1 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
3 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
5 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
7 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
8 ("enabled" attribute): Handle "isa" attribute.
9 (*movdi_internal): Substitute Yi and Yj constraint with x
10 and Ym and Yn constraint with y constraint. Update "isa"
11 attribute and set "preferred_for_speed" attribute from
12 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
13 (*movsi_internal): Ditto.
14 (*movdf_internal): Ditto.
15 (*movsf_internal): Ditto.
16 (*zero_extendsidi2): Ditto.
17 * config/i386/sse.md (vec_set<mode>_0): Ditto.
19 (*vec_extract<ssevecmodelower>_0): Ditto.
20 (*vec_extractv4si_0_zext_sse4): Ditto.
21 (vec_concatv2di): Ditto.
22 (*vec_dup<mode>): Ditto.
23 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
24 * config/i386/constraints.md (Yi): Remove.
29 2018-04-26 Nathan Sidwell <nathan@acm.org>
31 * dumpfile.c (dump_open): New.
32 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
33 (dump_finish): Detect stdio/stderr by value not name.
35 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
37 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
39 2018-04-26 Tom de Vries <tom@codesourcery.com>
42 * config/nvptx/nvptx.c (verify_neutering_jumps)
43 (verify_neutering_labels): New function
44 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
46 2018-04-26 Tom de Vries <tom@codesourcery.com>
49 * config/nvptx/nvptx.c (needs_neutering_p): New function.
50 (nvptx_single): Use needs_neutering_p to skip over insns that do not
53 2018-04-26 Richard Biener <rguenther@suse.de>
54 Tom de Vries <tom@codesourcery.com>
57 * lto-streamer-out.c (output_function): Fixup loops if required to match
58 discovery done in the reader.
60 2018-04-26 Richard Biener <rguenther@suse.de>
62 PR tree-optimization/85116
63 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
64 have a loop exit from the single latch predecessor. Remove
65 case of header with just condition.
66 (ch_base::copy_headers): Exclude infinite loops from any
68 (pass_ch::execute): Record exits.
70 2018-04-26 Richard Biener <rguenther@suse.de>
72 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
73 prologue cost vector and pass it to vect_get_load_cost.
74 (vect_get_peeling_costs_all_drs): Likewise.
75 (vect_peeling_hash_get_lowest_cost): Likewise.
76 (vect_enhance_data_refs_alignment): Likewise.
78 2018-04-26 Richard Biener <rguenther@suse.de>
81 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
82 checking of integer<->pointer conversions.
83 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
84 sign-/zero-extending pointer types.
85 (expand_omp_for_static_chunk): Likewise.
87 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
88 Jean Lee <xiaoyur347@gmail.com>
90 * config/mips/mips.c (mips_asan_shadow_offset): New function.
91 (TARGET_ASAN_SHADOW_OFFSET): Define.
92 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
93 true for -fsanitize=address.
95 2018-04-25 Mark Wielaard <mark@klomp.org>
97 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
100 2018-04-25 Jakub Jelinek <jakub@redhat.com>
102 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
103 than "alu", remove explicit "memory" and "imm_disp" attributes.
104 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
107 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
108 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
111 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
114 * config/i386/i386.c (ix86_expand_builtin): Change memory
115 operand to XI, extend p0 to Pmode.
116 * config/i386/i386.md: Change unspec volatile and operand
117 1 mode to XI, change operand 0 mode to P.
119 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
121 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
122 GET_MODE_MASK before any checking.
123 (nds32_can_use_bset_p): Likewise.
124 (nds32_can_use_btgl_p): Likewise.
126 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
128 * config/nds32/nds32-doubleword.md: New define_split pattern for
129 illegal register number.
131 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
133 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
135 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
137 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
139 2018-04-25 Richard Biener <rguenther@suse.de>
141 * lto-streamer.h (LTO_major_version): Bump to 8.
143 2018-04-25 Jakub Jelinek <jakub@redhat.com>
145 * BASE-VER: Set to 9.0.0.
147 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
149 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
150 in __abskf2 and __powikf2.
152 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
155 * config/aarch64/constraints.md (Usg, Usj): New constraints.
156 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
157 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
158 Use the above on operand 2. Reindent.
159 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
161 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
164 * common/config/i386/i386-common.c (ix86_handle_option): Don't
166 * config/i386/i386.opt (mcet): Removed.
