1 2018-04-27 Jeff Law <law@redhat.com>
3 * config.gcc: Mark tile* targets as deprecated/obsolete.
5 2018-04-27 Richard Biener <rguenther@suse.de>
7 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
10 2018-04-27 Richard Biener <rguenther@suse.de>
12 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
14 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
16 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
17 with Yd constraint. Set "preferred_for_speed" attribute from
18 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
20 (*movdi_internal): Ditto.
21 (movti_interunit splitters): Remove
22 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
23 (movdi_interunit splitters): Ditto.
24 * config/i386/constraints.md (Ye): Remove.
25 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
27 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30 * config/aarch64/constraints.md (Usg): Limit to 31.
33 2018-04-27 Jakub Jelinek <jakub@redhat.com>
35 PR tree-optimization/85529
36 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
37 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
38 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
39 zero extension or masking of the MSB bit.
40 (optimize_range_tests): Add FIRST_BB argument, pass it through
41 to optimize_range_tests_var_bound.
42 (maybe_optimize_range_tests, reassociate_bb): Adjust
43 optimize_range_tests callers.
45 2018-04-26 Richard Biener <rguenther@suse.de>
46 Jakub Jelinek <jakub@redhat.com>
48 * cgraph.h (symbol_table): Just declare debug method here.
49 * symtab.c (symbol_table::debug): Define.
51 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
53 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
55 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
57 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
58 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
59 (*movdi_internal): Substitute Yi and Yj constraint with x
60 and Ym and Yn constraint with y constraint. Update "isa"
61 attribute and set "preferred_for_speed" attribute from
62 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
63 (*movsi_internal): Ditto.
64 (*movdf_internal): Ditto.
65 (*movsf_internal): Ditto.
66 (*zero_extendsidi2): Ditto.
67 * config/i386/sse.md (vec_set<mode>_0): Ditto.
69 (*vec_extract<ssevecmodelower>_0): Ditto.
70 (*vec_extractv4si_0_zext_sse4): Ditto.
71 (vec_concatv2di): Ditto.
72 (*vec_dup<mode>): Ditto.
73 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
74 * config/i386/constraints.md (Yi): Remove.
79 2018-04-26 Nathan Sidwell <nathan@acm.org>
81 * dumpfile.c (dump_open): New.
82 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
83 (dump_finish): Detect stdio/stderr by value not name.
85 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
87 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
89 2018-04-26 Tom de Vries <tom@codesourcery.com>
92 * config/nvptx/nvptx.c (verify_neutering_jumps)
93 (verify_neutering_labels): New function
94 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
96 2018-04-26 Tom de Vries <tom@codesourcery.com>
99 * config/nvptx/nvptx.c (needs_neutering_p): New function.
100 (nvptx_single): Use needs_neutering_p to skip over insns that do not
103 2018-04-26 Richard Biener <rguenther@suse.de>
104 Tom de Vries <tom@codesourcery.com>
107 * lto-streamer-out.c (output_function): Fixup loops if required to match
108 discovery done in the reader.
110 2018-04-26 Richard Biener <rguenther@suse.de>
112 PR tree-optimization/85116
113 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
114 have a loop exit from the single latch predecessor. Remove
115 case of header with just condition.
116 (ch_base::copy_headers): Exclude infinite loops from any
118 (pass_ch::execute): Record exits.
120 2018-04-26 Richard Biener <rguenther@suse.de>
122 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
123 prologue cost vector and pass it to vect_get_load_cost.
124 (vect_get_peeling_costs_all_drs): Likewise.
125 (vect_peeling_hash_get_lowest_cost): Likewise.
126 (vect_enhance_data_refs_alignment): Likewise.
128 2018-04-26 Richard Biener <rguenther@suse.de>
131 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
132 checking of integer<->pointer conversions.
133 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
134 sign-/zero-extending pointer types.
135 (expand_omp_for_static_chunk): Likewise.
137 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
138 Jean Lee <xiaoyur347@gmail.com>
140 * config/mips/mips.c (mips_asan_shadow_offset): New function.
141 (TARGET_ASAN_SHADOW_OFFSET): Define.
142 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
143 true for -fsanitize=address.
145 2018-04-25 Mark Wielaard <mark@klomp.org>
147 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
150 2018-04-25 Jakub Jelinek <jakub@redhat.com>
152 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
153 than "alu", remove explicit "memory" and "imm_disp" attributes.
154 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
157 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
158 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
161 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
164 * config/i386/i386.c (ix86_expand_builtin): Change memory
165 operand to XI, extend p0 to Pmode.
166 * config/i386/i386.md: Change unspec volatile and operand
167 1 mode to XI, change operand 0 mode to P.
169 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
171 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
172 GET_MODE_MASK before any checking.
173 (nds32_can_use_bset_p): Likewise.
174 (nds32_can_use_btgl_p): Likewise.
176 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
178 * config/nds32/nds32-doubleword.md: New define_split pattern for
179 illegal register number.
181 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
183 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
185 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
187 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
189 2018-04-25 Richard Biener <rguenther@suse.de>
191 * lto-streamer.h (LTO_major_version): Bump to 8.
193 2018-04-25 Jakub Jelinek <jakub@redhat.com>
195 * BASE-VER: Set to 9.0.0.
197 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
199 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
200 in __abskf2 and __powikf2.
202 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
205 * config/aarch64/constraints.md (Usg, Usj): New constraints.
206 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
207 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
208 Use the above on operand 2. Reindent.
209 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
211 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
214 * common/config/i386/i386-common.c (ix86_handle_option): Don't
216 * config/i386/i386.opt (mcet): Removed.
217 * doc/install.texi: Remove -mcet documentation.
218 * doc/invoke.texi: Likewise.
220 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
223 * doc/install.texi: Remove -mcet from bootstrap-cet.
225 2018-04-24 Jakub Jelinek <jakub@redhat.com>
228 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
229 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
233 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
234 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
235 containing a CONST_VECTOR.
237 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
239 * doc/install.texi: Update newlib dependency for nvptx.
241 2018-04-24 Jakub Jelinek <jakub@redhat.com>
244 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
245 instead of INTVAL when shifting x left.
247 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
249 PR tree-optimization/85478
250 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
251 vect_grouped_store_supported for single element vectors.
253 2018-04-24 Richard Biener <rguenther@suse.de>
256 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
257 load cost increase to the case of non-constant step.
259 2018-04-24 Jakub Jelinek <jakub@redhat.com>
262 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
263 destination if any_malformed_asm.
265 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
268 * expr.c (store_field): In the bitfield case, if the value comes from
269 a function call and is returned in registers by means of a PARALLEL,
270 do not change the mode of the temporary unless BLKmode and VOIDmode.
272 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
274 PR rtl-optimization/85423
275 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
276 dependencies to debug insns when the previous insn is non-debug.
278 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
280 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
281 enums into a single definition.
282 (fls): Fix predicates and printing.
285 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
287 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
288 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
289 and short u6 immediate.
290 (check_if_valid_sleep_operand): Remove.
291 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
293 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
295 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
296 flag_always_save_lp condition.
297 * config/nds32/nds32.opt (malways-save-lp): New option.
299 2018-04-22 Shiva Chen <shiva0217@gmail.com>
301 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
302 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
303 * config/nds32/nds32.h
304 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
305 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
307 2018-04-22 Shiva Chen <shiva0217@gmail.com>
309 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
310 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
312 2018-04-22 Shiva Chen <shiva0217@gmail.com>
313 Chung-Ju Wu <jasonwucj@gmail.com>
315 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
317 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
318 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
320 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
322 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
324 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
326 * config/nds32/nds32-protos.h (nds32_data_alignment,
327 nds32_local_alignment): Declare.
328 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
329 nds32_local_alignment): New functions.
330 (TARGET_CONSTANT_ALIGNMENT): Define.
331 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
333 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
335 * config/nds32/nds32.c
336 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
337 (TARGET_MODES_TIEABLE_P): Likewise.
339 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
341 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
344 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
345 Chung-Ju Wu <jasonwucj@gmail.com>
347 * config/nds32/constants.md (unspec_volatile_element): Add enum values
348 for unaligned access.
349 * config/nds32/nds32-intrinsic.c: Implementation of expanding
351 * config/nds32/nds32-intrinsic.md: Likewise.
352 * config/nds32/nds32_intrinsic.h: Likewise.
353 * config/nds32/nds32.h (nds32_builtins): Likewise.
354 * config/nds32/nds32.opt (munaligned-access): New option.
355 * config/nds32/nds32.c (nds32_asm_file_start): Display
356 flag_unaligned_access status.
358 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
360 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
361 -mno-relax is present.
362 * config/riscv/linux.h (LINK_SPEC): Ditto.
364 2018-04-20 Martin Sebor <msebor@redhat.com>
367 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
369 (gimple_fold_builtin_stxcpy_chk): Same.
370 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
372 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
375 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
376 __powikf2 when long double is IEEE 128-bit.
378 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
380 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
381 step to make sure stack always aligned.
383 2018-04-20 Carl Love <cel@us.ibm.com>
386 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
389 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
390 Tom de Vries <tom@codesourcery.com>
393 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
394 Emit insns for calls too.
395 (nvptx_find_par): Always look for worker-level predecessor insn.
396 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
398 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
399 (nvptx_process_pars): Propagate frames for calls.
401 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
404 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
406 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
407 (ix86_handle_option): Don't handle OPT_mibt.
408 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
410 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
412 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
413 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
414 (ix86_target_macros): Define __CET__ with flag_cf_protection
416 * config/i386/i386.c (isa2_opts): Remove -mibt.
417 * config/i386/i386.h (TARGET_IBT): Removed.
418 (TARGET_IBT_P): Likewise.
419 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
420 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
421 * config/i386/i386.opt (mcet): Update help message.
424 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
425 -mcet as an alias for -mshstk.
427 2018-04-20 Richard Biener <rguenther@suse.de>
430 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
431 complexity by forcing a single use of the multiply operand.
433 2018-04-20 Martin Jambor <mjambor@suse.cz>
436 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
437 recursion dependency to only apply to non-clones.
439 2018-04-20 Martin Jambor <mjambor@suse.cz>
442 * ipa-cp.c (create_specialized_node): Check that clones of
443 self-recursive edges exist during IPA-CP.
445 2018-04-19 Toon Moene <toon@moene.org>
447 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
450 2018-04-19 Jakub Jelinek <jakub@redhat.com>
452 PR tree-optimization/85467
453 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
454 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
455 VECTOR_CST element to type.
457 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
460 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
461 * config/i386/i386.md (builtin_setjmp_setup): Removed.
462 (builtin_longjmp): Likewise.
463 (save_stack_nonlocal): New pattern.
464 (restore_stack_nonlocal): Likewise.
466 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
469 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
470 Replace ASM_OUTPUT_LABEL with fprintf.
472 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
475 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
476 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
477 * config/i386/i386-c.c (ix86_target_macros_internal): Also
478 define __IBT__ and __SHSTK__ for -fcf-protection.
479 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
481 (ix86_trampoline_init): Likewise.
482 (x86_output_mi_thunk): Likewise.
483 (ix86_notrack_prefixed_insn_p): Likewise.
484 (ix86_option_override_internal): Don't disallow -fcf-protection.
485 * config/i386/i386.md (rdssp<mode>): Also enable for
487 (incssp<mode>): Likewise.
488 (nop_endbr): Likewise.
489 * config/i386/i386.opt (mcet): Change help message to built-in
493 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
494 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
495 enable CET built-in functions.
497 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
499 * common/config/i386/i386-common.c
500 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
501 OPTION_MASK_ISA_MOVDIRI_UNSET,
502 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
503 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
504 * config.gcc (movdirintrin.h): New header.
505 * config/i386/cpuid.h (bit_MOVDIRI,
506 bit_MOVDIR64B): New bits.
507 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
509 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
510 (VOID, PVOID, PCVOID)): New function types.
511 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
512 __builtin_ia32_directstoreu_u64,
513 __builtin_ia32_movdir64b): New builtins.
514 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
515 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
517 (ix86_valid_target_attribute_inner_p): Ditto.
518 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
519 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
520 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
521 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
522 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
523 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
524 (movdiri<mode>, movdir64b_<mode>): New.
525 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
526 * config/i386/immintrin.h: Include movdirintrin.h.
527 * config/i386/movdirintrin.h: New file.
528 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
530 2018-04-19 Richard Biener <rguenther@suse.de>
533 * cfg.c (clear_bb_flags): When loop state says we have
534 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
536 2018-04-19 Richard Biener <rguenther@suse.de>
538 PR tree-optimization/84737
539 * tree-vect-data-refs.c (vect_copy_ref_info): New function
540 copying restrict info.
541 (vect_setup_realignment): Use it.
542 * tree-vectorizer.h (vect_copy_ref_info): Declare.
543 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
544 the first DR to all generated stores.
545 (vectorizable_load): Likewise for loads.
547 2018-04-19 Jakub Jelinek <jakub@redhat.com>
549 PR tree-optimization/85446
550 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
551 the integral and pointer types to have the same precision.
553 * doc/install.texi: Document --disable-cet being the default and
556 2018-04-18 Martin Liska <mliska@suse.cz>
558 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
561 2018-04-18 Martin Liska <mliska@suse.cz>
564 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
567 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
568 arguments if they are comparable.
570 2018-04-18 Martin Liska <mliska@suse.cz>
573 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
576 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
579 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
582 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
583 ENDBR after calling __morestack.
585 2018-04-18 David Malcolm <dmalcolm@redhat.com>
588 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
589 by using gcc_base_ver to generate a gcc_driver_version, and use
590 it when generating GCC_DRIVER_NAME.
591 * configure: Regenerate.
593 2018-04-18 Jakub Jelinek <jakub@redhat.com>
596 * config.gcc: Obsolete powerpc*-*-*spe*.
598 2018-04-17 Jakub Jelinek <jakub@redhat.com>
601 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
602 (stabstr_D): Change type of unum from unsigned int to
603 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
606 2018-04-17 Jim Wilson <jimw@sifive.com>
609 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
610 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
611 Set arg_pointer_offset after using pretend_args_size.
613 2018-04-17 Jakub Jelinek <jakub@redhat.com>
615 PR rtl-optimization/85431
616 * dse.c (record_store): Ignore zero width stores.
619 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
620 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
621 __builtin_stack_restore rather than after it.
622 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
623 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
624 argument instead of virtual_dynamic_stack_rtx.
626 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
628 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
630 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
631 Add note to error message to explain internal mapping of overloaded
632 built-in function name to non-overloaded built-in function name.
633 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
636 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
639 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
640 where the inputs overlap with the output.
642 2018-04-17 Jakub Jelinek <jakub@redhat.com>
644 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
645 (=v, v) alternative and explicit "memory" attribute.
646 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
647 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
649 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
650 "sselog1" type instead of "sselog".
651 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
652 "sselog". Remove explicit "memory" attribute.
653 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
654 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
656 (vec_extract_hi_v32hi): Merge all alternatives into one, use
657 "sselog1" type instead of "sselog". Remove explicit "memory"
659 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
660 use "sselog1" type instead of "sselog". Remove explicit "memory"
662 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
663 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
665 (vec_extract_hi_v64qi): Merge all alternatives into one, use
666 "sselog1" type instead of "sselog". Remove explicit "memory"
668 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
669 use "sselog1" type instead of "sselog". Remove explicit "memory"
673 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
676 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
679 2018-04-17 Martin Jambor <mjambor@suse.cz>
682 * ipa-cp.c (create_specialized_node): Call
683 expand_all_artificial_thunks if necessary.
685 2018-04-17 Martin Liska <mliska@suse.cz>
688 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
689 in message, remote space in between '_G' and '('.
691 2018-04-17 Jakub Jelinek <jakub@redhat.com>
694 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
695 avx512f_vmcmp<mode>3<round_saeonly_name>,
696 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
697 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
698 avx512f_rndscale<mode><round_saeonly_name>,
699 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
700 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
701 Use %<iptr>2 instead of %2 for -masm=intel.
702 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
703 avx512f_vcvttss2usi<round_saeonly_name>,
704 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
706 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
707 avx512f_vcvttsd2usi<round_saeonly_name>,
708 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
709 Use %q1 instead of %1 for -masm=intel.
710 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
711 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
712 of %3 for -masm=intel.
713 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
715 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
717 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
719 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
720 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
722 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
724 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
725 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
726 %g1 and one with %0 and %1.
727 (avx512er_vmrcp28<mode><round_saeonly_name>,
728 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
730 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
731 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
732 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
733 of %0 and %{%4%} for -masm=intel.
734 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
735 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
736 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
737 order of %0 and %{%5%}%{z%} for -masm=intel.
739 2018-04-17 Jan Hubicka <jh@suse.cz>
742 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
744 2018-04-17 Martin Liska <mliska@suse.cz>
747 * multiple_target.c (create_dispatcher_calls): Set apostrophes
748 for target_clone error message. Make default implementation
749 clone to be a local declaration.
750 (separate_attrs): Add new argument and check for an empty
752 (expand_target_clones): Handle it.
753 (ipa_target_clone): Make redirection just for target_clones
756 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
757 Tom de Vries <tom@codesourcery.com>
760 * omp-expand.c (expand_oacc_for): Add dummy false branch for
761 tiled basic blocks without omp continue statements.
763 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
766 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
767 vec_extract expression as having side effects to make sure it gets
770 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
773 * config/i386/i386.c (get_builtin_code_for_version): Check
776 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
779 * gcc/config.gcc: Support "skylake".
780 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
782 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
783 (processor_target_table): Add "skylake".
784 (ix86_option_override_internal): Add "skylake".
785 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
786 PROCESSOR_CANNONLAKE.
787 (get_builtin_code_for_version): Fix priority for
788 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
789 PROCESSOR_SKYLAKE-AVX512.
790 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
791 (processor_type): Add PROCESSOR_SKYLAKE.
793 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
794 Jason Merrill <jason@redhat.com>
797 * convert.c (convert_to_integer_1): Use direct recursion for
798 enumeral types and types with a precision less than the number
799 of bits in their mode.
801 2018-04-16 Julia Koval <julia.koval@intel.com>
804 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
805 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
807 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
810 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
811 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
812 and -mno-direct-move.
814 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
817 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
818 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
819 Ensure negative shifts result in {0}.
821 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
823 PR rtl-optimization/79916
824 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
825 regs (if any) to define how to gnerate SD moves when LRA is in
828 2018-04-13 Jakub Jelinek <jakub@redhat.com>
830 PR rtl-optimization/85393
831 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
832 * except.c (expand_dw2_landing_pad_for_region): Make static.
833 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
834 a label and unconditional jump to old_bb, rather than
835 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
838 PR rtl-optimization/85376
839 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
840 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
841 instead of a specific value.
843 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
844 Bin Cheng <bin.cheng@arm.com>
846 PR tree-optimization/82965
847 PR tree-optimization/83991
848 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
849 by_profile_only parameter.
850 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
851 information if the loop was predicted to iterate too many times.
852 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
854 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
857 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
860 2018-04-13 Martin Liska <mliska@suse.cz>
861 Jakub Jelinek <jakub@redhat.com>
864 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
865 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
866 * builtins.c (expand_builtin_memory_copy_args): Use
867 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
868 handle dest_addr == pc_rtx.
870 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
873 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
874 asked to not generate direct moves.
875 (fix_trunc<mode>si2_stfiwx): Similar.
876 (fix_trunc<mode>si2_internal): Similar.
878 2018-04-12 Jakub Jelinek <jakub@redhat.com>
881 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
882 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
883 lookup if dest in some wider mode is known to be const0_rtx and
884 if so, record permanent equivalence for it to be ZERO_EXTEND of
885 the narrower mode destination.
887 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
889 * lto-streamer-out.c (output_function): Revert 259346.
890 * omp-expand.c (expand_oacc_for): Likewise.
892 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
894 PR rtl-optimization/85354
895 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
896 * sel-sched.c (sel_global_init): ... here.
898 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
901 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
902 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
903 mode for PE-COFF targets.
904 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
905 (i386_pe_asm_lto_end): Likewise.
906 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
907 (TARGET_ASM_LTO_END): Likewise.
908 * config/i386/winnt.c (saved_debug_info_level): New static variable.
909 (i386_pe_asm_lto_start): New function.
910 (i386_pe_asm_lto_end): Likewise.
912 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
913 Richard Biener <rguenther@suse.de>
916 * lto-streamer-out.c (output_function): Fix CFG loop state before
918 * omp-expand.c (expand_oacc_for): Handle calls to internal
919 functions like regular functions.
921 2018-04-12 Richard Biener <rguenther@suse.de>
924 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
925 for the early LTO debug to properly generate references to it
926 during DIE emission. Do not re-use that for the skeleton for
928 (dwarf2out_early_finish): Likewise.
930 2018-04-12 Jakub Jelinek <jakub@redhat.com>
934 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
935 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
936 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
937 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
938 and output is a reg, avoid creating invalid lowpart subreg, but
939 instead split into a 512-bit move. Don't split if not AVX512VL,
940 input is xmm16+ reg and output is a mem.
941 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
942 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
943 xmm16+ reg and output is a mem.
945 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
947 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
948 also for flag_dwarf2_cfi_asm.
950 2018-04-12 Jakub Jelinek <jakub@redhat.com>
952 PR rtl-optimization/85342
953 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
954 a bool scalar var inside of the loop instead. Don't try to update
955 recog_data.operand after failed apply_change_group.
957 2018-04-12 Tom de Vries <tom@codesourcery.com>
960 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
961 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
962 array with flexible array member as array without given dimension.
963 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
964 argument for undefined param to true.
966 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
969 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
970 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
971 from PowerPC section.
972 * config/rs6000/sysv4.opt (mcall-): Improve help text.
973 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
974 help text that is too long.
975 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
976 help text that is too long.
977 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
978 help text that is too long.
980 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
982 * config/alpha/alpha.md (stack_probe_internal): Rename
983 from "probe_stack". Update all callers.
985 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
987 PR rtl-optimization/84566
988 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
989 sched_macro_fuse_insns.
991 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
994 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
995 (compute_block_dependences): ... from here.
997 2018-04-11 Jakub Jelinek <jakub@redhat.com>
999 PR tree-optimization/85331
1000 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
1001 from int to HOST_WIDE_INT.
1003 2018-04-11 Martin Jambor <mjambor@suse.cz>
1006 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
1007 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
1008 not the same as the source val.
1009 (cgraph_edge_brings_value_p): New parameter.
1010 (gather_edges_for_value): Pass destination value to
1011 cgraph_edge_brings_value_p.
1012 (perhaps_add_new_callers): Likewise.
1013 (get_info_about_necessary_edges): Likewise and exclude values brought
1014 only by self-recursive edges.
1015 (create_specialized_node): Redirect only clones of self-calling edges.
1016 (+self_recursive_pass_through_p): New function.
1017 (find_more_scalar_values_for_callers_subset): Use it.
1018 (find_aggregate_values_for_callers_subset): Likewise.
1019 (known_aggs_to_agg_replacement_list): Removed.
1020 (decide_whether_version_node): Re-calculate known constants for all
1021 remaining context clones.
1023 2018-04-11 Richard Biener <rguenther@suse.de>
1026 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
1027 from early DWARF output.
1028 (dwarf2out_early_finish): Output line info unconditionally into
1029 early DWARF and add reference to it.
1031 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1034 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
1035 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
1036 other than V2DFmode using iptr mode attribute.
1037 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
1039 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1041 PR rtl-optimization/84659
1042 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
1044 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1047 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
1049 (output_loc_list): Pass address of a dummy size variable even in the
1050 locview handling loop.
1051 (index_location_lists): Add comment on why skip_loc_list_entry can't
1054 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
1057 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
1060 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
1063 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1064 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
1065 and -mstring-compare-inline-limit.
1067 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1070 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
1071 for stack clash protection in a register whenever we need it to be in
1074 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1076 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
1077 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
1079 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1082 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
1084 (mlong-double-): Ditto.
1085 * config/rs6000/sysv4.opt (msdata=): Ditto.
1086 (mtls-size=): Ditto.
1088 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
1090 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1091 erroneous entries for
1092 "vector int vec_ldl (int, long int *)", and
1093 "vector unsigned int vec_ldl (int, unsigned long int *)".
1094 Add comments and entries for
1095 "vector bool char vec_ldl (int, bool char *)",
1096 "vector bool short vec_ldl (int, bool short *)",
1097 "vector bool int vec_ldl (int, bool int *)",
1098 "vector bool long long vec_ldl (int, bool long long *)",
1099 "vector pixel vec_ldl (int, pixel *)",
1100 "vector long long vec_ldl (int, long long *)",
1101 "vector unsigned long long vec_ldl (int, unsigned long long *)".
1102 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
1103 type tree bool_long_long_type_node and correct definition of
1104 bool_V2DI_type_node to make reference to this new type tree.
1105 (rs6000_mangle_type): Replace erroneous reference to
1106 bool_long_type_node with bool_long_long_type_node.
1107 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
1108 comments to emphasize sign distinctions for char and int types and
1109 replace RS6000_BTI_bool_long constant with
1110 RS6000_BTI_bool_long_long constant. Also add comment to restrict
1111 use of RS6000_BTI_pixel.
1112 (bool_long_type_node): Remove this macro definition.
1113 (bool_long_long_type_node): New macro definition
1115 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1117 PR rtl-optimization/85300
1118 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
1119 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
1120 simplify_unary_operation fails.
1122 2018-04-10 Martin Liska <mliska@suse.cz>
1124 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
1125 cgraph_edge and ipa_ref.
1127 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1131 * config/i386/sse.md
1132 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
1133 computation of the VEC_MERGE selector from mask.
1134 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
1135 Fix decoding of the VEC_MERGE selector into mask.
1137 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1139 PR tree-optimization/85286
1140 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
1142 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1144 * final.c (final_1): Set insn_last_address as well as
1145 insn_current_address.
1147 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1150 * explow.c (emit_stack_probe): Call validize_mem on memory location
1151 before passing it to gen_probe_stack. Create address operand and
1152 legitimize it for the probe_stack_address case.
1154 2018-04-09 Jan Hubicka <jh@suse.cz>
1157 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
1158 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
1159 * tree.c (free_lang_data_in_type): Fix handling of binfos;
1161 (free_lang_data): Rebuild type inheritance graph.
1163 2018-04-09 Martin Sebor <msebor@redhat.com>
1165 * invoke.texi (-finline-small-functions): Mention other optimization
1167 (-findirect-inlining, -fpartial-inlining): Same.
1168 (-finline-functions-called-once): Same.
1169 (-freorder-blocks-and-partition): Same.
1171 2018-04-09 Jan Hubicka <jh@suse.cz>
1174 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
1175 jumps; choose last target that matches the criteria (i.e.
1176 no partition changes for non-crossing jumps).
1177 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
1178 support for redirecting crossing jumps to non-crossing.
1180 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
1182 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
1183 also for naked functions.
1185 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
1187 * config/arc/arc.md (add_shift): New pattern.
1188 (add_shift2): Likewise.
1189 (sub_shift): Likewise.
1190 (sub_shift_cmp0_noout): Likewise.
1191 (compare_si_ashiftsi): Likewise.
1192 (xbfu_cmp0_noout): New combine pattern.
1193 (xbfu_cmp0"): Likewise.
1194 (movsi_set_cc_insn): Place the predicable variant first.
1195 (commutative_binary_cmp0_noout): Remove clobber.
1196 (commutative_binary_cmp0): New pattern.
1197 (noncommutative_binary_cmp0): Likewise.
1198 (noncommutative_binary_cmp0_noout): Likewise.
1199 (noncommutative_binary_comparison_result_used): Removed.
1200 (rsub_cmp0): New pattern.
1201 (rsub_cmp0_noout): Likewise.
1202 (extzvsi): Changed, keep only meaningful variants.
1203 (SQH, SEZ): New iterators.
1204 (SQH_postfix): New mode attribute.
1205 (SEZ_prefix): New code attribute.
1206 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
1207 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
1208 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
1210 (noncommutative_operator): Check the availability of barrel
1213 2018-04-09 Richard Biener <rguenther@suse.de>
1215 PR tree-optimization/85284
1216 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
1217 Only use the niter constraining form of simple_iv when the exit
1220 2018-04-09 Tom de Vries <tom@codesourcery.com>
1223 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1224 (define_expand "*memory_barrier"): New define_expand.
1225 (define_insn "memory_barrier"): New insn.
1227 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1229 PR rtl-optimization/80463
1230 PR rtl-optimization/83972
1231 PR rtl-optimization/83480
1233 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
1234 correct producer for the insn.
1235 (tidy_control_flow): Fixup seqnos in case of debug insns.
1237 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1239 PR rtl-optimization/83913
1241 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
1242 different sched-times when merging exprs.
1244 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1246 PR rtl-optimization/83962
1248 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
1249 tidy_fallthru_edge and tidy_control_flow.
1251 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1253 PR rtl-optimization/83530
1255 * sel-sched.c (force_next_insn): New global variable.
1256 (remove_insn_for_debug): When force_next_insn is true, also leave only
1257 next insn in the ready list.
1258 (sel_sched_region): When the region wasn't scheduled, make another pass
1259 over it with force_next_insn set to 1.
1261 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1263 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
1265 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1266 for interrupt control.
1267 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1268 functions for interrupt control.
1269 * config/nds32/nds32-intrinsic.md: Likewise.
1270 * config/nds32/nds32_intrinsic.h: Likewise.
1271 * config/nds32/nds32.h (nds32_builtins): Likewise.
1273 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1275 * config/nds32/nds32.c (nds32_init_machine_status,
1276 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
1277 strict_aligned_p field.
1278 (nds32_expand_to_rtl_hook): New function.
1279 (TARGET_EXPAND_TO_RTL_HOOK): Define.
1280 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
1282 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1283 Chung-Ju Wu <jasonwucj@gmail.com>
1285 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
1286 * config/nds32/nds32-n7.md: New file.
1287 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
1288 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
1290 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
1291 * config/nds32/nds32.md (pipeline_model): Add n7.
1292 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
1293 * config/nds32/pipelines.md: Include n7 settings.
1295 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1296 Chung-Ju Wu <jasonwucj@gmail.com>
1298 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
1299 * config/nds32/nds32-e8.md: New file.
1300 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
1301 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
1303 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
1304 * config/nds32/nds32.md (pipeline_model): Add e8.
1305 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
1306 * config/nds32/pipelines.md: Include e8 settings.
1308 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1309 Chung-Ju Wu <jasonwucj@gmail.com>
1311 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
1312 * config/nds32/nds32-n8.md: New file.
1313 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
1314 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
1316 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
1317 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
1318 * config/nds32/nds32.md (pipeline_model): Add n8.
1319 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
1320 * config/nds32/pipelines.md: Include n8 settings.
1322 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1323 Chung-Ju Wu <jasonwucj@gmail.com>
1325 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
1326 * config/nds32/nds32-n9-2r1w.md: New file.
1327 * config/nds32/nds32-n9-3r2w.md: New file.
1328 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
1329 nds32_register_ports): New or modify for cpu n9.
1330 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
1332 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
1333 * config/nds32/nds32-utils.c: New file.
1334 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
1335 TARGET_MUL_SLOW): Define.
1336 * config/nds32/nds32.md (pipeline_model): New attribute.
1337 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
1338 New options that support cpu n9.
1339 * config/nds32/pipelines.md: Include n9 settings.
1340 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
1342 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1344 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
1345 information if necessary.
1346 (output_cond_branch_compare_zero): Likewise.
1347 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
1348 (nds32_target_alignment): Refine for alignment.
1349 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
1350 (FUNCTION_BOUNDARY): Modify.
1351 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
1353 * config/nds32/nds32.opt (malways-align, malign-functions): New.
1355 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1357 * config/nds32/constants.md (unspec_volatile_element): Add values for
1358 TLB operation and data prefetch.
1359 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1360 functions for TLB operation and data prefetch.
1361 * config/nds32/nds32-intrinsic.md: Likewise.
1362 * config/nds32/nds32_intrinsic.h: Likewise.
1363 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
1364 (nds32_print_operand): Likewise.
1365 * config/nds32/nds32.h (nds32_builtins): Likewise.
1367 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
1368 Andrew Pinski <pinsika@gcc.gnu.org>
1371 * match.pd: Use constant_boolean_node of correct type instead of
1372 boolean_true_node or boolean_false_node for simplifying
1373 pointer comparisons to zero.
1375 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1377 PR tree-optimization/80021
1378 * tree.c (verify_type_variant): Make error call in verify_variant_match
1379 translatable and remove final full stop.
1381 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1383 * config/nds32/constants.md (unspec_volatile_element): Add
1384 UNSPEC_VOLATILE_EH_RETURN.
1385 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
1386 nds32_output_stack_pop): Support dwarf exception handling process.
1387 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
1388 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
1389 exception handling process.
1390 (nds32_compute_stack_frame): Likewise.
1391 (nds32_return_addr_rtx): Likewise.
1392 (nds32_initial_elimination_offset): Likewise.
1393 (nds32_expand_prologue): Likewise.
1394 (nds32_expand_epilogue): Likewise.
1395 (nds32_dynamic_chain_address): New function.
1396 * config/nds32/nds32.h (machine_function): Add fields for dwarf
1398 (DYNAMIC_CHAIN_ADDRESS): Define.
1399 (EH_RETURN_DATA_REGNO): Define.
1400 (EH_RETURN_STACKADJ_RTX): Define.
1401 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
1402 patterns for dwarf exception handling.
1404 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1406 * config/nds32/nds32.h: Clean up obsolete macros.
1408 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1410 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1411 Add enum values for particular instructions.
1412 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1413 particular intrinsic functions.
1414 * config/nds32/nds32-intrinsic.md: Likewise.
1415 * config/nds32/nds32_intrinsic.h: Likewise.
1416 * config/nds32/nds32.h (nds32_builtins): Likewise.
1417 * config/nds32/nds32.md (type): Add pbsad and pbsada.
1418 (btst, ave): New patterns for particular instructions.
1420 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1422 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1423 Add enum values for atomic load/store and memory sync.
1424 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
1426 * config/nds32/nds32-intrinsic.md: Likewise.
1427 * config/nds32/nds32_intrinsic.h: Likewise.
1428 * config/nds32/nds32.h (nds32_builtins): Likewise.
1430 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1432 PR tree-optimization/85257
1433 * fold-const.c (native_encode_vector): If not all elts could fit
1434 and off is -1, return 0 rather than offset.
1435 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
1436 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
1437 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
1438 adjust buffer in native_interpret_expr call.
1440 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1442 * config/nds32/constants.md (unspec_volatile_element): Add cache
1443 control enum values.
1444 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
1445 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
1446 * config/nds32/nds32.c (nds32_cctl_names): New.
1447 (nds32_print_operand): Handle cache control register names.
1448 * config/nds32/nds32.h (nds32_builtins): New enum values.
1449 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
1451 * config/nds32/nds32.md (type): Add mmu.
1452 * config/nds32/pipelines.md (simple_insn): Add mmu.
1454 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1456 * config/nds32/nds32.md (type): Remove call.
1457 * config/nds32/pipelines.md (simple_insn): Likewise.
1459 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1461 * config/nds32/constants.md (unspec_volatile_element): Add
1462 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
1463 UNSPEC_VOLATILE_FMFCFG.
1464 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
1465 description for fmfcfg and fmfcsr.
1466 (bdesc_1arg): Add fmtcsr.
1467 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
1468 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
1469 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
1470 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
1471 unspec_fmfcfg): New patterns.
1472 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
1473 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
1474 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
1475 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
1476 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
1477 __nds32__fmfcfg): Define.
1479 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1481 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
1482 intrinsic register names.
1483 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
1484 intrinsic register enum values and macros.
1486 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1488 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
1489 for load/store addressing form.
1490 (nds32_print_operand_address): Likewise.
1492 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
1495 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
1496 based on LABEL_REF. Remove useless assertion.
1497 (pic_address_needs_scratch): Fix formatting.
1498 (sparc_legitimize_pic_address): Minor tweaks.
1499 (sparc_delegitimize_address): Adjust assertion accordingly.
1500 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
1501 into symbolic_operand.
1502 (movsi_high_pic_label_ref): Likewise.
1503 (movsi_lo_sum_pic_label_ref): Likewise.
1504 (movdi_pic_label_ref): Likewise.
1505 (movdi_high_pic_label_ref): Likewise.
1506 (movdi_lo_sum_pic_label_ref): Likewise.
1508 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
1510 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
1511 custom LIB_SPEC setup.
1513 2018-04-06 Ruslan Bukin <br@bsdpad.com>
1514 Kito Cheng <kito.cheng@gmail.com>
1516 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
1517 * config/riscv/freebsd.h: New.
1519 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1521 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
1522 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
1525 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1526 Kito Cheng <kito.cheng@gmail.com>
1528 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
1529 nds32_output_call, nds32_symbol_binds_local_p): New functions.
1530 * config/nds32/nds32-protos.h (nds32_output_call,
1531 nds32_output_return): Declare.
1532 * config/nds32/nds32.md: Refine all the call and return patterns.
1534 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1537 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
1538 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
1540 PR rtl-optimization/84872
1541 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
1542 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
1545 2018-04-06 Tamar Christina <tamar.christina@arm.com>
1547 * expr.c (copy_blkmode_to_reg): Revert 254862.
1548 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
1550 2018-04-06 Richard Biener <rguenther@suse.de>
1553 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
1554 after seeing a component reference with an adjacent field. Treat
1555 refs to arrays at struct end of external decls similar to
1556 refs to unconstrained commons.
1558 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1561 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
1562 look through SAVE_EXPRs with non-side-effects argument. Adjust
1564 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
1567 2018-04-06 Richard Biener <rguenther@suse.de>
1570 * alias.c (find_base_term): New wrapper around find_base_term
1571 unwinding CSELIB_VAL_PTR changes.
1572 (find_base_term): Do not restore CSELIB_VAL_PTR during the
1575 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1577 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
1579 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
1580 constant definitions.
1581 ("nop"): lr 0,0 -> nopr r0
1582 ("nop_lr0", "nop_lr1"): New insn definitions.
1584 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1586 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
1587 NDS32_V3PUSH_AVAILABLE_P macro.
1589 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
1590 Chung-Ju Wu <jasonwucj@gmail.com>
1592 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
1593 (nds32*-*-*): Add float and fpu_config into supported_defaults.
1594 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
1595 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
1596 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
1597 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
1598 * config/nds32/constraints.md: New constraints and checking for hard
1599 float configuration.
1600 * config/nds32/iterators.md: New mode iterator and attribute for hard
1601 float configuration.
1602 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
1604 * config/nds32/nds32-fpu.md: New file.
1605 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
1606 deal with hard float code generation.
1607 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
1609 (abi_type, float_reg_number): New enum type.
1610 * config/nds32/nds32-predicates.c: New predicates for hard float.
1611 * config/nds32/nds32-protos.h: Declare functions for hard float.
1612 * config/nds32/nds32.c: Implementation for hard float configuration.
1613 * config/nds32/nds32.h: Definitions for hard float configuration.
1614 * config/nds32/nds32.md: Include hard float machine description and
1615 modify patterns for hard float configuration.
1616 * config/nds32/nds32.opt: New options for hard float configuration.
1617 * config/nds32/predicates.md: New predicates for hard float
1620 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1622 * common/config/nds32/nds32-common.c
1623 (nds32_option_optimization_table): Enable -mreleax-hint by default.
1625 2018-04-05 Jakub Jelinek <jakub@redhat.com>
1628 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
1629 CONSTRUCTOR_ELT (ctor, ...)->value.
1631 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
1634 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
1636 2018-04-05 Tom de Vries <tom@codesourcery.com>
1639 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
1642 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1643 Kito Cheng <kito.cheng@gmail.com>
1645 * config/nds32/constraints.md (U33): Fine-tune checking condition.
1646 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
1647 * config/nds32/nds32.h (nds32_16bit_address_type): Add
1648 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
1650 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1651 Kito Cheng <kito.cheng@gmail.com>
1653 * config/nds32/constraints.md (Ufe): New memory constraint.
1654 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
1655 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
1656 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
1658 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
1659 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
1661 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1663 * config/nds32/nds32.md: Use optimize_size in the condition for
1664 alu-shift instructions.
1666 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1668 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
1670 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1672 * config/nds32/nds32.md (negsi2): Refine pattern.
1674 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1675 Chung-Ju Wu <jasonwucj@gmail.com>
1677 * config/nds32/iterators.md (shift_rotate): New code iterator.
1678 (shift): New code attribute.
1679 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
1680 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
1681 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
1682 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
1683 bit-wise operations.
1684 (andsi3, *andsi3): Ditto.
1685 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
1686 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
1687 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
1688 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
1689 nds32_ior_operand, nds32_xor_operand): New predicates.
1691 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1693 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
1694 (addsi3, subsi3): ... this.
1696 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1698 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
1700 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1702 * config/nds32/nds32.md: Adjust indention.
1704 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1706 * config/nds32/nds32.md (feature): New attribute.
1708 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1710 * config/nds32/nds32.md (subtype): New attribute.
1712 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
1715 * config/arm/arm-builtins.c (arm_expand_builtin): Change
1716 expansion to perform a bitwise AND of the argument followed by a
1717 boolean negation of the result.
1719 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
1721 PR rtl-optimization/84878
1722 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
1723 the basic block. Assert the use reference is not artificial and that
1724 it has an associated insn.
