_const_vec_series): Likewise.
11203 (gen_rtx_CONST_VECTOR): Likewise.
11204 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
11205 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11206 Build a new vector rather than modifying a CONST_VECTOR in-place.
11207 (handle_special_swappables): Update call accordingly.
11208 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
11209 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11210 Build a new vector rather than modifying a CONST_VECTOR in-place.
11211 (handle_special_swappables): Update call accordingly.
11213 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11215 * simplify-rtx.c (simplify_const_binary_operation): Use
11216 CONST_VECTOR_ELT instead of XVECEXP.
11218 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11220 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
11221 the selector elements to be different from the data elements
11222 if the selector is a VECTOR_CST.
11223 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
11224 ssizetype for the selector.
11226 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11228 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
11229 before testing each element individually.
11230 * tree-vect-generic.c (lower_vec_perm): Likewise.
11232 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11234 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
11235 * selftest-run-tests.c (selftest::run_tests): Call it.
11236 * vector-builder.h (vector_builder::operator ==): New function.
11237 (vector_builder::operator !=): Likewise.
11238 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
11239 (vec_perm_indices::all_from_input_p): New function.
11240 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11241 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
11242 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
11243 instead of reading the VECTOR_CST directly. Detect whether both
11244 vector inputs are the same before constructing the vec_perm_indices,
11245 and update the number of inputs argument accordingly. Use the
11246 utility functions added above. Only construct sel2 if we need to.
11248 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11250 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
11251 the broadcast of the low byte.
11252 (expand_mult_highpart): Use an explicit encoding for the permutes.
11253 * optabs-query.c (can_mult_highpart_p): Likewise.
11254 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
11255 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11256 (vectorizable_bswap): Likewise.
11257 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
11258 explicit encoding for the power-of-2 permutes.
11259 (vect_permute_store_chain): Likewise.
11260 (vect_grouped_load_supported): Likewise.
11261 (vect_permute_load_chain): Likewise.
11263 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11265 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
11266 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
11267 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
11268 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11269 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
11270 (vect_gen_perm_mask_any): Likewise.
11272 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11274 * int-vector-builder.h: New file.
11275 * vec-perm-indices.h: Include int-vector-builder.h.
11276 (vec_perm_indices): Redefine as an int_vector_builder.
11277 (auto_vec_perm_indices): Delete.
11278 (vec_perm_builder): Redefine as a stand-alone class.
11279 (vec_perm_indices::vec_perm_indices): New function.
11280 (vec_perm_indices::clamp): Likewise.
11281 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
11282 (vec_perm_indices::new_vector): New function.
11283 (vec_perm_indices::new_expanded_vector): Update for new
11284 vec_perm_indices class.
11285 (vec_perm_indices::rotate_inputs): New function.
11286 (vec_perm_indices::all_in_range_p): Operate directly on the
11287 encoded form, without computing elided elements.
11288 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
11289 encoding. Update for new vec_perm_indices class.
11290 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
11291 the given vec_perm_builder.
11292 (expand_vec_perm_var): Update vec_perm_builder constructor.
11293 (expand_mult_highpart): Use vec_perm_builder instead of
11294 auto_vec_perm_indices.
11295 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
11296 vec_perm_indices instead of auto_vec_perm_indices. Use a single
11297 or double series encoding as appropriate.
11298 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
11299 vec_perm_indices instead of auto_vec_perm_indices.
11300 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11301 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11302 (vect_permute_store_chain): Likewise.
11303 (vect_grouped_load_supported): Likewise.
11304 (vect_permute_load_chain): Likewise.
11305 (vect_shift_permute_load_chain): Likewise.
11306 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11307 (vect_transform_slp_perm_load): Likewise.
11308 (vect_schedule_slp_instance): Likewise.
11309 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11310 (vectorizable_mask_load_store): Likewise.
11311 (vectorizable_bswap): Likewise.
11312 (vectorizable_store): Likewise.
11313 (vectorizable_load): Likewise.
11314 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
11315 vec_perm_indices instead of auto_vec_perm_indices. Use
11316 tree_to_vec_perm_builder to read the vector from a tree.
11317 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
11318 vec_perm_builder instead of a vec_perm_indices.
11319 (have_whole_vector_shift): Use vec_perm_builder and
11320 vec_perm_indices instead of auto_vec_perm_indices. Leave the
11321 truncation to calc_vec_perm_mask_for_shift.