167 * doc/install.texi: Remove -mcet documentation.
168 * doc/invoke.texi: Likewise.
170 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
173 * doc/install.texi: Remove -mcet from bootstrap-cet.
175 2018-04-24 Jakub Jelinek <jakub@redhat.com>
178 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
179 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
183 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
184 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
185 containing a CONST_VECTOR.
187 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
189 * doc/install.texi: Update newlib dependency for nvptx.
191 2018-04-24 Jakub Jelinek <jakub@redhat.com>
194 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
195 instead of INTVAL when shifting x left.
197 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
199 PR tree-optimization/85478
200 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
201 vect_grouped_store_supported for single element vectors.
203 2018-04-24 Richard Biener <rguenther@suse.de>
206 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
207 load cost increase to the case of non-constant step.
209 2018-04-24 Jakub Jelinek <jakub@redhat.com>
212 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
213 destination if any_malformed_asm.
215 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
218 * expr.c (store_field): In the bitfield case, if the value comes from
219 a function call and is returned in registers by means of a PARALLEL,
220 do not change the mode of the temporary unless BLKmode and VOIDmode.
222 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
224 PR rtl-optimization/85423
225 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
226 dependencies to debug insns when the previous insn is non-debug.
228 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
230 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
231 enums into a single definition.
232 (fls): Fix predicates and printing.
235 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
237 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
238 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
239 and short u6 immediate.
240 (check_if_valid_sleep_operand): Remove.
241 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
243 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
245 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
246 flag_always_save_lp condition.
247 * config/nds32/nds32.opt (malways-save-lp): New option.
249 2018-04-22 Shiva Chen <shiva0217@gmail.com>
251 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
252 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
253 * config/nds32/nds32.h
254 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
255 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
257 2018-04-22 Shiva Chen <shiva0217@gmail.com>
259 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
260 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
262 2018-04-22 Shiva Chen <shiva0217@gmail.com>
263 Chung-Ju Wu <jasonwucj@gmail.com>
265 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
267 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
268 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
270 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
272 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
274 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
276 * config/nds32/nds32-protos.h (nds32_data_alignment,
277 nds32_local_alignment): Declare.
278 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
279 nds32_local_alignment): New functions.
280 (TARGET_CONSTANT_ALIGNMENT): Define.
281 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
283 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
285 * config/nds32/nds32.c
286 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
287 (TARGET_MODES_TIEABLE_P): Likewise.
289 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
291 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
294 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
295 Chung-Ju Wu <jasonwucj@gmail.com>
297 * config/nds32/constants.md (unspec_volatile_element): Add enum values
298 for unaligned access.
299 * config/nds32/nds32-intrinsic.c: Implementation of expanding
301 * config/nds32/nds32-intrinsic.md: Likewise.
302 * config/nds32/nds32_intrinsic.h: Likewise.
303 * config/nds32/nds32.h (nds32_builtins): Likewise.
304 * config/nds32/nds32.opt (munaligned-access): New option.
305 * config/nds32/nds32.c (nds32_asm_file_start): Display
306 flag_unaligned_access status.
308 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
310 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
311 -mno-relax is present.
312 * config/riscv/linux.h (LINK_SPEC): Ditto.
314 2018-04-20 Martin Sebor <msebor@redhat.com>
317 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
319 (gimple_fold_builtin_stxcpy_chk): Same.
320 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
322 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
325 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
326 __powikf2 when long double is IEEE 128-bit.
328 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
330 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
331 step to make sure stack always aligned.
333 2018-04-20 Carl Love <cel@us.ibm.com>
336 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
339 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
340 Tom de Vries <tom@codesourcery.com>
343 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
344 Emit insns for calls too.
345 (nvptx_find_par): Always look for worker-level predecessor insn.
346 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
348 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
349 (nvptx_process_pars): Propagate frames for calls.
351 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
354 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
356 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
357 (ix86_handle_option): Don't handle OPT_mibt.
358 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
360 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
362 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
363 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
364 (ix86_target_macros): Define __CET__ with flag_cf_protection
366 * config/i386/i386.c (isa2_opts): Remove -mibt.
367 * config/i386/i386.h (TARGET_IBT): Removed.
368 (TARGET_IBT_P): Likewise.
369 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
370 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
371 * config/i386/i386.opt (mcet): Update help message.
374 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
375 -mcet as an alias for -mshstk.