1726 2018-04-04 Michael Matz <matz@suse.de>
1728 * builtins.c (compute_objsize): Pass correct operand
1729 to array_at_struct_end_p.
1731 2018-04-04 Richard Biener <rguenther@suse.de>
1734 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
1735 from contexts for DINFO_LEVEL_TERSE and below.
1737 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1739 * config/nds32/nds32-doubleword.md (move_<mode>): Require
1740 resiter_operand condition.
1741 * config/nds32/nds32.md (*move<mode>): Ditto.
1743 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1744 Monk Chiang <sh.chiang04@gmail.com>
1746 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
1748 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1750 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
1752 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1753 Kito Cheng <kito.cheng@gmail.com>
1755 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
1756 nds32_cond_code_str, output_cond_branch,
1757 output_cond_branch_compare_zero, nds32_expand_cbranch,
1758 nds32_expand_cstore, nds32_expand_movcc,
1759 nds32_output_cbranchsi4_equality_zero,
1760 nds32_output_cbranchsi4_equality_reg,
1761 nds32_output_cbranchsi4_equality_reg_or_const_int,
1762 nds32_output_cbranchsi4_greater_less_zero: New functions.
1763 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
1764 nds32_expand_cstore, nds32_expand_movcc,
1765 nds32_output_cbranchsi4_equality_zero,
1766 nds32_output_cbranchsi4_equality_reg,
1767 nds32_output_cbranchsi4_equality_reg_or_const_int,
1768 nds32_output_cbranchsi4_greater_less_zero): Declare.
1769 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
1770 nds32_rimm11s_operand): New predicates.
1771 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
1772 * config/nds32/nds32.md: Rewrite all the branch and conditional move
1775 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1777 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
1778 * config/nds32/nds32.md: Ditto.
1779 * config/nds32/pipelines.md: Ditto.
1781 2018-04-04 Richard Biener <rguenther@suse.de>
1783 PR tree-optimization/85168
1784 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
1785 propagating abnormals.
1787 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1789 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
1791 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1792 Kito Cheng <kito.cheng@gmail.com>
1794 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
1795 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
1796 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
1797 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
1798 * config/nds32/nds32.md (sibcall_internal): New.
1799 (sibcall_register): Remove.
1800 (sibcall_immediate): Remove.
1801 (sibcall_value_internal): New.
1802 (sibcall_value_register): Remove.
1803 (sibcall_value_immediate): Remove.
1804 * config/nds32/predicates.md (nds32_general_register_operand): New.
1805 (nds32_call_address_operand): New.
1807 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1809 PR rtl-optimization/85167
1810 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
1811 bb_defs if *split_p, instead preinitialize it to NULL.
1813 PR tree-optimization/85156
1814 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
1815 evaluating the argument multiple times.
1817 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
1819 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
1821 (_mm_cvtpd_ps): Likewise.
1822 (_mm_cvttpd_epi32): Likewise.
1823 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
1824 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
1825 vector, pixel, and bool following altivec.h include.
1827 2018-04-03 Martin Sebor <msebor@redhat.com>
1829 * doc/extend.texi (Common Function Attributes): Clarify.
1830 (const attribute): Likewise.
1831 (pure attribute): Likewise.
1833 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1836 * config/i386/i386.c (ix86_expand_vector_set): Use
1837 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
1839 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
1841 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
1842 instructions when changing rounding bits to preserve precision bits
1843 in the x87 control word.
1845 2018-04-03 Martin Liska <mliska@suse.cz>
1847 PR tree-optimization/82491
1848 * rtl.h (strip_offset_and_add): Replace += suboffset with
1849 poly_uint64 () + suboffset.
1851 2018-03-29 Martin Liska <mliska@suse.cz>
1852 Martin Jambor <mjambor@suse.cz>
1855 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
1856 param_type is not an integral or pointer type.
1858 2018-04-03 Richard Biener <rguenther@suse.de>
1860 * sese.h (recompute_all_dominators): Remove.
1862 2018-04-02 Martin Sebor <msebor@redhat.com>
1864 * doc/invoke.texi (-Wrestrict): Fix typos.
1866 2018-04-02 Jim Wilson <jimw@sifive.com>
1868 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
1869 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
1870 (<optab>di3, <optab>si3_extend): Likewise.
1871 (<optab>si3_mask, <optab>si3_mask_1): New.
1872 (<optab>di3_mask, <optab>di3_mask_1): New.
1873 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
1874 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
1875 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
1877 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
1879 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
1882 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
1884 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
1885 (nds32_canonicalize_comparison): New function.
1887 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1888 Kito Cheng <kito.cheng@gmail.com>
1889 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1891 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
1892 * config/nds32/constants.md (unspec_volatile_element): Add
1893 UNSPEC_VOLATILE_RELAX_GROUP.
1894 * config/nds32/nds32-relax-opt.c: New file.
1895 * config/nds32/nds32-predicates.c
1896 (nds32_symbol_load_store_p): New function.
1897 * config/nds32/nds32-protos.h
1898 (nds32_symbol_load_store_p): Declare function.
1899 (make_pass_nds32_relax_opt): Declare new rtl pass function.
1900 * config/nds32/nds32.c
1901 (nds32_register_pass): New function to register pass.
1902 (nds32_register_passes): New function to register passes.
1903 * config/nds32/nds32.md (relax_group): New pattern.
1904 * config/nds32/nds32.opt (mrelax-hint): New option.
1905 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
1907 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
1909 * config/nds32/t-nds32: Modify files dependency.
1911 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1913 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
1914 (PROFILE_HOOK): Define its implementation.
1916 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1918 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
1919 type and 32-bit size.
1921 2018-04-01 Jakub Jelinek <jakub@redhat.com>
1924 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
1925 (V_128_256): New mode iterator.
1926 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
1927 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
1928 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
1930 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
1931 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
1933 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
1936 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
1937 NaN inputs correctly.
1939 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
1942 * config/rs6000/vsx.md (??r): New mode attribute.
1943 (*vsx_mov<mode>_64bit): Use it.
1944 (*vsx_mov<mode>_32bit): Likewise.
1946 2018-03-30 Martin Sebor <msebor@redhat.com>
1948 PR tree-optimization/84818
1949 * builtins.c (check_access): Use warning_n.
1951 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1954 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
1956 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
1959 2018-03-30 Julia Koval <julia.koval@intel.com>
1962 * x86-tune.def (movx, partial_reg_dependency): Enable for
1965 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
1968 * lra-constraints.c (process_alt_operands): Move setting
1969 this_alternative_matches below.
1971 2018-03-29 Martin Liska <mliska@suse.cz>
1974 * doc/invoke.texi: Document how LTO works with debug info.
1975 Describe auto-load support of binutils. Mention 'x86-64'
1976 as valid option value of -march option.
1978 2018-03-29 Jakub Jelinek <jakub@redhat.com>
1980 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
1983 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
1984 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
1985 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
1988 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1991 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
1992 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
1993 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
1994 for RS6000_BTM_POWERPC64.
1995 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
1996 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
1997 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
2002 2018-03-28 Carl Love <cel@us.ibm.com>
2005 2017-09-27 Carl Love <cel@us.ibm.com>
2007 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
2008 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
2009 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
2012 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2014 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
2015 instead of __vector bool.
2016 (_mm_max_pu8): Likewise.
2017 (_mm_min_pi16): Likewise.
2019 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2022 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
2023 (DIVWEUO): Likewise.
2025 (DIVDEUO): Likewise.
2026 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
2027 DIVWEUO and DIVDEUO.
2028 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
2029 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
2030 (div_extend): Likewise.
2031 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
2033 (__builtin_divweuo): Likewise.
2034 (__builtin_divdeo): Likewise.
2035 (__builtin_divdeuo): Likewise.
2037 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2040 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
2041 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
2043 PR tree-optimization/82004
2044 * gimple-match-head.c (optimize_pow_to_exp): New function.
2045 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
2046 Don't fold to exp if optimize_pow_to_exp is false.
2048 2018-03-28 Martin Liska <mliska@suse.cz>
2051 * calls.c (initialize_argument_information): Fix trailing space.
2052 * common.opt: Fix typo and provide better explanation for
2053 -fsanitize-coverage option.
2054 * config/i386/i386.opt: Fix typo.
2056 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2057 Martin Liska <mliska@suse.cz>
2060 * gimplify.c (asan_poison_variable): Don't do the check for
2061 gimplify_omp_ctxp here.
2062 (gimplify_decl_expr): Do it here.
2063 (gimplify_target_expr): Likewise.
2065 2018-03-28 Martin Liska <mliska@suse.cz>
2068 * config/i386/i386.c (ix86_function_arg_advance): Do not call
2069 chkp_type_bounds_count if MPX is not enabled.
2071 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
2073 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
2075 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
2078 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
2079 function to create the function decl for complex long double
2080 multiply and divide for -mabi=ieeelongdouble.
2081 (init_float128_ieee): Call it.
2083 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
2086 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
2087 -fcf-protection=branch -mibt.
2088 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
2090 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2093 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
2095 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
2098 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
2099 extern array declarations.
2101 2018-03-27 Richard Biener <rguenther@suse.de>
2104 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
2105 explicit single_use checks.
2107 2018-03-27 Richard Biener <rguenther@suse.de>
2109 PR tree-optimization/85082
2110 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
2113 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2115 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
2116 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
2117 Turn on fasynchronous-unwind-tables and funwind-tables.
2119 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
2122 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
2123 (*bmi_blsr_<mode>_ccz): Ditto.
2125 2018-03-26 Tom de Vries <tom@codesourcery.com>
2127 PR tree-optimization/85063
2128 * omp-general.c (offloading_function_p): New function. Factor out
2130 * omp-offload.c (pass_omp_target_link::gate): ... here.
2131 * omp-general.h (offloading_function_p): Declare.
2132 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
2133 with attribute omp declare target for offloading functions.
2135 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
2137 PR tree-optimization/84005
2138 * tree-data-ref.h (get_base_for_alignment): Declare.
2139 * tree-data-ref.c (get_base_for_alignment_1): New function.
2140 (get_base_for_alignment): Likewise.
2141 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
2142 get_base_for_alignment to find a suitable base object, instead
2143 of always using drb->base_address.
2145 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2148 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
2149 known size by default.
2151 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
2154 * lra-constraints.c (process_alt_operands): Don't match BLKmode
2155 and non BLKmode operands.
2157 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2160 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
2161 Clean up attributes.
2163 2018-03-23 Richard Biener <rguenther@suse.de>
2166 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
2167 we are going to emit early debug for LTO.
2169 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2172 * function.c (match_asm_constraints_1): Don't optimize if input
2173 doesn't satisfy general_operand predicate for output's mode.
2176 * alias.c (write_dependence_p): Don't require for x_canonicalized
2177 non-VOIDmode if x has VOIDmode.
2180 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
2181 just don't try to optimize it rather than assert it never happens.
2183 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2185 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
2186 macro expansions for definition of ST_INTERNAL_<mode> and
2187 LD_INTERNAL_<mode> builtins.
2188 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
2190 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
2192 (altivec_expand_st_builtin): Likewise.
2193 (altivec_expand_builtin): Remove calls to deleted functions.
2194 (rs6000_address_for_altivec): Delete this function.
2195 * config/rs6000/vector.md: Remove expands for
2196 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
2198 2018-03-22 Sudakshina Das <sudi.das@arm.com>
2201 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
2202 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
2203 re-computing once computed.
2204 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
2205 (arm_init_machine_status): Initialize
2206 machine->static_chain_stack_bytes.
2208 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2211 * doc/extend.texi: Add four new prototypes for vec_ld.
2212 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
2213 definitions for more logical presentation.
2214 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
2215 entries for V1TI variants of __builtin_altivec_ld builtin.
2216 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
2217 handling of V1TI variant of LVX icode pattern.
2218 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
2219 (rs6000_gimple_fold_builtin): Likewise.
2220 (altivec_init_builtins): Add code to define
2221 __builtin_altivec_lvx_v1ti function.
2223 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2226 * function.c (match_asm_constraints_1): Don't do the optimization
2227 if input isn't a REG, SUBREG, MEM or constant.
2229 2018-03-22 Tom de Vries <tom@codesourcery.com>
2231 PR tree-optimization/84956
2232 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
2233 bb_has_abnormal_pred.
2235 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2238 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
2239 DECL_INITIAL (decl) to decl at the end.
2240 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
2243 2018-03-21 Joseph Myers <joseph@codesourcery.com>
2245 * doc/extend.texi (__builtin_tgmath): Document when complex
2246 integer types are treated as _Complex _Float64.
2248 2018-03-21 Tom de Vries <tom@codesourcery.com>
2250 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
2252 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2254 PR tree-optimization/84960
2255 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
2256 if it is ENTRY block, move them into single succ of ENTRY in that case.
2258 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
2260 PR tree-optimization/84811
2261 * poly-int.h (poly_span_traits): Remove the T3 parameter and
2262 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
2263 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
2264 (known_subrange_p): Update accordingly. Cast each value involved
2265 in the size comparison, rather than casting the result of the
2268 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2270 PR tree-optimization/84982
2271 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
2272 by flipping the least significant bit rather than all bits from
2273 bitpos to bitpos + bitsize - 1.
2275 2018-03-21 Nathan Sidwell <nathan@acm.org>
2277 * doc/extend.texi (Deprecated Features): Remove mention of
2278 long-deleted deprecations.
2280 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2283 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
2284 * configure: Regenerate.
2286 2018-03-21 Tom de Vries <tom@codesourcery.com>
2288 PR tree-optimization/83126
2289 * tree-parloops.c (num_phis): New function.
2290 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
2292 2018-03-21 Nathan Sidwell <nathan@acm.org>
2294 * doc/extend.texi (Deprecated Features): Update deprecated flags,
2295 mention anon-struct/union members and trailing attributes.
2297 2018-03-21 Bin Cheng <bin.cheng@arm.com>
2299 PR tree-optimization/84969
2300 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
2301 builtin memset partitions if they set different rhs values.
2303 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2305 PR rtl-optimization/84989
2306 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
2307 VEC_DUPLICATE with scalar result mode.
2309 2018-03-21 Martin Liska <mliska@suse.cz>
2312 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
2313 not intended return statement.
2315 2018-03-21 Martin Liska <mliska@suse.cz>
2318 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
2319 (chkp_find_bound_slots_1): Limit number of iterations.
2321 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
2324 * Minor grammar fixes for x86 options.
2326 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2329 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
2330 holding REG_CFA_RESTORE notes, instead turn them into a USE.
2332 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
2335 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
2336 (altivec_lvx_<mode>_1op): Likewise.
2337 (altivec_stvx_<mode>_2op): Likewise.
2338 (altivec_stvx_<mode>_1op): Likewise.
2339 (altivec_lvx_<VM2:mode>): New define_expand.
2340 (altivec_stvx_<VM2:mode>): Likewise.
2341 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
2342 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2343 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
2344 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2345 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
2346 (rs6000_gen_lvx): Likewise.
2347 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
2348 (altivec_expand_stv_builtin): Likewise.
2349 (altivec_expand_builtin): Likewise.
2350 * config/rs6000/vector.md: Likewise.
2352 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2355 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
2358 2018-03-20 Richard Biener <rguenther@suse.de>
2361 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
2362 sign-conversions as zero, fall back to standard scalar_stmt
2365 2018-03-20 Martin Liska <mliska@suse.cz>
2368 * predict.c (rebuild_frequencies): Handle case when we have
2369 PROFILE_ABSENT, but flag_guess_branch_prob is false.
2371 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2374 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
2375 flag_section_anchors.
2376 * varasm.c (use_blocks_for_decl_p): Remove hack for
2377 dw2_force_const_mem.
2380 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
2382 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
2383 be created, use lowpart_subreg of operands[0] rather than operands[0]
2385 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
2386 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
2387 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
2388 and n constraint instead of aarch64_shift_imm_di and Usd.
2389 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
2390 (*aarch64_<optab>_reg_minus<mode>3): ... this.
2392 2018-03-20 Sudakshina Das <sudi.das@arm.com>
2395 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
2396 to favor GPR over NEON registers.
2397 (<shift>di3_neon): Likewise.
2399 2018-03-20 Tom de Vries <tom@codesourcery.com>
2402 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
2403 (nvptx_process_pars): Emit bar.sync asap and alap.
2405 2018-03-20 Tom de Vries <tom@codesourcery.com>
2408 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
2409 seen_label if seen_label is already set.
2411 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2414 * config/i386/i386.c (fold_builtin_cpu): For features above 31
2415 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
2416 Use 1U instead of 1. Formatting fixes.
2419 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
2420 instead of TREE_TYPE (s1) for the return value.
2422 2018-03-19 Jakub Jelinek <jakub@redhat.com>
2424 PR tree-optimization/84946
2425 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
2426 bitsize + bitsize in poly_uint64 rather than poly_int64.
2429 * dwarf2asm.c: Include fold-const.c.
2430 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
2431 of decl rather than decl itself.
2433 PR rtl-optimization/84643
2434 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
2436 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
2439 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
2440 calling assemble_variable.
2442 2018-03-19 Sudakshina Das <sudi.das@arm.com>
2445 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
2446 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
2448 2018-03-19 Jim Wilson <jimw@sifive.com>
2451 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
2452 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
2453 (riscv_first_stack_step): Likewise.
2454 (riscv_option_override): Use STACK_BOUNDARY instead of
2456 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
2458 (BIGGEST_ALIGNMENT): Set to 128.
2459 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
2460 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
2463 2018-03-19 Richard Biener <rguenther@suse.de>
2465 PR tree-optimization/84933
2466 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
2467 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
2469 2018-03-19 Richard Biener <rguenther@suse.de>
2471 PR tree-optimization/84859
2472 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
2473 (cond_if_else_store_replacement): Perform sinking operation on
2474 single-store BBs regardless of MAX_STORES_TO_SINK setting.
2475 Generalize what a BB with a single eligible store is.
2477 2018-03-19 Richard Biener <rguenther@suse.de>
2479 PR tree-optimization/84929
2480 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
2481 chrec_is_positive against non-chrec arg.
2483 2018-03-19 Tamar Christina <tamar.christina@arm.com>
2486 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
2488 2018-03-18 Martin Liska <mliska@suse.cz>
2490 PR rtl-optimization/84635
2491 * regrename.c (build_def_use): Use matches_mode only when
2494 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
2496 PR tree-optimization/84913
2497 * tree-vect-loop.c (vectorizable_reduction): Don't try to
2498 vectorize chains of COND_EXPRs.
2500 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2502 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
2504 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2506 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
2508 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2510 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
2512 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2513 Kito Cheng <kito.cheng@gmail.com>
2515 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
2516 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
2517 (nds32_adjust_reg_alloc_order): New function.
2518 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
2520 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
2522 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
2523 nds32_print_operand, nds32_print_operand_address): Use
2524 HOST_WIDE_INT_PRINT_DEC instead.
2526 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2528 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
2530 2018-03-17 Jakub Jelinek <jakub@redhat.com>
2533 * config/i386/i386.c (initial_ix86_tune_features,
2534 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
2536 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
2537 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
2538 rather than 1u << ix86_tune. Formatting fix.
2539 (ix86_option_override_internal): Change ix86_arch_mask from
2540 unsigned int to unsigned HOST_WIDE_INT, initialize to
2541 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
2542 (ix86_function_specific_restore): Likewise.
2544 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2547 * postreload.c (reload_combine_recognize_pattern): Perform
2548 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
2549 truncate_int_for_mode the result for the destination's mode.
2552 * hsa-gen.c (mem_type_for_type): Fix comment typo.
2553 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
2555 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2558 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
2561 * lra-assigns.c (lra_split_hard_reg_for): Don't use
2562 regno_allocno_class_array and sorted_pseudos.
2563 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
2564 insns where regno is used.
2566 2018-03-16 Martin Liska <mliska@suse.cz>
2569 * multiple_target.c (create_dispatcher_calls): Redirect
2570 reference in the symbol table.
2572 2018-03-16 Martin Liska <mliska@suse.cz>
2575 * multiple_target.c (create_dispatcher_calls): Redirect also
2578 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2582 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
2583 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
2584 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
2586 2018-03-16 Julia Koval <julia.koval@intel.com>
2588 * doc/invoke.texi (Skylake Server): Add CLWB.
2589 Cannonlake): Remove CLWB.
2591 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2593 PR tree-optimization/84841
2594 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
2596 (FLOAT_ONE_CONST_TYPE): Define.
2597 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
2598 (sort_by_operand_rank): Put entries with higher constant_type last
2599 rather than first to match comments.
2601 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
2603 * config/nios2/nios2.md (movsi_internal): Fix thinko in
2606 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2609 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
2610 check and use address of target always.
2612 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
2615 * config/i386/i386.c (indirect_thunk_needed): Update comments.
2616 (indirect_thunk_bnd_needed): Likewise.
2617 (indirect_thunks_used): Likewise.
2618 (indirect_thunks_bnd_used): Likewise.
2619 (indirect_return_needed): New.
2620 (indirect_return_bnd_needed): Likewise.
2621 (output_indirect_thunk_function): Add a bool argument for
2623 (output_indirect_thunk_function): Don't generate alias for
2624 function return thunk.
2625 (ix86_code_end): Call output_indirect_thunk_function to generate
2626 function return thunks.
2627 (ix86_output_function_return): Set indirect_return_bnd_needed
2628 and indirect_return_needed instead of indirect_thunk_bnd_needed
2629 and indirect_thunk_needed.
2631 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
2633 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
2634 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
2635 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
2637 2018-03-15 David Malcolm <dmalcolm@redhat.com>
2638 Paul Hua <paul.hua.gm@gmail.com>
2641 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
2643 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
2645 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
2646 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
2649 2018-03-15 Tamar Christina <tamar.christina@arm.com>
2652 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
2653 instead of GET_MODE_SIZE when comparing Units.
2655 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2658 * varasm.c (hash_section): Return an unchangeble hash value
2659 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
2660 Return !aarch64_can_use_per_function_literal_pools_p ().
2662 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2665 * optabs.c (emit_conditional_move): Pass address of cmode's copy
2666 rather than address of cmode as last argument to prepare_cmp_insn.
2668 2018-03-15 Julia Koval <julia.koval@intel.com>
2670 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
2671 F_AVX512VNNI, F_AVX512BITALG): New.
2673 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
2676 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
2677 insn for floating-point loads and stores.
2679 2018-03-14 Carl Love <cel@us.ibm.com>
2681 * config/rs6000/rs6000-c.c: Add macro definitions for
2682 ALTIVEC_BUILTIN_VEC_PERMXOR.
2683 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
2684 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
2685 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
2686 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
2688 * config/doc/extend.texi: Add prototypes for vec_permxor.
2690 2018-03-14 David Malcolm <dmalcolm@redhat.com>
2693 * diagnostic-show-locus.c (class layout_point): Convert m_line
2694 from int to linenum_type.
2695 (line_span::comparator): Use linenum "compare" function when
2696 comparing line numbers.
2697 (test_line_span): New function.
2698 (layout_range::contains_point): Convert param "row" from int to
2700 (layout_range::intersects_line_p): Likewise.
2701 (layout::will_show_line_p): Likewise.
2702 (layout::print_source_line): Likewise.
2703 (layout::should_print_annotation_line_p): Likewise.
2704 (layout::print_annotation_line): Likewise.
2705 (layout::print_leading_fixits): Likewise.
2706 (layout::annotation_line_showed_range_p): Likewise.
2707 (struct line_corrections): Likewise for field m_row.
2708 (line_corrections::line_corrections): Likewise for param "row".
2709 (layout::print_trailing_fixits): Likewise.
2710 (layout::get_state_at_point): Likewise.
2711 (layout::get_x_bound_for_row): Likewise.
2712 (layout::print_line): Likewise.
2713 (diagnostic_show_locus): Likewise for locals "last_line" and
2715 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
2716 * input.c (selftest::test_linenum_comparisons): New function.
2717 (selftest::input_c_tests): Call it.
2718 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
2719 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
2720 * selftest.h (ASSERT_GT): New macro.
2721 (ASSERT_GT_AT): New macro.
2722 (ASSERT_LT): New macro.
2723 (ASSERT_LT_AT): New macro.
2725 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
2727 PR rtl-optimization/84780
2728 * combine.c (distribute_links): Don't make a link based on pc_rtx.
2730 2018-03-14 Martin Liska <mliska@suse.cz>
2732 * tree.c (record_node_allocation_statistics): Use
2733 get_stats_node_kind.
2734 (get_stats_node_kind): New function extracted from
2735 record_node_allocation_statistics.
2736 (free_node): Use get_stats_node_kind.
2738 2018-03-14 Richard Biener <rguenther@suse.de>
2740 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
2741 that the value-set of ANTIC_IN doesn't grow.
2744 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2746 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2747 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2749 2018-03-14 Julia Koval <julia.koval@intel.com>
2751 * config.gcc (icelake-client, icelake-server): New.
2753 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
2754 (initial_ix86_arch_features): Ditto.
2755 (PTA_SKYLAKE): Add SGX.
2756 (PTA_ICELAKE): Remove.
2757 (PTA_ICELAKE_CLIENT): New.
2758 (PTA_ICELAKE_SERVER): New.
2759 (ix86_option_override_internal): Split up icelake on icelake client and
2761 (get_builtin_code_for_version): Ditto.
2762 (fold_builtin_cpu): Ditto.
2763 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
2764 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
2765 * config/i386/i386.h (processor_type): Ditto.
2766 * doc/invoke.texi: Ditto.
2768 2018-03-14 Jakub Jelinek <jakub@redhat.com>
2771 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
2772 INTEGER_CST offset, add it together with bitpos / 8 and
2773 sign extend based on POINTER_SIZE.
2777 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
2780 * config/i386/constraints.md (Yc): New register constraint.
2781 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
2782 Use Yc constraint for alternative 2 of operand 0. Remove
2783 preferred_for_speed attribute.
2785 2018-03-14 Richard Biener <rguenther@suse.de>
2787 PR tree-optimization/84830
2788 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
2789 with the old one to avoid oscillations.
2791 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
2794 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
2796 (assign_by_spills): Return a flag of reload assignment failure.
2797 Do not process the reload assignment failures. Do not spill other
2798 reload pseudos if they has the same reg class. Update n if
2800 (lra_assign): Add a return arg. Set up from the result of
2801 assign_by_spills call.
2802 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2803 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2804 usage_insns if it is not NULL.
2805 (spill_hard_reg_in_range): New function.
2806 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2807 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2808 function prototypes.
2809 (lra_assign): Change prototype.
2810 * lra.c (lra): Add code to deal with fails by splitting hard reg
2813 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
2815 * config/riscv/riscv.opt (mrelax): New option.
2816 * config/riscv/riscv.c (riscv_file_start): Emit ".option
2817 "norelax" when riscv_mrelax is disabled.
2818 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
2820 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2823 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
2824 reassociation for int modes.
2826 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2828 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
2829 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
2831 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
2832 * config/aarch64/aarch64-sve.md
2833 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
2834 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
2835 (*extend<mode><Vwide>2): Rename to...
2836 (aarch64_sve_extend<mode><Vwide>2): ...this.
2837 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
2838 renaming the old pattern to...
2839 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
2841 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
2842 define_expand, renaming the old pattern to...
2843 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
2844 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
2845 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
2846 account when deciding which SVE instruction the optab should use.
2847 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
2849 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2851 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
2852 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
2853 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
2854 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
2855 (tlsdesc_small_<mode>): Turn a define_expand and use
2856 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
2857 (tlsdesc_small_advsimd_<mode>): ...this.
2858 (tlsdesc_small_sve_<mode>): New pattern.
2860 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2862 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
2863 (UNSPEC_UMUL_HIGHPART): New constants.
2864 (MUL_HIGHPART): New int iteraor.
2865 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
2866 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
2868 (*<su>mul<mode>3_highpart): New define_insn.
2870 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2873 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2876 2018-03-13 Martin Liska <mliska@suse.cz>
2879 * (sem_item_optimizer::sem_item_optimizer): Initialize new
2881 (sem_item_optimizer::~sem_item_optimizer): Release it.
2882 (sem_item_optimizer::merge_classes): Register variable aliases.
2883 (sem_item_optimizer::fixup_pt_set): New function.
2884 (sem_item_optimizer::fixup_points_to_sets): Likewise.
2885 * ipa-icf.h: Declare new variables and functions.
2887 2018-03-13 Jakub Jelinek <jakub@redhat.com>
2890 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
2891 integer_pow2p@2 and test integer_pow2p in condition.
2892 (A < 0 ? C : 0): Similarly for @1.
2895 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
2896 characters starting at p contain '\0' character, don't look beyond
2900 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
2901 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
2904 * reg-stack.c (change_stack): Change update_end var from int to
2905 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
2906 also call set_block_for_insn on the newly added insns and rescan.
2909 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
2910 on the last operand.
2913 * tree.c (stabilize_reference_1): Return save_expr (e) for
2914 STATEMENT_LIST even if it doesn't have side-effects.
2916 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
2918 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
2920 2018-03-12 Renlin Li <renlin.li@arm.com>
2922 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
2923 aarch64_output_scalar_simd_mov_immediate.
2925 2018-03-12 Martin Sebor <msebor@redhat.com>
2927 PR tree-optimization/83456
2928 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
2929 for perfectly overlapping calls to memcpy.
2930 (gimple_fold_builtin_memory_chk): Same.
2931 (gimple_fold_builtin_strcpy): Handle no-warning.
2932 (gimple_fold_builtin_stxcpy_chk): Same.
2933 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
2935 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2937 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
2938 parameter. Use it for SFmode.
2939 (rs6000_function_arg_advance_1): Adjust.
2940 (rs6000_function_arg): Adjust.
2941 (rs6000_gimplify_va_arg): Pass false for that new parameter.
2943 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2945 PR rtl-optimization/84169
2946 PR rtl-optimization/84780
2947 * combine.c (can_combine_p): Check for a 2-insn combination whether
2948 the destination register is used between the two insns, too.
2950 2018-03-12 Richard Biener <rguenther@suse.de>
2952 PR tree-optimization/84803
2953 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
2954 for refs DR analysis didn't process.
2956 2018-03-12 Richard Biener <rguenther@suse.de>
2958 PR tree-optimization/84777
2959 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
2960 force-vectorize loops ignore whether we are optimizing for size.
2962 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
2964 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
2965 (TARGET_MD_ASM_ADJUST): Define.
2967 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
2968 Kito Cheng <kito.cheng@gmail.com>
2969 Chung-Ju Wu <jasonwucj@gmail.com>
2971 * config/nds32/nds32.c (nds32_compute_stack_frame,
2972 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
2973 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
2974 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
2975 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
2976 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
2977 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
2978 * config/nds32/nds32.md (prologue, epilogue): Use macro
2979 NDS32_V3PUSH_AVAILABLE_P to do checking.
2981 2018-03-11 Jakub Jelinek <jakub@redhat.com>
2984 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
2985 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
2986 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
2987 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
2988 addition of most attributes on !orig_type_die or the attribute not
2989 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
2991 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2992 Chung-Ju Wu <jasonwucj@gmail.com>
2994 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
2996 * config/nds32/nds32.opt (mvh): New option.
2998 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2999 Chung-Ju Wu <jasonwucj@gmail.com>
3001 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
3003 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
3004 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
3007 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3008 Chung-Ju Wu <jasonwucj@gmail.com>
3010 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
3012 * config/nds32/nds32-multiple.md (strlensi): New pattern.
3013 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
3015 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
3016 Kito Cheng <kito.cheng@gmail.com>
3017 Chung-Ju Wu <jasonwucj@gmail.com>
3019 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
3020 UNSPEC_FFMISM and UNSPEC_FLMISM.
3021 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
3022 for ffb, ffmism and flmism.
3023 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
3024 (unspec_ffmism): Ditto.
3025 (unspec_flmism): Ditto.
3026 (nds32_expand_builtin_impl): Check if string extension is available.
3027 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
3028 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
3030 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
3033 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3036 * lra-assigns.c (assign_by_spills): Return a flag of reload
3037 assignment failure. Do not process the reload assignment
3038 failures. Do not spill other reload pseudos if they has the same
3040 (lra_assign): Add a return arg. Set up from the result of
3041 assign_by_spills call.
3042 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3043 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3044 usage_insns if it is not NULL.
3045 (spill_hard_reg_in_range): New function.
3046 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3047 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3048 function prototypes.
3049 (lra_assign): Change prototype.
3050 * lra.c (lra): Add code to deal with fails by splitting hard reg
3053 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
3056 * config/i386/i386.opt: Replace Enforcment with Enforcement.
3058 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
3061 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
3062 (dw_val_node): Add val_symbolic_view.
3063 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
3064 (symview_upper_bound): New.
3065 (new_line_info_table): Initialize symviews_since_reset.
3066 (dwarf2out_source_line): Count symviews_since_reset and set
3067 symview_upper_bound.
3068 (dw_val_equal_p): Handle symview.
3069 (add_AT_symview): New.
3070 (print_dw_val): Handle symview.
3071 (attr_checksum, attr_checksum_ordered): Likewise.
3072 (same_dw_val_p, size_of_die): Likewise.
3073 (value_format, output_die): Likewise.
3074 (add_high_low_attributes): Use add_AT_symview for entry_view.
3075 (dwarf2out_finish): Reset symview_upper_bound, clear
3078 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
3081 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
3082 Add strict argument and use it.
3083 (rs6000_split_multireg_move): Update for new strict argument.
3084 (mem_operand_gpr): Disallow all non-offsettable addresses.
3085 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
3087 2018-03-09 Jakub Jelinek <jakub@redhat.com>
3090 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
3091 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
3092 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
3095 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
3096 decl, use remap_type if we want to use the type.
3098 2018-03-09 Martin Sebor <msebor@redhat.com>
3100 PR tree-optimization/84526
3101 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3103 (builtin_access::generic_overlap): Be prepared to handle non-array
3106 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
3108 PR rtl-optimization/84682
3109 * lra-constraints.c (process_address_1): Check is_address flag
3110 for address constraints.
3111 (process_alt_operands): Likewise.
3112 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
3113 preprocess_constraints.
3114 * recog.h (preprocess_constraints): Add oploc parameter.
3116 * recog.c (preprocess_constraints): Test address_operand for
3117 CT_ADDRESS constraints.
3119 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3122 * lra-assigns.c (assign_by_spills): Return a flag of reload
3123 assignment failure. Do not process the reload assignment
3124 failures. Do not spill other reload pseudos if they has the same
3126 (lra_assign): Add a return arg. Set up from the result of
3127 assign_by_spills call.
3128 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3129 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3130 usage_insns if it is not NULL.
3131 (spill_hard_reg_in_range): New function.
3132 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3133 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3134 function prototypes.
3135 (lra_assign): Change prototype.
3136 * lra.c (lra): Add code to deal with fails by splitting hard reg
3139 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3142 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
3143 Accept complain bool parameter. Only emit errors if it is true.
3144 (arm_parse_cpu_option_name): Likewise.
3145 (arm_target_thumb_only): Adjust callers of the above.
3146 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
3147 prototype to take a default true bool parameter.
3148 (arm_parse_arch_option_name): Likewise.
3150 2018-03-09 David Malcolm <dmalcolm@redhat.com>
3151 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
3155 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
3156 * configure: Regenerate.
3157 * configure.ac ("linker --version-script option"): New.
3158 ("linker soname option"): New.
3160 2018-03-09 Richard Biener <rguenther@suse.de>
3162 PR tree-optimization/84775
3163 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
3164 immediate uses of predicate stmts and mark them modified.
3167 PR tree-optimization/84178
3168 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3170 (version_loop_for_if_conversion): Delay update_ssa call.
3171 (tree_if_conversion): Delay update_ssa until after predicate
3174 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
3177 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
3178 when the function accesses prior frames.
3180 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3183 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
3184 gen_llsym, otherwise call maybe_gen_llsym.
3187 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
3188 has ',' character inside of it.
3190 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3193 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
3194 as clobbering CC_REGNUM.
3196 2018-03-08 Richard Biener <rguenther@suse.de>
3199 * tree-scalar-evolution.c: Include tree-into-ssa.h.
3200 (follow_copies_to_constant): Do not follow SSA names registered
3203 2018-03-08 Richard Biener <rguenther@suse.de>
3205 PR tree-optimization/84178
3206 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3208 (version_loop_for_if_conversion): Delay update_ssa call.
3209 (tree_if_conversion): Delay update_ssa until after predicate
3212 2018-03-08 David Malcolm <dmalcolm@redhat.com>
3214 PR tree-optimization/84178
3215 * tree-if-conv.c (release_bb_predicate): Remove the
3216 the assertion that the stmts have NULL use_ops.
3217 Discard the statements, asserting that they haven't
3218 yet been added to a BB.
3220 2018-03-08 Richard Biener <rguenther@suse.de>
3222 PR tree-optimization/84746
3223 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
3224 (phi_translate): Pass in destination ANTIC_OUT set.
3225 (phi_translate_1): Likewise. For a simplified result lookup
3226 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
3227 (phi_translate_set): Adjust.
3228 (do_pre_regular_insertion): Likewise.
3229 (do_pre_partial_partial_insertion): Likewise.
3231 2018-03-08 Martin Liska <mliska@suse.cz>
3233 PR gcov-profile/84735
3234 * doc/gcov.texi: Document usage of profile files.
3235 * gcov-io.h: Document changes in the format.
3237 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
3241 * dwarf2out.c (struct dw_line_info_table): Update comments for
3243 (FORCE_RESET_NEXT_VIEW): New.
3244 (FORCE_RESETTING_VIEW_P): New.
3245 (RESETTING_VIEW_P): Check for -1 too.
3246 (ZERO_VIEW_P): Likewise.
3247 (new_line_info_table): Force-reset next view.
3248 (dwarf2out_begin_function): Likewise.
3249 (dwarf2out_source_line): Simplify zero_view_p initialization.
3250 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
3251 view directly. Omit view when omitting .loc at line 0.
3253 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3255 PR tree-optimization/84740
3256 * tree-switch-conversion.c (process_switch): Call build_constructors
3257 only if info.phi_count is non-zero.
3259 PR tree-optimization/84739
3260 * tree-tailcall.c (find_tail_calls): Check call arguments against
3261 DECL_ARGUMENTS (current_function_decl) rather than
3262 DECL_ARGUMENTS (func) when checking for tail recursion.
3264 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3266 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
3267 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
3268 Volker Reichelt's entry and add entries for people that perform
3269 GCC fuzzy testing and report numerous bugs.
3271 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
3274 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
3275 readonly data in sdata, if that is disabled.
3276 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
3277 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3278 -mreadonly-in-sdata option.
3280 2018-03-07 Martin Sebor <msebor@redhat.com>
3282 PR tree-optimization/84468
3283 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
3284 basic block when looking for nul assignment.
3286 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
3289 * except.h (output_function_exception_table): Adjust prototype.
3290 * except.c (output_function_exception_table): Remove FNNAME parameter
3291 and add SECTION parameter. Ouput one part of the table at a time.
3292 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
3293 the first part of the exception table and emit unwind directives.
3294 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
3295 (i386_pe_seh_cold_init): Likewise.
3296 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
3297 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
3298 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
3299 (ix86_output_call_insn): Emit a nop in one more case for SEH.
3300 * config/i386/winnt.c: Include except.h.
3301 (struct seh_frame_state): Add reg_offset, after_prologue and
3302 in_cold_section fields.
3303 (i386_pe_seh_end_prologue): Set seh->after_prologue.
3304 (i386_pe_seh_cold_init): New function.
3305 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
3306 to seh->in_cold_section.
3307 (seh_emit_push): Record the offset of the push.
3308 (seh_emit_save): Record the offet of the save.
3309 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
3310 Test seh->after_prologue to disregard the epilogue.
3311 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
3312 (i386_pe_end_cold_function): New function.
3314 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3317 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
3318 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
3321 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
3323 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
3324 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
3327 * multiple_target.c: Include tree-inline.h and intl.h.
3328 (expand_target_clones): Diagnose and fail if node->definition and
3329 !tree_versionable_function_p (node->decl).
3331 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
3333 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
3335 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
3336 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
3337 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
3339 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3342 * combine.c (try_combine): Use reg_or_subregno instead of handling
3343 just paradoxical SUBREGs and REGs.
3345 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
3347 * config/arc/arc.c (arc_finalize_pic): Remove function.
3348 (arc_must_save_register): We use single base PIC register, remove
3349 checks to save/restore the PIC register.
3350 (arc_expand_prologue): Likewise.
3351 * config/arc/arc-protos.h (arc_set_default_type_attributes):
3353 (arc_verify_short): Likewise.
3354 (arc_attr_type): Likewise.
3355 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
3356 (walk_stores): Likewise.
3357 (arc_address_cost): Make it static.
3358 (arc_verify_short): Likewise.
3359 (branch_dest): Likewise.
3360 (arc_attr_type): Likewise.
3361 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
3362 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
3363 (arc_final_prescan_insn): Remove inserting the nops due to
3364 hardware hazards. It is done in reorg step.
3365 (insn_length_variant_t): Remove.
3366 (insn_length_parameters_t): Likewise.
3367 (arc_insn_length_parameters): Likewise.
3368 (arc_get_insn_variants): Likewise.
3369 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
3371 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3374 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
3377 PR tree-optimization/84687
3378 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
3380 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
3382 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3384 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
3385 Rename to ppc_speculation_barrier.
3386 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
3387 __builtin_ppc_speculation_barrier.