11322 (vect_create_epilog_for_reduction): Likewise.
11323 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
11324 from auto_vec_perm_indices to vec_perm_indices.
11325 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11326 instead of changing individual elements.
11327 (aarch64_vectorize_vec_perm_const): Use new_vector to install
11328 the vector in d.perm.
11329 * config/arm/arm.c (expand_vec_perm_d::perm): Change
11330 from auto_vec_perm_indices to vec_perm_indices.
11331 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11332 instead of changing individual elements.
11333 (arm_vectorize_vec_perm_const): Use new_vector to install
11334 the vector in d.perm.
11335 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
11336 Update vec_perm_builder constructor.
11337 (rs6000_expand_interleave): Likewise.
11338 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
11339 (rs6000_expand_interleave): Likewise.
11341 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11343 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
11344 to qimode could truncate the indices.
11345 * optabs.c (expand_vec_perm_var): Likewise.
11347 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11349 * Makefile.in (OBJS): Add vec-perm-indices.o.
11350 * vec-perm-indices.h: New file.
11351 * vec-perm-indices.c: Likewise.
11352 * target.h (vec_perm_indices): Replace with a forward class
11354 (auto_vec_perm_indices): Move to vec-perm-indices.h.
11355 * optabs.h: Include vec-perm-indices.h.
11356 (expand_vec_perm): Delete.
11357 (selector_fits_mode_p, expand_vec_perm_var): Declare.
11358 (expand_vec_perm_const): Declare.
11359 * target.def (vec_perm_const_ok): Replace with...
11360 (vec_perm_const): ...this new hook.
11361 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
11362 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
11363 * doc/tm.texi: Regenerate.
11364 * optabs.def (vec_perm_const): Delete.
11365 * doc/md.texi (vec_perm_const): Likewise.
11366 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
11367 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
11368 expand_vec_perm for constant permutation vectors. Assert that
11369 the mode of variable permutation vectors is the integer equivalent
11370 of the mode that is being permuted.
11371 * optabs-query.h (selector_fits_mode_p): Declare.
11372 * optabs-query.c: Include vec-perm-indices.h.
11373 (selector_fits_mode_p): New function.
11374 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
11375 is defined, instead of checking whether the vec_perm_const_optab
11376 exists. Use targetm.vectorize.vec_perm_const instead of
11377 targetm.vectorize.vec_perm_const_ok. Check whether the indices
11378 fit in the vector mode before using a variable permute.
11379 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
11380 vec_perm_indices instead of an rtx.
11381 (expand_vec_perm): Replace with...
11382 (expand_vec_perm_const): ...this new function. Take the selector
11383 as a vec_perm_indices rather than an rtx. Also take the mode of
11384 the selector. Update call to shift_amt_for_vec_perm_mask.
11385 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
11386 Use vec_perm_indices::new_expanded_vector to expand the original
11387 selector into bytes. Check whether the indices fit in the vector
11388 mode before using a variable permute.
11389 (expand_vec_perm_var): Make global.
11390 (expand_mult_highpart): Use expand_vec_perm_const.
11391 * fold-const.c: Includes vec-perm-indices.h.
11392 * tree-ssa-forwprop.c: Likewise.
11393 * tree-vect-data-refs.c: Likewise.
11394 * tree-vect-generic.c: Likewise.
11395 * tree-vect-loop.c: Likewise.
11396 * tree-vect-slp.c: Likewise.
11397 * tree-vect-stmts.c: Likewise.
11398 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
11400 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
11401 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
11402 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
11403 (aarch64_vectorize_vec_perm_const): ...this new function.
11404 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11405 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11406 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
11407 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
11408 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11409 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11410 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
11412 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
11413 check for NEON modes.
11414 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
11415 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
11416 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
11417 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
11419 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
11420 the old VEC_PERM_CONST conditions.
11421 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
11422 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
11423 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
11424 (ia64_vectorize_vec_perm_const_ok): Merge into...
11425 (ia64_vectorize_vec_perm_const): ...this new function.
11426 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
11427 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
11428 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
11429 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
11430 * config/mips/mips.c (mips_expand_vec_perm_const)
11431 (mips_vectorize_vec_perm_const_ok): Merge into...
11432 (mips_vectorize_vec_perm_const): ...this new function.