377 2018-04-20 Richard Biener <rguenther@suse.de>
380 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
381 complexity by forcing a single use of the multiply operand.
383 2018-04-20 Martin Jambor <mjambor@suse.cz>
386 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
387 recursion dependency to only apply to non-clones.
389 2018-04-20 Martin Jambor <mjambor@suse.cz>
392 * ipa-cp.c (create_specialized_node): Check that clones of
393 self-recursive edges exist during IPA-CP.
395 2018-04-19 Toon Moene <toon@moene.org>
397 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
400 2018-04-19 Jakub Jelinek <jakub@redhat.com>
402 PR tree-optimization/85467
403 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
404 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
405 VECTOR_CST element to type.
407 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
410 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
411 * config/i386/i386.md (builtin_setjmp_setup): Removed.
412 (builtin_longjmp): Likewise.
413 (save_stack_nonlocal): New pattern.
414 (restore_stack_nonlocal): Likewise.
416 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
419 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
420 Replace ASM_OUTPUT_LABEL with fprintf.
422 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
425 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
426 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
427 * config/i386/i386-c.c (ix86_target_macros_internal): Also
428 define __IBT__ and __SHSTK__ for -fcf-protection.
429 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
431 (ix86_trampoline_init): Likewise.
432 (x86_output_mi_thunk): Likewise.
433 (ix86_notrack_prefixed_insn_p): Likewise.
434 (ix86_option_override_internal): Don't disallow -fcf-protection.
435 * config/i386/i386.md (rdssp<mode>): Also enable for
437 (incssp<mode>): Likewise.
438 (nop_endbr): Likewise.
439 * config/i386/i386.opt (mcet): Change help message to built-in
443 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
444 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
445 enable CET built-in functions.
447 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
449 * common/config/i386/i386-common.c
450 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
451 OPTION_MASK_ISA_MOVDIRI_UNSET,
452 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
453 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
454 * config.gcc (movdirintrin.h): New header.
455 * config/i386/cpuid.h (bit_MOVDIRI,
456 bit_MOVDIR64B): New bits.
457 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
459 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
460 (VOID, PVOID, PCVOID)): New function types.
461 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
462 __builtin_ia32_directstoreu_u64,
463 __builtin_ia32_movdir64b): New builtins.
464 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
465 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
467 (ix86_valid_target_attribute_inner_p): Ditto.
468 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
469 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
470 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
471 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
472 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
473 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
474 (movdiri<mode>, movdir64b_<mode>): New.
475 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
476 * config/i386/immintrin.h: Include movdirintrin.h.
477 * config/i386/movdirintrin.h: New file.
478 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
480 2018-04-19 Richard Biener <rguenther@suse.de>
483 * cfg.c (clear_bb_flags): When loop state says we have
484 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
486 2018-04-19 Richard Biener <rguenther@suse.de>
488 PR tree-optimization/84737
489 * tree-vect-data-refs.c (vect_copy_ref_info): New function
490 copying restrict info.
491 (vect_setup_realignment): Use it.
492 * tree-vectorizer.h (vect_copy_ref_info): Declare.
493 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
494 the first DR to all generated stores.
495 (vectorizable_load): Likewise for loads.
497 2018-04-19 Jakub Jelinek <jakub@redhat.com>
499 PR tree-optimization/85446
500 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
501 the integral and pointer types to have the same precision.
503 * doc/install.texi: Document --disable-cet being the default and
506 2018-04-18 Martin Liska <mliska@suse.cz>
508 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
511 2018-04-18 Martin Liska <mliska@suse.cz>
514 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
517 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
518 arguments if they are comparable.
520 2018-04-18 Martin Liska <mliska@suse.cz>
523 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
526 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
529 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
532 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
533 ENDBR after calling __morestack.
535 2018-04-18 David Malcolm <dmalcolm@redhat.com>
538 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
539 by using gcc_base_ver to generate a gcc_driver_version, and use
540 it when generating GCC_DRIVER_NAME.
541 * configure: Regenerate.
543 2018-04-18 Jakub Jelinek <jakub@redhat.com>
546 * config.gcc: Obsolete powerpc*-*-*spe*.
548 2018-04-17 Jakub Jelinek <jakub@redhat.com>
551 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
552 (stabstr_D): Change type of unum from unsigned int to
553 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
556 2018-04-17 Jim Wilson <jimw@sifive.com>
559 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
560 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
561 Set arg_pointer_offset after using pretend_args_size.
563 2018-04-17 Jakub Jelinek <jakub@redhat.com>
565 PR rtl-optimization/85431
566 * dse.c (record_store): Ignore zero width stores.