3389 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3392 * combine.c (combine_simplify_rtx): Don't try to simplify if
3393 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
3396 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
3398 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
3399 to 32 bytes when compiling for POWER9.
3401 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3404 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
3405 regparm >= 3 with no arg reg available also for calls with
3406 flag_force_indirect_call. Pass decl to ix86_function_regparm.
3409 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
3411 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
3413 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
3416 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
3418 2018-03-05 Richard Biener <rguenther@suse.de>
3420 PR tree-optimization/84486
3421 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
3422 When inserting a __builtin_assume_aligned call set the LHS
3423 SSA name alignment info accordingly.
3425 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
3427 PR tree-optimization/84114
3428 * config/aarch64/aarch64.c (aarch64_reassociation_width)
3429 Avoid reassociation of FLOAT_MODE addition.
3431 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
3433 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
3434 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
3435 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
3436 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
3437 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
3438 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
3439 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
3441 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
3442 __builtin_ia32_wbinvd): New builtins.
3443 (SPECIAL_ARGS2): New.
3444 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
3445 (SPECIAL_ARGS2): New.
3446 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
3447 (ix86_valid_target_attribute_inner_p): Ditto.
3448 (ix86_init_mmx_sse_builtins): Add special_args2.
3449 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
3450 TARGET_WBNOINVD_P): New.
3451 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
3452 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
3453 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
3454 * config/i386/immintrin.h (_wbinvd): New intrinsic.
3455 * config/i386/pconfigintrin.h: New file.
3456 * config/i386/wbnoinvdintrin.h: Ditto.
3457 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
3458 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
3460 2018-03-05 Richard Biener <rguenther@suse.de>
3462 PR tree-optimization/84670
3463 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3465 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3466 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3467 (compute_antic_aux): Only assert the number of values in ANTIC_IN
3468 doesn't grow if all successors (recursively) were visited at least
3471 2018-03-05 Richard Biener <rguenther@suse.de>
3473 PR tree-optimization/84650
3474 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
3475 if executed in the loop pipeline.
3477 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
3479 * doc/configfiles.texi (Configuration Files): Move info about
3480 conditionalizing $target-protos.h to...
3481 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
3482 differs from $target-protos.h.
3484 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
3485 Chung-Ju Wu <jasonwucj@gmail.com>
3487 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
3488 * config/nds32/nds32-multiple.md (setmemsi): Define.
3489 * config/nds32/nds32-memory-manipulation.c
3490 (nds32_gen_dup_4_byte_to_word_value): New.
3491 (emit_setmem_word_loop): New.
3492 (emit_setmem_byte_loop): New.
3493 (nds32_expand_setmem_loop): New.
3494 (nds32_expand_setmem_loop_v3m): New.
3495 (nds32_expand_setmem_unroll): New.
3496 (nds32_expand_setmem): New.
3498 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3499 Chung-Ju Wu <jasonwucj@gmail.com>
3501 * config/nds32/nds32-memory-manipulation.c
3502 (nds32_emit_load_store): New.
3503 (nds32_emit_post_inc_load_store): New.
3504 (nds32_emit_mem_move): New.
3505 (nds32_emit_mem_move_block): New.
3506 (nds32_expand_movmemsi_loop_unknown_size): New.
3507 (nds32_expand_movmemsi_loop_known_size): New.
3508 (nds32_expand_movmemsi_loop): New.
3509 (nds32_expand_movmemsi_unroll): New.
3510 (nds32_expand_movmemqi): Rename ...
3511 (nds32_expand_movmemsi): ... to this.
3512 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
3513 (movmemsi): ... to this.
3514 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
3515 (nds32_expand_movmemsi): ... to this.
3517 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3518 Monk Chiang <sh.chiang04@gmail.com>
3519 Chung-Ju Wu <jasonwucj@gmail.com>
3521 * config/nds32/nds32-protos.h
3522 (nds32_expand_load_multiple): New arguments.
3523 (nds32_expand_store_multiple): Ditto.
3524 (nds32_valid_multiple_load_store): Rename ...
3525 (nds32_valid_multiple_load_store_p): ... to this.
3526 * config/nds32/nds32-memory-manipulation.c
3527 (nds32_expand_load_multiple): Refine implementation.
3528 (nds32_expand_store_multiple): Ditto.
3529 * config/nds32/nds32-multiple.md
3530 (load_multiple): Update nds32_expand_load_multiple interface.
3531 (store_multiple): Update nds32_expand_store_multiple interface.
3532 * config/nds32/nds32-predicates.c
3533 (nds32_valid_multiple_load_store): Rename ...
3534 (nds32_valid_multiple_load_store_p): ... to this and refine
3536 * config/nds32/predicates.md
3537 (nds32_load_multiple_and_update_address_operation): New predicate.
3538 (nds32_store_multiple_and_update_address_operation): New predicate.
3540 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3541 Chung-Ju Wu <jasonwucj@gmail.com>
3543 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
3544 (combo): New attribute.
3545 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
3547 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
3549 * config/nds32/nds32.opt: Change -mcmodel= default value.
3551 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
3552 Monk Chiang <sh.chiang04@gmail.com>
3553 Chung-Ju Wu <jasonwucj@gmail.com>
3555 * config/nds32/constants.md (unspec_element): New enum.
3556 * config/nds32/constraints.md (Umw): New constraint.
3557 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
3558 * config/nds32/nds32-intrinsic.md: Likewise.
3559 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
3560 (nds32_valid_smw_lwm_base_p): New.
3561 (nds32_output_smw_single_word): New.
3562 (nds32_output_lmw_single_word): New.
3563 (nds32_expand_unaligned_load): New.
3564 (nds32_expand_unaligned_store): New.
3565 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
3566 (nds32_output_smw_single_word): Declare.
3567 (nds32_output_lmw_single_word): Declare.
3568 (nds32_expand_unaligned_load): Declare.
3569 (nds32_expand_unaligned_store): Declare.
3570 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
3571 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
3572 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
3573 NDS32_BUILTIN_UASTORE_DW.
3574 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
3577 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
3578 Kito Cheng <kito.cheng@gmail.com>
3579 Chung-Ju Wu <jasonwucj@gmail.com>
3581 * config/nds32/nds32-intrinsic.c
3582 (nds32_expand_builtin_null_ftype_reg): Delete.
3583 (nds32_expand_builtin_reg_ftype_imm): Ditto.
3584 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
3585 (nds32_read_argument): New.
3586 (nds32_legitimize_target): Ditto.
3587 (nds32_legitimize_argument): Ditto.
3588 (nds32_check_constant_argument): Ditto.
3589 (nds32_expand_unop_builtin): Ditto.
3590 (nds32_expand_unopimm_builtin): Ditto.
3591 (nds32_expand_binop_builtin): Ditto.
3592 (nds32_builtin_decl_impl): Ditto.
3593 (builtin_description): Ditto.
3594 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
3595 (nds32_init_builtins_impl): Ditto.
3596 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
3597 (nds32_builtin_decl): New.
3598 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
3599 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
3601 2018-03-02 Jeff Law <law@redhat.com>
3603 * reorg.c (stop_search_p): Handle DEBUG_INSN.
3604 (redundant_insn, fill_simple_delay_slots): Likewise.
3605 (fill_slots_from_thread): Likewise.
3606 * resource.c (mark_referenced_resources): Likewise.
3607 (mark_set_resources, find_dead_or_set_registers): Likewise.
3609 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3611 * substring-locations.h (format_warning_va): Formatting fix for
3613 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
3615 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
3616 * substring-locations.c: Include intl.h.
3617 (format_warning_va): Turned into small wrapper around
3618 format_warning_n_va, renamed to ...
3619 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
3620 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
3622 (format_warning_at_substring_n): New function.
3623 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
3624 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
3625 format_warning_at_substring with just a shorter name instead of
3626 const function pointer.
3627 (fmtwarn_n): New function.
3628 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
3629 appropriate, get rid of all the fmtstr temporaries, move conditionals
3630 with G_() wrapped string literals directly into fmtwarn arguments,
3631 cast dir.len to (int), formatting fixes.
3633 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
3635 * doc/invoke.texi: Remove "Cilk Plus" references.
3637 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3638 Richard Biener <rguenther@suse.de>
3641 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
3642 for error or warning attributes if CALL_FROM_THUNK_P is set.
3645 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3648 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
3649 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
3652 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
3653 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
3655 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
3656 instead of -1U in last predictors element's probability member.
3658 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3661 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3662 arguments if they are comparable.
3664 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3666 PR tree-optimization/84634
3667 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
3668 masks and masked_loop_p with a single loop_masks, making sure it's
3669 null for bb vectorization.
3671 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3673 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3674 (vect_analyze_data_ref_access): Use loop->safe_len rather than
3675 loop->force_vectorize to check whether there is no alias.
3677 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3680 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
3682 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
3684 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
3685 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
3686 instead of a loop around prev_real_insn.
3687 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
3691 * config/i386/i386.c (ix86_print_operand): Use conditional
3692 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
3695 2018-03-02 Richard Biener <rguenther@suse.de>
3697 PR tree-optimization/84427
3698 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
3699 (bitmap_set_subtract_values): Rewrite to handle multiple
3702 (prune_clobbered_mems): Likewise.
3703 (phi_translate): Take edge instead of pred/phiblock.
3704 (phi_translate_1): Likewise.
3705 (phi_translate_set): Likewise. Insert all translated
3706 exprs for a value into the set, keeping possibly multiple
3707 expressions per value.
3708 (compute_antic_aux): Adjust for phi_translate changes.
3709 When intersecting union the expressions and prune those
3710 not in the final value set, keeping possibly multiple
3711 expressions per value. Do not use value-insertion
3712 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
3713 all expressions. Add verification that the value-sets
3714 only shrink during iteration.
3715 (compute_partial_antic_aux): Adjust for the phi_translate changes.
3716 (do_pre_regular_insertion): Likewise.
3717 (do_pre_partial_partial_insertion): Likewise.
3719 2018-03-02 Richard Biener <rguenther@suse.de>
3722 * config/darwin.c (saved_debug_info_level): New static global.
3723 (darwin_asm_lto_start): Disable debug info generation for LTO out.
3724 (darwin_asm_lto_end): Restore debug info generation settings.
3726 2018-03-01 Martin Liska <mliska@suse.cz>
3729 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
3732 2018-03-01 Richard Biener <rguenther@suse.de>
3735 * dwarf2out.c (gen_variable_die): Properly handle late VLA
3736 type annotation with LTO when debug was disabled at compile-time.
3738 2018-03-01 Matthew Fortune <mfortune@gmail.com>
3740 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
3742 (mips_final_postscan_insn): Likewise.
3744 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
3746 PR rtl-optimization/84528
3747 * alias.c (init_alias_target): Add commentary.
3748 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
3749 a unique base value if the frame pointer is not eliminated
3750 to the stack pointer.
3752 2018-03-01 Tom de Vries <tom@codesourcery.com>
3754 PR rtl-optimization/83327
3755 * lra-int.h (hard_regs_spilled_into): Declare.
3756 * lra.c (hard_regs_spilled_into): Define.
3757 (init_reg_info): Init hard_regs_spilled_into.
3758 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
3759 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
3760 (process_bb_lives): Handle hard_regs_spilled_into.
3761 (lra_create_live_ranges_1): Before doing liveness propagation, clear
3762 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
3764 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
3766 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
3767 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
3768 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
3769 * config/rs6000/aix72.h: New file.
3771 2018-02-28 Jakub Jelinek <jakub@redhat.com>
3773 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
3774 instead of warning_at with conditional singular and plural messages
3778 * stor-layout.c (update_alignment_for_field): For
3779 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
3780 && !DECL_PACKED (field), do the alignment update, just use
3781 only desired_align instead of MAX (type_align, desired_align)
3783 (place_field): Don't do known_align < desired_align handling
3784 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
3785 is non-NULL, instead do it after rli->prev_field handling and
3786 only if not within a bitfield word. For DECL_PACKED (field)
3787 use type_align of BITS_PER_UNIT.
3789 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
3791 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
3792 superfluous parentheses and trailing spaces.
3794 2018-02-28 Richard Biener <rguenther@suse.de>
3796 PR tree-optimization/84584
3797 * graphite-scop-detection.c (scop_detection::add_scop): Discard
3798 SCoPs with fake exit edge.
3800 2018-02-28 Martin Liska <mliska@suse.cz>
3803 * timevar.c (timer::print): Fix format to properly print 100%
3806 2018-02-28 Richard Biener <rguenther@suse.de>
3809 * genmatch.c (capture_info::walk_match): Do not mark
3810 captured expressions without operands as expr_p given
3811 they act more like predicates and should be subject to
3812 "lost tail" side-effect preserving.
3814 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
3816 PR rtl-optimization/81611
3817 * auto-inc-dec.c (attempt_change): Move dead note from
3818 mem_insn if it's the next use of regno
3819 (find_address): Take address use of reg holding
3820 non-incremented value. Add parm to limit search to the named
3822 (merge_in_block): Attempt to use a mem insn that is the next
3823 use of the original regno.
3825 2018-02-27 Martin Sebor <msebor@redhat.com>
3828 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
3829 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
3831 2018-02-27 Martin Sebor <msebor@redhat.com>
3833 PR translation/84207
3834 * diagnostic-core.h (warning_n, error_n, inform_n): Change
3835 n argument to unsigned HOST_WIDE_INT.
3836 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
3837 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
3838 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
3839 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
3841 2018-02-27 Richard Biener <rguenther@suse.de>
3843 PR tree-optimization/84512
3844 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
3845 Do not use the estimate returned from record_stmt_cost for
3846 the scalar iteration cost but sum properly using add_stmt_cost.
3848 2018-02-27 Richard Biener <rguenther@suse.de>
3850 PR tree-optimization/84466
3851 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3852 Adjust last change to less strictly validate use operands.
3854 2018-02-27 Martin Liska <mliska@suse.cz>
3856 PR gcov-profile/84548
3857 * gcov.c (process_file): Allow partial overlap and consider it
3858 also as group functions.
3859 (output_lines): Properly calculate range of lines for a group.
3861 2018-02-27 Martin Liska <mliska@suse.cz>
3863 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
3864 'ggc' suffixes. Change first column width.
3865 (timer::print): Fix formatting of the column.
3867 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
3869 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
3870 preserve inline entry blocks for the sake of debug inline
3871 entry point markers alone.
3872 (remove_unused_locals): Suggest in comments a better place to
3873 force the preservation of inline entry blocks that are
3874 otherwise unused, but do not preserve them.
3876 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3878 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
3880 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3883 * config/i386/constraints.md (Bs): Replace
3884 ix86_indirect_branch_register with
3885 TARGET_INDIRECT_BRANCH_REGISTER.
3887 * config/i386/i386.md (indirect_jump): Likewise.
3888 (tablejump): Likewise.
3889 (*sibcall_memory): Likewise.
3890 (*sibcall_value_memory): Likewise.
3891 Peepholes of indirect call and jump via memory: Likewise.
3892 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
3893 (*sibcall_value_GOT_32): Likewise.
3894 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3895 (GOT_memory_operand): Likewise.
3896 (call_insn_operand): Likewise.
3897 (sibcall_insn_operand): Likewise.
3898 (GOT32_symbol_operand): Likewise.
3899 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
3901 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3903 PR rtl-optimization/83496
3904 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
3905 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
3906 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
3907 redundant insn, if any.
3908 (relax_delay_slots): Likewise.
3909 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
3911 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
3913 PR tree-optimization/83965
3914 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
3915 that grouped statements are part of a reduction chain. Return
3916 true if the statement is not marked as a reduction itself but
3918 (vect_recog_dot_prod_pattern): Don't check whether the statement
3919 is part of a group here.
3920 (vect_recog_sad_pattern): Likewise.
3921 (vect_recog_widen_sum_pattern): Likewise.
3923 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3926 * final.c (rest_of_clean_state): Also look for calls inside sequences.
3928 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3931 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
3933 (ix86_output_indirect_function_return): New prototype.
3934 (ix86_split_simple_return_pop_internal): Likewise.
3935 * config/i386/i386.c (indirect_return_via_cx): New.
3936 (indirect_return_via_cx_bnd): Likewise.
3937 (indirect_thunk_name): Handle return va CX_REG.
3938 (output_indirect_thunk_function): Create alias for
3939 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
3940 (ix86_output_indirect_jmp): Remove the bool argument.
3941 (ix86_output_indirect_function_return): New function.
3942 (ix86_split_simple_return_pop_internal): Likewise.
3943 * config/i386/i386.md (*indirect_jump): Don't pass false
3944 to ix86_output_indirect_jmp.
3945 (*tablejump_1): Likewise.
3946 (simple_return_pop_internal): Change it to define_insn_and_split.
3947 Call ix86_split_simple_return_pop_internal to split it for
3949 (simple_return_indirect_internal): Call
3950 ix86_output_indirect_function_return instead of
3951 ix86_output_indirect_jmp.
3953 2018-02-26 Jakub Jelinek <jakub@redhat.com>
3956 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
3957 memset and value initialization afterwards.
3959 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
3961 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
3963 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3966 * common/config/aarch64/aarch64-common.c
3967 (aarch_option_optimization_table[]): Switch
3968 off fomit-frame-pointer
3970 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3971 Chung-Ju Wu <jasonwucj@gmail.com>
3973 * config/nds32/nds32-multiple.md (load_multiple): Disallow
3975 (store_multiple): Ditto.
3977 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3979 * config.gcc: Add --with-cpu support for nds32 target.
3980 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
3981 * config/nds32/nds32.opt: Add -mcpu= option.
3983 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
3985 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
3986 isel=yes): Warn for these deprecated options.
3988 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
3990 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
3991 ISA_2_5_MASKS_EMBEDDED.
3993 2018-02-23 Jakub Jelinek <jakub@redhat.com>
3995 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
3996 p->max as pointers rather than using iterative_hash_expr.
3998 2018-02-23 Carl Love <cel@us.ibm.com>
4000 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
4001 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
4003 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
4004 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
4005 P8V_BUILTIN_VEC_VUNSIGNED2.
4007 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
4010 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
4011 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
4013 * lra-constraints.c (curr_insn_transform): Set up
4014 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
4016 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
4017 * lra-eliminations.c (spill_pseudos): Ditto.
4018 (process_insn_for_elimination): Ditto.
4019 * lra-lives.c (reg_early_clobber_p): Use the new macros.
4020 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
4021 LRA_NON_CLOBBERED_ALT.
4023 2018-02-22 Martin Sebor <msebor@redhat.com>
4025 PR tree-optimization/84480
4026 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
4027 to maybe_diag_stxncpy_trunc. Call it.
4028 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
4029 from gimple_fold_builtin_strcpy. Print inlining stack.
4030 (handle_builtin_stxncpy): Print inlining stack.
4031 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
4033 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
4036 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
4037 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
4038 and -fcheck-pointer-bounds are used together.
4039 (indirect_thunk_prefix): New enum.
4040 (indirect_thunk_need_prefix): New function.
4041 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
4042 "_nt" instead of "_bnd" for NOTRACK prefix.
4043 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
4044 (output_indirect_thunk_function): Likewise.
4046 (ix86_code_end): Update output_indirect_thunk_function calls.
4047 (ix86_output_indirect_branch_via_reg): Replace
4048 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
4049 (ix86_output_indirect_branch_via_push): Likewise.
4050 (ix86_output_function_return): Likewise.
4051 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
4052 incompatible with -fcf-protection=branch and
4053 -fcheck-pointer-bounds.
4055 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4058 * config/aarch64/aarch64.c (aarch64_print_address_internal):
4059 Change gcc_assert call to output_operand_lossage.
4061 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4063 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
4065 2018-02-22 DJ Delorie <dj@redhat.com>
4066 Sebastian Perta <sebastian.perta@renesas.com>
4067 Oleg Endo <olegendo@gcc.gnu.org>
4069 * config/rx/rx.c (rx_rtx_costs): New function.
4070 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
4072 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
4074 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
4076 2018-02-22 Martin Liska <mliska@suse.cz>
4079 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
4080 Add "native" as a possible value.
4082 2018-02-22 Martin Liska <mliska@suse.cz>
4085 * config/i386/i386.c (ix86_option_override_internal):
4086 Add "native" as a possible value for -march and -mtune.
4088 2018-02-22 Jakub Jelinek <jakub@redhat.com>
4091 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
4092 to all type variants.
4094 PR tree-optimization/84503
4095 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
4096 width as info->bitpos + info->bitsize - start.
4097 (merged_store_group::merge_overlapping): Simplify width computation.
4098 (check_no_overlap): New function.
4099 (imm_store_chain_info::try_coalesce_bswap): Compute expected
4100 start + width and last_order of the group, fail if check_no_overlap
4102 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
4103 to group if check_no_overlap fails.
4105 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4107 * config/rs6000/altivec.md: Delete contraint arguments to
4108 define_expand, define_split, and define_peephole2, and in
4109 define_insn_and_split if always unused.
4110 * config/rs6000/darwin.md: Ditto.
4111 * config/rs6000/dfp.md: Ditto.
4112 * config/rs6000/rs6000.md: Ditto.
4113 * config/rs6000/sync.md: Ditto.
4114 * config/rs6000/vector.md: Ditto.
4115 * config/rs6000/vsx.md: Ditto.
4117 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4119 * config/rs6000/altivec.md: Write output control strings as braced
4120 blocks instead of double-quoted strings.
4121 * config/rs6000/darwin.md: Ditto.
4122 * config/rs6000/rs6000.md: Ditto.
4123 * config/rs6000/vector.md: Ditto.
4124 * config/rs6000/vsx.md: Ditto.
4126 2018-02-21 Jason Merrill <jason@redhat.com>
4128 PR c++/84314 - ICE with templates and fastcall attribute.
4129 * attribs.c (build_type_attribute_qual_variant): Remove assert.
4131 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4133 * ipa-cp.c (determine_versionability): Fix comment typos.
4135 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4138 * ipa-cp.c (determine_versionability): Do not version functions caling
4141 2018-02-21 Martin Liska <mliska@suse.cz>
4144 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
4145 Add "native" as a possible value.
4146 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
4147 the macro when native cpu detection is available.
4149 2018-02-21 Martin Liska <mliska@suse.cz>
4152 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
4153 Add "native" as a possible value.
4154 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
4155 when native cpu detection is available.
4157 2018-02-21 Jakub Jelinek <jakub@redhat.com>
4158 Martin Sebor <msebor@redhat.com>
4160 PR tree-optimization/84478
4161 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
4163 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
4164 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
4165 support which is conservatively correct, for 2 only stay conservative
4166 for maxlen. Formatting and comment capitalization fixes. Add STRICT
4167 argument to the 2 argument get_range_strlen, adjust 6 arg
4168 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
4170 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
4171 (gimple_fold_builtin_strlen): Pass true as last argument to
4174 2018-02-20 Martin Sebor <msebor@redhat.com>
4177 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
4178 (builtin_memref::set_base_and_offset): Same. Handle inner references.
4179 (builtin_memref::builtin_memref): Factor out parts into
4180 set_base_and_offset and call it.
4182 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4185 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
4186 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
4187 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
4188 search at the associated MODE_INT.
4190 2018-02-20 Jeff Law <law@redhat.com>
4193 PR tree-optimization/81592
4195 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
4196 for range data rather than using global data.
4197 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
4198 range data rather than using global data.
4199 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
4200 pass it to children as needed.
4201 (struct directive::fmtresult): Similarly.
4202 (struct directive::set_width): Similarly.
4203 (struct directive::set_precision): Similarly.
4204 (format_integer, format_directive, parse_directive): Similarly.
4205 (format_none): Accept unnamed vr_values parameter.
4206 (format_percent, format_floating, format_character): Similarly.
4207 (format_string, format_plain): Similarly.
4208 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
4209 the EVRP range analyzer for range data rather than using global data.
4210 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
4211 gimple-ssa-evrp-analyze.h
4212 (class sprintf_dom_walker): Add after_dom_children member function.
4213 Add evrp_range_analyzer member.
4214 (sprintf_dom_walker::before_dom_children): Call into the EVRP
4215 range analyzer as needed.
4216 (sprintf_dom_walker::after_dom_children): New member function.
4217 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
4219 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
4220 (evrp_range_analyzer::pop_to_marker): Likewise.
4222 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4224 PR tree-optimization/84419
4225 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
4226 with the required type if its current type is compatible but
4229 2018-02-20 Jakub Jelinek <jakub@redhat.com>
4232 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
4233 after vectorization.
4235 2018-02-20 Martin Liska <mliska@suse.cz>
4238 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
4239 possible values if we don't have a hint.
4241 2018-02-20 Martin Liska <mliska@suse.cz>
4245 * final.c (shorten_branches): Build align_tab array with one
4247 * opts.c (finish_options): Add alignment option limit check.
4248 (MAX_CODE_ALIGN): Likewise.
4249 (MAX_CODE_ALIGN_VALUE): Likewise.
4250 * doc/invoke.texi: Document maximum allowed option value for
4251 all -falign-* options.
4253 2018-02-19 Jakub Jelinek <jakub@redhat.com>
4256 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
4257 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
4258 * var-tracking.c (emit_note_insn_var_location): Remove all references
4259 to NOTE_INSN_CALL_ARG_LOCATION.
4260 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
4261 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
4262 Use copy_rtx_if_shared.
4263 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
4264 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
4265 (dwarf2out_var_location): Remove handling of
4266 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
4268 * final.c (final_scan_insn): Remove all references to
4269 NOTE_INSN_CALL_ARG_LOCATION.
4270 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
4271 before dumping final insns.
4272 * except.c (emit_note_eh_region_end): Remove all references to
4273 NOTE_INSN_CALL_ARG_LOCATION.
4274 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
4275 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
4276 * config/arc/arc.c (hwloop_optimize): Likewise.
4277 * config/arm/arm.c (create_fix_barrier): Likewise.
4278 * config/s390/s390.c (s390_chunkify_start): Likewise.
4279 * config/sh/sh.c (find_barrier): Likewise.
4280 * config/i386/i386.c (rest_of_insert_endbranch,
4281 ix86_seh_fixup_eh_fallthru): Likewise.
4282 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
4283 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
4284 * config/frv/frv.c (frv_function_prologue): Likewise.
4285 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
4287 (note_outside_basic_block_p): Remove all references to
4288 NOTE_INSN_CALL_ARG_LOCATION.
4289 * gengtype.c (adjust_field_rtx_def): Likewise.
4290 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
4292 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
4293 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
4296 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
4299 PR tree-optimization/84452
4300 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
4301 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
4304 2018-02-19 Martin Liska <mliska@suse.cz>
4307 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
4309 2018-02-19 Martin Liska <mliska@suse.cz>
4310 Richard Sandiford <richard.sandiford@linaro.org>
4312 PR tree-optimization/82491
4313 * gimple-fold.c (get_base_constructor): Make earlier bail out
4316 2018-02-19 Carl Love <cel@us.ibm.com>
4318 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
4319 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
4321 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
4322 P8V_BUILTIN_VEC_NEG.
4324 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
4326 * config/rl78/rl78.md (movdf): New define expand.
4328 2018-02-19 Martin Liska <mliska@suse.cz>
4331 * doc/invoke.texi: Fix typo.
4332 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
4334 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
4336 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
4337 handle rs6000_single_float and rs6000_double_float specially for
4340 2018-02-16 Jeff Law <law@redhat.com>
4342 * config/rx/rx.c (add_pop_cfi_notes): New function.;
4345 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4348 * ipa-inline.c (inline_small_functions): Fix a typo.
4350 2018-02-16 Nathan Sidwell <nathan@acm.org>
4352 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
4354 2018-02-16 Carl Love <cel@us.ibm.com>
4356 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
4357 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
4358 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
4359 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
4360 expansion to P8V_BUILTIN_VEC_FLOAT2.
4362 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
4364 PR rtl-optimization/70023
4365 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
4366 src_regno into account.
4368 2018-02-16 Carl Love <cel@us.ibm.com>
4370 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
4371 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
4372 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
4373 * config/rs6000/rs6000.c: Remove case statements for
4374 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
4375 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
4376 and P9V_BUILTIN_VEC_VINSERT4B.
4377 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
4378 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
4379 * config/rs6000/vsx.md:
4380 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
4383 2018-02-16 Carl Love <cel@us.ibm.com>
4385 * config/rs6000/altivec.h: Add builtin names vec_extract4b
4387 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
4389 * config/rs6000/rs6000-c.c: Add the definitions for
4390 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
4391 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
4392 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
4393 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
4394 definition for insert4b and define insn *insert3b_internal.
4395 * doc/extend.texi: Add documentation for vec_extract4b.
4397 2018-02-16 Nathan Sidwell <nathan@acm.org>
4399 * doc/extend.texi (Backwards Compatibility): Mention friend
4400 injection. Note for-scope is deprecated.
4401 * doc/invoke.texi (-ffriend-injection): Deprecate.
4403 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
4405 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
4406 that moved to I2, also allow destinations that are a paradoxical
4407 subreg (instead of a normal reg).
4409 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
4412 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
4415 2018-02-16 Richard Biener <rguenther@suse.de>
4417 PR tree-optimization/84037
4418 PR tree-optimization/84016
4420 * config/i386/i386.c (ix86_builtin_vectorization_cost):
4421 Adjust vec_construct for the fact we need additional higher latency
4422 128bit inserts for AVX256 and AVX512 vector builds.
4423 (ix86_add_stmt_cost): Scale vector construction cost for
4426 2018-02-16 Richard Biener <rguenther@suse.de>
4428 PR tree-optimization/84417
4429 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
4430 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
4431 (non_rewritable_lvalue_p): Likewise, use poly-ints.
4433 2018-02-16 Martin Liska <mliska@suse.cz>
4436 * internal-fn.def (ASAN_CHECK): Set proper flags.
4437 (ASAN_MARK): Likewise.
4439 2018-02-16 Julia Koval <julia.koval@intel.com>
4441 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
4442 from PTA_CANNONLAKE.
4444 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4447 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
4448 Use ++iter rather than iter++ for std::list iterators.
4449 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
4450 defer deleting them until all nodes in the forest are processed. Do
4451 free even leaf nodes. Change to_process into auto_vec.
4454 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
4455 * vec.h (vec_default_construct): Use memset instead of placement new
4456 if BROKEN_VALUE_INITIALIZATION is defined.
4457 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
4458 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
4461 PR rtl-optimization/83723
4462 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
4463 * lra.c (lra_substitute_pseudo): Likewise. If true, use
4464 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
4466 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
4468 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
4470 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
4472 PR rtl-optimization/81443
4473 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
4474 from inner REGs to paradoxical SUBREGs.
4476 2018-02-16 Richard Biener <rguenther@suse.de>
4478 PR tree-optimization/84399
4479 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4480 For operands we can analyze at their definition make sure we can
4481 analyze them at each use as well.
4483 2018-02-16 Richard Biener <rguenther@suse.de>
4485 PR tree-optimization/84190
4486 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
4487 volatile accesses if the decl isn't volatile.
4489 2018-02-15 Jason Merrill <jason@redhat.com>
4491 PR c++/84314 - ICE with templates and fastcall attribute.
4492 * attribs.c (build_type_attribute_qual_variant): Don't clobber
4493 TYPE_CANONICAL on an existing type.
4495 2018-02-15 Jakub Jelinek <jakub@redhat.com>
4497 PR tree-optimization/84383
4498 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
4499 dstoff nor call operand_equal_p if dstbase is NULL.
4501 PR tree-optimization/84334
4502 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
4503 also a CONSTANT_CLASS_P, punt.
4505 2018-02-14 Jim Wilson <jimw@sifive.com>
4507 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
4508 first SMALL_OPERAND check. New local min_second_step. Move assert
4509 to where locals are set. Add TARGET_RVC support.
4510 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
4512 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
4514 * doc/invoke.texi: Correct -Wformat-overflow code sample.
4516 2018-02-14 Martin Sebor <msebor@redhat.com>
4518 PR tree-optimization/83698
4519 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
4520 arrays constrain the offset range to their bounds.
4521 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
4522 (builtin_access::overlap): Avoid setting the size of overlap if it's
4524 (maybe_diag_overlap): Also consider arrays when deciding what values
4525 of offsets to include in diagnostics.
4527 2018-02-14 Martin Sebor <msebor@redhat.com>
4530 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
4531 that correspond to the kind of a declaration.
4533 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
4536 * config/pa/pa.md: Load address of PIC label using the linkage table
4537 if the label is nonlocal.
4539 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
4541 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
4542 warning message if user requests -maltivec=be.
4543 * doc/invoke.texi: Document deprecation of -maltivec=be.
4545 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
4548 * config/rs6000/rs6000-c.c: Update definitions for
4549 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
4550 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
4552 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4555 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
4556 add _get_ssp intrinsics. Remove argument from
4557 __builtin_ia32_rdssp[d|q].
4558 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
4559 * config/i386/i386-builtin.def: Remove argument from
4560 __builtin_ia32_rdssp[d|q].
4561 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
4562 ix86_expand_special_args_builtin for _rdssp[d|q].
4563 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
4564 Clear register before usage.
4565 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
4566 Add documentation for new _get_ssp and _inc_ssp intrinsics.
4568 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
4570 PR tree-optimization/84357
4571 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
4572 operand 1 of an ARRAY_REF too.
4574 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
4577 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
4578 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4580 (set_of_reg): New struct.
4581 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
4582 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
4583 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4585 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
4586 Split into bitclr, bitset, bitinvert patterns if appropriate.
4587 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
4588 use rx_fuse_in_memory_bitop.
4589 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
4590 to named insn, correct maximum insn length.
4592 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
4595 * machmode.def: Define a complex mode for PARTIAL_INT.
4596 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
4598 * doc/rtl.texi: Document CSPImode.
4599 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
4601 (msp430_hard_regno_nregs_with_padding): Likewise.
4603 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
4606 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
4608 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
4610 PR rtl-optimization/84169
4611 * combine.c (try_combine): New variable split_i2i3. Set it to true if
4612 we generated a parallel as new i3 and we split that to new i2 and i3
4613 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
4614 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
4615 those to i2, not i1. Partially rewrite this scan code.
4617 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4620 * stor-layout.c (place_field): For variable length fields, adjust
4621 offset_align afterwards not just based on the field's alignment,
4622 but also on the size.
4625 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
4626 of exps and logs in the use_exp2 case.
4628 2018-02-13 Jeff Law <law@redhat.com>
4630 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
4633 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
4636 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
4640 * common.opt (gas-loc-support, gas-locview-support): New.
4641 (ginline-points, ginternal-reset-location-views): New.
4642 * doc/invoke.texi: Document them. Use @itemx where intended.
4643 (gvariable-location-views): Adjust.
4644 * target.def (reset_location_view): New.
4645 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
4646 (TARGET_RESET_LOCATION_VIEW): New.
4647 * doc/tm.texi: Rebuilt.
4648 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
4649 (dwarf2out_default_as_locview_support): New.
4650 (output_asm_line_debug_info): Use option variables.
4651 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
4652 (output_loc_list): Likewise.
4653 (add_high_low_attributes): Check option variables.
4654 Don't output entry view attribute in strict mode.
4655 (gen_inlined_subroutine_die): Check option variables.
4656 (dwarf2out_inline_entry): Likewise.
4657 (init_sections_and_labels): Likewise.
4658 (dwarf2out_early_finish): Likewise.
4659 (maybe_reset_location_view): New, from...
4660 (dwarf2out_var_location): ... here. Call it.
4661 * debug.h (dwarf2out_default_as_loc_support): Declare.
4662 (dwarf2out_default_as_locview_support): Declare.
4663 * hooks.c (hook_int_rtx_insn_0): New.
4664 * hooks.h (hook_int_rtx_insn_0): Declare.
4665 * toplev.c (process_options): Take -gas-loc-support and
4666 -gas-locview-support from dwarf2out. Enable
4667 -gvariable-location-views by default only with locview
4668 assembler support. Enable -ginternal-reset-location-views by
4669 default only if the target defines the corresponding hook.
4670 Enable -ginline-points by default if location views are
4671 enabled; force it disabled if statement frontiers are
4673 * tree-inline.c (expand_call_inline): Check option variables.
4674 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
4676 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
4678 PR tree-optimization/84321
4679 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
4680 handling. Also check whether the anti-range contains any values
4681 that satisfy the mask; switch to a VR_RANGE if not.
4683 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
4686 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
4688 2018-02-13 Martin Jambor <mjambor@suse.cz>
4691 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
4692 of call statements, also set location of a load to a temporary.
4694 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
4696 * config/rl78/rl78.c (add_vector_labels): New function.
4697 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
4698 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
4699 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
4700 which checks that no arguments are passed.
4701 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
4702 * doc/extend.texi: Documentation for the new attribute.
4704 2018-02-13 Andreas Schwab <schwab@suse.de>
4706 * config/riscv/linux.h (CPP_SPEC): Define.
4708 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4711 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
4712 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
4713 OPTION_MASK_ISA_AES as first argument to def_builtin_const
4714 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
4715 instead of OPTION_MASK_ISA_PCLMUL as first argument to
4716 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
4717 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
4718 temporarily for AES and PCLMUL builtins.
4720 PR tree-optimization/84339
4721 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
4722 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
4726 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
4727 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
4728 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
4730 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
4732 * omp-simd-clone.h: New file.
4733 * omp-simd-clone.c: Include omp-simd-clone.h.
4734 (expand_simd_clones): No longer static.
4735 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
4736 cgraph.h and omp-simd-clone.h.
4737 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
4738 (vect_recog_widen_shift_pattern): Formatting fix.
4739 (vect_pattern_recog_1): Don't check optab for calls.
4742 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
4743 operands[2] into a REG before using gen_lowpart on it.
4745 2018-02-12 Jeff Law <law@redhat.com>
4748 * config/sh/sh.c (find_barrier): Consider a sibling call
4751 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
4752 successfully back substituting a reg.
4754 2018-02-12 Richard Biener <rguenther@suse.de>
4756 PR tree-optimization/84037
4757 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
4758 parameter, move visited init to caller.
4759 (vect_slp_analyze_operations): Separate cost from validity
4760 check, initialize visited once for all instances.
4761 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
4763 * tree-vect-stmts.c (vect_model_simple_cost): Make early
4765 (vect_model_promotion_demotion_cost): Likewise.
4766 (vectorizable_bswap): Guard cost modeling with !slp_node
4767 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
4769 (vectorizable_call): Likewise.
4770 (vectorizable_conversion): Likewise.
4771 (vectorizable_assignment): Likewise.
4772 (vectorizable_shift): Likewise.
4773 (vectorizable_operation): Likewise.
4774 (vectorizable_store): Likewise.
4775 (vectorizable_load): Likewise.
4776 (vectorizable_condition): Likewise.
4777 (vectorizable_comparison): Likewise.
4779 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
4782 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
4783 (ASAN_MARK): Fix fnspec to account for return value, change pointer
4784 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
4786 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4789 * params.def (inline-min-speedup): Increase from 8 to 15.
4790 (max-inline-insns-auto): Decrease from 40 to 30.
4791 * ipa-split.c (consider_split): Add some buffer for function to
4792 be considered inlining candidate.
4793 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
4796 2018-02-12 Richard Biener <rguenther@suse.de>
4798 PR tree-optimization/84037
4799 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
4800 matched stmts if we cannot swap the non-matched ones.
4802 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4804 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
4805 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
4806 _mm_maskz_scalef_round_ss): New intrinsics.
4807 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
4808 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
4809 __builtin_ia32_scalefss_round): Remove.
4810 (__builtin_ia32_scalefsd_mask_round,
4811 __builtin_ia32_scalefss_mask_round): New intrinsics.
4812 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
4813 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
4814 ((match_operand:VF_128 2 "<round_nimm_predicate>"
4815 "<round_constraint>")): Changed to ...
4816 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
4817 "<round_scalar_constraint>")): ... this.
4818 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
4819 %0, %1, %2<round_op3>}"): Changed to ...
4820 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
4821 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
4822 %2<round_scalar_mask_op3>}"): ... this.
4823 * config/i386/subst.md (round_scalar_nimm_predicate): New.
4825 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4827 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
4828 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
4829 (_mm_maskz_sqrt_round_ss): New intrinsics.
4830 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
4831 (__builtin_ia32_sqrtsd_mask_round)
4832 (__builtin_ia32_sqrtss_mask_round): New builtins.
4833 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
4834 (__builtin_ia32_sqrtss_round): Remove.
4835 (__builtin_ia32_sqrtsd_mask_round)
4836 (__builtin_ia32_sqrtss_mask_round): New builtins.
4837 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
4838 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
4839 ((match_operand:VF_128 1 "vector_operand"
4840 "xBm,<round_constraint>")): Changed to ...
4841 ((match_operand:VF_128 1 "vector_operand"
4842 "xBm,<round_scalar_constraint>")): ... this.
4843 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
4844 %0, %2, %<iptr>1<round_op3>}): Changed to ...
4845 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
4846 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
4847 %<iptr>1<round_scalar_mask_op3>}): ... this.
4848 ((set_attr "prefix" "<round_prefix>")): Changed to ...
4849 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
4851 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
4854 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
4855 Cast vec_cmpeq result to correct type.
4856 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
4857 Cast vec_cmpgt result to correct type.
4859 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
4861 * final.c (final_scan_insn_1): Renamed from...
4862 (final_scan_insn): ... this. New wrapper, to recover
4863 seen from the outermost call in recursive ones.