11433 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
11434 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
11435 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
11436 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
11437 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
11438 (rs6000_expand_vec_perm_const): Delete.
11439 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
11441 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11442 (altivec_expand_vec_perm_const_le): Take each operand individually.
11443 Operate on constant selectors rather than rtxes.
11444 (altivec_expand_vec_perm_const): Likewise. Update call to
11445 altivec_expand_vec_perm_const_le.
11446 (rs6000_expand_vec_perm_const): Delete.
11447 (rs6000_vectorize_vec_perm_const_ok): Delete.
11448 (rs6000_vectorize_vec_perm_const): New function.
11449 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11450 an element count and rtx array.
11451 (rs6000_expand_extract_even): Update call accordingly.
11452 (rs6000_expand_interleave): Likewise.
11453 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
11454 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
11455 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
11456 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
11457 (rs6000_expand_vec_perm_const): Delete.
11458 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11459 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11460 (altivec_expand_vec_perm_const_le): Take each operand individually.
11461 Operate on constant selectors rather than rtxes.
11462 (altivec_expand_vec_perm_const): Likewise. Update call to
11463 altivec_expand_vec_perm_const_le.
11464 (rs6000_expand_vec_perm_const): Delete.
11465 (rs6000_vectorize_vec_perm_const_ok): Delete.
11466 (rs6000_vectorize_vec_perm_const): New function. Remove stray
11467 reference to the SPE evmerge intructions.
11468 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11469 an element count and rtx array.
11470 (rs6000_expand_extract_even): Update call accordingly.
11471 (rs6000_expand_interleave): Likewise.
11472 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
11473 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
11475 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11477 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11479 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
11480 vector mode and that that mode matches the mode of the data
11482 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
11483 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
11484 directly using expand_vec_perm_1 when forcing selectors into
11486 (expand_vec_perm_var): New function, split out from expand_vec_perm.
11488 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11490 * optabs-query.h (can_vec_perm_p): Delete.
11491 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
11492 * optabs-query.c (can_vec_perm_p): Split into...
11493 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
11494 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
11495 particular selector is valid.
11496 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11497 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11498 (vect_grouped_load_supported): Likewise.
11499 (vect_shift_permute_load_chain): Likewise.
11500 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11501 (vect_transform_slp_perm_load): Likewise.
11502 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11503 (vectorizable_bswap): Likewise.
11504 (vect_gen_perm_mask_checked): Likewise.
11505 * fold-const.c (fold_ternary_loc): Likewise. Don't take
11506 implementations of variable permutation vectors into account
11507 when deciding which selector to use.
11508 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
11509 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
11510 with a false third argument.
11511 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
11512 to test whether the constant selector is valid and can_vec_perm_var_p
11513 to test whether a variable selector is valid.
11515 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11517 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
11518 * optabs-query.c (can_vec_perm_p): Likewise.
11519 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
11520 instead of vec_perm_indices.
11521 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
11522 (vect_gen_perm_mask_checked): Likewise,
11523 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
11524 (vect_gen_perm_mask_checked): Likewise,
11526 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11528 * optabs-query.h (qimode_for_vec_perm): Declare.
11529 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
11530 (qimode_for_vec_perm): ...this new function.
11531 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
11533 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11535 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
11536 does not have a conditional at the top.
11538 2018-01-02 Richard Biener <rguenther@suse.de>
11540 * ipa-inline.c (big_speedup_p): Fix expression.
11542 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11545 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
11548 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11552 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
11553 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
11554 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
11555 cond_taken_branch_cost 3->4.
11557 2018-01-01 Jakub Jelinek <jakub@redhat.com>
11559 PR tree-optimization/83581
11560 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
11561 TODO_cleanup_cfg if any changes have been made.
11563 PR middle-end/83608
11564 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
11565 convert_modes if target mode has the right side, but different mode
11568 PR middle-end/83609
11569 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
11570 last argument when extracting from CONCAT. If either from_real or
11571 from_imag is NULL, use expansion through memory. If result is not
11572 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
11573 the parts directly to inner mode, if even that fails, use expansion
11576 PR middle-end/83623
11577 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
11578 check for bswap in mode rather than HImode and use that in expand_unop
11581 Copyright (C) 2018 Free Software Foundation, Inc.
11583 Copying and distribution of this file, with or without modification,
11584 are permitted in any medium without royalty provided the copyright
11585 notice and this notice are preserved.