569 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
570 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
571 __builtin_stack_restore rather than after it.
572 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
573 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
574 argument instead of virtual_dynamic_stack_rtx.
576 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
578 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
580 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
581 Add note to error message to explain internal mapping of overloaded
582 built-in function name to non-overloaded built-in function name.
583 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
586 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
589 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
590 where the inputs overlap with the output.
592 2018-04-17 Jakub Jelinek <jakub@redhat.com>
594 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
595 (=v, v) alternative and explicit "memory" attribute.
596 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
597 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
599 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
600 "sselog1" type instead of "sselog".
601 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
602 "sselog". Remove explicit "memory" attribute.
603 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
604 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
606 (vec_extract_hi_v32hi): Merge all alternatives into one, use
607 "sselog1" type instead of "sselog". Remove explicit "memory"
609 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
610 use "sselog1" type instead of "sselog". Remove explicit "memory"
612 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
613 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
615 (vec_extract_hi_v64qi): Merge all alternatives into one, use
616 "sselog1" type instead of "sselog". Remove explicit "memory"
618 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
619 use "sselog1" type instead of "sselog". Remove explicit "memory"
623 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
626 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
629 2018-04-17 Martin Jambor <mjambor@suse.cz>
632 * ipa-cp.c (create_specialized_node): Call
633 expand_all_artificial_thunks if necessary.
635 2018-04-17 Martin Liska <mliska@suse.cz>
638 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
639 in message, remote space in between '_G' and '('.
641 2018-04-17 Jakub Jelinek <jakub@redhat.com>
644 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
645 avx512f_vmcmp<mode>3<round_saeonly_name>,
646 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
647 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
648 avx512f_rndscale<mode><round_saeonly_name>,
649 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
650 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
651 Use %<iptr>2 instead of %2 for -masm=intel.
652 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
653 avx512f_vcvttss2usi<round_saeonly_name>,
654 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
656 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
657 avx512f_vcvttsd2usi<round_saeonly_name>,
658 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
659 Use %q1 instead of %1 for -masm=intel.
660 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
661 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
662 of %3 for -masm=intel.
663 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
665 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
667 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
669 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
670 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
672 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
674 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
675 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
676 %g1 and one with %0 and %1.
677 (avx512er_vmrcp28<mode><round_saeonly_name>,
678 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
680 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
681 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
682 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
683 of %0 and %{%4%} for -masm=intel.
684 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
685 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
686 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
687 order of %0 and %{%5%}%{z%} for -masm=intel.
689 2018-04-17 Jan Hubicka <jh@suse.cz>
692 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
694 2018-04-17 Martin Liska <mliska@suse.cz>
697 * multiple_target.c (create_dispatcher_calls): Set apostrophes
698 for target_clone error message. Make default implementation
699 clone to be a local declaration.
700 (separate_attrs): Add new argument and check for an empty
702 (expand_target_clones): Handle it.
703 (ipa_target_clone): Make redirection just for target_clones
706 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
707 Tom de Vries <tom@codesourcery.com>
710 * omp-expand.c (expand_oacc_for): Add dummy false branch for
711 tiled basic blocks without omp continue statements.
713 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
716 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
717 vec_extract expression as having side effects to make sure it gets
720 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
723 * config/i386/i386.c (get_builtin_code_for_version): Check
726 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
729 * gcc/config.gcc: Support "skylake".
730 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
732 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
733 (processor_target_table): Add "skylake".
734 (ix86_option_override_internal): Add "skylake".
735 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
736 PROCESSOR_CANNONLAKE.
737 (get_builtin_code_for_version): Fix priority for
738 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
739 PROCESSOR_SKYLAKE-AVX512.
740 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
741 (processor_type): Add PROCESSOR_SKYLAKE.
743 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
744 Jason Merrill <jason@redhat.com>
747 * convert.c (convert_to_integer_1): Use direct recursion for
748 enumeral types and types with a precision less than the number
749 of bits in their mode.
751 2018-04-16 Julia Koval <julia.koval@intel.com>
754 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
755 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
757 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
760 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
761 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
762 and -mno-direct-move.