4864 * config/sparc/sparc.c (output_return): Drop seen from call.
4865 (output_sibcall): Likewise.
4866 * config/visium/visium.c (output_branch): Likewise.
4868 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
4870 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
4873 2018-02-10 Alan Modra <amodra@gmail.com>
4876 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
4877 Specify LR as an input.
4879 2018-02-10 Jakub Jelinek <jakub@redhat.com>
4882 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
4883 remove_member_access_dummy_vars): New functions.
4884 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
4885 lower_omp_1, execute_lower_omp): Use them.
4887 PR rtl-optimization/84308
4888 * shrink-wrap.c (spread_components): Release todo vector.
4890 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
4892 PR rtl-optimization/57193
4893 * ira-color.c (struct allocno_color_data): Add member
4894 conflict_allocno_hard_prefs.
4895 (update_conflict_allocno_hard_prefs): New.
4896 (bucket_allocno_compare_func): Add a preference based on
4897 conflict_allocno_hard_prefs.
4898 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
4899 (color_allocnos): Remove a dead code. Initiate
4900 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
4902 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4905 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
4906 constraint from =wa to wa. Avoid a subreg on the output operand,
4907 instead use a pseudo and subreg it in a move.
4908 (p9_xxbrd_<mode>): Changed to ...
4909 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
4910 (p9_xxbrd_v2df): New expander.
4911 (p9_xxbrw_<mode>): Changed to ...
4912 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
4913 (p9_xxbrw_v4sf): New expander.
4915 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4917 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
4919 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
4922 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
4923 multiply in 32-bit mode.
4924 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
4925 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
4928 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4930 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
4931 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
4932 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
4933 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
4935 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
4938 * dwarf2out.c (is_trivial_indirect_ref): New function.
4939 (dwarf2out_late_global_decl): Do not generate a location
4940 attribute for variables that have a non-trivial DECL_VALUE_EXPR
4941 and that are not defined in the current unit.
4943 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4945 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
4946 instead of a libcall for UNORDERED.
4948 2018-02-09 Tamar Christina <tamar.christina@arm.com>
4951 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
4952 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
4954 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4957 * config/s390/s390.c (s390_set_current_function): Invoke
4958 s390_indirect_branch_settings also if fndecl didn't change.
4960 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4962 * config/rs6000/rs6000.md (blockage): Set length to zero.
4964 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4966 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
4968 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4971 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
4972 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
4976 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
4977 PARALLEL incoming that failed vt_get_decl_and_offset check.
4980 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
4981 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
4983 (get_variable_section): For decls in named .bss* sections pass true as
4984 second argument to bss_initializer_p.
4986 2018-02-09 Marek Polacek <polacek@redhat.com>
4987 Jakub Jelinek <jakub@redhat.com>
4990 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
4991 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
4992 Sync some changes from cxx_fold_indirect_ref.
4994 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4996 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
4998 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
4999 (BLOCK_INLINE_ENTRY_LABEL): New.
5000 (dwarf2out_var_location): Disregard inline entry markers.
5001 (inline_entry_data): New struct.
5002 (inline_entry_data_hasher): New hashtable type.
5003 (inline_entry_data_hasher::hash): New.
5004 (inline_entry_data_hasher::equal): New.
5005 (inline_entry_data_table): New variable.
5006 (add_high_low_attributes): Add DW_AT_entry_pc and
5007 DW_AT_GNU_entry_view attributes if a pending entry is found
5008 in inline_entry_data_table. Add old entry_pc attribute only
5009 if debug nonbinding markers are disabled.
5010 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
5011 markers are enabled.
5012 (block_within_block_p, dwarf2out_inline_entry): New.
5013 (dwarf2out_finish): Check that no entries remained in
5014 inline_entry_data_table.
5015 * final.c (reemit_insn_block_notes): Handle inline entry notes.
5016 (final_scan_insn, notice_source_line): Likewise.
5017 (rest_of_clean_state): Skip inline entry markers.
5018 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
5020 * gimple.c (gimple_build_debug_inline_entry): New.
5021 * gimple.h (enum gimple_debug_subcode): Add
5022 GIMPLE_DEBUG_INLINE_ENTRY.
5023 (gimple_build_debug_inline_entry): Declare.
5024 (gimple_debug_inline_entry_p): New.
5025 (gimple_debug_nonbind_marker_p): Adjust.
5026 * insn-notes.def (INLINE_ENTRY): New.
5027 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
5028 inline entry marker notes.
5029 (print_insn): Likewise.
5030 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
5031 (INSN_DEBUG_MARKER_KIND): Likewise.
5032 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
5033 * tree-inline.c (expand_call_inline): Build and insert
5034 debug_inline_entry stmt.
5035 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
5036 inline entry blocks early, if nonbind markers are enabled.
5037 (dump_scope_block): Dump fragment info.
5038 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
5039 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
5040 (gimple_build_debug_inline_entry): New.
5041 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
5042 Enable/disable inline entry points too.
5043 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
5044 (DEBUG_INSN): Describe inline entry markers.
5046 * common.opt (gvariable-location-views): New.
5047 (gvariable-location-views=incompat5): New.
5048 * config.in: Rebuilt.
5049 * configure: Rebuilt.
5050 * configure.ac: Test assembler for view support.
5051 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
5052 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
5053 * dwarf2out.c (var_loc_view): New typedef.
5054 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
5055 (dwarf2out_locviews_in_attribute): New.
5056 (dwarf2out_locviews_in_loclist): New.
5057 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
5058 (enum dw_line_info_opcode): Add LI_adv_address.
5059 (struct dw_line_info_table): Add view.
5060 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
5061 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
5062 (zero_view_p): New variable.
5063 (ZERO_VIEW_P): New macro.
5064 (output_asm_line_debug_info): New.
5065 (struct var_loc_node): Add view.
5066 (add_AT_view_list, AT_loc_list): New.
5067 (add_var_loc_to_decl): Add view param. Test it against last.
5068 (new_loc_list): Add view params. Record them.
5069 (AT_loc_list_ptr): Handle loc and view lists.
5070 (view_list_to_loc_list_val_node): New.
5071 (print_dw_val): Handle dw_val_class_view_list.
5072 (size_of_die): Likewise.
5073 (value_format): Likewise.
5074 (loc_list_has_views): New.
5075 (gen_llsym): Set vl_symbol too.
5076 (maybe_gen_llsym, skip_loc_list_entry): New.
5077 (dwarf2out_maybe_output_loclist_view_pair): New.
5078 (output_loc_list): Output view list or entries too.
5079 (output_view_list_offset): New.
5080 (output_die): Handle dw_val_class_view_list.
5081 (output_dwarf_version): New.
5082 (output_compilation_unit_header): Use it.
5083 (output_skeleton_debug_sections): Likewise.
5084 (output_rnglists, output_line_info): Likewise.
5085 (output_pubnames, output_aranges): Update version comments.
5086 (output_one_line_info_table): Output view numbers in asm comments.
5087 (dw_loc_list): Determine current endview, pass it to new_loc_list.
5088 Call maybe_gen_llsym.
5089 (loc_list_from_tree_1): Adjust.
5090 (add_AT_location_description): Create view list attribute if
5091 needed, check it's absent otherwise.
5092 (convert_cfa_to_fb_loc_list): Adjust.
5093 (maybe_emit_file): Call output_asm_line_debug_info for test.
5094 (dwarf2out_var_location): Reset views as needed. Precompute
5095 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
5096 attribute. Set view.
5097 (new_line_info_table): Reset next view.
5098 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
5099 (dwarf2out_source_line): Likewise. Output view resets and labels to
5100 the assembler, or select appropriate line info opcodes.
5101 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
5102 (optimize_string_length): Catch it. Adjust.
5103 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
5104 dw_val_class_view_list, and remove it if no longer needed.
5105 (hash_loc_list): Hash view numbers.
5106 (loc_list_hasher::equal): Compare them.
5107 (optimize_location_lists): Check whether a view list symbol is
5108 needed, and whether the locview attribute is present, and
5109 whether they match. Remove the locview attribute if no longer
5111 (index_location_lists): Call skip_loc_list_entry for test.
5112 (dwarf2out_finish): Call output_asm_line_debug_info for test.
5113 Use output_dwarf_version.
5114 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
5115 (struct dw_val_node): Add val_view_list.
5116 * final.c (SEEN_NEXT_VIEW): New.
5117 (set_next_view_needed): New.
5118 (clear_next_view_needed): New.
5119 (maybe_output_next_view): New.
5120 (final_start_function): Rename to...
5121 (final_start_function_1): ... this. Take pointer to FIRST,
5122 add SEEN parameter. Emit param bindings in the initial view.
5123 (final_start_function): Reintroduce SEEN-less interface.
5124 (final): Rename to...
5125 (final_1): ... this. Take SEEN parameter. Output final pending
5126 next view at the end.
5127 (final): Reintroduce seen-less interface.
5128 (final_scan_insn): Output pending next view before switching
5129 sections or ending a block. Mark the next view as needed when
5130 outputting variable locations. Notify debug backend of section
5131 changes, and of location view changes.
5132 (rest_of_handle_final): Adjust.
5133 * toplev.c (process_options): Autodetect value for debug variable
5134 location views option. Warn on incompat5 without -gdwarf-5.
5135 * doc/invoke.texi (gvariable-location-views): New.
5136 (gvariable-location-views=incompat5): New.
5137 (gno-variable-location-views): New.
5139 2018-02-08 David Malcolm <dmalcolm@redhat.com>
5141 PR tree-optimization/84136
5142 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
5143 that the result of find_edge is non-NULL.
5145 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5148 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
5149 storing integer register in SImode. Fix cost of 256 and 512
5150 byte aligned SSE register store.
5152 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5154 * config/i386/i386.c (ix86_multiplication_cost): Fix
5155 multiplication cost for TARGET_AVX512DQ.
5157 2018-02-08 Marek Polacek <polacek@redhat.com>
5159 PR tree-optimization/84238
5160 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
5163 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5165 PR tree-optimization/84265
5166 * tree-vect-stmts.c (vectorizable_store): Don't treat
5167 VMAT_CONTIGUOUS accesses as grouped.
5168 (vectorizable_load): Likewise.
5170 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5172 PR tree-optimization/81635
5173 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
5174 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
5175 (test_round_for_mask): New functions.
5176 (wide_int_cc_tests): Call test_round_for_mask.
5177 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
5178 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
5179 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
5180 range returned by get_range_info.
5182 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5185 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
5186 * symtab.c: Include builtins.h
5187 (symtab_node::output_to_lto_symbol_table_p): Move here
5188 from lto-streamer-out.c:output_symbol_p.
5189 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
5190 (output_symbol_p): Move all logic to symtab.c
5191 (produce_symtab): Update.
5193 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5195 * config/s390/s390-opts.h (enum indirect_branch): Define.
5196 * config/s390/s390-protos.h (s390_return_addr_from_memory)
5197 (s390_indirect_branch_via_thunk)
5198 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
5199 (enum s390_indirect_branch_type): Define.
5200 * config/s390/s390.c (struct s390_frame_layout, struct
5201 machine_function): Remove.
5202 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
5203 (indirect_branch_table_label_no, indirect_branch_table_name):
5205 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
5206 (enum s390_indirect_branch_option): Define.
5207 (s390_return_addr_from_memory): New function.
5208 (s390_handle_string_attribute): New function.
5209 (s390_attribute_table): Add new attribute handler.
5210 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
5211 (s390_indirect_branch_via_thunk): New function.
5212 (s390_indirect_branch_via_inline_thunk): New function.
5213 (s390_function_ok_for_sibcall): When jumping via thunk disallow
5214 sibling call optimization for non z10 compiles.
5215 (s390_emit_call): Force indirect branch target to be a single
5216 register. Add r1 clobber for non-z10 compiles.
5217 (s390_emit_epilogue): Emit return jump via return_use expander.
5218 (s390_reorg): Handle JUMP_INSNs as execute targets.
5219 (s390_option_override_internal): Perform validity checks for the
5220 new command line options.
5221 (s390_indirect_branch_attrvalue): New function.
5222 (s390_indirect_branch_settings): New function.
5223 (s390_set_current_function): Invoke s390_indirect_branch_settings.
5224 (s390_output_indirect_thunk_function): New function.
5225 (s390_code_end): Implement target hook.
5226 (s390_case_values_threshold): Implement target hook.
5227 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
5229 * config/s390/s390.h (struct s390_frame_layout)
5230 (struct machine_function): Move here from s390.c.
5231 (TARGET_INDIRECT_BRANCH_NOBP_RET)
5232 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
5233 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
5234 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
5235 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
5236 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
5237 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
5238 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
5239 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
5240 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
5241 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
5242 (mnemonic attribute): Add values which aren't recognized
5244 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
5245 pattern for branch conversion. Fix mnemonic attribute.
5246 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
5247 indirect branch via thunk if requested.
5248 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
5249 ("*indirect_jump"): Disable for branch conversion using out of
5251 ("indirect_jump_via_thunk<mode>_z10")
5252 ("indirect_jump_via_thunk<mode>")
5253 ("indirect_jump_via_inlinethunk<mode>_z10")
5254 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
5255 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
5256 ("casesi_jump_via_inlinethunk<mode>_z10")
5257 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
5258 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
5259 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
5260 ("*indirect2_jump"): Disable for branch conversion.
5261 ("casesi_jump"): Turn into expander and expand patterns for branch
5263 ("return_use"): New expander.
5264 ("*return"): Emit return via thunk and rename it to ...
5265 ("*return<mode>"): ... this one.
5266 * config/s390/s390.opt: Add new options and and enum for the
5269 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5271 * lra-constraints.c (match_reload): Unconditionally use
5272 gen_lowpart_SUBREG, rather than selecting between that
5273 and equivalent gen_rtx_SUBREG code.
5275 2018-02-08 Richard Biener <rguenther@suse.de>
5277 PR tree-optimization/84233
5278 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
5279 changed flag instead of boguously re-using phi_inserted.
5281 2018-02-08 Martin Jambor <mjambor@suse.cz>
5283 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
5284 static local variables.
5286 2018-02-08 Richard Biener <rguenther@suse.de>
5288 PR tree-optimization/84278
5289 * tree-vect-stmts.c (vectorizable_store): When looking for
5290 smaller vector types to perform grouped strided loads/stores
5291 make sure the mode is supported by the target.
5292 (vectorizable_load): Likewise.
5294 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5296 * config/aarch64/aarch64.c (aarch64_components_for_bb):
5297 Increase LDP/STP opportunities by adding adjacent callee-saves.
5299 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5301 PR rtl-optimization/84068
5302 PR rtl-optimization/83459
5303 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
5305 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
5307 PR tree-optimization/84224
5308 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
5309 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
5312 2018-02-07 Iain Sandoe <iain@codesourcery.com>
5315 * config/rs6000/altivec.md (*restore_world): Remove LR use.
5316 * config/rs6000/predicates.md (restore_world_operation): Adjust op
5317 count, remove one USE.
5319 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
5321 * doc/install.texi (Configuration): Document the
5322 --with-long-double-format={ibm,ieee} PowerPC configuration
5326 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
5327 Convert from define_expand to be define_insn_and_split. Rework
5328 float/double/_Float128 conversions to QI/HI/SImode to work with
5329 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
5330 conversions to QI/HImode types did a store and then a load to
5331 truncate the value. For conversions to VSX registers, don't split
5332 the insn, instead emit the code directly. Use the code iterator
5333 any_fix to combine signed and unsigned conversions.
5334 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
5335 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
5336 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5337 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
5338 (fix_<mode>di2_hw): Likewise.
5339 (fixuns_<mode>di2_hw): Likewise.
5340 (fix_<mode>si2_hw): Likewise.
5341 (fixuns_<mode>si2_hw): Likewise.
5342 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
5343 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5344 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
5345 fix<uns>_trunc<SFDF:mode>si2_p8.
5346 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
5348 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
5349 (fix<uns>_<mode>_mem): Likewise.
5350 (fctiw<u>z_<mode>_mem): Likewise.
5351 (fix<uns>_<mode>_mem): Likewise.
5352 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
5353 the register allocator from doing a direct move to the GPRs to do
5354 a store, and instead use the ISA 3.0 store byte/half-word from
5355 vector register instruction. For IEEE 128-bit floating point,
5356 also optimize stores of 32-bit ints.
5357 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
5359 2018-02-07 Alan Hayward <alan.hayward@arm.com>
5361 * genextract.c (push_pathstr_operand): New function to support
5363 (walk_rtx): Call push_pathstr_operand.
5364 (print_path): Support [a-zA-Z].
5366 2018-02-07 Richard Biener <rguenther@suse.de>
5368 PR tree-optimization/84037
5369 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
5370 (cse_and_gimplify_to_preheader): Declare.
5371 (vect_get_place_in_interleaving_chain): Likewise.
5372 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5374 (_loop_vec_info::~_loop_vec_info): Delete it.
5375 (cse_and_gimplify_to_preheader): New function.
5376 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
5377 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
5378 (vectorizable_load): Likewise. For grouped stores always base
5379 the IV on the first element.
5380 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
5381 condition before gimplifying.
5383 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5385 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
5386 *DIV_EXPR and *MOD_EXPR.
5388 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
5391 * config/i386/i386.c (ix86_option_override_internal): Mask out
5392 the CF_SET bit when checking -fcf-protection.
5394 2018-02-07 Tom de Vries <tom@codesourcery.com>
5397 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
5400 2018-02-07 Richard Biener <rguenther@suse.de>
5402 PR tree-optimization/84204
5403 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
5406 PR tree-optimization/84205
5407 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
5408 special-case isl_ast_op_zdiv_r.
5410 PR tree-optimization/84223
5411 * graphite-scop-detection.c (gather_bbs::before_dom_children):
5412 Only add conditions from within the region.
5413 (gather_bbs::after_dom_children): Adjust.
5415 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
5418 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
5419 * config/avr/avr.md: Only post-reload split REG-REG moves if
5420 either register is GENERAL_REG_P.
5422 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5424 PR tree-optimization/84235
5425 * tree-ssa-scopedtables.c
5426 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
5427 if the subtraction is performed in floating point type where NaNs are
5428 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
5429 build 1. Formatting fix.
5431 2018-02-06 Jakub Jelinek <jakub@redhat.com>
5434 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
5435 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
5436 and skip it regardless of bb boundaries. Use CALL_P macro,
5437 don't test INSN_P (insn) together with CALL_P or JUMP_P check
5438 unnecessarily, formatting fix.
5440 2018-02-06 Michael Collison <michael.collison@arm.com>
5442 * config/arm/thumb2.md:
5443 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
5444 (*thumb_mov_notscc): Ditto.
5446 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
5449 * config/rs6000/rs6000.md (su code attribute): Use "u" for
5450 unsigned_fix, not "s".
5452 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5454 * configure.ac (gcc_fn_eh_frame_ro): New function.
5455 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
5456 correct .eh_frame permissions.
5457 * configure: Regenerate.
5459 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
5461 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
5464 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5466 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5467 Display warning message for -mno-speculate-indirect-jumps.
5469 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
5471 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
5473 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
5475 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
5477 PR tree-optimization/84225
5478 * tree-eh.c (find_trapping_overflow): Only call
5479 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
5481 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5484 * config/i386/i386.c: Reimplement the check of possible options
5485 -mibt/-mshstk conbination. Change error messages.
5486 * doc/invoke.texi: Fix a typo: remove extra '='.
5488 2018-02-06 Marek Polacek <polacek@redhat.com>
5490 PR tree-optimization/84228
5491 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
5493 2018-02-06 Tamar Christina <tamar.christina@arm.com>
5496 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
5497 emitted arch directives.
5498 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
5499 __ARM_FEATURE_COPROC before changing architectures.
5501 2018-02-06 Richard Biener <rguenther@suse.de>
5503 * config/i386/i386.c (print_reg): Fix typo.
5504 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
5506 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
5508 * configure: Regenerate.
5510 2018-02-05 Martin Sebor <msebor@redhat.com>
5512 PR tree-optimization/83369
5513 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
5516 2018-02-05 Martin Liska <mliska@suse.cz>
5518 * doc/invoke.texi: Cherry-pick upstream r323995.
5520 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
5522 * ira.c (ira_init_register_move_cost): Adjust comment.
5524 2018-02-05 Martin Liska <mliska@suse.cz>
5526 PR gcov-profile/84137
5527 * doc/gcov.texi: Fix typo in documentation.
5529 2018-02-05 Martin Liska <mliska@suse.cz>
5531 PR gcov-profile/83879
5532 * doc/gcov.texi: Document necessity of --dynamic-list-data when
5533 using dlopen functionality.
5535 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
5537 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
5538 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
5539 _mm_maskz_range_ss, _mm_mask_range_round_ss,
5540 _mm_maskz_range_round_ss): New intrinsics.
5541 (__builtin_ia32_rangesd128_round)
5542 (__builtin_ia32_rangess128_round): Remove.
5543 (__builtin_ia32_rangesd128_mask_round,
5544 __builtin_ia32_rangess128_mask_round): New builtins.
5545 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
5546 __builtin_ia32_rangess128_round): Remove.
5547 (__builtin_ia32_rangesd128_mask_round,
5548 __builtin_ia32_rangess128_mask_round): New builtins.
5549 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
5550 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
5551 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
5552 "<round_saeonly_constraint>")): Changed to ...
5553 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
5554 "<round_saeonly_scalar_constraint>")): ... this.
5555 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
5556 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
5557 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
5558 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
5559 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
5561 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
5563 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
5565 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
5566 Remove all values except native, 8540 and 8548.
5568 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
5570 * config/i386/i386.c (ix86_output_function_return): Pass
5571 INVALID_REGNUM, instead of -1, as invalid register number to
5572 indirect_thunk_name and output_indirect_thunk.
5574 2018-02-02 Julia Koval <julia.koval@intel.com>
5576 * config.gcc: Add -march=icelake.
5577 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
5578 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
5579 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
5580 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
5581 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
5582 (processor_target_table): Add icelake.
5583 (ix86_option_override_internal): Handle new PTAs.
5584 (get_builtin_code_for_version): Handle icelake.
5585 (M_INTEL_COREI7_ICELAKE): New.
5586 (fold_builtin_cpu): Handle icelake.
5587 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
5588 * doc/invoke.texi: Add -march=icelake.
5590 2018-02-02 Julia Koval <julia.koval@intel.com>
5592 * config/i386/i386.c (ix86_option_override_internal): Change flags type
5593 to wide_int_bitmask.
5594 * wide-int-bitmask.h: New.
5596 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5599 * config/i386/i386.md: Replace Pmode with word_mode in
5600 builtin_setjmp_setup and builtin_longjmp to support x32.
5602 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
5606 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
5608 (rs6000_supported_cpu_names): New static variable.
5609 (linux_cpu_translation_table): Likewise.
5610 (elf_platform) <cpu>: Define new static variable and use it.
5611 Translate kernel AT_PLATFORM name to canonical name if needed.
5612 Error if platform name is unknown.
5614 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
5617 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
5619 2018-02-01 Jeff Law <law@redhat.com>
5622 * config/i386/i386.c (release_scratch_register_on_entry): Add new
5623 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
5624 the scratch if RELEASE_VIA_POP is false.
5625 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
5626 If we have to save a temporary register, decrement SIZE appropriately.
5627 Pass new arguments to release_scratch_register_on_entry.
5628 (ix86_adjust_stack_and_probe): Likewise.
5629 (ix86_emit_probe_stack_range): Pass new arguments to
5630 release_scratch_register_on_entry.
5632 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
5634 PR rtl-optimization/84157
5635 * combine.c (change_zero_ext): Use REG_P predicate in
5636 front of HARD_REGISTER_P predicate.
5638 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
5640 * config/avr/avr.c (avr_option_override): Move disabling of
5641 -fdelete-null-pointer-checks to...
5642 * common/config/avr/avr-common.c (avr_option_optimization_table):
5645 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5647 PR tree-optimization/81635
5648 * tree-data-ref.c (split_constant_offset_1): For types that
5649 wrap on overflow, try to use range info to prove that wrapping
5652 2018-02-01 Renlin Li <renlin.li@arm.com>
5655 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
5657 (aarch64_register_move_cost): Likewise.
5658 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
5660 (REG_CLASS_NAMES): Likewise.
5661 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
5662 TAILCALL_ADDR_REGS. Remove IP registers.
5663 * config/aarch64/aarch64.md (Ucs): Update register constraint.
5665 2018-02-01 Richard Biener <rguenther@suse.de>
5667 * domwalk.h (dom_walker::dom_walker): Add additional constructor
5668 for specifying RPO order and allow NULL for that.
5669 * domwalk.c (dom_walker::dom_walker): Likewise.
5670 (dom_walker::walk): Handle NULL RPO order.
5671 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
5673 (rewrite_update_dom_walker): Likewise.
5674 (mark_def_dom_walker): Likewise.
5676 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5678 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
5679 (aarch64_maybe_expand_sve_subreg_move): Declare.
5680 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
5681 * config/aarch64/predicates.md (aarch64_any_register_operand): New
5683 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
5684 that are semantically a reverse operation.
5685 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
5686 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
5687 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
5689 (aarch64_can_change_mode_class): For big-endian, forbid changes
5690 between two SVE modes if they have different element sizes.
5692 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5694 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
5695 the TImode handling for big-endian targets.
5697 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5699 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
5700 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
5702 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
5703 Remove BSWAP handing for big-endian targets and use the form of
5704 LD1RQ appropariate for the mode.
5706 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5708 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
5709 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
5712 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5715 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
5716 check for operands that need to go through aarch64_sve_reload_be.
5718 2018-02-01 Jakub Jelinek <jakub@redhat.com>
5720 PR tree-optimization/81661
5721 PR tree-optimization/84117
5722 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
5723 * tree-eh.c: Include gimplify.h.
5724 (find_trapping_overflow, replace_trapping_overflow,
5725 rewrite_to_non_trapping_overflow): New functions.
5726 * tree-vect-loop.c: Include tree-eh.h.
5727 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
5728 * tree-data-ref.c: Include tree-eh.h.
5729 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
5731 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
5733 PR rtl-optimization/84123
5734 * combine.c (change_zero_ext): Check if hard register satisfies
5735 can_change_dest_mode before calling gen_lowpart_SUBREG.
5737 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
5740 * ira.c (ira_init_register_move_cost): Remove assert.
5742 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5744 PR rtl-optimization/84071
5745 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
5746 * doc/tm.texi: Regenerate.
5748 2018-01-31 Richard Biener <rguenther@suse.de>
5750 PR tree-optimization/84132
5751 * tree-data-ref.c (analyze_miv_subscript): Properly
5752 check whether evolution_function_is_affine_multivariate_p
5753 before calling gcd_of_steps_may_divide_p.
5755 2018-01-31 Julia Koval <julia.koval@intel.com>
5758 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
5759 * config/i386/i386.md (rdpid_rex64) New.
5760 (rdpid): Make 32bit only.
5762 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
5765 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
5766 an IDENTIFIER_NODE for FUNCTION_TYPE's.
5768 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5771 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5773 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5775 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5777 PR rtl-optimization/84071
5778 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
5779 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
5781 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5783 * config/arc/arc.c (arc_handle_aux_attribute): New function.
5784 (arc_attribute_table): Add 'aux' attribute.
5785 (arc_in_small_data_p): Consider aux like variables.
5786 (arc_is_aux_reg_p): New function.
5787 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
5788 (arc_get_aux_arg): New function.
5789 (prepare_move_operands): Handle aux-register access.
5790 (arc_handle_aux_attribute): New function.
5791 * doc/extend.texi (ARC Variable attributes): Add subsection.
5793 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5795 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
5796 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
5797 (arc_attribute_table): Add 'uncached' attribute.
5798 (arc_print_operand): Print '.di' flag for uncached memory
5800 (arc_in_small_data_p): Do not consider for small data the uncached
5802 (arc_is_uncached_mem_p): New function.
5803 * config/arc/predicates.md (compact_store_memory_operand): Check
5804 for uncached memory accesses.
5805 (nonvol_nonimm_operand): Likewise.
5806 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
5808 2018-01-31 Jakub Jelinek <jakub@redhat.com>
5811 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
5812 falign-loops=): Add Optimization flag.
5814 2018-01-30 Jeff Law <law@redhat.com>
5817 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
5818 INT_REGISTERS_SAVED. Check it prior to calling
5819 get_scratch_register_on_entry.
5820 (ix86_adjust_stack_and_probe): Similarly.
5821 (ix86_emit_probe_stack_range): Similarly.
5822 (ix86_expand_prologue): Corresponding changes.
5824 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5827 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
5828 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
5830 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
5833 * lra-constraints.c (curr_insn_transform): Process AND in the
5836 2018-01-30 Jakub Jelinek <jakub@redhat.com>
5838 PR rtl-optimization/83986
5839 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
5840 dependence against last_pending_memory_flush in addition to
5843 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
5845 PR tree-optimization/81611
5846 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
5849 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5852 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
5855 2018-01-30 Richard Biener <rguenther@suse.de>
5856 Jakub Jelinek <jakub@redhat.com>
5858 PR tree-optimization/84111
5859 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
5860 inner loops added during recursion, as they don't have up-to-date
5863 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5866 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
5867 (can_inline_edge_by_limits_p): ... here.
5868 (can_early_inline_edge_p, check_callers,
5869 update_caller_keys, update_callee_keys, recursive_inlining,
5870 add_new_edges_to_heap, speculation_useful_p,
5871 inline_small_functions,
5872 inline_small_functions, flatten_function,
5873 inline_to_all_callers_1): Update.
5875 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5877 * profile-count.c (profile_count::combine_with_ipa_count): Handle
5880 2018-01-30 Richard Biener <rguenther@suse.de>
5882 PR tree-optimization/83008
5883 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
5884 invariant and constant vector uses in stmts when they need
5887 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5890 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
5891 * configure: Regenerate.
5893 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5895 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
5897 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
5898 Use gen_rtx_REG rather than gen_lowpart.
5900 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5902 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
5903 rather than 0 when creating partial subregs.
5905 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5907 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
5910 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
5913 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
5914 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
5915 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
5916 flags. This restores the settings used before the 2017-07-24.
5917 Turning off pre increment/decrement/modify allows IVOPTS to
5918 optimize DF/SF loops where the index is an int.
5920 2018-01-29 Richard Biener <rguenther@suse.de>
5921 Kelvin Nilsen <kelvin@gcc.gnu.org>
5924 * tree-vect-stmts.c (vectorizable_call): Don't call
5925 targetm.vectorize_builtin_md_vectorized_function if callee is
5928 2018-01-22 Carl Love <cel@us.ibm.com>
5930 * doc/extend.tex: Fix typo in second arg in
5931 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
5933 2018-01-29 Richard Biener <rguenther@suse.de>
5935 PR tree-optimization/84086
5936 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
5937 (flush_ssaname_freelist): When SSA names were released reset
5938 the SCEV hash table.
5940 2018-01-29 Richard Biener <rguenther@suse.de>
5942 PR tree-optimization/84057
5943 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
5944 removed paths when removing edges.
5946 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
5948 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
5949 -mfunction-return=@var{choice}.
5951 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
5954 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
5956 (layout::print_source_line): Likewise.
5957 (test_get_line_width_without_trailing_whitespace): Add test cases.
5959 2018-01-27 Jakub Jelinek <jakub@redhat.com>
5962 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
5965 2018-01-26 Jim Wilson <jimw@sifive.com>
5967 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
5969 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
5972 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5974 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
5975 and CMP + SUB-immediate -> SUBS.
5977 2018-01-26 Martin Sebor <msebor@redhat.com>
5979 PR tree-optimization/83896
5980 * tree-ssa-strlen.c (get_string_len): Rename...
5981 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
5982 Avoid assuming length is constant.
5983 (handle_char_store): Use HOST_WIDE_INT for string length.
5985 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
5988 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
5989 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
5991 2018-01-26 Richard Biener <rguenther@suse.de>
5993 PR rtl-optimization/84003
5994 * dse.c (record_store): Only record redundant stores when
5995 the earlier store aliases at least all accesses the later one does.
5997 2018-01-26 Jakub Jelinek <jakub@redhat.com>
5999 PR rtl-optimization/83985
6000 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
6001 REG_CFA_RESTORE insns.
6002 (delete_unmarked_insns): Don't ignore separate shrink wrapping
6003 REG_CFA_RESTORE insns here.
6006 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
6007 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
6009 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6011 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
6012 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
6013 (arc_init): Likewise.
6014 (arc_override_options): Likewise.
6015 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
6017 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
6019 * config/arc/arc.h (TARGET_DBNZ): Define.
6020 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
6021 properly set the tune attribute.
6022 (dbnz): Use TARGET_DBNZ guard.
6023 * config/arc/arc.opt (mtune): Add core3 option.
6025 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6027 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
6028 recognize new pic like addresses.
6029 (arc_delegitimize_address): Clean up.
6031 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6033 * config/arc/arc-arches.def: Option mrf16 valid for all
6035 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
6036 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
6037 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
6038 * config/arc/arc-tables.opt: Regenerate.
6039 * config/arc/arc.c (arc_conditional_register_usage): Handle
6040 reduced register file case.
6041 (arc_file_start): Set must have build attributes.
6042 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
6044 * config/arc/arc.opt (mrf16): Add new option.
6045 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
6046 * config/arc/genmultilib.awk: Handle new mrf16 option.
6047 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
6048 * config/arc/t-multilib: Regenerate.
6049 * doc/invoke.texi (ARC Options): Document mrf16 option.
6051 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6053 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
6054 * config/arc/arc.c (arc_handle_secure_attribute): New function.
6055 (arc_attribute_table): Add 'secure_call' attribute.
6056 (arc_print_operand): Print secure call operand.
6057 (arc_function_ok_for_sibcall): Don't optimize tail calls when
6059 (arc_is_secure_call_p): New function. * config/arc/arc.md
6060 (call_i): Add support for sjli instruction.
6061 (call_value_i): Likewise.
6062 * config/arc/constraints.md (Csc): New constraint.
6064 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6065 John Eric Martin <John.Martin@emmicro-us.com>
6067 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
6068 * config/arc/arc.c (_arc_jli_section): New struct.
6069 (arc_jli_section): New type.
6070 (rc_jli_sections): New static variable.
6071 (arc_handle_jli_attribute): New function.
6072 (arc_attribute_table): Add jli_always and jli_fixed attribute.
6073 (arc_file_end): New function.
6074 (TARGET_ASM_FILE_END): Define.
6075 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
6076 (arc_add_jli_section): New function.
6077 (jli_call_scan): Likewise.
6078 (arc_reorg): Call jli_call_scan.
6079 (arc_output_addsi): Remove 'S' from printing asm operand.
6080 (arc_is_jli_call_p): New function.
6081 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
6083 (movhi_insn): Likewise.
6084 (movsi_insn): Likewise.
6085 (movsi_set_cc_insn): Likewise.
6086 (loadqi_update): Likewise.
6087 (load_zeroextendqisi_update): Likewise.
6088 (load_signextendqisi_update): Likewise.
6089 (loadhi_update): Likewise.
6090 (load_zeroextendhisi_update): Likewise.
6091 (load_signextendhisi_update): Likewise.
6092 (loadsi_update): Likewise.
6093 (loadsf_update): Likewise.
6094 (movsicc_insn): Likewise.
6095 (bset_insn): Likewise.
6096 (bxor_insn): Likewise.
6097 (bclr_insn): Likewise.
6098 (bmsk_insn): Likewise.
6099 (bicsi3_insn): Likewise.
6100 (cmpsi_cc_c_insn): Likewise.
6101 (movsi_ne): Likewise.
6102 (movsi_cond_exec): Likewise.
6103 (clrsbsi2): Likewise.
6114 (call_i): Remove 'S' asm letter, add jli instruction.
6115 (call_value_i): Likewise.
6116 * config/arc/arc.op (mjli-always): New option.
6117 * config/arc/constraints.md (Cji): New constraint.
6118 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
6120 (subsf3_fpx): Likewise.
6121 (mulsf3_fpx): Likewise.
6122 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
6124 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
6126 * doc/invoke.texi (ARC): Document mjli-always option.
6128 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
6130 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
6131 avoid addition with 0 and use incw and decw where possible.
6133 2018-01-26 Richard Biener <rguenther@suse.de>
6135 PR tree-optimization/81082
6136 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
6137 association if it requires casting to unsigned.
6138 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
6139 from fold_plusminus_mult_expr to catch important cases late when
6140 range info is available.
6142 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6144 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
6145 * configure.ac (hidden_linkonce): New test.
6146 * configure: Regenerate.
6147 * config.in: Regenerate.
6149 2018-01-26 Julia Koval <julia.koval@intel.com>
6151 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
6152 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
6153 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
6154 _mm_mask_bitshuffle_epi64_mask): Fix type.
6155 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
6156 USI_FTYPE_V4DI_V4DI_USI): Remove.
6157 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
6158 __builtin_ia32_vpshufbitqmb256_mask,
6159 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
6160 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
6161 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
6163 2018-01-26 Alan Modra <amodra@gmail.com>
6166 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
6167 UNSPEC_VBPERMQ. Sort other unspecs.
6169 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
6171 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
6173 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
6176 * predict.c (drop_profile): Do not push/pop cfun; update also
6178 (handle_missing_profiles): Fix logic looking for zero profiles.
6180 2018-01-25 Jakub Jelinek <jakub@redhat.com>
6183 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
6184 on functions with #pragma omp declare simd or functions with simd
6186 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
6187 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
6188 Remove trailing \n from warning_at calls.
6190 2018-01-25 Tom de Vries <tom@codesourcery.com>
6193 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6194 for neutered workers.
6196 2018-01-24 Joseph Myers <joseph@codesourcery.com>
6199 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
6200 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
6202 2018-01-24 Jeff Law <law@redhat.com>
6205 * i386.c (get_probe_interval): Move to earlier point.
6206 (ix86_compute_frame_layout): If -fstack-clash-protection and
6207 the frame is larger than the probe interval, then use pushes
6208 to save registers rather than reg->mem moves.
6209 (ix86_expand_prologue): Remove conditional for int_registers_saved
6212 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
6215 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
6216 min/max for never referenced object.
6218 2018-01-24 Jakub Jelinek <jakub@redhat.com>
6221 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
6223 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
6224 attributes from DECL_ATTRIBUTES (decl) without affecting
6225 DECL_ATTRIBUTES (current_function_decl).
6226 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
6227 functions with non-NULL DECL_ABSTRACT_ORIGIN.
6229 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
6231 PR tree-optimization/83979
6232 * fold-const.c (fold_comparison): Use constant_boolean_node
6233 instead of boolean_{true,false}_node.
6235 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
6237 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
6240 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6242 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6243 Simplify the clause that sets the length attribute.
6244 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6245 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
6246 clause that sets the length attribute.
6247 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6249 2018-01-24 Tom de Vries <tom@codesourcery.com>
6252 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
6253 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
6254 Add strict parameter.
6255 (prevent_branch_around_nothing): Insert dummy insn between branch to
6256 label and label with no ptx insn inbetween.
6257 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
6259 2018-01-24 Tom de Vries <tom@codesourcery.com>
6262 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6263 for neutered threads in warp.
6264 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
6266 2018-01-24 Richard Biener <rguenther@suse.de>
6268 PR tree-optimization/83176
6269 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
6272 2018-01-24 Richard Biener <rguenther@suse.de>
6274 PR tree-optimization/82819
6275 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
6276 code generating pluses that are no-ops in the target precision.
6278 2018-01-24 Richard Biener <rguenther@suse.de>
6281 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
6283 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6285 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
6286 to merge probabilities.
6287 * predict.c (probably_never_executed): Also mark as cold functions
6288 with global 0 profile and guessed local profile.
6289 * profile-count.c (profile_probability::combine_with_count): New
6291 * profile-count.h (profile_probability::operator*,
6292 profile_probability::operator*=, profile_probability::operator/,
6293 profile_probability::operator/=): Reduce precision to adjusted
6294 and set value to guessed on contradictory divisions.
6295 (profile_probability::combine_with_freq): Remove.
6296 (profile_probability::combine_wiht_count): Declare.
6297 (profile_count::force_nonzero):: Set to adjusted.
6298 (profile_count::probability_in):: Set quality to adjusted.
6299 * tree-ssa-tail-merge.c (replace_block_by): Use
6302 2018-01-23 Andrew Waterman <andrew@sifive.com>
6303 Jim Wilson <jimw@sifive.com>
6305 * config/riscv/riscv.c (riscv_stack_boundary): New.
6306 (riscv_option_override): Set riscv_stack_boundary. Handle
6307 riscv_preferred_stack_boundary_arg.
6308 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
6309 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
6310 (STACK_BOUNDARY): Set to riscv_stack_boundary.
6311 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
6312 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
6313 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
6315 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
6318 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
6319 of struct ix86_frame.
6320 (ix86_expand_epilogue): Likewise. Add a local variable for
6321 the reg_save_offset field in struct ix86_frame.
6323 2018-01-23 Bin Cheng <bin.cheng@arm.com>
6325 PR tree-optimization/82604
6326 * tree-loop-distribution.c (enum partition_kind): New enum item
6327 PKIND_PARTIAL_MEMSET.
6328 (partition_builtin_p): Support above new enum item.
6329 (generate_code_for_partition): Ditto.
6330 (compute_access_range): Differentiate cases that equality can be
6331 proven at all loops, the innermost loops or no loops.