764 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
767 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
768 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
769 Ensure negative shifts result in {0}.
771 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
773 PR rtl-optimization/79916
774 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
775 regs (if any) to define how to gnerate SD moves when LRA is in
778 2018-04-13 Jakub Jelinek <jakub@redhat.com>
780 PR rtl-optimization/85393
781 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
782 * except.c (expand_dw2_landing_pad_for_region): Make static.
783 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
784 a label and unconditional jump to old_bb, rather than
785 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
788 PR rtl-optimization/85376
789 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
790 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
791 instead of a specific value.
793 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
794 Bin Cheng <bin.cheng@arm.com>
796 PR tree-optimization/82965
797 PR tree-optimization/83991
798 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
799 by_profile_only parameter.
800 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
801 information if the loop was predicted to iterate too many times.
802 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
804 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
807 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
810 2018-04-13 Martin Liska <mliska@suse.cz>
811 Jakub Jelinek <jakub@redhat.com>
814 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
815 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
816 * builtins.c (expand_builtin_memory_copy_args): Use
817 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
818 handle dest_addr == pc_rtx.
820 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
823 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
824 asked to not generate direct moves.
825 (fix_trunc<mode>si2_stfiwx): Similar.
826 (fix_trunc<mode>si2_internal): Similar.
828 2018-04-12 Jakub Jelinek <jakub@redhat.com>
831 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
832 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
833 lookup if dest in some wider mode is known to be const0_rtx and
834 if so, record permanent equivalence for it to be ZERO_EXTEND of
835 the narrower mode destination.
837 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
839 * lto-streamer-out.c (output_function): Revert 259346.
840 * omp-expand.c (expand_oacc_for): Likewise.
842 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
844 PR rtl-optimization/85354
845 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
846 * sel-sched.c (sel_global_init): ... here.
848 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
851 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
852 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
853 mode for PE-COFF targets.
854 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
855 (i386_pe_asm_lto_end): Likewise.
856 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
857 (TARGET_ASM_LTO_END): Likewise.
858 * config/i386/winnt.c (saved_debug_info_level): New static variable.
859 (i386_pe_asm_lto_start): New function.
860 (i386_pe_asm_lto_end): Likewise.
862 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
863 Richard Biener <rguenther@suse.de>
866 * lto-streamer-out.c (output_function): Fix CFG loop state before
868 * omp-expand.c (expand_oacc_for): Handle calls to internal
869 functions like regular functions.
871 2018-04-12 Richard Biener <rguenther@suse.de>
874 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
875 for the early LTO debug to properly generate references to it
876 during DIE emission. Do not re-use that for the skeleton for
878 (dwarf2out_early_finish): Likewise.
880 2018-04-12 Jakub Jelinek <jakub@redhat.com>
884 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
885 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
886 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
887 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
888 and output is a reg, avoid creating invalid lowpart subreg, but
889 instead split into a 512-bit move. Don't split if not AVX512VL,
890 input is xmm16+ reg and output is a mem.
891 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
892 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
893 xmm16+ reg and output is a mem.
895 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
897 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
898 also for flag_dwarf2_cfi_asm.
900 2018-04-12 Jakub Jelinek <jakub@redhat.com>
902 PR rtl-optimization/85342
903 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
904 a bool scalar var inside of the loop instead. Don't try to update
905 recog_data.operand after failed apply_change_group.
907 2018-04-12 Tom de Vries <tom@codesourcery.com>
910 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
911 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
912 array with flexible array member as array without given dimension.
913 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
914 argument for undefined param to true.
916 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
919 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
920 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
921 from PowerPC section.
922 * config/rs6000/sysv4.opt (mcall-): Improve help text.
923 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
924 help text that is too long.
925 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
926 help text that is too long.
927 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
928 help text that is too long.
930 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
932 * config/alpha/alpha.md (stack_probe_internal): Rename
933 from "probe_stack". Update all callers.
935 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
937 PR rtl-optimization/84566
938 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
939 sched_macro_fuse_insns.
941 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
944 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
945 (compute_block_dependences): ... from here.
947 2018-04-11 Jakub Jelinek <jakub@redhat.com>
949 PR tree-optimization/85331
950 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
951 from int to HOST_WIDE_INT.
953 2018-04-11 Martin Jambor <mjambor@suse.cz>
956 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
957 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
958 not the same as the source val.
959 (cgraph_edge_brings_value_p): New parameter.