6332 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
6333 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
6334 (finalize_partitions, distribute_loop): Don't fuse partition of
6335 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
6336 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
6339 2018-01-23 Martin Liska <mliska@suse.cz>
6341 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
6342 order to ignore the predictor.
6343 (PRED_POLYMORPHIC_CALL): Likewise.
6344 (PRED_RECURSIVE_CALL): Likewise.
6346 2018-01-23 Martin Liska <mliska@suse.cz>
6348 * tree-profile.c (tree_profiling): Print function header to
6349 aware reader which function we are working on.
6350 * value-prof.c (gimple_find_values_to_profile): Do not print
6351 not interesting value histograms.
6353 2018-01-23 Martin Liska <mliska@suse.cz>
6355 * profile-count.h (enum profile_quality): Add
6356 profile_uninitialized as the first value. Do not number values
6357 as they are zero based.
6358 (profile_count::verify): Update sanity check.
6359 (profile_probability::verify): Likewise.
6361 2018-01-23 Nathan Sidwell <nathan@acm.org>
6363 * doc/invoke.texi (ffor-scope): Deprecate.
6365 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6367 PR tree-optimization/83510
6368 * domwalk.c (set_all_edges_as_executable): New function.
6369 (dom_walker::dom_walker): Convert bool param
6370 "skip_unreachable_blocks" to enum reachability. Move setup of
6371 edge flags to set_all_edges_as_executable and only do it when
6372 reachability is REACHABLE_BLOCKS.
6373 * domwalk.h (enum dom_walker::reachability): New enum.
6374 (dom_walker::dom_walker): Convert bool param
6375 "skip_unreachable_blocks" to enum reachability.
6376 (set_all_edges_as_executable): New decl.
6377 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
6378 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
6380 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
6381 but converting true to REACHABLE_BLOCKS.
6382 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
6384 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
6385 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
6386 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
6388 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
6389 if check_all_array_refs will be called.
6391 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6393 * tree.c (selftest::test_location_wrappers): Add more test
6396 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6398 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
6399 (selftest::test_bit_in_range): Likewise.
6401 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6404 * doc/sourcebuild.texi (vect_float): Say that the selector
6405 only describes the situation when -funsafe-math-optimizations is on.
6406 (vect_float_strict): Document.
6408 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6410 PR tree-optimization/83965
6411 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
6412 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
6413 instead of checking only for a reduction.
6414 (vect_recog_widen_sum_pattern): Likewise.
6416 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6418 * predict.c (probably_never_executed): Only use precise profile info.
6419 (compute_function_frequency): Skip after inlining hack since we now
6420 have quality checking.
6422 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6424 * profile-count.h (profile_probability::very_unlikely,
6425 profile_probability::unlikely, profile_probability::even): Set
6426 precision to guessed.
6428 2018-01-23 Richard Biener <rguenther@suse.de>
6430 PR tree-optimization/83963
6431 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
6432 Properly terminate dominator walk when crossing the exit edge not
6433 when visiting its source block.
6435 2018-01-23 Jakub Jelinek <jakub@redhat.com>
6438 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
6439 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
6441 2018-01-22 Jakub Jelinek <jakub@redhat.com>
6443 PR tree-optimization/83957
6444 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
6445 semicolon after for body surrounded by braces.
6447 PR tree-optimization/83081
6448 * profile-count.h (profile_probability::split): New method.
6449 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
6450 Use profile_probability::split.
6451 (do_compare_rtx_and_jump): Fix adjustment of probabilities
6452 when splitting a single conditional jump into 2.
6454 2018-01-22 David Malcolm <dmalcolm@redhat.com>
6456 PR tree-optimization/69452
6457 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
6460 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6462 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
6463 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
6464 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
6466 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6468 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
6469 * config/rl78/rl78.md (movdi): New define_expand.
6470 * config/rl78/rl78.c (rl78_split_movdi): New function.
6472 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
6475 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
6477 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
6478 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
6479 128-bit to produce an UNSPEC move to get the double word with the
6480 signbit and then a shift directly to do signbit.
6481 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
6482 implementation with a new version that just does either a direct
6483 move or a regular move. Move memory interface to separate insns.
6484 Move insns so they are next to the expander.
6485 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
6486 with signbit move. Split big and little endian case.
6487 (signbit<mode>2_dm_mem_le): Likewise.
6488 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
6489 (signbit<mode>2_dm2): Likewise.
6491 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6493 * config/rl78/rl78.md (anddi3): New define_expand.
6495 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6497 * config/rl78/rl78.md (umindi3): New define_expand.
6499 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6501 * config/rl78/rl78.md (smindi3): New define_expand.
6503 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6505 * config/rl78/rl78.md (smaxdi3): New define_expand.
6507 2018-01-22 Carl Love <cel@us.ibm.com>
6509 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
6510 LVX_V1TI): Add macro expansion.
6511 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
6512 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
6513 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
6514 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
6515 Change check to determine if the instruction is a byte reversing
6516 entry. Fix typo in comment.
6517 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
6518 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
6519 Add def_builtin calls for new builtins.
6520 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
6521 Add define_insn expansion.
6523 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6525 * config/rl78/rl78.md (umaxdi3): New define_expand.
6527 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6529 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
6530 for non-QImode registers.
6532 2018-01-22 Richard Biener <rguenther@suse.de>
6534 PR tree-optimization/83963
6535 * graphite-scop-detection.c (scop_detection::get_sese): Delay
6536 including the loop exit block.
6537 (scop_detection::merge_sese): Likewise.
6538 (scop_detection::add_scop): Do it here instead.
6540 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6542 * doc/sourcebuild.texi (arm_softfloat): Document.
6544 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
6547 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
6548 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
6549 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
6551 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6552 David Edelsohn <dje.gcc@gmail.com>
6555 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6556 Change "crset eq" to "crset 2".
6557 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6558 (*call_indirect_aix<mode>_nospec): Likewise.
6559 (*call_value_indirect_aix<mode>_nospec): Likewise.
6560 (*call_indirect_elfv2<mode>_nospec): Likewise.
6561 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
6562 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
6563 change assembly output from . to $.
6564 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6565 (indirect_jump<mode>_nospec): Change assembly output from . to $.
6566 (*tablejump<mode>_internal1_nospec): Likewise.
6568 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
6571 * config/sh/sh_optimize_sett_clrt.cc:
6572 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
6574 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
6576 PR tree-optimization/83940
6577 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
6578 offset_dt to vect_constant_def rather than vect_unknown_def_type.
6579 (vect_check_load_store_mask): Add a mask_dt_out parameter and
6580 use it to pass back the definition type.
6581 (vect_check_store_rhs): Likewise rhs_dt_out.
6582 (vect_build_gather_load_calls): Add a mask_dt argument and use
6583 it instead of a call to vect_is_simple_use.
6584 (vectorizable_store): Update calls to vect_check_load_store_mask
6585 and vect_check_store_rhs. Use the dt returned by the latter instead
6586 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
6587 instead of calls to vect_is_simple_use. Pass the scalar rather
6588 than the vector operand to vect_is_simple_use when handling
6589 second and subsequent copies of an rhs value.
6590 (vectorizable_load): Update calls to vect_check_load_store_mask
6591 and vect_build_gather_load_calls. Use the cached mask_dt and
6592 gs_info.offset_dt instead of calls to vect_is_simple_use.
6594 2018-01-20 Jakub Jelinek <jakub@redhat.com>
6597 * tree-emutls.c: Include gimplify.h.
6598 (lower_emutls_2): New function.
6599 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
6600 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
6601 it before further processing.
6604 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
6605 UINTVAL (trueop1) instead of INTVAL (op1).
6607 2018-01-19 Jakub Jelinek <jakub@redhat.com>
6611 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
6612 INCOMING_FRAME_SP_OFFSET if not defined.
6613 (scan_trace): Add ENTRY argument. If true and
6614 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
6615 emit a note to adjust the CFA offset.
6616 (create_cfi_notes): Adjust scan_trace callers.
6617 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
6618 INCOMING_FRAME_SP_OFFSET in the CIE.
6619 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
6620 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
6622 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
6623 * doc/tm.texi: Regenerated.
6625 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6627 PR rtl-optimization/83147
6628 * lra-constraints.c (remove_inheritance_pseudos): Use
6629 lra_substitute_pseudo_within_insn.
6631 2018-01-19 Tom de Vries <tom@codesourcery.com>
6632 Cesar Philippidis <cesar@codesourcery.com>
6635 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
6637 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
6640 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
6641 spaces for function labels.
6643 2018-01-19 Martin Liska <mliska@suse.cz>
6645 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
6646 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
6647 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
6648 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
6649 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
6650 (PRED_CONST_RETURN): Change from 69 to 65.
6651 (PRED_NULL_RETURN): Change from 91 to 71.
6652 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
6653 (PRED_LOOP_GUARD): Change from 66 to 73.
6655 2018-01-19 Martin Liska <mliska@suse.cz>
6657 * predict.c (predict_insn_def): Add new assert.
6658 (struct branch_predictor): Change type to signed integer.
6659 (test_prediction_value_range): Amend test to cover
6661 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
6662 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
6663 (PRED_LOOP_ITERATIONS_MAX): Likewise.
6664 (PRED_LOOP_IV_COMPARE): Likewise.
6665 * predict.h (PROB_UNINITIALIZED): Define new constant.
6667 2018-01-19 Martin Liska <mliska@suse.cz>
6669 * predict.c (dump_prediction): Add new format for
6670 analyze_brprob.py script which is enabled with -details
6672 * profile-count.h (precise_p): New function.
6674 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6676 PR tree-optimization/83922
6677 * tree-vect-loop.c (vect_verify_full_masking): Return false if
6678 there are no statements that need masking.
6679 (vect_active_double_reduction_p): New function.
6680 (vect_analyze_loop_operations): Use it when handling phis that
6681 are not in the loop header.
6683 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6685 PR tree-optimization/83914
6686 * tree-vect-loop.c (vectorizable_induction): Don't convert
6687 init_expr or apply the peeling adjustment for inductions
6688 that are nested within the vectorized loop.
6690 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6692 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
6695 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6699 * function.h (gimplify_parameters): Add gimple_seq * argument.
6700 * function.c: Include gimple.h and options.h.
6701 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
6702 for the added local temporaries if needed.
6703 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
6704 if there are any parameter cleanups, wrap whole body into a
6705 try/finally with the cleanups.
6707 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
6710 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
6711 Use GET_MODE_CLASS for scalar floating point.
6713 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6717 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
6718 Fix call of call_cgraph_insertion_hooks.
6720 2018-01-18 Martin Sebor <msebor@redhat.com>
6722 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
6724 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6727 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
6730 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
6733 * common.opt: (-ffile-prefix-map): New option.
6734 * opts.c (common_handle_option): Defer it.
6735 * opts-global.c (handle_common_deferred_options): Handle it.
6736 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
6737 * file-prefix-map.h: New file.
6738 (remap_debug_filename, add_debug_prefix_map): ...here.
6739 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
6740 * final.c (debug_prefix_map, add_debug_prefix_map
6741 remap_debug_filename): Move to...
6742 * file-prefix-map.c: New file.
6743 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
6744 generalize, get rid of alloca(), use strrchr() instead of strchr().
6745 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
6746 Implement in terms of add_prefix_map().
6747 (remap_macro_filename, remap_debug_filename): Implement in term of
6749 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
6750 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
6751 * dbxout.c: Include file-prefix-map.h.
6752 * varasm.c: Likewise.
6753 * vmsdbgout.c: Likewise.
6754 * xcoffout.c: Likewise.
6755 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
6756 * doc/cppopts.texi (-fmacro-prefix-map): Document.
6757 * doc/invoke.texi (-ffile-prefix-map): Document.
6758 (-fdebug-prefix-map): Update description.
6760 2018-01-18 Martin Liska <mliska@suse.cz>
6762 * config/i386/i386.c (indirect_thunk_name): Document that also
6764 (output_indirect_thunk): Document why both instructions
6765 (pause and lfence) are generated.
6767 2018-01-18 Richard Biener <rguenther@suse.de>
6769 PR tree-optimization/83887
6770 * graphite-scop-detection.c
6771 (scop_detection::get_nearest_dom_with_single_entry): Remove.
6772 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
6773 (scop_detection::merge_sese): Re-implement with a flood-fill
6774 algorithm that properly finds a SESE region if it exists.
6776 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6779 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
6780 pointer_diff optimizations use view_convert instead of convert.
6782 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6784 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6785 Generate different code for -mno-speculate-indirect-jumps.
6786 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6787 (*call_indirect_aix<mode>): Disable for
6788 -mno-speculate-indirect-jumps.
6789 (*call_indirect_aix<mode>_nospec): New define_insn.
6790 (*call_value_indirect_aix<mode>): Disable for
6791 -mno-speculate-indirect-jumps.
6792 (*call_value_indirect_aix<mode>_nospec): New define_insn.
6793 (*sibcall_nonlocal_sysv<mode>): Generate different code for
6794 -mno-speculate-indirect-jumps.
6795 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6797 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6799 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
6800 long double type, set the flags for noting the default long double
6801 type, even if we don't pass or return a long double type.
6803 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
6806 * ipa-inline.c (flatten_function): Do not overwrite final inlining
6809 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
6811 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
6812 support for merge[hl].
6813 (fold_mergehl_helper): New helper function.
6814 (tree-vector-builder.h): New #include for tree_vector_builder usage.
6815 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
6816 (altivec_vmrglw_direct): Add xxmrglw insn.
6818 2018-01-17 Andrew Waterman <andrew@sifive.com>
6820 * config/riscv/riscv.c (riscv_conditional_register_usage): If
6821 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
6823 2018-01-17 David Malcolm <dmalcolm@redhat.com>
6826 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
6827 call the lto_location_cache before reading the
6828 DECL_SOURCE_LOCATION of the types.
6830 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
6831 Richard Sandiford <richard.sandiford@linaro.org>
6833 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
6834 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
6835 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
6836 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
6837 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
6839 * config/aarch64/constraints.md (aarch64_movti_operand):
6841 * config/aarch64/predicates.md (Uti): Add new constraint.
6843 2018-01-17 Carl Love <cel@us.ibm.com>
6845 * config/rs6000/vsx.md (define_expand xl_len_r,
6846 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
6847 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
6849 (define_expand, define_insn): Move the shift left from the
6850 define_insn to the define_expand for lxvl and stxvl instructions.
6851 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
6852 and XL_LEN_R definitions to PURE.
6854 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
6856 * config/i386/i386.c (indirect_thunk_name): Declare regno
6857 as unsigned int. Compare regno with INVALID_REGNUM.
6858 (output_indirect_thunk): Ditto.
6859 (output_indirect_thunk_function): Ditto.
6860 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
6861 in the call to output_indirect_thunk_function.
6863 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
6866 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
6867 rather than the size of inner_type to determine the stack slot size
6868 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
6870 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
6873 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
6876 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
6878 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
6879 endian Linux systems to optionally enable multilibs for selecting
6880 the long double type if the user configured an explicit type.
6881 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
6882 have no long double multilibs if not defined.
6883 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
6884 warn if the user used -mabi={ieee,ibm}longdouble and we built
6885 multilibs for long double.
6886 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
6887 appropriate multilib option.
6888 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
6890 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
6891 for building long double multilibs.
6892 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
6894 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
6896 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
6899 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
6901 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
6904 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
6907 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
6910 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6912 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
6913 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
6915 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
6917 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
6918 different rtl trees depending on TARGET_64BIT.
6919 (rs6000_gen_lvx): Likewise.
6921 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6923 * config/visium/visium.md (nop): Tweak comment.
6924 (hazard_nop): Likewise.
6926 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6928 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
6929 -mspeculate-indirect-jumps.
6930 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
6931 for -mno-speculate-indirect-jumps.
6932 (*call_indirect_elfv2<mode>_nospec): New define_insn.
6933 (*call_value_indirect_elfv2<mode>): Disable for
6934 -mno-speculate-indirect-jumps.
6935 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
6936 (indirect_jump): Emit different RTL for
6937 -mno-speculate-indirect-jumps.
6938 (*indirect_jump<mode>): Disable for
6939 -mno-speculate-indirect-jumps.
6940 (*indirect_jump<mode>_nospec): New define_insn.
6941 (tablejump): Emit different RTL for
6942 -mno-speculate-indirect-jumps.
6943 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
6944 (tablejumpsi_nospec): New define_expand.
6945 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
6946 (tablejumpdi_nospec): New define_expand.
6947 (*tablejump<mode>_internal1): Disable for
6948 -mno-speculate-indirect-jumps.
6949 (*tablejump<mode>_internal1_nospec): New define_insn.
6950 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
6953 2018-01-16 Artyom Skrobov tyomitch@gmail.com
6955 * caller-save.c (insert_save): Drop unnecessary parameter. All
6958 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6959 Richard Biener <rguenth@suse.de>
6962 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
6963 return early, inline manually is_gimple_sizepos. Make sure if we
6964 call gimplify_expr we don't end up with a gimple constant.
6965 * tree.c (variably_modified_type_p): Don't return true for
6966 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
6967 * gimplify.h (is_gimple_sizepos): Remove.
6969 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6971 PR tree-optimization/83857
6972 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
6973 vectorizable_live_operation for pure SLP statements.
6974 (vectorizable_live_operation): Handle PHIs.
6976 2018-01-16 Richard Biener <rguenther@suse.de>
6978 PR tree-optimization/83867
6979 * tree-vect-stmts.c (vect_transform_stmt): Precompute
6980 nested_in_vect_loop_p since the scalar stmt may get invalidated.
6982 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6985 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
6986 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
6987 If off is not INTEGER_CST, issue a may not be aligned warning
6988 rather than isn't aligned. Use isn%'t rather than isn't.
6989 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
6991 <case MULT_EXPR>: Improve the case when bottom and one of the
6992 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
6993 operand, in that case check if the other operand is multiple of
6994 bottom divided by the INTEGER_CST operand.
6996 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6999 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
7000 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
7001 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
7002 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
7003 * config/pa/pa.c (pa_function_arg_advance): Likewise.
7004 (pa_function_arg, pa_arg_partial_bytes): Likewise.
7005 (pa_function_arg_size): New function.
7007 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7009 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
7010 in a separate statement.
7012 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7014 PR tree-optimization/83847
7015 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
7016 group gathers and scatters.
7018 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7020 PR rtl-optimization/86620
7021 * params.def (max-sched-ready-insns): Bump minimum value to 1.
7023 PR rtl-optimization/83213
7024 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
7025 to last if both are JUMP_INSNs.
7027 PR tree-optimization/83843
7028 * gimple-ssa-store-merging.c
7029 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
7030 store_immediate_info for bswap/nop orig_stores.
7032 2018-01-15 Andrew Waterman <andrew@sifive.com>
7034 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
7036 <UDIV>: Increase cost if !TARGET_DIV.
7038 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
7040 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
7041 (define_attr "cr_logical_3op"): New.
7042 (cceq_ior_compare): Adjust.
7043 (cceq_ior_compare_complement): Adjust.
7044 (*cceq_rev_compare): Adjust.
7045 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
7046 (is_cracked_insn): Adjust.
7047 (insn_must_be_first_in_group): Adjust.
7048 * config/rs6000/40x.md: Adjust.
7049 * config/rs6000/440.md: Adjust.
7050 * config/rs6000/476.md: Adjust.
7051 * config/rs6000/601.md: Adjust.
7052 * config/rs6000/603.md: Adjust.
7053 * config/rs6000/6xx.md: Adjust.
7054 * config/rs6000/7450.md: Adjust.
7055 * config/rs6000/7xx.md: Adjust.
7056 * config/rs6000/8540.md: Adjust.
7057 * config/rs6000/cell.md: Adjust.
7058 * config/rs6000/e300c2c3.md: Adjust.
7059 * config/rs6000/e500mc.md: Adjust.
7060 * config/rs6000/e500mc64.md: Adjust.
7061 * config/rs6000/e5500.md: Adjust.
7062 * config/rs6000/e6500.md: Adjust.
7063 * config/rs6000/mpc.md: Adjust.
7064 * config/rs6000/power4.md: Adjust.
7065 * config/rs6000/power5.md: Adjust.
7066 * config/rs6000/power6.md: Adjust.
7067 * config/rs6000/power7.md: Adjust.
7068 * config/rs6000/power8.md: Adjust.
7069 * config/rs6000/power9.md: Adjust.
7070 * config/rs6000/rs64.md: Adjust.
7071 * config/rs6000/titan.md: Adjust.
7073 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7075 * config/i386/predicates.md (indirect_branch_operand): Rewrite
7076 ix86_indirect_branch_register logic.
7078 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7080 * config/i386/constraints.md (Bs): Update
7081 ix86_indirect_branch_register check. Don't check
7082 ix86_indirect_branch_register with GOT_memory_operand.
7084 * config/i386/predicates.md (GOT_memory_operand): Don't check
7085 ix86_indirect_branch_register here.
7086 (GOT32_symbol_operand): Likewise.
7088 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7090 * config/i386/predicates.md (constant_call_address_operand):
7091 Rewrite ix86_indirect_branch_register logic.
7092 (sibcall_insn_operand): Likewise.
7094 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7096 * config/i386/constraints.md (Bs): Replace
7097 ix86_indirect_branch_thunk_register with
7098 ix86_indirect_branch_register.
7100 * config/i386/i386.md (indirect_jump): Likewise.
7101 (tablejump): Likewise.
7102 (*sibcall_memory): Likewise.
7103 (*sibcall_value_memory): Likewise.
7104 Peepholes of indirect call and jump via memory: Likewise.
7105 * config/i386/i386.opt: Likewise.
7106 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7107 (GOT_memory_operand): Likewise.
7108 (call_insn_operand): Likewise.
7109 (sibcall_insn_operand): Likewise.
7110 (GOT32_symbol_operand): Likewise.
7112 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7115 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
7116 type rather than type addr's type points to.
7117 (expand_omp_atomic_mutex): Likewise.
7118 (expand_omp_atomic): Likewise.
7120 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7123 * config/i386/i386.c (output_indirect_thunk_function): Use
7124 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
7125 for __x86_return_thunk.
7127 2018-01-15 Richard Biener <rguenther@suse.de>
7130 * expmed.c (extract_bit_field_1): Fix typo.
7132 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7135 * config/arm/iterators.md (VF): New mode iterator.
7136 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
7137 Remove integer-related logic from pattern.
7138 (neon_vabd<mode>_3): Likewise.
7140 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7143 * common.opt (fstrict-overflow): No longer an alias.
7144 (fwrapv-pointer): New option.
7145 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
7146 also for pointer types based on flag_wrapv_pointer.
7147 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
7148 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
7149 opts->x_flag_wrapv got set.
7150 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
7151 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
7152 POINTER_TYPE_OVERFLOW_UNDEFINED.
7153 * match.pd: Likewise in address comparison pattern.
7154 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
7156 2018-01-15 Richard Biener <rguenther@suse.de>
7159 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
7160 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
7161 Reset type names to their identifier if their TYPE_DECL doesn't
7162 have linkage (and thus is used for ODR and devirt).
7163 (save_debug_info_for_decl): Remove.
7164 (save_debug_info_for_type): Likewise.
7165 (add_tree_to_fld_list): Adjust.
7166 * tree-pretty-print.c (dump_generic_node): Make dumping of
7167 type names more robust.
7169 2018-01-15 Richard Biener <rguenther@suse.de>
7171 * BASE-VER: Bump to 8.0.1.
7173 2018-01-14 Martin Sebor <msebor@redhat.com>
7176 * builtins.c (check_access): Avoid warning when the no-warning bit
7179 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
7181 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
7182 * ira-color (allocno_hard_regs_compare): Likewise.
7184 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
7187 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
7188 Use .pushsection/.popsection.
7190 2018-01-14 Martin Sebor <msebor@redhat.com>
7193 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
7195 2018-01-14 Jakub Jelinek <jakub@redhat.com>
7197 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
7198 entry from extra_headers.
7199 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
7200 extra_headers, make the list bitwise identical to the i?86-*-* one.
7202 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7204 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
7205 -mcmodel=large with -mindirect-branch=thunk,
7206 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
7207 -mfunction-return=thunk-extern.
7208 * doc/invoke.texi: Document -mcmodel=large is incompatible with
7209 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
7210 -mfunction-return=thunk and -mfunction-return=thunk-extern.
7212 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7214 * config/i386/i386.c (print_reg): Print the name of the full
7215 integer register without '%'.
7216 (ix86_print_operand): Handle 'V'.
7217 * doc/extend.texi: Document 'V' modifier.
7219 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7221 * config/i386/constraints.md (Bs): Disallow memory operand for
7222 -mindirect-branch-register.
7224 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7225 (GOT_memory_operand): Likewise.
7226 (call_insn_operand): Likewise.
7227 (sibcall_insn_operand): Likewise.
7228 (GOT32_symbol_operand): Likewise.
7229 * config/i386/i386.md (indirect_jump): Call convert_memory_address
7230 for -mindirect-branch-register.
7231 (tablejump): Likewise.
7232 (*sibcall_memory): Likewise.
7233 (*sibcall_value_memory): Likewise.
7234 Disallow peepholes of indirect call and jump via memory for
7235 -mindirect-branch-register.
7236 (*call_pop): Replace m with Bw.
7237 (*call_value_pop): Likewise.
7238 (*sibcall_pop_memory): Replace m with Bs.
7239 * config/i386/i386.opt (mindirect-branch-register): New option.
7240 * doc/invoke.texi: Document -mindirect-branch-register option.
7242 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7244 * config/i386/i386-protos.h (ix86_output_function_return): New.
7245 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
7246 set function_return_type.
7247 (indirect_thunk_name): Add ret_p to indicate thunk for function
7249 (output_indirect_thunk_function): Pass false to
7250 indirect_thunk_name.
7251 (ix86_output_indirect_branch_via_reg): Likewise.
7252 (ix86_output_indirect_branch_via_push): Likewise.
7253 (output_indirect_thunk_function): Create alias for function
7254 return thunk if regno < 0.
7255 (ix86_output_function_return): New function.
7256 (ix86_handle_fndecl_attribute): Handle function_return.
7257 (ix86_attribute_table): Add function_return.
7258 * config/i386/i386.h (machine_function): Add
7259 function_return_type.
7260 * config/i386/i386.md (simple_return_internal): Use
7261 ix86_output_function_return.
7262 (simple_return_internal_long): Likewise.
7263 * config/i386/i386.opt (mfunction-return=): New option.
7264 (indirect_branch): Mention -mfunction-return=.
7265 * doc/extend.texi: Document function_return function attribute.
7266 * doc/invoke.texi: Document -mfunction-return= option.
7268 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7270 * config/i386/i386-opts.h (indirect_branch): New.
7271 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
7272 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
7273 with local indirect jump when converting indirect call and jump.
7274 (ix86_set_indirect_branch_type): New.
7275 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
7276 (indirectlabelno): New.
7277 (indirect_thunk_needed): Likewise.
7278 (indirect_thunk_bnd_needed): Likewise.
7279 (indirect_thunks_used): Likewise.
7280 (indirect_thunks_bnd_used): Likewise.
7281 (INDIRECT_LABEL): Likewise.
7282 (indirect_thunk_name): Likewise.
7283 (output_indirect_thunk): Likewise.
7284 (output_indirect_thunk_function): Likewise.
7285 (ix86_output_indirect_branch_via_reg): Likewise.
7286 (ix86_output_indirect_branch_via_push): Likewise.
7287 (ix86_output_indirect_branch): Likewise.
7288 (ix86_output_indirect_jmp): Likewise.
7289 (ix86_code_end): Call output_indirect_thunk_function if needed.
7290 (ix86_output_call_insn): Call ix86_output_indirect_branch if
7292 (ix86_handle_fndecl_attribute): Handle indirect_branch.
7293 (ix86_attribute_table): Add indirect_branch.
7294 * config/i386/i386.h (machine_function): Add indirect_branch_type
7295 and has_local_indirect_jump.
7296 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
7298 (tablejump): Likewise.
7299 (*indirect_jump): Use ix86_output_indirect_jmp.
7300 (*tablejump_1): Likewise.
7301 (simple_return_indirect_internal): Likewise.
7302 * config/i386/i386.opt (mindirect-branch=): New option.
7303 (indirect_branch): New.
7306 (thunk-inline): Likewise.
7307 (thunk-extern): Likewise.
7308 * doc/extend.texi: Document indirect_branch function attribute.
7309 * doc/invoke.texi: Document -mindirect-branch= option.
7311 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
7314 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
7316 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
7318 * ipa-inline.c (want_inline_small_function_p): Return false if
7319 inlining has already failed with CIF_FINAL_ERROR.
7320 (update_caller_keys): Call want_inline_small_function_p before
7322 (update_callee_keys): Likewise.
7324 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7326 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
7328 (rs6000_quadword_masked_address_p): Likewise.
7329 (quad_aligned_load_p): Likewise.
7330 (quad_aligned_store_p): Likewise.
7331 (const_load_sequence_p): Add comment to describe the outer-most loop.
7332 (mimic_memory_attributes_and_flags): New function.
7333 (rs6000_gen_stvx): Likewise.
7334 (replace_swapped_aligned_store): Likewise.
7335 (rs6000_gen_lvx): Likewise.
7336 (replace_swapped_aligned_load): Likewise.
7337 (replace_swapped_load_constant): Capitalize argument name in
7338 comment describing this function.
7339 (rs6000_analyze_swaps): Add a third pass to search for vector loads
7340 and stores that access quad-word aligned addresses and replace
7341 with stvx or lvx instructions when appropriate.
7342 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
7343 New function prototype.
7344 (rs6000_quadword_masked_address_p): Likewise.
7345 (rs6000_gen_lvx): Likewise.
7346 (rs6000_gen_stvx): Likewise.
7347 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
7348 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
7349 when memory address is aligned.
7350 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
7351 this split to select lvx instruction when memory address is aligned.
7352 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
7353 instruction when memory address is aligned.
7354 (*vsx_le_perm_load_v16qi): Likewise.
7355 (four unnamed splitters): Modify to select the stvx instruction
7356 when memory is aligned.
7358 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
7360 * predict.c (determine_unlikely_bbs): Handle correctly BBs
7361 which appears in the queue multiple times.
7363 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7364 Alan Hayward <alan.hayward@arm.com>
7365 David Sherwood <david.sherwood@arm.com>
7367 * tree-vectorizer.h (vec_lower_bound): New structure.
7368 (_loop_vec_info): Add check_nonzero and lower_bounds.
7369 (LOOP_VINFO_CHECK_NONZERO): New macro.
7370 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
7371 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
7372 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
7373 fields. Make seg_len the distance travelled, not including the
7375 (dr_direction_indicator): Declare.
7376 (dr_zero_step_indicator): Likewise.
7377 (dr_known_forward_stride_p): Likewise.
7378 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
7380 (runtime_alias_check_p): Allow runtime alias checks with
7382 (operator ==): Compare access_size and align.
7383 (prune_runtime_alias_test_list): Rework for new distinction between
7384 the access_size and seg_len.
7385 (create_intersect_range_checks_index): Likewise. Cope with polynomial
7387 (get_segment_min_max): New function.
7388 (create_intersect_range_checks): Use it.
7389 (dr_step_indicator): New function.
7390 (dr_direction_indicator): Likewise.
7391 (dr_zero_step_indicator): Likewise.
7392 (dr_known_forward_stride_p): Likewise.
7393 * tree-loop-distribution.c (data_ref_segment_size): Return
7394 DR_STEP * (niters - 1).
7395 (compute_alias_check_pairs): Update call to the dr_with_seg_len
7397 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
7398 (vect_preserves_scalar_order_p): New function, split out from...
7399 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
7400 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
7401 (vect_vfa_access_size): New function.
7402 (vect_vfa_align): Likewise.
7403 (vect_compile_time_alias): Take access_size_a and access_b arguments.
7404 (dump_lower_bound): New function.
7405 (vect_check_lower_bound): Likewise.
7406 (vect_small_gap_p): Likewise.
7407 (vectorizable_with_step_bound_p): Likewise.
7408 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
7409 depencies if the vectorization factor is 1. Convert the checks
7410 for nonzero steps into checks on the bounds of DR_STEP. Try using
7411 a bunds check for variable steps if the minimum required step is
7412 relatively small. Update calls to the dr_with_seg_len
7413 constructor and to vect_compile_time_alias.
7414 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
7416 (vect_loop_versioning): Call it.
7417 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
7419 (vect_estimate_min_profitable_iters): Account for any bounds checks.
7421 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7422 Alan Hayward <alan.hayward@arm.com>
7423 David Sherwood <david.sherwood@arm.com>
7425 * doc/sourcebuild.texi (vect_scatter_store): Document.
7426 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
7428 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
7430 * genopinit.c (main): Add supports_vec_scatter_store and
7431 supports_vec_scatter_store_cached to target_optabs.
7432 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
7433 IFN_MASK_SCATTER_STORE.
7434 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
7436 * internal-fn.h (internal_store_fn_p): Declare.
7437 (internal_fn_stored_value_index): Likewise.
7438 * internal-fn.c (scatter_store_direct): New macro.
7439 (expand_scatter_store_optab_fn): New function.
7440 (direct_scatter_store_optab_supported_p): New macro.
7441 (internal_store_fn_p): New function.
7442 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
7443 IFN_MASK_SCATTER_STORE.
7444 (internal_fn_mask_index): Likewise.
7445 (internal_fn_stored_value_index): New function.
7446 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
7448 * optabs-query.h (supports_vec_scatter_store_p): Declare.
7449 * optabs-query.c (supports_vec_scatter_store_p): New function.
7450 * tree-vectorizer.h (vect_get_store_rhs): Declare.
7451 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
7452 true for scatter stores.
7453 (vect_gather_scatter_fn_p): Handle scatter stores too.
7454 (vect_check_gather_scatter): Consider using scatter stores if
7455 supports_vec_scatter_store_p.
7456 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
7458 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7459 internal_fn_stored_value_index.
7460 (check_load_store_masking): Handle scatter stores too.
7461 (vect_get_store_rhs): Make public.
7462 (vectorizable_call): Use internal_store_fn_p.
7463 (vectorizable_store): Handle scatter store internal functions.
7464 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
7465 when deciding whether the end of the group has been reached.
7466 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
7467 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
7468 (mask_scatter_store<mode>): New insns.
7470 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7471 Alan Hayward <alan.hayward@arm.com>
7472 David Sherwood <david.sherwood@arm.com>
7474 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
7475 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
7476 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
7478 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
7479 Use vect_truncate_gather_scatter_offset if we can't treat the
7480 operation as a normal gather load or scatter store.
7481 (get_group_load_store_type): Take the gather_scatter_info
7482 as argument. Try using a gather load or scatter store for
7483 single-element groups.
7484 (get_load_store_type): Update calls to get_group_load_store_type
7485 and vect_use_strided_gather_scatters_p.
7487 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7488 Alan Hayward <alan.hayward@arm.com>
7489 David Sherwood <david.sherwood@arm.com>
7491 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
7492 optional tree argument.
7493 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
7495 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
7496 but continue to use the current value as a fallback.
7497 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
7498 to compare the updates.
7499 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
7500 (get_load_store_type): Use it when handling a strided access.
7501 (vect_get_strided_load_store_ops): New function.
7502 (vect_get_data_ptr_increment): Likewise.
7503 (vectorizable_load): Handle strided gather loads. Always pass
7504 a step to vect_create_data_ref_ptr and bump_vector_ptr.
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 * doc/md.texi (gather_load@var{m}): Document.
7511 (mask_gather_load@var{m}): Likewise.
7512 * genopinit.c (main): Add supports_vec_gather_load and
7513 supports_vec_gather_load_cached to target_optabs.
7514 * optabs-tree.c (init_tree_optimization_optabs): Use
7515 ggc_cleared_alloc to allocate target_optabs.
7516 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
7517 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
7519 * internal-fn.h (internal_load_fn_p): Declare.
7520 (internal_gather_scatter_fn_p): Likewise.
7521 (internal_fn_mask_index): Likewise.
7522 (internal_gather_scatter_fn_supported_p): Likewise.
7523 * internal-fn.c (gather_load_direct): New macro.
7524 (expand_gather_load_optab_fn): New function.
7525 (direct_gather_load_optab_supported_p): New macro.
7526 (direct_internal_fn_optab): New function.
7527 (internal_load_fn_p): Likewise.
7528 (internal_gather_scatter_fn_p): Likewise.
7529 (internal_fn_mask_index): Likewise.
7530 (internal_gather_scatter_fn_supported_p): Likewise.
7531 * optabs-query.c (supports_at_least_one_mode_p): New function.
7532 (supports_vec_gather_load_p): Likewise.
7533 * optabs-query.h (supports_vec_gather_load_p): Declare.
7534 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
7535 and memory_type field.
7536 (NUM_PATTERNS): Bump to 15.
7537 * tree-vect-data-refs.c: Include internal-fn.h.
7538 (vect_gather_scatter_fn_p): New function.
7539 (vect_describe_gather_scatter_call): Likewise.
7540 (vect_check_gather_scatter): Try using internal functions for
7541 gather loads. Recognize existing calls to a gather load function.
7542 (vect_analyze_data_refs): Consider using gather loads if
7543 supports_vec_gather_load_p.
7544 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
7545 (vect_get_gather_scatter_offset_type): Likewise.
7546 (vect_convert_mask_for_vectype): Likewise.
7547 (vect_add_conversion_to_patterm): Likewise.
7548 (vect_try_gather_scatter_pattern): Likewise.
7549 (vect_recog_gather_scatter_pattern): New pattern recognizer.
7550 (vect_vect_recog_func_ptrs): Add it.
7551 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7552 internal_fn_mask_index and internal_gather_scatter_fn_p.
7553 (check_load_store_masking): Take the gather_scatter_info as an
7554 argument and handle gather loads.
7555 (vect_get_gather_scatter_ops): New function.
7556 (vectorizable_call): Check internal_load_fn_p.
7557 (vectorizable_load): Likewise. Handle gather load internal
7559 (vectorizable_store): Update call to check_load_store_masking.
7560 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
7561 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
7562 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
7563 (aarch64_gather_scale_operand_d): New predicates.
7564 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
7565 (mask_gather_load<mode>): New insns.
7567 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7568 Alan Hayward <alan.hayward@arm.com>
7569 David Sherwood <david.sherwood@arm.com>
7571 * optabs.def (fold_left_plus_optab): New optab.
7572 * doc/md.texi (fold_left_plus_@var{m}): Document.
7573 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
7574 * internal-fn.c (fold_left_direct): Define.
7575 (expand_fold_left_optab_fn): Likewise.
7576 (direct_fold_left_optab_supported_p): Likewise.
7577 * fold-const-call.c (fold_const_fold_left): New function.
7578 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
7579 * tree-parloops.c (valid_reduction_p): New function.
7580 (gather_scalar_reductions): Use it.
7581 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
7582 (vect_finish_replace_stmt): Declare.
7583 * tree-vect-loop.c (fold_left_reduction_fn): New function.
7584 (needs_fold_left_reduction_p): New function, split out from...
7585 (vect_is_simple_reduction): ...here. Accept reductions that
7586 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
7587 (vect_force_simple_reduction): Also store the reduction type in
7588 the assignment's STMT_VINFO_REDUC_TYPE.
7589 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
7590 (merge_with_identity): New function.
7591 (vect_expand_fold_left): Likewise.
7592 (vectorize_fold_left_reduction): Likewise.
7593 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
7594 scalar phi in place for it. Check for target support and reject
7595 cases that would reassociate the operation. Defer the transform
7596 phase to vectorize_fold_left_reduction.
7597 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
7598 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
7599 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
7601 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7603 * tree-if-conv.c (predicate_mem_writes): Remove redundant
7604 call to ifc_temp_var.
7606 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7607 Alan Hayward <alan.hayward@arm.com>
7608 David Sherwood <david.sherwood@arm.com>
7610 * target.def (legitimize_address_displacement): Take the original
7611 offset as a poly_int.
7612 * targhooks.h (default_legitimize_address_displacement): Update
7614 * targhooks.c (default_legitimize_address_displacement): Likewise.
7615 * doc/tm.texi: Regenerate.
7616 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
7617 as an argument, moving assert of ad->disp == ad->disp_term to...
7618 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
7619 Try calling targetm.legitimize_address_displacement before expanding
7620 the address rather than afterwards, and adjust for the new interface.
7621 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
7622 Match the new hook interface. Handle SVE addresses.
7623 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
7626 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7628 * Makefile.in (OBJS): Add early-remat.o.
7629 * target.def (select_early_remat_modes): New hook.
7630 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
7631 * doc/tm.texi: Regenerate.
7632 * targhooks.h (default_select_early_remat_modes): Declare.
7633 * targhooks.c (default_select_early_remat_modes): New function.
7634 * timevar.def (TV_EARLY_REMAT): New timevar.
7635 * passes.def (pass_early_remat): New pass.
7636 * tree-pass.h (make_pass_early_remat): Declare.
7637 * early-remat.c: New file.
7638 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
7640 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
7642 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7643 Alan Hayward <alan.hayward@arm.com>
7644 David Sherwood <david.sherwood@arm.com>
7646 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
7647 vfm1 with a bound_epilog parameter.
7648 (vect_do_peeling): Update calls accordingly, and move the prologue
7649 call earlier in the function. Treat the base bound_epilog as 0 for
7650 fully-masked loops and retain vf - 1 for other loops. Add 1 to
7651 this base when peeling for gaps.