960 (gather_edges_for_value): Pass destination value to
961 cgraph_edge_brings_value_p.
962 (perhaps_add_new_callers): Likewise.
963 (get_info_about_necessary_edges): Likewise and exclude values brought
964 only by self-recursive edges.
965 (create_specialized_node): Redirect only clones of self-calling edges.
966 (+self_recursive_pass_through_p): New function.
967 (find_more_scalar_values_for_callers_subset): Use it.
968 (find_aggregate_values_for_callers_subset): Likewise.
969 (known_aggs_to_agg_replacement_list): Removed.
970 (decide_whether_version_node): Re-calculate known constants for all
971 remaining context clones.
973 2018-04-11 Richard Biener <rguenther@suse.de>
976 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
977 from early DWARF output.
978 (dwarf2out_early_finish): Output line info unconditionally into
979 early DWARF and add reference to it.
981 2018-04-11 Jakub Jelinek <jakub@redhat.com>
984 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
985 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
986 other than V2DFmode using iptr mode attribute.
987 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
989 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
991 PR rtl-optimization/84659
992 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
994 2018-04-11 Jakub Jelinek <jakub@redhat.com>
997 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
999 (output_loc_list): Pass address of a dummy size variable even in the
1000 locview handling loop.
1001 (index_location_lists): Add comment on why skip_loc_list_entry can't
1004 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
1007 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
1010 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
1013 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1014 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
1015 and -mstring-compare-inline-limit.
1017 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1020 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
1021 for stack clash protection in a register whenever we need it to be in
1024 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1026 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
1027 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
1029 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1032 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
1034 (mlong-double-): Ditto.
1035 * config/rs6000/sysv4.opt (msdata=): Ditto.
1036 (mtls-size=): Ditto.
1038 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
1040 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1041 erroneous entries for
1042 "vector int vec_ldl (int, long int *)", and
1043 "vector unsigned int vec_ldl (int, unsigned long int *)".
1044 Add comments and entries for
1045 "vector bool char vec_ldl (int, bool char *)",
1046 "vector bool short vec_ldl (int, bool short *)",
1047 "vector bool int vec_ldl (int, bool int *)",
1048 "vector bool long long vec_ldl (int, bool long long *)",
1049 "vector pixel vec_ldl (int, pixel *)",
1050 "vector long long vec_ldl (int, long long *)",
1051 "vector unsigned long long vec_ldl (int, unsigned long long *)".
1052 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
1053 type tree bool_long_long_type_node and correct definition of
1054 bool_V2DI_type_node to make reference to this new type tree.
1055 (rs6000_mangle_type): Replace erroneous reference to
1056 bool_long_type_node with bool_long_long_type_node.
1057 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
1058 comments to emphasize sign distinctions for char and int types and
1059 replace RS6000_BTI_bool_long constant with
1060 RS6000_BTI_bool_long_long constant. Also add comment to restrict
1061 use of RS6000_BTI_pixel.
1062 (bool_long_type_node): Remove this macro definition.
1063 (bool_long_long_type_node): New macro definition
1065 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1067 PR rtl-optimization/85300
1068 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
1069 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
1070 simplify_unary_operation fails.
1072 2018-04-10 Martin Liska <mliska@suse.cz>
1074 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
1075 cgraph_edge and ipa_ref.
1077 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1081 * config/i386/sse.md
1082 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
1083 computation of the VEC_MERGE selector from mask.
1084 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
1085 Fix decoding of the VEC_MERGE selector into mask.
1087 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1089 PR tree-optimization/85286
1090 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
1092 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1094 * final.c (final_1): Set insn_last_address as well as
1095 insn_current_address.
1097 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1100 * explow.c (emit_stack_probe): Call validize_mem on memory location
1101 before passing it to gen_probe_stack. Create address operand and
1102 legitimize it for the probe_stack_address case.
1104 2018-04-09 Jan Hubicka <jh@suse.cz>
1107 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
1108 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
1109 * tree.c (free_lang_data_in_type): Fix handling of binfos;
1111 (free_lang_data): Rebuild type inheritance graph.
1113 2018-04-09 Martin Sebor <msebor@redhat.com>
1115 * invoke.texi (-finline-small-functions): Mention other optimization
1117 (-findirect-inlining, -fpartial-inlining): Same.
1118 (-finline-functions-called-once): Same.
1119 (-freorder-blocks-and-partition): Same.