7652 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
7653 with fully-masked loops.
7654 (vect_estimate_min_profitable_iters): Handle the single peeled
7655 iteration in that case.
7657 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7658 Alan Hayward <alan.hayward@arm.com>
7659 David Sherwood <david.sherwood@arm.com>
7661 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
7662 single-element interleaving even if the size is not a power of 2.
7663 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
7664 accesses for single-element interleaving if the group size is
7667 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7668 Alan Hayward <alan.hayward@arm.com>
7669 David Sherwood <david.sherwood@arm.com>
7671 * doc/md.texi (fold_extract_last_@var{m}): Document.
7672 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
7673 * optabs.def (fold_extract_last_optab): New optab.
7674 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
7675 * internal-fn.c (fold_extract_direct): New macro.
7676 (expand_fold_extract_optab_fn): Likewise.
7677 (direct_fold_extract_optab_supported_p): Likewise.
7678 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
7679 * tree-vect-loop.c (vect_model_reduction_cost): Handle
7680 EXTRACT_LAST_REDUCTION.
7681 (get_initial_def_for_reduction): Do not create an initial vector
7682 for EXTRACT_LAST_REDUCTION reductions.
7683 (vectorizable_reduction): Leave the scalar phi in place for
7684 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
7685 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
7686 epilogue code for EXTRACT_LAST_REDUCTION and defer the
7687 transform phase to vectorizable_condition.
7688 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
7690 (vect_finish_stmt_generation): ...here.
7691 (vect_finish_replace_stmt): New function.
7692 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
7693 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
7695 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
7697 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7698 Alan Hayward <alan.hayward@arm.com>
7699 David Sherwood <david.sherwood@arm.com>
7701 * doc/md.texi (extract_last_@var{m}): Document.
7702 * optabs.def (extract_last_optab): New optab.
7703 * internal-fn.def (EXTRACT_LAST): New internal function.
7704 * internal-fn.c (cond_unary_direct): New macro.
7705 (expand_cond_unary_optab_fn): Likewise.
7706 (direct_cond_unary_optab_supported_p): Likewise.
7707 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
7708 loops using EXTRACT_LAST.
7709 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
7710 (extract_last_<mode>): ...this optab.
7711 (vec_extract<mode><Vel>): Update accordingly.
7713 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7714 Alan Hayward <alan.hayward@arm.com>
7715 David Sherwood <david.sherwood@arm.com>
7717 * target.def (empty_mask_is_expensive): New hook.
7718 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
7719 * doc/tm.texi: Regenerate.
7720 * targhooks.h (default_empty_mask_is_expensive): Declare.
7721 * targhooks.c (default_empty_mask_is_expensive): New function.
7722 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
7723 if the target says that empty masks are expensive.
7724 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
7726 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
7728 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7729 Alan Hayward <alan.hayward@arm.com>
7730 David Sherwood <david.sherwood@arm.com>
7732 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
7733 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
7734 (vect_use_loop_mask_for_alignment_p): New function.
7735 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
7736 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
7737 niters_skip argument. Make sure that the first niters_skip elements
7738 of the first iteration are inactive.
7739 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
7740 Update call to vect_set_loop_masks_directly.
7741 (get_misalign_in_elems): New function, split out from...
7742 (vect_gen_prolog_loop_niters): ...here.
7743 (vect_update_init_of_dr): Take a code argument that specifies whether
7744 the adjustment should be added or subtracted.
7745 (vect_update_init_of_drs): Likewise.
7746 (vect_prepare_for_masked_peels): New function.
7747 (vect_do_peeling): Skip prologue peeling if we're using a mask
7748 instead. Update call to vect_update_inits_of_drs.
7749 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7751 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
7752 alignment. Do not include the number of peeled iterations in
7753 the minimum threshold in that case.
7754 (vectorizable_induction): Adjust the start value down by
7755 LOOP_VINFO_MASK_SKIP_NITERS iterations.
7756 (vect_transform_loop): Call vect_prepare_for_masked_peels.
7757 Take the number of skipped iterations into account when calculating
7759 * tree-vect-stmts.c (vect_gen_while_not): New function.
7761 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7762 Alan Hayward <alan.hayward@arm.com>
7763 David Sherwood <david.sherwood@arm.com>
7765 * doc/sourcebuild.texi (vect_fully_masked): Document.
7766 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
7768 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
7770 (vect_analyze_loop_2): ...here. Don't check the vectorization
7771 factor against the number of loop iterations if the loop is
7774 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7775 Alan Hayward <alan.hayward@arm.com>
7776 David Sherwood <david.sherwood@arm.com>
7778 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
7779 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
7780 (dump_groups): Update accordingly.
7781 (iv_use::mem_type): New member variable.
7782 (address_p): New function.
7783 (record_use): Add a mem_type argument and initialize the new
7785 (record_group_use): Add a mem_type argument. Use address_p.
7786 Remove obsolete null checks of base_object. Update call to record_use.
7787 (find_interesting_uses_op): Update call to record_group_use.
7788 (find_interesting_uses_cond): Likewise.
7789 (find_interesting_uses_address): Likewise.
7790 (get_mem_type_for_internal_fn): New function.
7791 (find_address_like_use): Likewise.
7792 (find_interesting_uses_stmt): Try find_address_like_use before
7793 calling find_interesting_uses_op.
7794 (addr_offset_valid_p): Use the iv mem_type field as the type
7795 of the addressed memory.
7796 (add_autoinc_candidates): Likewise.
7797 (get_address_cost): Likewise.
7798 (split_small_address_groups_p): Use address_p.
7799 (split_address_groups): Likewise.
7800 (add_iv_candidate_for_use): Likewise.
7801 (autoinc_possible_for_pair): Likewise.
7802 (rewrite_groups): Likewise.
7803 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
7804 (determine_group_iv_cost): Update after split of USE_ADDRESS.
7805 (get_alias_ptr_type_for_ptr_address): New function.
7806 (rewrite_use_address): Rewrite address uses in calls that were
7807 identified by find_address_like_use.
7809 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7810 Alan Hayward <alan.hayward@arm.com>
7811 David Sherwood <david.sherwood@arm.com>
7813 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
7815 * gimple-expr.h (is_gimple_addressable: Likewise.
7816 * gimple-expr.c (is_gimple_address): Likewise.
7817 * internal-fn.c (expand_call_mem_ref): New function.
7818 (expand_mask_load_optab_fn): Use it.
7819 (expand_mask_store_optab_fn): Likewise.
7821 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7822 Alan Hayward <alan.hayward@arm.com>
7823 David Sherwood <david.sherwood@arm.com>
7825 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
7826 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
7827 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
7828 (cond_umax@var{mode}): Document.
7829 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
7830 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
7831 (cond_umin_optab, cond_umax_optab): New optabs.
7832 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
7833 (COND_IOR, COND_XOR): New internal functions.
7834 * internal-fn.h (get_conditional_internal_fn): Declare.
7835 * internal-fn.c (cond_binary_direct): New macro.
7836 (expand_cond_binary_optab_fn): Likewise.
7837 (direct_cond_binary_optab_supported_p): Likewise.
7838 (get_conditional_internal_fn): New function.
7839 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
7840 Cope with reduction statements that are vectorized as calls rather
7842 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
7843 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
7844 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
7845 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
7846 (UNSPEC_COND_EOR): New unspecs.
7847 (optab): Add mappings for them.
7848 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
7849 (sve_int_op, sve_fp_op): New int attributes.
7851 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7852 Alan Hayward <alan.hayward@arm.com>
7853 David Sherwood <david.sherwood@arm.com>
7855 * optabs.def (while_ult_optab): New optab.
7856 * doc/md.texi (while_ult@var{m}@var{n}): Document.
7857 * internal-fn.def (WHILE_ULT): New internal function.
7858 * internal-fn.h (direct_internal_fn_supported_p): New override
7859 that takes two types as argument.
7860 * internal-fn.c (while_direct): New macro.
7861 (expand_while_optab_fn): New function.
7862 (convert_optab_supported_p): Likewise.
7863 (direct_while_optab_supported_p): New macro.
7864 * wide-int.h (wi::udiv_ceil): New function.
7865 * tree-vectorizer.h (rgroup_masks): New structure.
7866 (vec_loop_masks): New typedef.
7867 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
7869 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
7870 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
7871 (vect_max_vf): New function.
7872 (slpeel_make_loop_iterate_ntimes): Delete.
7873 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
7874 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
7875 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
7876 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
7877 internal-fn.h, stor-layout.h and optabs-query.h.
7878 (vect_set_loop_mask): New function.
7879 (add_preheader_seq): Likewise.
7880 (add_header_seq): Likewise.
7881 (interleave_supported_p): Likewise.
7882 (vect_maybe_permute_loop_masks): Likewise.
7883 (vect_set_loop_masks_directly): Likewise.
7884 (vect_set_loop_condition_masked): Likewise.
7885 (vect_set_loop_condition_unmasked): New function, split out from
7886 slpeel_make_loop_iterate_ntimes.
7887 (slpeel_make_loop_iterate_ntimes): Rename to..
7888 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
7889 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
7890 (vect_do_peeling): Update call accordingly.
7891 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
7893 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7894 mask_compare_type, can_fully_mask_p and fully_masked_p.
7895 (release_vec_loop_masks): New function.
7896 (_loop_vec_info): Use it to free the loop masks.
7897 (can_produce_all_loop_masks_p): New function.
7898 (vect_get_max_nscalars_per_iter): Likewise.
7899 (vect_verify_full_masking): Likewise.
7900 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
7901 retries, and free the mask rgroups before retrying. Check loop-wide
7902 reasons for disallowing fully-masked loops. Make the final decision
7903 about whether use a fully-masked loop or not.
7904 (vect_estimate_min_profitable_iters): Do not assume that peeling
7905 for the number of iterations will be needed for fully-masked loops.
7906 (vectorizable_reduction): Disable fully-masked loops.
7907 (vectorizable_live_operation): Likewise.
7908 (vect_halve_mask_nunits): New function.
7909 (vect_double_mask_nunits): Likewise.
7910 (vect_record_loop_mask): Likewise.
7911 (vect_get_loop_mask): Likewise.
7912 (vect_transform_loop): Handle the case in which the final loop
7913 iteration might handle a partial vector. Call vect_set_loop_condition
7914 instead of slpeel_make_loop_iterate_ntimes.
7915 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
7916 (check_load_store_masking): New function.
7917 (prepare_load_store_mask): Likewise.
7918 (vectorizable_store): Handle fully-masked loops.
7919 (vectorizable_load): Likewise.
7920 (supportable_widening_operation): Use vect_halve_mask_nunits for
7922 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
7923 (vect_gen_while): New function.
7924 * config/aarch64/aarch64.md (umax<mode>3): New expander.
7925 (aarch64_uqdec<mode>): New insn.
7927 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7928 Alan Hayward <alan.hayward@arm.com>
7929 David Sherwood <david.sherwood@arm.com>
7931 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
7932 (reduc_xor_scal_optab): New optabs.
7933 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
7934 (reduc_xor_scal_@var{m}): Document.
7935 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
7936 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
7938 * fold-const-call.c (fold_const_call): Handle them.
7939 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
7940 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
7941 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
7942 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
7943 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
7944 (UNSPEC_XORV): New unspecs.
7945 (optab): Add entries for them.
7946 (BITWISEV): New int iterator.
7947 (bit_reduc_op): New int attributes.
7949 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7950 Alan Hayward <alan.hayward@arm.com>
7951 David Sherwood <david.sherwood@arm.com>
7953 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
7954 * internal-fn.def (VEC_SHL_INSERT): New internal function.
7955 * optabs.def (vec_shl_insert_optab): New optab.
7956 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
7957 (duplicate_and_interleave): Likewise.
7958 * tree-vect-loop.c: Include internal-fn.h.
7959 (neutral_op_for_slp_reduction): New function, split out from
7960 get_initial_defs_for_reduction.
7961 (get_initial_def_for_reduction): Handle option 2 for variable-length
7962 vectors by loading the neutral value into a vector and then shifting
7963 the initial value into element 0.
7964 (get_initial_defs_for_reduction): Replace the code argument with
7965 the neutral value calculated by neutral_op_for_slp_reduction.
7966 Use gimple_build_vector for constant-length vectors.
7967 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
7968 but the first group_size elements have a neutral value.
7969 Use duplicate_and_interleave otherwise.
7970 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
7971 Update call to get_initial_defs_for_reduction. Handle SLP
7972 reductions for variable-length vectors by creating one vector
7973 result for each scalar result, with the elements associated
7974 with other scalar results stubbed out with the neutral value.
7975 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
7976 Require IFN_VEC_SHL_INSERT for double reductions on
7977 variable-length vectors, or SLP reductions that have
7978 a neutral value. Require can_duplicate_and_interleave_p
7979 support for variable-length unchained SLP reductions if there
7980 is no neutral value, such as for MIN/MAX reductions. Also require
7981 the number of vector elements to be a multiple of the number of
7982 SLP statements when doing variable-length unchained SLP reductions.
7983 Update call to vect_create_epilog_for_reduction.
7984 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
7985 and remove initial values.
7986 (duplicate_and_interleave): Make public.
7987 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
7988 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
7990 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7991 Alan Hayward <alan.hayward@arm.com>
7992 David Sherwood <david.sherwood@arm.com>
7994 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
7995 (can_duplicate_and_interleave_p): New function.
7996 (vect_get_and_check_slp_defs): Take the vector of statements
7997 rather than just the current one. Remove excess parentheses.
7998 Restriction rejectinon of vect_constant_def and vect_external_def
7999 for variable-length vectors to boolean types, or types for which
8000 can_duplicate_and_interleave_p is false.
8001 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
8002 (duplicate_and_interleave): New function.
8003 (vect_get_constant_vectors): Use gimple_build_vector for
8004 constant-length vectors and suitable variable-length constant
8005 vectors. Use duplicate_and_interleave for other variable-length
8006 vectors. Don't defer the update when inserting new statements.
8008 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8009 Alan Hayward <alan.hayward@arm.com>
8010 David Sherwood <david.sherwood@arm.com>
8012 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
8013 min_profitable_iters doesn't go negative.
8015 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8016 Alan Hayward <alan.hayward@arm.com>
8017 David Sherwood <david.sherwood@arm.com>
8019 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
8020 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
8021 * optabs.def (vec_mask_load_lanes_optab): New optab.
8022 (vec_mask_store_lanes_optab): Likewise.
8023 * internal-fn.def (MASK_LOAD_LANES): New internal function.
8024 (MASK_STORE_LANES): Likewise.
8025 * internal-fn.c (mask_load_lanes_direct): New macro.
8026 (mask_store_lanes_direct): Likewise.
8027 (expand_mask_load_optab_fn): Handle masked operations.
8028 (expand_mask_load_lanes_optab_fn): New macro.
8029 (expand_mask_store_optab_fn): Handle masked operations.
8030 (expand_mask_store_lanes_optab_fn): New macro.
8031 (direct_mask_load_lanes_optab_supported_p): Likewise.
8032 (direct_mask_store_lanes_optab_supported_p): Likewise.
8033 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
8035 (vect_load_lanes_supported): Likewise.
8036 * tree-vect-data-refs.c (strip_conversion): New function.
8037 (can_group_stmts_p): Likewise.
8038 (vect_analyze_data_ref_accesses): Use it instead of checking
8039 for a pair of assignments.
8040 (vect_store_lanes_supported): Take a masked_p parameter.
8041 (vect_load_lanes_supported): Likewise.
8042 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
8043 vect_store_lanes_supported and vect_load_lanes_supported.
8044 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8045 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
8046 parameter. Don't allow gaps for masked accesses.
8047 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
8048 and vect_load_lanes_supported.
8049 (get_load_store_type): Take a masked_p parameter and update
8050 call to get_group_load_store_type.
8051 (vectorizable_store): Update call to get_load_store_type.
8052 Handle IFN_MASK_STORE_LANES.
8053 (vectorizable_load): Update call to get_load_store_type.
8054 Handle IFN_MASK_LOAD_LANES.
8056 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8057 Alan Hayward <alan.hayward@arm.com>
8058 David Sherwood <david.sherwood@arm.com>
8060 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
8062 * config/aarch64/aarch64-protos.h
8063 (aarch64_sve_struct_memory_operand_p): Declare.
8064 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
8065 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
8066 (VPRED, vpred): Handle SVE structure modes.
8067 * config/aarch64/constraints.md (Utx): New constraint.
8068 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
8069 (aarch64_sve_struct_nonimmediate_operand): New predicates.
8070 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
8071 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
8072 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
8073 structure modes. Split into pieces after RA.
8074 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
8075 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
8077 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
8078 SVE structure modes.
8079 (aarch64_classify_address): Likewise.
8080 (sizetochar): Move earlier in file.
8081 (aarch64_print_operand): Handle SVE register lists.
8082 (aarch64_array_mode): New function.
8083 (aarch64_sve_struct_memory_operand_p): Likewise.
8084 (TARGET_ARRAY_MODE): Redefine.
8086 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8087 Alan Hayward <alan.hayward@arm.com>
8088 David Sherwood <david.sherwood@arm.com>
8090 * target.def (array_mode): New target hook.
8091 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
8092 * doc/tm.texi: Regenerate.
8093 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
8094 * hooks.c (hook_optmode_mode_uhwi_none): New function.
8095 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
8097 * stor-layout.c (mode_for_array): Likewise. Support polynomial
8100 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8101 Alan Hayward <alan.hayward@arm.com>
8102 David Sherwood <david.sherwood@arm.com>
8104 * fold-const.c (fold_binary_loc): Check the argument types
8105 rather than the result type when testing for a vector operation.
8107 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8109 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
8110 * doc/tm.texi: Regenerate.
8112 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8113 Alan Hayward <alan.hayward@arm.com>
8114 David Sherwood <david.sherwood@arm.com>
8116 * doc/invoke.texi (-msve-vector-bits=): Document new option.
8117 (sve): Document new AArch64 extension.
8118 * doc/md.texi (w): Extend the description of the AArch64
8119 constraint to include SVE vectors.
8120 (Upl, Upa): Document new AArch64 predicate constraints.
8121 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
8123 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
8124 (msve-vector-bits=): New option.
8125 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
8126 SVE when these are disabled.
8127 (sve): New extension.
8128 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
8129 modes. Adjust their number of units based on aarch64_sve_vg.
8130 (MAX_BITSIZE_MODE_ANY_MODE): Define.
8131 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
8132 aarch64_addr_query_type.
8133 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
8134 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
8135 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
8136 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
8137 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
8138 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
8139 (aarch64_simd_imm_zero_p): Delete.
8140 (aarch64_check_zero_based_sve_index_immediate): Declare.
8141 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8142 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8143 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8144 (aarch64_sve_float_mul_immediate_p): Likewise.
8145 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8147 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
8148 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
8149 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
8150 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
8151 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
8152 (aarch64_regmode_natural_size): Likewise.
8153 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
8154 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
8156 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
8157 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
8158 for VG and the SVE predicate registers.
8159 (V_ALIASES): Add a "z"-prefixed alias.
8160 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
8161 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
8162 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
8163 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
8164 (REG_CLASS_NAMES): Add entries for them.
8165 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
8166 and the predicate registers.
8167 (aarch64_sve_vg): Declare.
8168 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
8169 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
8170 (REGMODE_NATURAL_SIZE): Define.
8171 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
8173 * config/aarch64/aarch64.c: Include cfgrtl.h.
8174 (simd_immediate_info): Add a constructor for series vectors,
8175 and an associated step field.
8176 (aarch64_sve_vg): New variable.
8177 (aarch64_dbx_register_number): Handle VG and the predicate registers.
8178 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
8179 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
8180 (VEC_ANY_DATA, VEC_STRUCT): New constants.
8181 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
8182 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
8183 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
8184 (aarch64_get_mask_mode): New functions.
8185 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
8186 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
8187 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
8188 predicate modes and predicate registers. Explicitly restrict
8189 GPRs to modes of 16 bytes or smaller. Only allow FP registers
8190 to store a vector mode if it is recognized by
8191 aarch64_classify_vector_mode.
8192 (aarch64_regmode_natural_size): New function.
8193 (aarch64_hard_regno_caller_save_mode): Return the original mode
8195 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
8196 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
8197 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
8198 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
8200 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
8201 does not overlap dest if the function is frame-related. Handle
8203 (aarch64_split_add_offset): New function.
8204 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
8205 them aarch64_add_offset.
8206 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
8207 and update call to aarch64_sub_sp.
8208 (aarch64_add_cfa_expression): New function.
8209 (aarch64_expand_prologue): Pass extra temporary registers to the
8210 functions above. Handle the case in which we need to emit new
8211 DW_CFA_expressions for registers that were originally saved
8212 relative to the stack pointer, but now have to be expressed
8213 relative to the frame pointer.
8214 (aarch64_output_mi_thunk): Pass extra temporary registers to the
8216 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
8217 IP0 and IP1 values for SVE frames.
8218 (aarch64_expand_vec_series): New function.
8219 (aarch64_expand_sve_widened_duplicate): Likewise.
8220 (aarch64_expand_sve_const_vector): Likewise.
8221 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
8222 Handle SVE constants. Use emit_move_insn to move a force_const_mem
8223 into the register, rather than emitting a SET directly.
8224 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
8225 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
8226 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
8227 (offset_9bit_signed_scaled_p): New functions.
8228 (aarch64_replicate_bitmask_imm): New function.
8229 (aarch64_bitmask_imm): Use it.
8230 (aarch64_cannot_force_const_mem): Reject expressions involving
8231 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
8232 (aarch64_classify_index): Handle SVE indices, by requiring
8233 a plain register index with a scale that matches the element size.
8234 (aarch64_classify_address): Handle SVE addresses. Assert that
8235 the mode of the address is VOIDmode or an integer mode.
8236 Update call to aarch64_classify_symbol.
8237 (aarch64_classify_symbolic_expression): Update call to
8238 aarch64_classify_symbol.
8239 (aarch64_const_vec_all_in_range_p): New function.
8240 (aarch64_print_vector_float_operand): Likewise.
8241 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
8242 "vN" for FP registers with SVE modes. Handle (const ...) vectors
8243 and the FP immediates 1.0 and 0.5.
8244 (aarch64_print_address_internal): Handle SVE addresses.
8245 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
8246 (aarch64_regno_regclass): Handle predicate registers.
8247 (aarch64_secondary_reload): Handle big-endian reloads of SVE
8249 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
8250 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
8251 (aarch64_convert_sve_vector_bits): New function.
8252 (aarch64_override_options): Use it to handle -msve-vector-bits=.
8253 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8255 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
8256 Handle SVE vector and predicate modes. Accept VL-based constants
8257 that need only one temporary register, and VL offsets that require
8258 no temporary registers.
8259 (aarch64_conditional_register_usage): Mark the predicate registers
8260 as fixed if SVE isn't available.
8261 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
8262 Return true for SVE vector and predicate modes.
8263 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
8264 rather than an unsigned int. Handle SVE modes.
8265 (aarch64_preferred_simd_mode): Update call accordingly. Handle
8267 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
8269 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8270 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8271 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8272 (aarch64_sve_float_mul_immediate_p): New functions.
8273 (aarch64_sve_valid_immediate): New function.
8274 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
8275 Explicitly reject structure modes. Check for INDEX constants.
8276 Handle PTRUE and PFALSE constants.
8277 (aarch64_check_zero_based_sve_index_immediate): New function.
8278 (aarch64_simd_imm_zero_p): Delete.
8279 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
8280 vector modes. Accept constants in the range of CNT[BHWD].
8281 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
8282 ask for an Advanced SIMD mode.
8283 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
8284 (aarch64_simd_vector_alignment): Handle SVE predicates.
8285 (aarch64_vectorize_preferred_vector_alignment): New function.
8286 (aarch64_simd_vector_alignment_reachable): Use it instead of
8288 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
8289 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
8291 (MAX_VECT_LEN): Delete.
8292 (expand_vec_perm_d): Add a vec_flags field.
8293 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
8294 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
8295 (aarch64_evpc_ext): Don't apply a big-endian lane correction
8297 (aarch64_evpc_rev): Rename to...
8298 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
8299 (aarch64_evpc_rev_global): New function.
8300 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
8301 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
8303 (aarch64_evpc_sve_tbl): New function.
8304 (aarch64_expand_vec_perm_const_1): Update after rename of
8305 aarch64_evpc_rev. Handle SVE permutes too, trying
8306 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
8307 than aarch64_evpc_tbl.
8308 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
8309 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
8310 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
8311 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
8312 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
8313 (aarch64_expand_sve_vcond): New functions.
8314 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
8315 of aarch64_vector_mode_p.
8316 (aarch64_dwarf_poly_indeterminate_value): New function.
8317 (aarch64_compute_pressure_classes): Likewise.
8318 (aarch64_can_change_mode_class): Likewise.
8319 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
8320 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
8321 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
8322 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
8323 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
8324 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
8325 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
8326 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
8328 (Dn, Dl, Dr): Accept const as well as const_vector.
8329 (Dz): Likewise. Compare against CONST0_RTX.
8330 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
8331 of "vector" where appropriate.
8332 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
8333 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
8334 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
8335 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
8336 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
8337 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
8338 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
8339 (v_int_equiv): Extend to SVE modes.
8340 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
8342 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
8343 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
8344 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
8345 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
8346 (SVE_COND_FP_CMP): New int iterators.
8347 (perm_hilo): Handle the new unpack unspecs.
8348 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
8350 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
8351 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
8352 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
8353 (aarch64_equality_operator, aarch64_constant_vector_operand)
8354 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
8355 (aarch64_sve_nonimmediate_operand): Likewise.
8356 (aarch64_sve_general_operand): Likewise.
8357 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
8358 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
8359 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
8360 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
8361 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
8362 (aarch64_sve_float_arith_immediate): Likewise.
8363 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
8364 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
8365 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
8366 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
8367 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
8368 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
8369 (aarch64_sve_float_arith_operand): Likewise.
8370 (aarch64_sve_float_arith_with_sub_operand): Likewise.
8371 (aarch64_sve_float_mul_operand): Likewise.
8372 (aarch64_sve_vec_perm_operand): Likewise.
8373 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
8374 (aarch64_mov_operand): Accept const_poly_int and const_vector.
8375 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
8376 as well as const_vector.
8377 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
8378 in file. Use CONST0_RTX and CONSTM1_RTX.
8379 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
8380 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
8381 Use aarch64_simd_imm_zero.
8382 * config/aarch64/aarch64-sve.md: New file.
8383 * config/aarch64/aarch64.md: Include it.
8384 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
8385 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
8386 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
8387 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
8388 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
8389 (sve): New attribute.
8390 (enabled): Disable instructions with the sve attribute unless
8392 (movqi, movhi): Pass CONST_POLY_INT operaneds through
8393 aarch64_expand_mov_immediate.
8394 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
8395 CNT[BHSD] immediates.
8396 (movti): Split CONST_POLY_INT moves into two halves.
8397 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
8398 Split additions that need a temporary here if the destination
8399 is the stack pointer.
8400 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
8401 (*add<mode>3_poly_1): New instruction.
8402 (set_clobber_cc): New expander.
8404 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8406 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
8407 parameter and use it instead of GET_MODE_SIZE (innermode). Use
8408 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
8409 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
8410 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
8411 Change innermode from fixed_mode_size to machine_mode.
8412 (simplify_subreg): Update call accordingly. Handle a constant-sized
8413 subreg of a variable-length CONST_VECTOR.
8415 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8416 Alan Hayward <alan.hayward@arm.com>
8417 David Sherwood <david.sherwood@arm.com>
8419 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
8420 (add_offset_to_base): New function, split out from...
8421 (create_mem_ref): ...here. When handling a scale other than 1,
8422 check first whether the address is valid without the offset.
8423 Add it into the base if so, leaving the index and scale as-is.
8425 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8428 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
8429 fold_for_warn before checking if arg2 is INTEGER_CST.
8431 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
8433 * config/rs6000/predicates.md (load_multiple_operation): Delete.
8434 (store_multiple_operation): Delete.
8435 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
8436 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
8437 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
8438 guarded by TARGET_STRING.
8439 (rs6000_output_load_multiple): Delete.
8440 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
8441 OPTION_MASK_STRING / TARGET_STRING handling.
8442 (print_operand) <'N', 'O'>: Add comment that these are unused now.
8443 (const rs6000_opt_masks) <"string">: Change mask to 0.
8444 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
8445 (MASK_STRING): Delete.
8446 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
8448 (load_multiple): Delete.
8455 (store_multiple): Delete.
8462 (movmemsi_8reg): Delete.
8463 (corresponding unnamed define_insn): Delete.
8464 (movmemsi_6reg): Delete.
8465 (corresponding unnamed define_insn): Delete.
8466 (movmemsi_4reg): Delete.
8467 (corresponding unnamed define_insn): Delete.
8468 (movmemsi_2reg): Delete.
8469 (corresponding unnamed define_insn): Delete.
8470 (movmemsi_1reg): Delete.
8471 (corresponding unnamed define_insn): Delete.
8472 * config/rs6000/rs6000.opt (mno-string): New.
8473 (mstring): Replace by deprecation warning stub.
8474 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
8476 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8478 * regrename.c (regrename_do_replace): If replacing the same
8479 reg multiple times, try to reuse last created gen_raw_REG.
8482 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
8483 main to workaround a bug in GDB.
8485 2018-01-12 Tom de Vries <tom@codesourcery.com>
8488 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
8490 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
8492 PR rtl-optimization/80481
8493 * ira-color.c (get_cap_member): New function.
8494 (allocnos_conflict_by_live_ranges_p): Use it.
8495 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
8496 (setup_slot_coalesced_allocno_live_ranges): Ditto.
8498 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
8501 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
8502 (*saddl_se_1): Ditto.
8504 (*ssubl_se_1): Ditto.
8506 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8508 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
8509 rather than wi::to_widest for DR_INITs.
8510 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
8511 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
8512 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
8514 (vect_analyze_group_access_1): Note that here.
8516 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8518 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
8519 polynomial type sizes.
8521 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8523 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
8524 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
8525 (gimple_add_tmp_var): Likewise.
8527 2018-01-12 Martin Liska <mliska@suse.cz>
8529 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
8530 (gimple_alloc_sizes): Likewise.
8531 (dump_gimple_statistics): Use PRIu64 in printf format.
8532 * gimple.h: Change uint64_t to int.
8534 2018-01-12 Martin Liska <mliska@suse.cz>
8536 * tree-core.h: Use uint64_t instead of int.
8537 * tree.c (tree_node_counts): Likewise.
8538 (tree_node_sizes): Likewise.
8539 (dump_tree_statistics): Use PRIu64 in printf format.
8541 2018-01-12 Martin Liska <mliska@suse.cz>
8543 * Makefile.in: As qsort_chk is implemented in vec.c, add
8544 vec.o to linkage of gencfn-macros.
8545 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
8546 passing the info to record_node_allocation_statistics.
8547 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
8549 * ggc-common.c (struct ggc_usage): Add operator== and use
8550 it in operator< and compare function.
8551 * mem-stats.h (struct mem_usage): Likewise.
8552 * vec.c (struct vec_usage): Remove operator< and compare
8553 function. Can be simply inherited.
8555 2018-01-12 Martin Jambor <mjambor@suse.cz>
8558 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
8559 * tree-ssa-math-opts.c: Include domwalk.h.
8560 (convert_mult_to_fma_1): New function.
8561 (fma_transformation_info): New type.
8562 (fma_deferring_state): Likewise.
8563 (cancel_fma_deferring): New function.
8564 (result_of_phi): Likewise.
8565 (last_fma_candidate_feeds_initial_phi): Likewise.
8566 (convert_mult_to_fma): Added deferring logic, split actual
8567 transformation to convert_mult_to_fma_1.
8568 (math_opts_dom_walker): New type.
8569 (math_opts_dom_walker::after_dom_children): New method, body moved
8570 here from pass_optimize_widening_mul::execute, added deferring logic
8572 (pass_optimize_widening_mul::execute): Moved most of code to
8573 math_opts_dom_walker::after_dom_children.
8574 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
8575 * config/i386/i386.c (ix86_option_override_internal): Added
8576 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
8578 2018-01-12 Richard Biener <rguenther@suse.de>
8581 * dwarf2out.c (gen_variable_die): Do not reset old_die for
8582 inline instance vars.
8584 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
8587 * config/rx/rx.c (rx_is_restricted_memory_address):
8590 2018-01-12 Richard Biener <rguenther@suse.de>
8592 PR tree-optimization/80846
8593 * target.def (split_reduction): New target hook.
8594 * targhooks.c (default_split_reduction): New function.
8595 * targhooks.h (default_split_reduction): Declare.
8596 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
8597 target requests first reduce vectors by combining low and high
8599 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
8600 (get_vectype_for_scalar_type_and_size): Export.
8601 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
8602 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
8603 * doc/tm.texi: Regenerate.
8604 * config/i386/i386.c (ix86_split_reduction): Implement
8605 TARGET_VECTORIZE_SPLIT_REDUCTION.
8607 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8610 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
8611 in PIC mode except for TARGET_VXWORKS_RTP.
8612 * config/sparc/sparc.c: Include cfgrtl.h.
8613 (TARGET_INIT_PIC_REG): Define.
8614 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
8615 (sparc_pic_register_p): New predicate.
8616 (sparc_legitimate_address_p): Use it.
8617 (sparc_legitimize_pic_address): Likewise.
8618 (sparc_delegitimize_address): Likewise.
8619 (sparc_mode_dependent_address_p): Likewise.
8620 (gen_load_pcrel_sym): Remove 4th parameter.
8621 (load_got_register): Adjust call to above. Remove obsolete stuff.
8622 (sparc_expand_prologue): Do not call load_got_register here.
8623 (sparc_flat_expand_prologue): Likewise.
8624 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
8625 (sparc_use_pseudo_pic_reg): New function.
8626 (sparc_init_pic_reg): Likewise.
8627 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8628 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
8630 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
8632 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
8633 Add item for branch_cost.
8635 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8637 PR rtl-optimization/83565
8638 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
8639 not extend the result to a larger mode for rotate operations.
8640 (num_sign_bit_copies1): Likewise.
8642 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8645 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
8647 Use values-Xc.o for -pedantic.
8648 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
8650 2018-01-12 Martin Liska <mliska@suse.cz>
8653 * ipa-devirt.c (final_warning_record::grow_type_warnings):
8655 (possible_polymorphic_call_targets): Use it.
8656 (ipa_devirt): Likewise.
8658 2018-01-12 Martin Liska <mliska@suse.cz>
8660 * profile-count.h (enum profile_quality): Use 0 as invalid
8661 enum value of profile_quality.
8663 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
8665 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
8666 -mext-string options.
8668 2018-01-12 Richard Biener <rguenther@suse.de>
8670 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
8671 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
8672 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
8674 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
8676 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
8678 * configure.ac (--with-long-double-format): Add support for the
8679 configuration option to change the default long double format on
8681 * config.gcc (powerpc*-linux*-*): Likewise.
8682 * configure: Regenerate.
8683 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
8684 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
8685 used without modification.
8687 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8689 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
8690 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
8691 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
8692 MISC_BUILTIN_SPEC_BARRIER.
8693 (rs6000_init_builtins): Likewise.
8694 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
8696 (speculation_barrier): New define_insn.
8697 * doc/extend.texi: Document __builtin_speculation_barrier.
8699 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8702 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
8703 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
8704 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
8706 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
8707 integral modes instead of "ss" and "sd".
8708 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
8709 vectors with 32-bit and 64-bit elements.
8710 (vecdupssescalarmodesuffix): New mode attribute.
8711 (vec_dup<mode>): Use it.
8713 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
8716 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
8717 frame if argument is passed on stack.
8719 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8722 * ree.c (combine_reaching_defs): Optimize also
8723 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
8724 reg2=any_extend(exp); reg1=reg2;, formatting fix.
8726 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8729 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
8731 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8734 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
8735 after they are computed.
8737 2018-01-11 Bin Cheng <bin.cheng@arm.com>
8739 PR tree-optimization/83695
8740 * gimple-loop-linterchange.cc
8741 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
8742 reset cached scev information after interchange.
8743 (pass_linterchange::execute): Remove call to scev_reset_htab.
8745 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8747 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
8748 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
8749 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
8750 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
8751 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
8752 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
8753 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
8754 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
8755 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
8756 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
8757 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
8758 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
8759 (V_lane_reg): Likewise.
8760 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
8762 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
8763 (vfmal_lane_low<mode>_intrinsic,
8764 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
8765 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
8766 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
8767 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
8768 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
8769 vfmsl_lane_high<mode>_intrinsic): New define_insns.
8771 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8773 * config/arm/arm-cpus.in (fp16fml): New feature.
8774 (ALL_SIMD): Add fp16fml.
8775 (armv8.2-a): Add fp16fml as an option.
8776 (armv8.3-a): Likewise.
8777 (armv8.4-a): Add fp16fml as part of fp16.
8778 * config/arm/arm.h (TARGET_FP16FML): Define.
8779 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
8781 * config/arm/arm-modes.def (V2HF): Define.
8782 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
8783 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
8784 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
8785 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
8786 vfmsl_low, vfmsl_high): New set of builtins.
8787 * config/arm/iterators.md (PLUSMINUS): New code iterator.
8788 (vfml_op): New code attribute.
8789 (VFMLHALVES): New int iterator.
8790 (VFML, VFMLSEL): New mode attributes.
8791 (V_reg): Define mapping for V2HF.
8792 (V_hi, V_lo): New mode attributes.
8793 (VF_constraint): Likewise.
8794 (vfml_half, vfml_half_selector): New int attributes.
8795 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
8797 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
8798 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
8800 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
8801 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
8802 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
8803 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
8805 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
8806 Document new effective target and option set.
8808 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8810 * config/arm/arm-cpus.in (armv8_4): New feature.
8811 (ARMv8_4a): New fgroup.
8812 (armv8.4-a): New arch.
8813 * config/arm/arm-tables.opt: Regenerate.
8814 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
8815 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
8816 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
8817 Add matching rules for -march=armv8.4-a and extensions.
8818 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
8820 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
8823 * config/rx/rx.md (BW): New mode attribute.
8824 (sync_lock_test_and_setsi): Add mode suffix to insn output.
8826 2018-01-11 Richard Biener <rguenther@suse.de>
8828 PR tree-optimization/83435
8829 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
8830 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
8831 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
8833 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8834 Alan Hayward <alan.hayward@arm.com>
8835 David Sherwood <david.sherwood@arm.com>
8837 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
8839 (aarch64_classify_address): Initialize it. Track polynomial offsets.
8840 (aarch64_print_address_internal): Use it to check for a zero offset.
8842 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8843 Alan Hayward <alan.hayward@arm.com>
8844 David Sherwood <david.sherwood@arm.com>
8846 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
8847 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
8848 Return a poly_int64 rather than a HOST_WIDE_INT.
8849 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
8850 rather than a HOST_WIDE_INT.
8851 * config/aarch64/aarch64.h (aarch64_frame): Protect with
8852 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
8853 hard_fp_offset, frame_size, initial_adjust, callee_offset and
8854 final_offset from HOST_WIDE_INT to poly_int64.
8855 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
8856 to_constant when getting the number of units in an Advanced SIMD
8858 (aarch64_builtin_vectorized_function): Check for a constant number
8860 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
8862 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
8863 attribute instead of GET_MODE_NUNITS.
8864 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
8865 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
8866 GET_MODE_SIZE for fixed-size registers.
8867 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
8868 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
8869 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
8870 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
8871 (aarch64_print_operand, aarch64_print_address_internal)
8872 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
8873 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
8874 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
8875 Handle polynomial GET_MODE_SIZE.
8876 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
8877 wider than SImode without modification.
8878 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
8879 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
8880 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
8881 passing and returning SVE modes.
8882 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
8883 rather than GEN_INT.
8884 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
8885 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
8886 (aarch64_allocate_and_probe_stack_space): Likewise.
8887 (aarch64_layout_frame): Cope with polynomial offsets.
8888 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
8889 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
8891 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
8892 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
8893 poly_int64 rather than a HOST_WIDE_INT.
8894 (aarch64_get_separate_components, aarch64_process_components)
8895 (aarch64_expand_prologue, aarch64_expand_epilogue)
8896 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
8897 (aarch64_anchor_offset): New function, split out from...
8898 (aarch64_legitimize_address): ...here.
8899 (aarch64_builtin_vectorization_cost): Handle polynomial
8900 TYPE_VECTOR_SUBPARTS.
8901 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
8903 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
8904 number of elements from the PARALLEL rather than the mode.
8905 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
8906 rather than GET_MODE_BITSIZE.
8907 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
8908 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
8909 (aarch64_expand_vec_perm_const_1): Handle polynomial
8910 d->perm.length () and d->perm elements.
8911 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
8912 Apply to_constant to d->perm elements.
8913 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
8914 polynomial CONST_VECTOR_NUNITS.
8915 (aarch64_move_pointer): Take amount as a poly_int64 rather
8917 (aarch64_progress_pointer): Avoid temporary variable.
8918 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
8919 the mode attribute instead of GET_MODE.
8921 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8922 Alan Hayward <alan.hayward@arm.com>
8923 David Sherwood <david.sherwood@arm.com>
8925 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
8926 x exists before using it.
8927 (aarch64_add_constant_internal): Rename to...
8928 (aarch64_add_offset_1): ...this. Replace regnum with separate
8929 src and dest rtxes. Handle the case in which they're different,
8930 including when the offset is zero. Replace scratchreg with an rtx.
8931 Use 2 additions if there is no spare register into which we can
8932 move a 16-bit constant.
8933 (aarch64_add_constant): Delete.
8934 (aarch64_add_offset): Replace reg with separate src and dest
8935 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
8936 Use aarch64_add_offset_1.
8937 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
8938 an rtx rather than an int. Take the delta as a poly_int64
8939 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
8940 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
8941 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
8942 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
8943 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
8945 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
8946 aarch64_add_constant.
8948 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8950 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
8951 Use scalar_float_mode.
8953 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8955 * config/aarch64/aarch64-simd.md
8956 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
8957 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
8958 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
8959 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
8960 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
8961 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
8962 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
8963 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
8964 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
8965 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
8967 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8970 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
8971 targ_options->x_arm_arch_string is non NULL.
8973 2018-01-11 Tamar Christina <tamar.christina@arm.com>
8975 * config/aarch64/aarch64.h
8976 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
8978 2018-01-11 Sudakshina Das <sudi.das@arm.com>
8981 * expmed.c (emit_store_flag_force): Swap if const op0
8982 and change VOIDmode to mode of op0.
8984 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8986 PR rtl-optimization/83761
8987 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
8988 than bytes to mode_for_size.
8990 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8993 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
8994 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
8997 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9000 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
9001 when in layout mode.
9002 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
9003 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
9006 2018-01-10 Michael Collison <michael.collison@arm.com>
9008 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
9009 * config/aarch64/aarch64-option-extension.def: Add
9010 AARCH64_OPT_EXTENSION of 'fp16fml'.
9011 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9012 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
9013 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
9014 * config/aarch64/constraints.md (Ui7): New constraint.
9015 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
9016 (VFMLA_SEL_W): Ditto.
9019 (VFMLA16_LOW): New int iterator.
9020 (VFMLA16_HIGH): Ditto.
9021 (UNSPEC_FMLAL): New unspec.
9022 (UNSPEC_FMLSL): Ditto.
9023 (UNSPEC_FMLAL2): Ditto.
9024 (UNSPEC_FMLSL2): Ditto.
9025 (f16mac): New code attribute.
9026 * config/aarch64/aarch64-simd-builtins.def
9027 (aarch64_fmlal_lowv2sf): Ditto.
9028 (aarch64_fmlsl_lowv2sf): Ditto.
9029 (aarch64_fmlalq_lowv4sf): Ditto.
9030 (aarch64_fmlslq_lowv4sf): Ditto.
9031 (aarch64_fmlal_highv2sf): Ditto.
9032 (aarch64_fmlsl_highv2sf): Ditto.
9033 (aarch64_fmlalq_highv4sf): Ditto.
9034 (aarch64_fmlslq_highv4sf): Ditto.
9035 (aarch64_fmlal_lane_lowv2sf): Ditto.
9036 (aarch64_fmlsl_lane_lowv2sf): Ditto.
9037 (aarch64_fmlal_laneq_lowv2sf): Ditto.
9038 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
9039 (aarch64_fmlalq_lane_lowv4sf): Ditto.
9040 (aarch64_fmlsl_lane_lowv4sf): Ditto.
9041 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
9042 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
9043 (aarch64_fmlal_lane_highv2sf): Ditto.
9044 (aarch64_fmlsl_lane_highv2sf): Ditto.
9045 (aarch64_fmlal_laneq_highv2sf): Ditto.
9046 (aarch64_fmlsl_laneq_highv2sf): Ditto.
9047 (aarch64_fmlalq_lane_highv4sf): Ditto.
9048 (aarch64_fmlsl_lane_highv4sf): Ditto.
9049 (aarch64_fmlalq_laneq_highv4sf): Ditto.
9050 (aarch64_fmlsl_laneq_highv4sf): Ditto.
9051 * config/aarch64/aarch64-simd.md:
9052 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
9053 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9054 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
9055 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9056 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
9057 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
9058 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
9059 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
9060 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
9061 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
9062 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
9063 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
9064 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
9065 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
9066 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
9067 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
9068 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
9069 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
9070 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
9071 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
9072 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
9073 (vfmlsl_low_u32): Ditto.
9074 (vfmlalq_low_u32): Ditto.
9075 (vfmlslq_low_u32): Ditto.
9076 (vfmlal_high_u32): Ditto.
9077 (vfmlsl_high_u32): Ditto.
9078 (vfmlalq_high_u32): Ditto.
9079 (vfmlslq_high_u32): Ditto.
9080 (vfmlal_lane_low_u32): Ditto.
9081 (vfmlsl_lane_low_u32): Ditto.
9082 (vfmlal_laneq_low_u32): Ditto.
9083 (vfmlsl_laneq_low_u32): Ditto.
9084 (vfmlalq_lane_low_u32): Ditto.
9085 (vfmlslq_lane_low_u32): Ditto.
9086 (vfmlalq_laneq_low_u32): Ditto.
9087 (vfmlslq_laneq_low_u32): Ditto.
9088 (vfmlal_lane_high_u32): Ditto.
9089 (vfmlsl_lane_high_u32): Ditto.
9090 (vfmlal_laneq_high_u32): Ditto.
9091 (vfmlsl_laneq_high_u32): Ditto.
9092 (vfmlalq_lane_high_u32): Ditto.
9093 (vfmlslq_lane_high_u32): Ditto.
9094 (vfmlalq_laneq_high_u32): Ditto.
9095 (vfmlslq_laneq_high_u32): Ditto.
9096 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
9097 (AARCH64_FL_FOR_ARCH8_4): New.
9098 (AARCH64_ISA_F16FML): New ISA flag.
9099 (TARGET_F16FML): New feature flag for fp16fml.
9100 (doc/invoke.texi): Document new fp16fml option.
9102 2018-01-10 Michael Collison <michael.collison@arm.com>
9104 * config/aarch64/aarch64-builtins.c:
9105 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
9106 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9107 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
9108 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
9109 (AARCH64_ISA_SHA3): New ISA flag.
9110 (TARGET_SHA3): New feature flag for sha3.
9111 * config/aarch64/iterators.md (sha512_op): New int attribute.
9112 (CRYPTO_SHA512): New int iterator.
9113 (UNSPEC_SHA512H): New unspec.
9114 (UNSPEC_SHA512H2): Ditto.
9115 (UNSPEC_SHA512SU0): Ditto.
9116 (UNSPEC_SHA512SU1): Ditto.
9117 * config/aarch64/aarch64-simd-builtins.def
9118 (aarch64_crypto_sha512hqv2di): New builtin.
9119 (aarch64_crypto_sha512h2qv2di): Ditto.
9120 (aarch64_crypto_sha512su0qv2di): Ditto.
9121 (aarch64_crypto_sha512su1qv2di): Ditto.
9122 (aarch64_eor3qv8hi): Ditto.
9123 (aarch64_rax1qv2di): Ditto.
9124 (aarch64_xarqv2di): Ditto.
9125 (aarch64_bcaxqv8hi): Ditto.
9126 * config/aarch64/aarch64-simd.md:
9127 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
9128 (aarch64_crypto_sha512su0qv2di): Ditto.
9129 (aarch64_crypto_sha512su1qv2di): Ditto.
9130 (aarch64_eor3qv8hi): Ditto.
9131 (aarch64_rax1qv2di): Ditto.
9132 (aarch64_xarqv2di): Ditto.
9133 (aarch64_bcaxqv8hi): Ditto.
9134 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
9135 (vsha512h2q_u64): Ditto.
9136 (vsha512su0q_u64): Ditto.
9137 (vsha512su1q_u64): Ditto.
9138 (veor3q_u16): Ditto.
9139 (vrax1q_u64): Ditto.
9141 (vbcaxq_u16): Ditto.
9142 * config/arm/types.md (crypto_sha512): New type attribute.
9143 (crypto_sha3): Ditto.
9144 (doc/invoke.texi): Document new sha3 option.
9146 2018-01-10 Michael Collison <michael.collison@arm.com>
9148 * config/aarch64/aarch64-builtins.c:
9149 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
9150 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9151 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
9152 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
9153 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
9154 (AARCH64_ISA_SM4): New ISA flag.
9155 (TARGET_SM4): New feature flag for sm4.
9156 * config/aarch64/aarch64-simd-builtins.def
9157 (aarch64_sm3ss1qv4si): Ditto.
9158 (aarch64_sm3tt1aq4si): Ditto.
9159 (aarch64_sm3tt1bq4si): Ditto.
9160 (aarch64_sm3tt2aq4si): Ditto.
9161 (aarch64_sm3tt2bq4si): Ditto.
9162 (aarch64_sm3partw1qv4si): Ditto.
9163 (aarch64_sm3partw2qv4si): Ditto.
9164 (aarch64_sm4eqv4si): Ditto.
9165 (aarch64_sm4ekeyqv4si): Ditto.
9166 * config/aarch64/aarch64-simd.md:
9167 (aarch64_sm3ss1qv4si): Ditto.
9168 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
9169 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
9170 (aarch64_sm4eqv4si): Ditto.
9171 (aarch64_sm4ekeyqv4si): Ditto.
9172 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
9173 (sm3part_op): Ditto.
9174 (CRYPTO_SM3TT): Ditto.
9175 (CRYPTO_SM3PART): Ditto.
9176 (UNSPEC_SM3SS1): New unspec.
9177 (UNSPEC_SM3TT1A): Ditto.
9178 (UNSPEC_SM3TT1B): Ditto.
9179 (UNSPEC_SM3TT2A): Ditto.
9180 (UNSPEC_SM3TT2B): Ditto.
9181 (UNSPEC_SM3PARTW1): Ditto.
9182 (UNSPEC_SM3PARTW2): Ditto.
9183 (UNSPEC_SM4E): Ditto.
9184 (UNSPEC_SM4EKEY): Ditto.
9185 * config/aarch64/constraints.md (Ui2): New constraint.
9186 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
9187 * config/arm/types.md (crypto_sm3): New type attribute.
9188 (crypto_sm4): Ditto.
9189 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
9190 (vsm3tt1aq_u32): Ditto.
9191 (vsm3tt1bq_u32): Ditto.
9192 (vsm3tt2aq_u32): Ditto.
9193 (vsm3tt2bq_u32): Ditto.
9194 (vsm3partw1q_u32): Ditto.
9195 (vsm3partw2q_u32): Ditto.
9196 (vsm4eq_u32): Ditto.
9197 (vsm4ekeyq_u32): Ditto.
9198 (doc/invoke.texi): Document new sm4 option.
9200 2018-01-10 Michael Collison <michael.collison@arm.com>
9202 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
9203 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
9204 (AARCH64_FL_FOR_ARCH8_4): New.
9205 (AARCH64_FL_V8_4): New flag.
9206 (doc/invoke.texi): Document new armv8.4-a option.
9208 2018-01-10 Michael Collison <michael.collison@arm.com>
9210 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9211 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
9212 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
9213 * config/aarch64/aarch64-option-extension.def: Add
9214 AARCH64_OPT_EXTENSION of 'sha2'.
9215 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
9216 (crypto): Disable sha2 and aes if crypto disabled.
9217 (crypto): Enable aes and sha2 if enabled.
9218 (simd): Disable sha2 and aes if simd disabled.
9219 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
9221 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
9222 (TARGET_SHA2): New feature flag for sha2.
9223 (TARGET_AES): New feature flag for aes.
9224 * config/aarch64/aarch64-simd.md:
9225 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
9226 conditional on TARGET_AES.
9227 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
9228 (aarch64_crypto_sha1hsi): Make pattern conditional
9230 (aarch64_crypto_sha1hv4si): Ditto.
9231 (aarch64_be_crypto_sha1hv4si): Ditto.
9232 (aarch64_crypto_sha1su1v4si): Ditto.
9233 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
9234 (aarch64_crypto_sha1su0v4si): Ditto.
9235 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
9236 (aarch64_crypto_sha256su0v4si): Ditto.
9237 (aarch64_crypto_sha256su1v4si): Ditto.
9238 (doc/invoke.texi): Document new aes and sha2 options.
9240 2018-01-10 Martin Sebor <msebor@redhat.com>
9242 PR tree-optimization/83781
9243 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
9246 2018-01-11 Martin Sebor <msebor@gmail.com>
9247 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9249 PR tree-optimization/83501
9250 PR tree-optimization/81703
9252 * tree-ssa-strlen.c (get_string_cst): Rename...
9253 (get_string_len): ...to this. Handle global constants.
9254 (handle_char_store): Adjust.
9256 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
9257 Jim Wilson <jimw@sifive.com>
9259 * config/riscv/riscv-protos.h (riscv_output_return): New.
9260 * config/riscv/riscv.c (struct machine_function): New naked_p field.
9261 (riscv_attribute_table, riscv_output_return),
9262 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
9263 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
9264 (riscv_compute_frame_info): Only compute frame->mask if not a naked
9266 (riscv_expand_prologue): Add early return for naked function.
9267 (riscv_expand_epilogue): Likewise.
9268 (riscv_function_ok_for_sibcall): Return false for naked function.
9269 (riscv_set_current_function): New.
9270 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
9271 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
9272 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
9273 * doc/extend.texi (RISC-V Function Attributes): New.
9275 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
9277 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
9278 check for 128-bit long double before checking TCmode.
9279 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
9280 128-bit long doubles before checking TFmode or TCmode.
9281 (FLOAT128_IBM_P): Likewise.
9283 2018-01-10 Martin Sebor <msebor@redhat.com>
9285 PR tree-optimization/83671
9286 * builtins.c (c_strlen): Unconditionally return zero for the empty
9288 Use -Warray-bounds for warnings.
9289 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
9290 for non-constant array indices with COMPONENT_REF, arrays of
9291 arrays, and pointers to arrays.
9292 (gimple_fold_builtin_strlen): Determine and set length range for
9293 non-constant character arrays.
9295 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
9298 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
9301 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
9303 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
9305 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9308 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
9309 VECTOR_MEM_ALTIVEC_OR_VSX_P.
9310 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
9311 indexed_or_indirect_operand predicate.
9312 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
9313 (*vsx_le_perm_load_v8hi): Likewise.
9314 (*vsx_le_perm_load_v16qi): Likewise.
9315 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
9316 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
9317 (*vsx_le_perm_store_v8hi): Likewise.
9318 (*vsx_le_perm_store_v16qi): Likewise.
9319 (eight unnamed splitters): Likewise.
9321 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9323 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
9324 * config/rs6000/emmintrin.h: Likewise.
9325 * config/rs6000/mmintrin.h: Likewise.
9326 * config/rs6000/xmmintrin.h: Likewise.
9328 2018-01-10 David Malcolm <dmalcolm@redhat.com>
9331 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
9333 * tree.c (tree_nop_conversion): Return true for location wrapper
9335 (maybe_wrap_with_location): New function.
9336 (selftest::check_strip_nops): New function.
9337 (selftest::test_location_wrappers): New function.
9338 (selftest::tree_c_tests): Call it.
9339 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
9340 (maybe_wrap_with_location): New decl.
9341 (EXPR_LOCATION_WRAPPER_P): New macro.
9342 (location_wrapper_p): New inline function.
9343 (tree_strip_any_location_wrapper): New inline function.
9345 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
9348 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
9349 stack_realign_offset for the largest alignment of stack slot
9351 (ix86_find_max_used_stack_alignment): New function.
9352 (ix86_finalize_stack_frame_flags): Use it. Set
9353 max_used_stack_alignment if we don't realign stack.
9354 * config/i386/i386.h (machine_function): Add
9355 max_used_stack_alignment.
9357 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
9359 * config/arm/arm.opt (-mbranch-cost): New option.
9360 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
9363 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
9366 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
9367 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
9369 2018-01-10 Richard Biener <rguenther@suse.de>
9372 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
9373 early out so it also covers the case where we have a non-NULL
9376 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9378 PR tree-optimization/83753
9379 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
9380 for non-strided grouped accesses if the number of elements is 1.
9382 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9385 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
9386 * i386.h (TARGET_USE_GATHER): Define.
9387 * x86-tune.def (X86_TUNE_USE_GATHER): New.
9389 2018-01-10 Martin Liska <mliska@suse.cz>
9392 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
9393 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
9395 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
9396 CLEANUP_NO_PARTITIONING is not set.
9398 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9400 * doc/rtl.texi: Remove documentation of (const ...) wrappers
9401 for vectors, as a partial revert of r254296.
9402 * rtl.h (const_vec_p): Delete.
9403 (const_vec_duplicate_p): Don't test for vector CONSTs.
9404 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
9405 * expmed.c (make_tree): Likewise.
9408 * common.md (E, F): Use CONSTANT_P instead of checking for
9410 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9411 checking for CONST_VECTOR.
9413 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9416 * predict.c (force_edge_cold): Handle in more sane way edges
9419 2018-01-09 Carl Love <cel@us.ibm.com>
9421 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
9423 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
9424 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
9425 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
9426 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
9427 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
9428 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
9429 * config/rs6000/rs6000-protos.h: Add extern defition for
9430 rs6000_generate_float2_double_code.
9431 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
9433 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
9434 (float2_v2df): Add define_expand.
9436 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
9439 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
9440 op_mode in the force_to_mode call.
9442 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9444 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
9445 instead of checking each element individually.
9446 (aarch64_evpc_uzp): Likewise.
9447 (aarch64_evpc_zip): Likewise.
9448 (aarch64_evpc_ext): Likewise.
9449 (aarch64_evpc_rev): Likewise.
9450 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
9451 instead of checking each element individually. Return true without
9453 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
9454 whether all selected elements come from the same input, instead of
9455 checking each element individually. Remove calls to gen_rtx_REG,
9456 start_sequence and end_sequence and instead assert that no rtl is
9459 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9461 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
9462 order of HIGH and CONST checks.
9464 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9466 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
9467 if the destination isn't an SSA_NAME.
9469 2018-01-09 Richard Biener <rguenther@suse.de>
9471 PR tree-optimization/83668
9472 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
9474 (canonicalize_loop_form): ... here, renamed from ...
9475 (canonicalize_loop_closed_ssa_form): ... this and amended to
9476 swap successor edges for loop exit blocks to make us use
9477 the RPO order we need for initial schedule generation.
9479 2018-01-09 Joseph Myers <joseph@codesourcery.com>
9481 PR tree-optimization/64811
9482 * match.pd: When optimizing comparisons with Inf, avoid
9483 introducing or losing exceptions from comparisons with NaN.
9485 2018-01-09 Martin Liska <mliska@suse.cz>
9488 * asan.c (shadow_mem_size): Add gcc_assert.
9490 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
9492 Don't save registers in main().
9495 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
9496 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
9497 * config/avr/avr.c (avr_set_current_function): Don't error if
9498 naked, OS_task or OS_main are specified at the same time.
9499 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
9501 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
9503 * common/config/avr/avr-common.c (avr_option_optimization_table):
9504 Switch on -mmain-is-OS_task for optimizing compilations.
9506 2018-01-09 Richard Biener <rguenther@suse.de>
9508 PR tree-optimization/83572
9509 * graphite.c: Include cfganal.h.
9510 (graphite_transform_loops): Connect infinite loops to exit
9511 and remove fake edges at the end.
9513 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9515 * ipa-inline.c (edge_badness): Revert accidental checkin.
9517 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9520 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
9521 symbols; not inline clones.
9523 2018-01-09 Jakub Jelinek <jakub@redhat.com>
9526 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
9527 hard registers. Formatting fixes.
9529 PR preprocessor/83722
9530 * gcc.c (try_generate_repro): Pass
9531 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
9532 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
9535 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
9536 Kito Cheng <kito.cheng@gmail.com>
9538 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
9539 (riscv_leaf_function_p): Delete.
9540 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
9542 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9544 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
9546 (do_ifelse): New function.
9547 (do_isel): New function.
9548 (do_sub3): New function.
9549 (do_add3): New function.
9550 (do_load_mask_compare): New function.
9551 (do_overlap_load_compare): New function.
9552 (expand_compare_loop): New function.
9553 (expand_block_compare): Call expand_compare_loop() when appropriate.
9554 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
9556 (-mblock-compare-inline-loop-limit): New option.
9558 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9561 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
9562 Reverse order of second and third operands in first alternative.
9563 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
9564 of first and second elements in UNSPEC_VPERMR vector.
9565 (altivec_expand_vec_perm_le): Likewise.
9567 2018-01-08 Jeff Law <law@redhat.com>
9569 PR rtl-optimizatin/81308
9570 * tree-switch-conversion.c (cfg_altered): New file scoped static.
9571 (process_switch): If group_case_labels makes a change, then set
9573 (pass_convert_switch::execute): If a switch is converted, then
9574 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
9576 PR rtl-optimization/81308
9577 * recog.c (split_all_insns): Conditionally cleanup the CFG after
9580 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
9582 PR target/83663 - Revert r255946
9583 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
9584 generation for cases where splatting a value is not useful.
9585 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
9586 across a vec_duplicate and a paradoxical subreg forming a vector
9587 mode to a vec_concat.
9589 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9591 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
9592 -march=armv8.3-a variants.
9593 * config/arm/t-multilib: Likewise.
9594 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
9596 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9598 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
9600 (cceq_ior_compare_complement): Give it a name so I can use it, and
9601 change boolean_or_operator predicate to boolean_operator so it can
9602 be used to generate a crand.
9603 (eqne): New code iterator.
9604 (bd/bd_neg): New code_attrs.
9605 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
9606 a single define_insn.
9607 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
9608 decrement (bdnzt/bdnzf/bdzt/bdzf).
9609 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
9610 with the new names of the branch decrement patterns, and added the
9611 names of the branch decrement conditional patterns.
9613 2018-01-08 Richard Biener <rguenther@suse.de>
9615 PR tree-optimization/83563
9616 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
9619 2018-01-08 Richard Biener <rguenther@suse.de>
9622 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
9624 2018-01-08 Richard Biener <rguenther@suse.de>
9626 PR tree-optimization/83685
9627 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
9628 references to abnormals.
9630 2018-01-08 Richard Biener <rguenther@suse.de>
9633 * dwarf2out.c (output_indirect_strings): Handle empty
9634 skeleton_debug_str_hash.
9635 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
9637 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9639 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
9640 (emit_store_direct): Likewise.
9641 (arc_trampoline_adjust_address): Likewise.
9642 (arc_asm_trampoline_template): New function.
9643 (arc_initialize_trampoline): Use asm_trampoline_template.
9644 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
9645 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
9646 * config/arc/arc.md (flush_icache): Delete pattern.
9648 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9650 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
9651 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
9654 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9657 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
9658 by not USED_FOR_TARGET.
9659 (make_pass_resolve_sw_modes): Likewise.
9661 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9663 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
9666 2018-01-08 Richard Biener <rguenther@suse.de>
9669 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
9671 2018-01-08 Richard Biener <rguenther@suse.de>
9674 * match.pd ((t * 2) / 2) -> t): Add missing :c.
9676 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
9679 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
9680 basic blocks with a small number of successors.
9681 (convert_control_dep_chain_into_preds): Improve handling of
9683 (dump_predicates): Split apart into...
9684 (dump_pred_chain): ...here...
9685 (dump_pred_info): ...and here.
9686 (can_one_predicate_be_invalidated_p): Add debugging printfs.
9687 (can_chain_union_be_invalidated_p): Improve check for invalidation
9689 (uninit_uses_cannot_happen): Avoid unnecessary if
9690 convert_control_dep_chain_into_preds yielded nothing.
9692 2018-01-06 Martin Sebor <msebor@redhat.com>
9694 PR tree-optimization/83640
9695 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
9696 subtracting negative offset from size.
9697 (builtin_access::overlap): Adjust offset bounds of the access to fall
9698 within the size of the object if possible.
9700 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
9702 PR rtl-optimization/83699
9703 * expmed.c (extract_bit_field_1): Restrict the vector usage of
9704 extract_bit_field_as_subreg to cases in which the extracted
9705 value is also a vector.
9707 * lra-constraints.c (process_alt_operands): Test for the equivalence
9708 substitutions when detecting a possible reload cycle.
9710 2018-01-06 Jakub Jelinek <jakub@redhat.com>
9713 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
9714 by default if flag_selective_schedling{,2}. Formatting fixes.
9716 PR rtl-optimization/83682
9717 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
9718 if it has non-VECTOR_MODE element mode.
9719 (vec_duplicate_p): Likewise.
9722 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
9723 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
9725 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9728 * config/i386/i386-builtin.def
9729 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
9730 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
9731 Require also OPTION_MASK_ISA_AVX512F in addition to
9732 OPTION_MASK_ISA_GFNI.
9733 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
9734 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
9735 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
9736 to OPTION_MASK_ISA_GFNI.
9737 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
9738 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
9739 OPTION_MASK_ISA_AVX512BW.
9740 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
9741 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
9742 addition to OPTION_MASK_ISA_GFNI.
9743 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
9744 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
9745 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
9746 to OPTION_MASK_ISA_GFNI.
9747 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
9748 a requirement for all ISAs rather than any of them with a few
9750 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
9752 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
9753 bitmasks to be enabled with 3 exceptions, instead of requiring any
9754 enabled ISA with lots of exceptions.
9755 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
9756 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
9757 Change avx512bw in isa attribute to avx512f.
9758 * config/i386/sgxintrin.h: Add license boilerplate.
9759 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
9760 to __AVX512F__ and __AVX512VL to __AVX512VL__.
9761 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
9762 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
9764 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
9765 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
9766 temporarily sse2 rather than sse if not enabled already.
9769 * config/i386/sse.md (VI248_VLBW): Rename to ...
9770 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
9771 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
9772 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
9773 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
9774 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
9775 mode iterator instead of VI248_VLBW.
9777 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
9779 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
9780 (record_modified): Skip clobbers; add debug output.
9781 (param_change_prob): Use sreal frequencies.
9783 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9785 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
9786 punt for user-aligned variables.
9788 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9790 * tree-chrec.c (chrec_contains_symbols): Return true for
9793 2018-01-05 Sudakshina Das <sudi.das@arm.com>
9796 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
9797 of (x|y) == x for BICS pattern.
9799 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9801 PR tree-optimization/83605
9802 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
9803 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
9806 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
9808 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
9809 * config/epiphany/rtems.h: New file.
9811 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9812 Uros Bizjak <ubizjak@gmail.com>
9815 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
9816 QIreg_operand instead of register_operand predicate.
9817 * config/i386/i386.c (ix86_rop_should_change_byte_p,
9818 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
9819 comments instead of -fmitigate[-_]rop.
9821 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9824 * cgraphunit.c (symbol_table::compile): Switch to text_section
9825 before calling assembly_start debug hook.
9826 * run-rtl-passes.c (run_rtl_passes): Likewise.
9829 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9831 * tree-vrp.c (extract_range_from_binary_expr_1): Check
9832 range_int_cst_p rather than !symbolic_range_p before calling
9833 extract_range_from_multiplicative_op_1.
9835 2018-01-04 Jeff Law <law@redhat.com>
9837 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
9838 redundant test in assertion.
9840 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9842 * doc/rtl.texi: Document machine_mode wrapper classes.
9844 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9846 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
9849 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9851 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
9852 the VEC_PERM_EXPR fold to fail.
9854 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9857 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
9858 to switched_sections.
9860 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9863 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
9866 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
9869 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
9870 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
9872 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9875 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
9876 is BLKmode and bitpos not zero or mode change is needed.
9878 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9881 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
9884 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
9887 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
9888 instead of MULT rtx. Update all corresponding splitters.
9890 (*ssub<modesuffix>): Ditto.
9892 (*cmp_sadd_di): Update split patterns.
9893 (*cmp_sadd_si): Ditto.
9894 (*cmp_sadd_sidi): Ditto.
9895 (*cmp_ssub_di): Ditto.
9896 (*cmp_ssub_si): Ditto.
9897 (*cmp_ssub_sidi): Ditto.
9898 * config/alpha/predicates.md (const23_operand): New predicate.
9899 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
9900 Look for ASHIFT, not MULT inner operand.
9901 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
9903 2018-01-04 Martin Liska <mliska@suse.cz>
9905 PR gcov-profile/83669
9906 * gcov.c (output_intermediate_file): Add version to intermediate
9908 * doc/gcov.texi: Document new field 'version' in intermediate
9909 file format. Fix location of '-k' option of gcov command.
9911 2018-01-04 Martin Liska <mliska@suse.cz>
9914 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
9916 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9918 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
9920 2018-01-03 Martin Sebor <msebor@redhat.com>
9922 PR tree-optimization/83655
9923 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
9924 checking calls with invalid arguments.
9926 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9928 * tree-vect-stmts.c (vect_get_store_rhs): New function.
9929 (vectorizable_mask_load_store): Delete.
9930 (vectorizable_call): Return false for masked loads and stores.
9931 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
9932 instead of gimple_assign_rhs1.
9933 (vectorizable_load): Handle IFN_MASK_LOAD.
9934 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
9936 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9938 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
9940 (vectorizable_mask_load_store): ...here.
9941 (vectorizable_load): ...and here.
9943 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9945 * tree-vect-stmts.c (vect_build_all_ones_mask)
9946 (vect_build_zero_merge_argument): New functions, split out from...
9947 (vectorizable_load): ...here.
9949 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9951 * tree-vect-stmts.c (vect_check_store_rhs): New function,
9953 (vectorizable_mask_load_store): ...here.
9954 (vectorizable_store): ...and here.
9956 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9958 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
9960 (vectorizable_mask_load_store): ...here.
9962 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9964 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
9965 (vect_model_store_cost): Take a vec_load_store_type instead of a
9967 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
9968 (vect_model_store_cost): Take a vec_load_store_type instead of a
9970 (vectorizable_mask_load_store): Update accordingly.
9971 (vectorizable_store): Likewise.
9972 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
9974 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9976 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
9977 IFN_MASK_LOAD calls here rather than...
9978 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
9980 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9981 Alan Hayward <alan.hayward@arm.com>
9982 David Sherwood <david.sherwood@arm.com>
9984 * expmed.c (extract_bit_field_1): For vector extracts,
9985 fall back to extract_bit_field_as_subreg if vec_extract
9988 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9989 Alan Hayward <alan.hayward@arm.com>
9990 David Sherwood <david.sherwood@arm.com>
9992 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
9993 they are variable or constant sized.
9994 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
9995 slots for constant-sized data.
9997 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9998 Alan Hayward <alan.hayward@arm.com>
9999 David Sherwood <david.sherwood@arm.com>
10001 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
10002 handling COND_EXPRs with boolean comparisons, try to find a better
10003 basis for the mask type than the boolean itself.
10005 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10007 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
10008 is calculated and how it can be overridden.
10009 * genmodes.c (max_bitsize_mode_any_mode): New variable.
10010 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
10012 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
10015 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10016 Alan Hayward <alan.hayward@arm.com>
10017 David Sherwood <david.sherwood@arm.com>
10019 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
10020 Remove the mode argument.
10021 (aarch64_simd_valid_immediate): Remove the mode and inverse
10023 * config/aarch64/iterators.md (bitsize): New iterator.
10024 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
10025 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
10026 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
10027 aarch64_simd_valid_immediate.
10028 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
10029 (aarch64_reg_or_bic_imm): Likewise.
10030 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
10031 with an insn_type enum and msl with a modifier_type enum.
10032 Replace element_width with a scalar_mode. Change the shift
10033 to unsigned int. Add constructors for scalar_float_mode and
10034 scalar_int_mode elements.
10035 (aarch64_vect_float_const_representable_p): Delete.
10036 (aarch64_can_const_movi_rtx_p)
10037 (aarch64_simd_scalar_immediate_valid_for_move)
10038 (aarch64_simd_make_constant): Update call to
10039 aarch64_simd_valid_immediate.
10040 (aarch64_advsimd_valid_immediate_hs): New function.
10041 (aarch64_advsimd_valid_immediate): Likewise.
10042 (aarch64_simd_valid_immediate): Remove mode and inverse
10043 arguments. Rewrite to use the above. Use const_vec_duplicate_p
10044 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
10045 and aarch64_float_const_representable_p on the result.
10046 (aarch64_output_simd_mov_immediate): Remove mode argument.
10047 Update call to aarch64_simd_valid_immediate and use of
10048 simd_immediate_info.
10049 (aarch64_output_scalar_simd_mov_immediate): Update call
10052 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10053 Alan Hayward <alan.hayward@arm.com>
10054 David Sherwood <david.sherwood@arm.com>
10056 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
10057 (mode_nunits): Likewise CONST_MODE_NUNITS.
10058 * machmode.def (ADJUST_NUNITS): Document.
10059 * genmodes.c (mode_data::need_nunits_adj): New field.
10060 (blank_mode): Update accordingly.
10061 (adj_nunits): New variable.
10062 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
10064 (emit_mode_size_inline): Set need_bytesize_adj for all modes
10065 listed in adj_nunits.
10066 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
10067 listed in adj_nunits. Don't emit case statements for such modes.
10068 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
10069 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
10070 nothing if adj_nunits is nonnull.
10071 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
10072 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
10073 (emit_mode_fbit): Update use of print_maybe_const_decl.
10074 (emit_move_size): Likewise. Treat the array as non-const
10076 (emit_mode_adjustments): Handle adj_nunits.
10078 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10080 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
10081 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
10082 (VECTOR_MODES): Use it.
10083 (make_vector_modes): Take the prefix as an argument.
10085 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10086 Alan Hayward <alan.hayward@arm.com>
10087 David Sherwood <david.sherwood@arm.com>
10089 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
10090 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
10091 for MODE_VECTOR_BOOL.
10092 * machmode.def (VECTOR_BOOL_MODE): Document.
10093 * genmodes.c (VECTOR_BOOL_MODE): New macro.
10094 (make_vector_bool_mode): New function.
10095 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
10097 * lto-streamer-in.c (lto_input_mode_table): Likewise.
10098 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
10100 * stor-layout.c (int_mode_for_mode): Likewise.
10101 * tree.c (build_vector_type_for_mode): Likewise.
10102 * varasm.c (output_constant_pool_2): Likewise.
10103 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
10104 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
10105 for MODE_VECTOR_BOOL.
10106 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
10107 of mode class checks.
10108 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
10109 instead of a list of mode class checks.
10110 (expand_vector_scalar_condition): Likewise.
10111 (type_for_widest_vector_mode): Handle BImode as an inner mode.
10113 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10114 Alan Hayward <alan.hayward@arm.com>
10115 David Sherwood <david.sherwood@arm.com>
10117 * machmode.h (mode_size): Change from unsigned short to
10119 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
10120 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10121 or if measurement_type is not polynomial.
10122 (fixed_size_mode::includes_p): Check for constant-sized modes.
10123 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
10124 return a poly_uint16 rather than an unsigned short.
10125 (emit_mode_size): Change the type of mode_size from unsigned short
10126 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
10127 (emit_mode_adjustments): Cope with polynomial vector sizes.
10128 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10130 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10132 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
10133 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
10134 * caller-save.c (setup_save_areas): Likewise.
10135 (replace_reg_with_saved_mem): Likewise.
10136 * calls.c (emit_library_call_value_1): Likewise.
10137 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
10138 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
10139 (gen_lowpart_for_combine): Likewise.
10140 * convert.c (convert_to_integer_1): Likewise.
10141 * cse.c (equiv_constant, cse_insn): Likewise.
10142 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
10143 (cselib_subst_to_values): Likewise.
10144 * dce.c (word_dce_process_block): Likewise.
10145 * df-problems.c (df_word_lr_mark_ref): Likewise.
10146 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
10147 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
10148 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
10149 (rtl_for_decl_location): Likewise.
10150 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
10151 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
10152 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
10153 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
10154 (expand_expr_real_1): Likewise.
10155 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
10156 (pad_below): Likewise.
10157 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10158 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
10159 * ira.c (get_subreg_tracking_sizes): Likewise.
10160 * ira-build.c (ira_create_allocno_objects): Likewise.
10161 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
10162 (ira_sort_regnos_for_alter_reg): Likewise.
10163 * ira-costs.c (record_operand_costs): Likewise.
10164 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
10165 (resolve_simple_move): Likewise.
10166 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
10167 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
10168 (lra_constraints): Likewise.
10169 (CONST_POOL_OK_P): Reject variable-sized modes.
10170 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
10171 (add_pseudo_to_slot, lra_spill): Likewise.
10172 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10173 * optabs-query.c (get_best_extraction_insn): Likewise.
10174 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10175 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
10176 (expand_mult_highpart, valid_multiword_target_p): Likewise.
10177 * recog.c (offsettable_address_addr_space_p): Likewise.
10178 * regcprop.c (maybe_mode_change): Likewise.
10179 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
10180 * regrename.c (build_def_use): Likewise.
10181 * regstat.c (dump_reg_info): Likewise.
10182 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
10183 (find_reloads, find_reloads_subreg_address): Likewise.
10184 * reload1.c (eliminate_regs_1): Likewise.
10185 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
10186 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
10187 (simplify_binary_operation_1, simplify_subreg): Likewise.
10188 * targhooks.c (default_function_arg_padding): Likewise.
10189 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
10190 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
10191 (verify_gimple_assign_ternary): Likewise.
10192 * tree-inline.c (estimate_move_cost): Likewise.
10193 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10194 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
10195 (get_address_cost_ainc): Likewise.
10196 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
10197 (vect_supportable_dr_alignment): Likewise.
10198 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10199 (vectorizable_reduction): Likewise.
10200 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
10201 (vectorizable_operation, vectorizable_load): Likewise.
10202 * tree.c (build_same_sized_truth_vector_type): Likewise.
10203 * valtrack.c (cleanup_auto_inc_dec): Likewise.
10204 * var-tracking.c (emit_note_insn_var_location): Likewise.
10205 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
10206 (ADDR_VEC_ALIGN): Likewise.
10208 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10209 Alan Hayward <alan.hayward@arm.com>
10210 David Sherwood <david.sherwood@arm.com>
10212 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
10214 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10215 or if measurement_type is polynomial.
10216 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
10217 * combine.c (make_extraction): Likewise.
10218 * dse.c (find_shift_sequence): Likewise.
10219 * dwarf2out.c (mem_loc_descriptor): Likewise.
10220 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
10221 (extract_bit_field, extract_low_bits): Likewise.
10222 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
10223 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
10224 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
10225 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
10226 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10227 * reload.c (find_reloads): Likewise.
10228 * reload1.c (alter_reg): Likewise.
10229 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
10230 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
10231 * tree-if-conv.c (predicate_mem_writes): Likewise.
10232 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
10233 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
10234 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
10235 * valtrack.c (dead_debug_insert_temp): Likewise.
10236 * varasm.c (mergeable_constant_section): Likewise.
10237 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
10239 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10240 Alan Hayward <alan.hayward@arm.com>
10241 David Sherwood <david.sherwood@arm.com>
10243 * expr.c (expand_assignment): Cope with polynomial mode sizes
10244 when assigning to a CONCAT.
10246 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10247 Alan Hayward <alan.hayward@arm.com>
10248 David Sherwood <david.sherwood@arm.com>
10250 * machmode.h (mode_precision): Change from unsigned short to
10252 (mode_to_precision): Return a poly_uint16 rather than an unsigned
10254 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
10255 or if measurement_type is not polynomial.
10256 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
10257 in which the mode is already known to be a scalar_int_mode.
10258 * genmodes.c (emit_mode_precision): Change the type of mode_precision
10259 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
10261 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10262 for GET_MODE_PRECISION.
10263 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10264 for GET_MODE_PRECISION.
10265 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
10267 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
10268 (expand_field_assignment, make_extraction): Likewise.
10269 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
10270 (get_last_value): Likewise.
10271 * convert.c (convert_to_integer_1): Likewise.
10272 * cse.c (cse_insn): Likewise.
10273 * expr.c (expand_expr_real_1): Likewise.
10274 * lra-constraints.c (simplify_operand_subreg): Likewise.
10275 * optabs-query.c (can_atomic_load_p): Likewise.
10276 * optabs.c (expand_atomic_load): Likewise.
10277 (expand_atomic_store): Likewise.
10278 * ree.c (combine_reaching_defs): Likewise.
10279 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
10280 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
10281 * tree.h (type_has_mode_precision_p): Likewise.
10282 * ubsan.c (instrument_si_overflow): Likewise.
10284 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10285 Alan Hayward <alan.hayward@arm.com>
10286 David Sherwood <david.sherwood@arm.com>
10288 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
10289 polynomial numbers of units.
10290 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
10291 (valid_vector_subparts_p): New function.
10292 (build_vector_type): Remove temporary shim and take the number
10293 of units as a poly_uint64 rather than an int.
10294 (build_opaque_vector_type): Take the number of units as a
10295 poly_uint64 rather than an int.
10296 * tree.c (build_vector_from_ctor): Handle polynomial
10297 TYPE_VECTOR_SUBPARTS.
10298 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
10299 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
10300 (build_vector_from_val): If the number of units is variable,
10301 use build_vec_duplicate_cst for constant operands and
10302 VEC_DUPLICATE_EXPR otherwise.
10303 (make_vector_type): Remove temporary is_constant ().
10304 (build_vector_type, build_opaque_vector_type): Take the number of
10305 units as a poly_uint64 rather than an int.
10306 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
10308 * cfgexpand.c (expand_debug_expr): Likewise.
10309 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
10310 (store_constructor, expand_expr_real_1): Likewise.
10311 (const_scalar_mask_from_tree): Likewise.
10312 * fold-const-call.c (fold_const_reduction): Likewise.
10313 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
10314 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
10315 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
10316 (fold_relational_const): Likewise.
10317 (native_interpret_vector): Likewise. Change the size from an
10318 int to an unsigned int.
10319 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
10320 TYPE_VECTOR_SUBPARTS.
10321 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
10322 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
10323 duplicating a non-constant operand into a variable-length vector.
10324 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
10325 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
10326 * ipa-icf.c (sem_variable::equals): Likewise.
10327 * match.pd: Likewise.
10328 * omp-simd-clone.c (simd_clone_subparts): Likewise.
10329 * print-tree.c (print_node): Likewise.
10330 * stor-layout.c (layout_type): Likewise.
10331 * targhooks.c (default_builtin_vectorization_cost): Likewise.
10332 * tree-cfg.c (verify_gimple_comparison): Likewise.
10333 (verify_gimple_assign_binary): Likewise.
10334 (verify_gimple_assign_ternary): Likewise.
10335 (verify_gimple_assign_single): Likewise.
10336 * tree-pretty-print.c (dump_generic_node): Likewise.
10337 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10338 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
10339 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
10340 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
10341 (vect_shift_permute_load_chain): Likewise.
10342 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
10343 (expand_vector_condition, optimize_vector_constructor): Likewise.
10344 (lower_vec_perm, get_compute_type): Likewise.
10345 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10346 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
10347 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
10348 (vect_recog_mask_conversion_pattern): Likewise.
10349 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
10350 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
10351 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10352 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
10353 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
10354 (vectorizable_shift, vectorizable_operation, vectorizable_store)
10355 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
10356 (supportable_widening_operation): Likewise.
10357 (supportable_narrowing_operation): Likewise.
10358 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
10360 * varasm.c (output_constant): Likewise.
10362 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10363 Alan Hayward <alan.hayward@arm.com>
10364 David Sherwood <david.sherwood@arm.com>
10366 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
10367 so that both the length == 3 and length != 3 cases set up their
10368 own permute vectors. Add comments explaining why we know the
10369 number of elements is constant.
10370 (vect_permute_load_chain): Likewise.
10372 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10373 Alan Hayward <alan.hayward@arm.com>
10374 David Sherwood <david.sherwood@arm.com>
10376 * machmode.h (mode_nunits): Change from unsigned char to
10378 (ONLY_FIXED_SIZE_MODES): New macro.
10379 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
10380 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
10381 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
10383 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
10384 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
10385 or if measurement_type is not polynomial.
10386 * genmodes.c (ZERO_COEFFS): New macro.
10387 (emit_mode_nunits_inline): Make mode_nunits_inline return a
10389 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
10390 Use ZERO_COEFFS when emitting initializers.
10391 * data-streamer.h (bp_pack_poly_value): New function.
10392 (bp_unpack_poly_value): Likewise.
10393 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10394 for GET_MODE_NUNITS.
10395 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10396 for GET_MODE_NUNITS.
10397 * tree.c (make_vector_type): Remove temporary shim and make
10398 the real function take the number of units as a poly_uint64
10399 rather than an int.
10400 (build_vector_type_for_mode): Handle polynomial nunits.
10401 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
10402 * emit-rtl.c (const_vec_series_p_1): Likewise.
10403 (gen_rtx_CONST_VECTOR): Likewise.
10404 * fold-const.c (test_vec_duplicate_folding): Likewise.
10405 * genrecog.c (validate_pattern): Likewise.
10406 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
10407 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10408 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
10409 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
10410 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
10411 * rtlanal.c (subreg_get_info): Likewise.
10412 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10413 (vect_grouped_load_supported): Likewise.
10414 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
10415 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
10416 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
10417 (simplify_const_unary_operation, simplify_binary_operation_1)
10418 (simplify_const_binary_operation, simplify_ternary_operation)
10419 (test_vector_ops_duplicate, test_vector_ops): Likewise.
10420 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
10421 instead of CONST_VECTOR_NUNITS.
10422 * varasm.c (output_constant_pool_2): Likewise.
10423 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
10424 explicit-encoded elements in the XVEC for variable-length vectors.
10426 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10428 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
10430 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10431 Alan Hayward <alan.hayward@arm.com>
10432 David Sherwood <david.sherwood@arm.com>
10434 * coretypes.h (fixed_size_mode): Declare.
10435 (fixed_size_mode_pod): New typedef.
10436 * builtins.h (target_builtins::x_apply_args_mode)
10437 (target_builtins::x_apply_result_mode): Change type to
10438 fixed_size_mode_pod.
10439 * builtins.c (apply_args_size, apply_result_size, result_vector)
10440 (expand_builtin_apply_args_1, expand_builtin_apply)
10441 (expand_builtin_return): Update accordingly.
10443 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10445 * cse.c (hash_rtx_cb): Hash only the encoded elements.
10446 * cselib.c (cselib_hash_rtx): Likewise.
10447 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
10448 CONST_VECTOR encoding.
10450 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10451 Jeff Law <law@redhat.com>
10454 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
10455 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
10456 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
10457 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
10460 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
10461 explicitly probe *sp in a noreturn function if there were any callee
10462 register saves or frame pointer is needed.
10464 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10467 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
10468 BLKmode for ternary, binary or unary expressions.
10471 * var-tracking.c (delete_vta_debug_insn): New inline function.
10472 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
10473 insns from get_insns () to NULL instead of each bb separately.
10474 Use delete_vta_debug_insn. No longer static.
10475 (vt_debug_insns_local, variable_tracking_main_1): Adjust
10476 delete_vta_debug_insns callers.
10477 * rtl.h (delete_vta_debug_insns): Declare.
10478 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
10479 instead of variable_tracking_main.
10481 2018-01-03 Martin Sebor <msebor@redhat.com>
10483 PR tree-optimization/83603
10484 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
10485 arguments past the endof the argument list in functions declared
10486 without a prototype.
10487 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
10488 Avoid checking when arguments are null.
10490 2018-01-03 Martin Sebor <msebor@redhat.com>
10493 * doc/extend.texi (attribute const): Fix a typo.
10494 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
10495 issuing -Wsuggest-attribute for void functions.
10497 2018-01-03 Martin Sebor <msebor@redhat.com>
10499 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
10500 offset_int::from instead of wide_int::to_shwi.
10501 (maybe_diag_overlap): Remove assertion.
10502 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
10503 * gimple-ssa-sprintf.c (format_directive): Same.
10504 (parse_directive): Same.
10505 (sprintf_dom_walker::compute_format_length): Same.
10506 (try_substitute_return_value): Same.
10508 2018-01-03 Jeff Law <law@redhat.com>
10510 PR middle-end/83654
10511 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
10512 non-constant residual for zero at runtime and avoid probing in
10513 that case. Reorganize code for trailing problem to mirror handling
10516 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10518 PR tree-optimization/83501
10519 * tree-ssa-strlen.c (get_string_cst): New.
10520 (handle_char_store): Call get_string_cst.
10522 2018-01-03 Martin Liska <mliska@suse.cz>
10524 PR tree-optimization/83593
10525 * tree-ssa-strlen.c: Include tree-cfg.h.
10526 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
10527 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
10528 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
10530 (strlen_dom_walker::before_dom_children): Call
10531 gimple_purge_dead_eh_edges. Dump tranformation with details
10533 (strlen_dom_walker::before_dom_children): Update call by adding
10534 new argument cleanup_eh.
10535 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
10537 2018-01-03 Martin Liska <mliska@suse.cz>
10540 * cif-code.def (VARIADIC_THUNK): New enum value.
10541 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
10544 2018-01-03 Jan Beulich <jbeulich@suse.com>
10546 * sse.md (mov<mode>_internal): Tighten condition for when to use
10547 vmovdqu<ssescalarsize> for TI and OI modes.
10549 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10551 Update copyright years.
10553 2018-01-03 Martin Liska <mliska@suse.cz>
10556 * ipa-visibility.c (function_and_variable_visibility): Skip
10557 functions with noipa attribure.
10559 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10561 * gcc.c (process_command): Update copyright notice dates.
10562 * gcov-dump.c (print_version): Ditto.
10563 * gcov.c (print_version): Ditto.
10564 * gcov-tool.c (print_version): Ditto.
10565 * gengtype.c (create_file): Ditto.
10566 * doc/cpp.texi: Bump @copying's copyright year.
10567 * doc/cppinternals.texi: Ditto.
10568 * doc/gcc.texi: Ditto.
10569 * doc/gccint.texi: Ditto.
10570 * doc/gcov.texi: Ditto.
10571 * doc/install.texi: Ditto.
10572 * doc/invoke.texi: Ditto.
10574 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10576 * vector-builder.h (vector_builder::m_full_nelts): Change from
10577 unsigned int to poly_uint64.
10578 (vector_builder::full_nelts): Update prototype accordingly.
10579 (vector_builder::new_vector): Likewise.
10580 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
10581 (vector_builder::operator ==): Likewise.
10582 (vector_builder::finalize): Likewise.
10583 * int-vector-builder.h (int_vector_builder::int_vector_builder):
10584 Take the number of elements as a poly_uint64 rather than an
10586 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
10587 from unsigned int to poly_uint64.
10588 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
10589 (vec_perm_indices::new_vector): Likewise.
10590 (vec_perm_indices::length): Likewise.
10591 (vec_perm_indices::nelts_per_input): Likewise.
10592 (vec_perm_indices::input_nelts): Likewise.
10593 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
10594 number of elements per input as a poly_uint64 rather than an
10595 unsigned int. Use the original encoding for variable-length
10596 vectors, rather than clamping each individual element.
10597 For the second and subsequent elements in each pattern,
10598 clamp the step and base before clamping their sum.
10599 (vec_perm_indices::series_p): Handle polynomial element counts.
10600 (vec_perm_indices::all_in_range_p): Likewise.
10601 (vec_perm_indices_to_tree): Likewise.
10602 (vec_perm_indices_to_rtx): Likewise.
10603 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
10604 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
10605 (tree_vector_builder::new_binary_operation): Handle polynomial
10606 element counts. Return false if we need to know the number
10607 of elements at compile time.
10608 * fold-const.c (fold_vec_perm): Punt if the number of elements
10609 isn't known at compile time.
10611 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10613 * vec-perm-indices.h (vec_perm_builder): Change element type
10614 from HOST_WIDE_INT to poly_int64.
10615 (vec_perm_indices::element_type): Update accordingly.
10616 (vec_perm_indices::clamp): Handle polynomial element_types.
10617 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10618 (vec_perm_indices::all_in_range_p): Likewise.
10619 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
10621 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
10622 polynomial vec_perm_indices element types.
10623 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
10624 * fold-const.c (fold_vec_perm): Likewise.
10625 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
10626 * tree-vect-generic.c (lower_vec_perm): Likewise.
10627 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10628 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
10629 element type to HOST_WIDE_INT.
10631 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10632 Alan Hayward <alan.hayward@arm.com>
10633 David Sherwood <david.sherwood@arm.com>
10635 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
10636 rather than an int. Use plus_constant.
10637 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
10638 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
10640 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10641 Alan Hayward <alan.hayward@arm.com>
10642 David Sherwood <david.sherwood@arm.com>
10644 * calls.c (emit_call_1, expand_call): Change struct_value_size from
10645 a HOST_WIDE_INT to a poly_int64.
10647 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10648 Alan Hayward <alan.hayward@arm.com>
10649 David Sherwood <david.sherwood@arm.com>
10651 * calls.c (load_register_parameters): Cope with polynomial
10652 mode sizes. Require a constant size for BLKmode parameters
10653 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
10654 forces a parameter to be padded at the lsb end in order to
10655 fill a complete number of words, require the parameter size
10656 to be ordered wrt UNITS_PER_WORD.
10658 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10659 Alan Hayward <alan.hayward@arm.com>
10660 David Sherwood <david.sherwood@arm.com>
10662 * reload1.c (spill_stack_slot_width): Change element type
10663 from unsigned int to poly_uint64_pod.
10664 (alter_reg): Treat mode sizes as polynomial.
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 * reload.c (complex_word_subreg_p): New function.
10671 (reload_inner_reg_of_subreg, push_reload): Use it.
10673 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10674 Alan Hayward <alan.hayward@arm.com>
10675 David Sherwood <david.sherwood@arm.com>
10677 * lra-constraints.c (process_alt_operands): Reject matched
10678 operands whose sizes aren't ordered.
10679 (match_reload): Refer to this check here.
10681 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10682 Alan Hayward <alan.hayward@arm.com>
10683 David Sherwood <david.sherwood@arm.com>
10685 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
10686 that the mode size is in the set {1, 2, 4, 8, 16}.
10688 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10689 Alan Hayward <alan.hayward@arm.com>
10690 David Sherwood <david.sherwood@arm.com>
10692 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
10693 Use plus_constant instead of gen_rtx_PLUS.
10695 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10696 Alan Hayward <alan.hayward@arm.com>
10697 David Sherwood <david.sherwood@arm.com>
10699 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
10700 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
10701 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
10702 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
10703 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
10704 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
10705 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
10706 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
10707 * config/i386/i386.c (ix86_push_rounding): ...this new function.
10708 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
10710 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
10711 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
10712 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
10713 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
10714 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
10715 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
10716 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
10717 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
10718 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
10719 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
10721 * expr.c (emit_move_resolve_push): Treat the input and result
10722 of PUSH_ROUNDING as a poly_int64.
10723 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
10724 (emit_push_insn): Likewise.
10725 * lra-eliminations.c (mark_not_eliminable): Likewise.
10726 * recog.c (push_operand): Likewise.
10727 * reload1.c (elimination_effects): Likewise.
10728 * rtlanal.c (nonzero_bits1): Likewise.
10729 * calls.c (store_one_arg): Likewise. Require the padding to be
10730 known at compile time.
10732 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10733 Alan Hayward <alan.hayward@arm.com>
10734 David Sherwood <david.sherwood@arm.com>
10736 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
10737 Use plus_constant instead of gen_rtx_PLUS.
10739 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10740 Alan Hayward <alan.hayward@arm.com>
10741 David Sherwood <david.sherwood@arm.com>
10743 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
10744 rather than an int.
10746 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10747 Alan Hayward <alan.hayward@arm.com>
10748 David Sherwood <david.sherwood@arm.com>
10750 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
10751 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
10752 via stack temporaries. Treat the mode size as polynomial too.
10754 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10755 Alan Hayward <alan.hayward@arm.com>
10756 David Sherwood <david.sherwood@arm.com>
10758 * expr.c (expand_expr_real_2): When handling conversions involving
10759 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
10760 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
10761 as a poly_uint64 too.
10763 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10764 Alan Hayward <alan.hayward@arm.com>
10765 David Sherwood <david.sherwood@arm.com>
10767 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
10769 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10770 Alan Hayward <alan.hayward@arm.com>
10771 David Sherwood <david.sherwood@arm.com>
10773 * combine.c (can_change_dest_mode): Handle polynomial
10774 REGMODE_NATURAL_SIZE.
10775 * expmed.c (store_bit_field_1): Likewise.
10776 * expr.c (store_constructor): Likewise.
10777 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
10778 and polynomial REGMODE_NATURAL_SIZE.
10779 (gen_lowpart_common): Likewise.
10780 * reginfo.c (record_subregs_of_mode): Likewise.
10781 * rtlanal.c (read_modify_subreg_p): Likewise.
10783 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10784 Alan Hayward <alan.hayward@arm.com>
10785 David Sherwood <david.sherwood@arm.com>
10787 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
10788 numbers of elements.
10790 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10791 Alan Hayward <alan.hayward@arm.com>
10792 David Sherwood <david.sherwood@arm.com>
10794 * match.pd: Cope with polynomial numbers of vector elements.
10796 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10797 Alan Hayward <alan.hayward@arm.com>
10798 David Sherwood <david.sherwood@arm.com>
10800 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
10801 in a POINTER_PLUS_EXPR.
10803 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10804 Alan Hayward <alan.hayward@arm.com>
10805 David Sherwood <david.sherwood@arm.com>
10807 * omp-simd-clone.c (simd_clone_subparts): New function.
10808 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
10809 (ipa_simd_modify_function_body): Likewise.
10811 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10812 Alan Hayward <alan.hayward@arm.com>
10813 David Sherwood <david.sherwood@arm.com>
10815 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
10816 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
10817 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
10818 (expand_vector_condition, vector_element): Likewise.
10819 (subparts_gt): New function.
10820 (get_compute_type): Use subparts_gt.
10821 (count_type_subparts): Delete.
10822 (expand_vector_operations_1): Use subparts_gt instead of
10823 count_type_subparts.
10825 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10826 Alan Hayward <alan.hayward@arm.com>
10827 David Sherwood <david.sherwood@arm.com>
10829 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
10830 (vect_compile_time_alias): ...this new function. Do the calculation
10831 on poly_ints rather than trees.
10832 (vect_prune_runtime_alias_test_list): Update call accordingly.
10834 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10835 Alan Hayward <alan.hayward@arm.com>
10836 David Sherwood <david.sherwood@arm.com>
10838 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
10840 (vect_schedule_slp_instance): Likewise.
10842 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10843 Alan Hayward <alan.hayward@arm.com>
10844 David Sherwood <david.sherwood@arm.com>
10846 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
10847 constant and extern definitions for variable-length vectors.
10848 (vect_get_constant_vectors): Note that the number of units
10849 is known to be constant.
10851 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10852 Alan Hayward <alan.hayward@arm.com>
10853 David Sherwood <david.sherwood@arm.com>
10855 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
10856 of units as polynomial. Choose between WIDE and NARROW based
10859 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10860 Alan Hayward <alan.hayward@arm.com>
10861 David Sherwood <david.sherwood@arm.com>
10863 * tree-vect-stmts.c (simd_clone_subparts): New function.
10864 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
10866 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10867 Alan Hayward <alan.hayward@arm.com>
10868 David Sherwood <david.sherwood@arm.com>
10870 * tree-vect-stmts.c (vectorizable_call): Treat the number of
10871 vectors as polynomial. Use build_index_vector for
10872 IFN_GOMP_SIMD_LANE.
10874 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10875 Alan Hayward <alan.hayward@arm.com>
10876 David Sherwood <david.sherwood@arm.com>
10878 * tree-vect-stmts.c (get_load_store_type): Treat the number of
10879 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
10880 for variable-length vectors.
10881 (vectorizable_mask_load_store): Treat the number of units as
10882 polynomial, asserting that it is constant if the condition has
10883 already been enforced.
10884 (vectorizable_store, vectorizable_load): Likewise.
10886 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10887 Alan Hayward <alan.hayward@arm.com>
10888 David Sherwood <david.sherwood@arm.com>
10890 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
10891 of units as polynomial. Punt if we can't tell at compile time
10892 which vector contains the final result.
10894 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10895 Alan Hayward <alan.hayward@arm.com>
10896 David Sherwood <david.sherwood@arm.com>
10898 * tree-vect-loop.c (vectorizable_induction): Treat the number
10899 of units as polynomial. Punt on SLP inductions. Use an integer
10900 VEC_SERIES_EXPR for variable-length integer reductions. Use a
10901 cast of such a series for variable-length floating-point
10904 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10905 Alan Hayward <alan.hayward@arm.com>
10906 David Sherwood <david.sherwood@arm.com>
10908 * tree.h (build_index_vector): Declare.
10909 * tree.c (build_index_vector): New function.
10910 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
10911 of units as polynomial, forcibly converting it to a constant if
10912 vectorizable_reduction has already enforced the condition.
10913 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
10914 to create a {1,2,3,...} vector.
10915 (vectorizable_reduction): Treat the number of units as polynomial.
10916 Choose vectype_in based on the largest scalar element size rather
10917 than the smallest number of units. Enforce the restrictions
10920 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10921 Alan Hayward <alan.hayward@arm.com>
10922 David Sherwood <david.sherwood@arm.com>
10924 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
10925 number of units as polynomial.
10927 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10928 Alan Hayward <alan.hayward@arm.com>
10929 David Sherwood <david.sherwood@arm.com>
10931 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
10932 * target.def (autovectorize_vector_sizes): Return the vector sizes
10933 by pointer, using vector_sizes rather than a bitmask.
10934 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
10935 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
10936 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
10938 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
10939 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
10940 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
10941 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
10942 * omp-general.c (omp_max_vf): Likewise.
10943 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10944 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
10945 * tree-vect-loop.c (vect_analyze_loop): Likewise.
10946 * tree-vect-slp.c (vect_slp_bb): Likewise.
10947 * doc/tm.texi: Regenerate.
10948 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
10950 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
10951 the vector size as a poly_uint64 rather than an unsigned int.
10952 (current_vector_size): Change from an unsigned int to a poly_uint64.
10953 (get_vectype_for_scalar_type): Update accordingly.
10954 * tree.h (build_truth_vector_type): Take the size and number of
10955 units as a poly_uint64 rather than an unsigned int.
10956 (build_vector_type): Add a temporary overload that takes
10957 the number of units as a poly_uint64 rather than an unsigned int.
10958 * tree.c (make_vector_type): Likewise.
10959 (build_truth_vector_type): Take the number of units as a poly_uint64
10960 rather than an unsigned int.
10962 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10963 Alan Hayward <alan.hayward@arm.com>
10964 David Sherwood <david.sherwood@arm.com>
10966 * target.def (get_mask_mode): Take the number of units and length
10967 as poly_uint64s rather than unsigned ints.
10968 * targhooks.h (default_get_mask_mode): Update accordingly.
10969 * targhooks.c (default_get_mask_mode): Likewise.
10970 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
10971 * doc/tm.texi: Regenerate.
10973 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10974 Alan Hayward <alan.hayward@arm.com>
10975 David Sherwood <david.sherwood@arm.com>
10977 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
10978 * omp-general.c (omp_max_vf): Likewise.
10979 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
10980 (expand_omp_simd): Handle polynomial safelen.
10981 * omp-low.c (omplow_simd_context): Add a default constructor.
10982 (omplow_simd_context::max_vf): Change from int to poly_uint64.
10983 (lower_rec_simd_input_clauses): Update accordingly.
10984 (lower_rec_input_clauses): Likewise.
10986 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10987 Alan Hayward <alan.hayward@arm.com>
10988 David Sherwood <david.sherwood@arm.com>
10990 * tree-vectorizer.h (vect_nunits_for_cost): New function.
10991 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
10992 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
10993 (vect_analyze_slp_cost): Likewise.
10994 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
10995 (vect_model_load_cost): Likewise.
10997 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10998 Alan Hayward <alan.hayward@arm.com>
10999 David Sherwood <david.sherwood@arm.com>
11001 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
11002 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
11003 from an unsigned int * to a poly_uint64_pod *.
11004 (calculate_unrolling_factor): New function.
11005 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
11007 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11008 Alan Hayward <alan.hayward@arm.com>
11009 David Sherwood <david.sherwood@arm.com>
11011 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
11012 from an unsigned int to a poly_uint64.
11013 (_loop_vec_info::slp_unrolling_factor): Likewise.
11014 (_loop_vec_info::vectorization_factor): Change from an int
11016 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
11017 (vect_get_num_vectors): New function.
11018 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
11019 (vect_get_num_copies): Use vect_get_num_vectors.
11020 (vect_analyze_data_ref_dependences): Change max_vf from an int *
11021 to an unsigned int *.
11022 (vect_analyze_data_refs): Change min_vf from an int * to a
11024 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11025 than an unsigned HOST_WIDE_INT.
11026 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
11027 (vect_analyze_data_ref_dependence): Change max_vf from an int *
11028 to an unsigned int *.
11029 (vect_analyze_data_ref_dependences): Likewise.
11030 (vect_compute_data_ref_alignment): Handle polynomial vf.
11031 (vect_enhance_data_refs_alignment): Likewise.
11032 (vect_prune_runtime_alias_test_list): Likewise.
11033 (vect_shift_permute_load_chain): Likewise.
11034 (vect_supportable_dr_alignment): Likewise.
11035 (dependence_distance_ge_vf): Take the vectorization factor as a
11036 poly_uint64 rather than an unsigned HOST_WIDE_INT.
11037 (vect_analyze_data_refs): Change min_vf from an int * to a
11039 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
11040 vfm1 as a poly_uint64 rather than an int. Make the same change
11041 for the returned bound_scalar.
11042 (vect_gen_vector_loop_niters): Handle polynomial vf.
11043 (vect_do_peeling): Likewise. Update call to
11044 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
11045 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
11047 * tree-vect-loop.c (vect_determine_vectorization_factor)
11048 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
11049 (vect_get_known_peeling_cost): Likewise.
11050 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
11051 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
11052 (vect_transform_loop): Likewise. Use the lowest possible VF when
11053 updating the upper bounds of the loop.
11054 (vect_min_worthwhile_factor): Make static. Return an unsigned int
11055 rather than an int.
11056 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
11057 polynomial unroll factors.
11058 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
11059 (vect_make_slp_decision): Likewise.
11060 (vect_supported_load_permutation_p): Likewise, and polynomial
11062 (vect_analyze_slp_cost): Handle polynomial vf.
11063 (vect_slp_analyze_node_operations): Likewise.
11064 (vect_slp_analyze_bb_1): Likewise.
11065 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11066 than an unsigned HOST_WIDE_INT.
11067 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
11068 (vectorizable_load): Handle polynomial vf.
11069 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
11071 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
11073 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11074 Alan Hayward <alan.hayward@arm.com>
11075 David Sherwood <david.sherwood@arm.com>
11077 * match.pd: Handle bit operations involving three constants
11078 and try to fold one pair.
11080 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11082 * tree-vect-loop-manip.c: Include gimple-fold.h.
11083 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
11084 niters_maybe_zero parameters. Handle other cases besides a step of 1.
11085 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
11086 Add a path that uses a step of VF instead of 1, but disable it
11088 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
11089 and niters_no_overflow parameters. Update calls to
11090 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
11091 Create a new SSA name if the latter choses to use a ste other
11092 than zero, and return it via niters_vector_mult_vf_var.
11093 * tree-vect-loop.c (vect_transform_loop): Update calls to
11094 vect_do_peeling, vect_gen_vector_loop_niters and
11095 slpeel_make_loop_iterate_ntimes.
11096 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
11097 (vect_gen_vector_loop_niters): Update declarations after above changes.
11099 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
11101 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
11102 128-bit round to integer instructions.
11103 (ceil<mode>2): Likewise.
11104 (btrunc<mode>2): Likewise.
11105 (round<mode>2): Likewise.
11107 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11109 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
11110 unaligned VSX load/store on P8/P9.
11111 (expand_block_clear): Allow the use of unaligned VSX
11112 load/store on P8/P9.
11114 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11116 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
11118 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
11119 swap associated with both a load and a store.
11121 2018-01-02 Andrew Waterman <andrew@sifive.com>
11123 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
11124 * config/riscv/riscv.md (clear_cache): Use it.
11126 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
11128 * web.c: Remove out-of-date comment.
11130 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11132 * expr.c (fixup_args_size_notes): Check that any existing
11133 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
11134 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
11135 (emit_single_push_insn): ...here.
11137 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11139 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
11140 (const_vector_encoded_nelts): New function.
11141 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
11142 (const_vector_int_elt, const_vector_elt): Declare.
11143 * emit-rtl.c (const_vector_int_elt_1): New function.
11144 (const_vector_elt): Likewise.
11145 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
11146 of CONST_VECTOR_ELT.
11148 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11150 * expr.c: Include rtx-vector-builder.h.
11151 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
11152 directly on the tree encoding.
11153 (const_vector_from_tree): Likewise.
11154 * optabs.c: Include rtx-vector-builder.h.
11155 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
11156 sequence of "u" values.
11157 * vec-perm-indices.c: Include rtx-vector-builder.h.
11158 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
11159 directly on the vec_perm_indices encoding.
11161 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11163 * doc/rtl.texi (const_vector): Describe new encoding scheme.
11164 * Makefile.in (OBJS): Add rtx-vector-builder.o.
11165 * rtx-vector-builder.h: New file.
11166 * rtx-vector-builder.c: Likewise.
11167 * rtl.h (rtx_def::u2): Add a const_vector field.
11168 (CONST_VECTOR_NPATTERNS): New macro.
11169 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
11170 (CONST_VECTOR_DUPLICATE_P): Likewise.
11171 (CONST_VECTOR_STEPPED_P): Likewise.
11172 (CONST_VECTOR_ENCODED_ELT): Likewise.
11173 (const_vec_duplicate_p): Check for a duplicated vector encoding.
11174 (unwrap_const_vec_duplicate): Likewise.
11175 (const_vec_series_p): Check for a non-duplicated vector encoding.
11176 Say that the function only returns true for integer vectors.
11177 * emit-rtl.c: Include rtx-vector-builder.h.
11178 (gen_const_vec_duplicate_1): Delete.
11179 (gen_const_vector): Call gen_const_vec_duplicate instead of
11180 gen_const_vec_duplicate_1.
11181 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
11182 (gen_const_vec_duplicate): Use rtx_vector_builder.
11183 (gen_const_vec_series): Likewise.
11184 (gen_rtx_CONST_VECTOR): Likewise.
11185 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
11186 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11187 Build a new vector rather than modifying a CONST_VECTOR in-place.
11188 (handle_special_swappables): Update call accordingly.
11189 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
11190 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11191 Build a new vector rather than modifying a CONST_VECTOR in-place.
11192 (handle_special_swappables): Update call accordingly.
11194 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11196 * simplify-rtx.c (simplify_const_binary_operation): Use
11197 CONST_VECTOR_ELT instead of XVECEXP.
11199 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11201 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
11202 the selector elements to be different from the data elements
11203 if the selector is a VECTOR_CST.
11204 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
11205 ssizetype for the selector.
11207 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11209 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
11210 before testing each element individually.
11211 * tree-vect-generic.c (lower_vec_perm): Likewise.
11213 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11215 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
11216 * selftest-run-tests.c (selftest::run_tests): Call it.
11217 * vector-builder.h (vector_builder::operator ==): New function.
11218 (vector_builder::operator !=): Likewise.
11219 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
11220 (vec_perm_indices::all_from_input_p): New function.
11221 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11222 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
11223 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
11224 instead of reading the VECTOR_CST directly. Detect whether both
11225 vector inputs are the same before constructing the vec_perm_indices,
11226 and update the number of inputs argument accordingly. Use the
11227 utility functions added above. Only construct sel2 if we need to.
11229 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11231 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
11232 the broadcast of the low byte.
11233 (expand_mult_highpart): Use an explicit encoding for the permutes.
11234 * optabs-query.c (can_mult_highpart_p): Likewise.
11235 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
11236 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11237 (vectorizable_bswap): Likewise.
11238 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
11239 explicit encoding for the power-of-2 permutes.
11240 (vect_permute_store_chain): Likewise.
11241 (vect_grouped_load_supported): Likewise.
11242 (vect_permute_load_chain): Likewise.
11244 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11246 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
11247 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
11248 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
11249 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11250 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
11251 (vect_gen_perm_mask_any): Likewise.
11253 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11255 * int-vector-builder.h: New file.
11256 * vec-perm-indices.h: Include int-vector-builder.h.
11257 (vec_perm_indices): Redefine as an int_vector_builder.
11258 (auto_vec_perm_indices): Delete.
11259 (vec_perm_builder): Redefine as a stand-alone class.
11260 (vec_perm_indices::vec_perm_indices): New function.
11261 (vec_perm_indices::clamp): Likewise.
11262 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
11263 (vec_perm_indices::new_vector): New function.
11264 (vec_perm_indices::new_expanded_vector): Update for new
11265 vec_perm_indices class.
11266 (vec_perm_indices::rotate_inputs): New function.
11267 (vec_perm_indices::all_in_range_p): Operate directly on the
11268 encoded form, without computing elided elements.
11269 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
11270 encoding. Update for new vec_perm_indices class.
11271 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
11272 the given vec_perm_builder.
11273 (expand_vec_perm_var): Update vec_perm_builder constructor.
11274 (expand_mult_highpart): Use vec_perm_builder instead of
11275 auto_vec_perm_indices.
11276 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
11277 vec_perm_indices instead of auto_vec_perm_indices. Use a single
11278 or double series encoding as appropriate.
11279 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
11280 vec_perm_indices instead of auto_vec_perm_indices.
11281 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11282 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11283 (vect_permute_store_chain): Likewise.
11284 (vect_grouped_load_supported): Likewise.
11285 (vect_permute_load_chain): Likewise.
11286 (vect_shift_permute_load_chain): Likewise.
11287 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11288 (vect_transform_slp_perm_load): Likewise.
11289 (vect_schedule_slp_instance): Likewise.
11290 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11291 (vectorizable_mask_load_store): Likewise.
11292 (vectorizable_bswap): Likewise.
11293 (vectorizable_store): Likewise.
11294 (vectorizable_load): Likewise.
11295 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
11296 vec_perm_indices instead of auto_vec_perm_indices. Use
11297 tree_to_vec_perm_builder to read the vector from a tree.
11298 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
11299 vec_perm_builder instead of a vec_perm_indices.
11300 (have_whole_vector_shift): Use vec_perm_builder and
11301 vec_perm_indices instead of auto_vec_perm_indices. Leave the
11302 truncation to calc_vec_perm_mask_for_shift.
11303 (vect_create_epilog_for_reduction): Likewise.
11304 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
11305 from auto_vec_perm_indices to vec_perm_indices.
11306 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11307 instead of changing individual elements.
11308 (aarch64_vectorize_vec_perm_const): Use new_vector to install
11309 the vector in d.perm.
11310 * config/arm/arm.c (expand_vec_perm_d::perm): Change
11311 from auto_vec_perm_indices to vec_perm_indices.
11312 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11313 instead of changing individual elements.
11314 (arm_vectorize_vec_perm_const): Use new_vector to install
11315 the vector in d.perm.
11316 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
11317 Update vec_perm_builder constructor.
11318 (rs6000_expand_interleave): Likewise.
11319 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
11320 (rs6000_expand_interleave): Likewise.
11322 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11324 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
11325 to qimode could truncate the indices.
11326 * optabs.c (expand_vec_perm_var): Likewise.
11328 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11330 * Makefile.in (OBJS): Add vec-perm-indices.o.
11331 * vec-perm-indices.h: New file.
11332 * vec-perm-indices.c: Likewise.
11333 * target.h (vec_perm_indices): Replace with a forward class
11335 (auto_vec_perm_indices): Move to vec-perm-indices.h.
11336 * optabs.h: Include vec-perm-indices.h.
11337 (expand_vec_perm): Delete.
11338 (selector_fits_mode_p, expand_vec_perm_var): Declare.
11339 (expand_vec_perm_const): Declare.
11340 * target.def (vec_perm_const_ok): Replace with...
11341 (vec_perm_const): ...this new hook.
11342 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
11343 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
11344 * doc/tm.texi: Regenerate.
11345 * optabs.def (vec_perm_const): Delete.
11346 * doc/md.texi (vec_perm_const): Likewise.
11347 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
11348 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
11349 expand_vec_perm for constant permutation vectors. Assert that
11350 the mode of variable permutation vectors is the integer equivalent
11351 of the mode that is being permuted.
11352 * optabs-query.h (selector_fits_mode_p): Declare.
11353 * optabs-query.c: Include vec-perm-indices.h.
11354 (selector_fits_mode_p): New function.
11355 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
11356 is defined, instead of checking whether the vec_perm_const_optab
11357 exists. Use targetm.vectorize.vec_perm_const instead of
11358 targetm.vectorize.vec_perm_const_ok. Check whether the indices
11359 fit in the vector mode before using a variable permute.
11360 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
11361 vec_perm_indices instead of an rtx.
11362 (expand_vec_perm): Replace with...
11363 (expand_vec_perm_const): ...this new function. Take the selector
11364 as a vec_perm_indices rather than an rtx. Also take the mode of
11365 the selector. Update call to shift_amt_for_vec_perm_mask.
11366 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
11367 Use vec_perm_indices::new_expanded_vector to expand the original
11368 selector into bytes. Check whether the indices fit in the vector
11369 mode before using a variable permute.
11370 (expand_vec_perm_var): Make global.
11371 (expand_mult_highpart): Use expand_vec_perm_const.
11372 * fold-const.c: Includes vec-perm-indices.h.
11373 * tree-ssa-forwprop.c: Likewise.
11374 * tree-vect-data-refs.c: Likewise.
11375 * tree-vect-generic.c: Likewise.
11376 * tree-vect-loop.c: Likewise.
11377 * tree-vect-slp.c: Likewise.
11378 * tree-vect-stmts.c: Likewise.
11379 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
11381 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
11382 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
11383 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
11384 (aarch64_vectorize_vec_perm_const): ...this new function.
11385 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11386 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11387 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
11388 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
11389 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11390 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11391 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
11393 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
11394 check for NEON modes.
11395 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
11396 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
11397 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
11398 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
11400 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
11401 the old VEC_PERM_CONST conditions.
11402 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
11403 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
11404 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
11405 (ia64_vectorize_vec_perm_const_ok): Merge into...
11406 (ia64_vectorize_vec_perm_const): ...this new function.
11407 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
11408 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
11409 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
11410 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
11411 * config/mips/mips.c (mips_expand_vec_perm_const)
11412 (mips_vectorize_vec_perm_const_ok): Merge into...
11413 (mips_vectorize_vec_perm_const): ...this new function.
11414 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
11415 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
11416 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
11417 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
11418 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
11419 (rs6000_expand_vec_perm_const): Delete.
11420 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
11422 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11423 (altivec_expand_vec_perm_const_le): Take each operand individually.
11424 Operate on constant selectors rather than rtxes.
11425 (altivec_expand_vec_perm_const): Likewise. Update call to
11426 altivec_expand_vec_perm_const_le.
11427 (rs6000_expand_vec_perm_const): Delete.
11428 (rs6000_vectorize_vec_perm_const_ok): Delete.
11429 (rs6000_vectorize_vec_perm_const): New function.
11430 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11431 an element count and rtx array.
11432 (rs6000_expand_extract_even): Update call accordingly.
11433 (rs6000_expand_interleave): Likewise.
11434 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
11435 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
11436 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
11437 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
11438 (rs6000_expand_vec_perm_const): Delete.
11439 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11440 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11441 (altivec_expand_vec_perm_const_le): Take each operand individually.
11442 Operate on constant selectors rather than rtxes.
11443 (altivec_expand_vec_perm_const): Likewise. Update call to
11444 altivec_expand_vec_perm_const_le.
11445 (rs6000_expand_vec_perm_const): Delete.
11446 (rs6000_vectorize_vec_perm_const_ok): Delete.
11447 (rs6000_vectorize_vec_perm_const): New function. Remove stray
11448 reference to the SPE evmerge intructions.
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/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
11454 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
11456 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11458 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11460 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
11461 vector mode and that that mode matches the mode of the data
11463 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
11464 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
11465 directly using expand_vec_perm_1 when forcing selectors into
11467 (expand_vec_perm_var): New function, split out from expand_vec_perm.
11469 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11471 * optabs-query.h (can_vec_perm_p): Delete.
11472 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
11473 * optabs-query.c (can_vec_perm_p): Split into...
11474 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
11475 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
11476 particular selector is valid.
11477 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11478 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11479 (vect_grouped_load_supported): Likewise.
11480 (vect_shift_permute_load_chain): Likewise.
11481 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11482 (vect_transform_slp_perm_load): Likewise.
11483 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11484 (vectorizable_bswap): Likewise.
11485 (vect_gen_perm_mask_checked): Likewise.
11486 * fold-const.c (fold_ternary_loc): Likewise. Don't take
11487 implementations of variable permutation vectors into account
11488 when deciding which selector to use.
11489 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
11490 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
11491 with a false third argument.
11492 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
11493 to test whether the constant selector is valid and can_vec_perm_var_p
11494 to test whether a variable selector is valid.
11496 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11498 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
11499 * optabs-query.c (can_vec_perm_p): Likewise.
11500 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
11501 instead of vec_perm_indices.
11502 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
11503 (vect_gen_perm_mask_checked): Likewise,
11504 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
11505 (vect_gen_perm_mask_checked): Likewise,
11507 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11509 * optabs-query.h (qimode_for_vec_perm): Declare.
11510 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
11511 (qimode_for_vec_perm): ...this new function.
11512 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
11514 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11516 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
11517 does not have a conditional at the top.
11519 2018-01-02 Richard Biener <rguenther@suse.de>
11521 * ipa-inline.c (big_speedup_p): Fix expression.
11523 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11526 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
11529 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11533 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
11534 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
11535 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
11536 cond_taken_branch_cost 3->4.
11538 2018-01-01 Jakub Jelinek <jakub@redhat.com>
11540 PR tree-optimization/83581
11541 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
11542 TODO_cleanup_cfg if any changes have been made.
11544 PR middle-end/83608
11545 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
11546 convert_modes if target mode has the right side, but different mode
11549 PR middle-end/83609
11550 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
11551 last argument when extracting from CONCAT. If either from_real or
11552 from_imag is NULL, use expansion through memory. If result is not
11553 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
11554 the parts directly to inner mode, if even that fails, use expansion
11557 PR middle-end/83623
11558 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
11559 check for bswap in mode rather than HImode and use that in expand_unop
11562 Copyright (C) 2018 Free Software Foundation, Inc.
11564 Copying and distribution of this file, with or without modification,
11565 are permitted in any medium without royalty provided the copyright
11566 notice and this notice are preserved.