1121 2018-04-09 Jan Hubicka <jh@suse.cz>
1124 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
1125 jumps; choose last target that matches the criteria (i.e.
1126 no partition changes for non-crossing jumps).
1127 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
1128 support for redirecting crossing jumps to non-crossing.
1130 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
1132 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
1133 also for naked functions.
1135 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
1137 * config/arc/arc.md (add_shift): New pattern.
1138 (add_shift2): Likewise.
1139 (sub_shift): Likewise.
1140 (sub_shift_cmp0_noout): Likewise.
1141 (compare_si_ashiftsi): Likewise.
1142 (xbfu_cmp0_noout): New combine pattern.
1143 (xbfu_cmp0"): Likewise.
1144 (movsi_set_cc_insn): Place the predicable variant first.
1145 (commutative_binary_cmp0_noout): Remove clobber.
1146 (commutative_binary_cmp0): New pattern.
1147 (noncommutative_binary_cmp0): Likewise.
1148 (noncommutative_binary_cmp0_noout): Likewise.
1149 (noncommutative_binary_comparison_result_used): Removed.
1150 (rsub_cmp0): New pattern.
1151 (rsub_cmp0_noout): Likewise.
1152 (extzvsi): Changed, keep only meaningful variants.
1153 (SQH, SEZ): New iterators.
1154 (SQH_postfix): New mode attribute.
1155 (SEZ_prefix): New code attribute.
1156 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
1157 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
1158 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
1160 (noncommutative_operator): Check the availability of barrel
1163 2018-04-09 Richard Biener <rguenther@suse.de>
1165 PR tree-optimization/85284
1166 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
1167 Only use the niter constraining form of simple_iv when the exit
1170 2018-04-09 Tom de Vries <tom@codesourcery.com>
1173 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1174 (define_expand "*memory_barrier"): New define_expand.
1175 (define_insn "memory_barrier"): New insn.
1177 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1179 PR rtl-optimization/80463
1180 PR rtl-optimization/83972
1181 PR rtl-optimization/83480
1183 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
1184 correct producer for the insn.
1185 (tidy_control_flow): Fixup seqnos in case of debug insns.
1187 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1189 PR rtl-optimization/83913
1191 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
1192 different sched-times when merging exprs.
1194 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1196 PR rtl-optimization/83962
1198 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
1199 tidy_fallthru_edge and tidy_control_flow.
1201 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1203 PR rtl-optimization/83530
1205 * sel-sched.c (force_next_insn): New global variable.
1206 (remove_insn_for_debug): When force_next_insn is true, also leave only
1207 next insn in the ready list.
1208 (sel_sched_region): When the region wasn't scheduled, make another pass
1209 over it with force_next_insn set to 1.
1211 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1213 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
1215 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1216 for interrupt control.
1217 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1218 functions for interrupt control.
1219 * config/nds32/nds32-intrinsic.md: Likewise.
1220 * config/nds32/nds32_intrinsic.h: Likewise.
1221 * config/nds32/nds32.h (nds32_builtins): Likewise.
1223 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1225 * config/nds32/nds32.c (nds32_init_machine_status,
1226 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
1227 strict_aligned_p field.
1228 (nds32_expand_to_rtl_hook): New function.
1229 (TARGET_EXPAND_TO_RTL_HOOK): Define.
1230 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
1232 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1233 Chung-Ju Wu <jasonwucj@gmail.com>
1235 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
1236 * config/nds32/nds32-n7.md: New file.
1237 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
1238 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
1240 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
1241 * config/nds32/nds32.md (pipeline_model): Add n7.
1242 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
1243 * config/nds32/pipelines.md: Include n7 settings.
1245 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1246 Chung-Ju Wu <jasonwucj@gmail.com>
1248 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
1249 * config/nds32/nds32-e8.md: New file.
1250 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
1251 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
1253 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
1254 * config/nds32/nds32.md (pipeline_model): Add e8.
1255 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
1256 * config/nds32/pipelines.md: Include e8 settings.
1258 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1259 Chung-Ju Wu <jasonwucj@gmail.com>
1261 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
1262 * config/nds32/nds32-n8.md: New file.
1263 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
1264 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
1266 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
1267 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
1268 * config/nds32/nds32.md (pipeline_model): Add n8.
1269 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
1270 * config/nds32/pipelines.md: Include n8 settings.
1272 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1273 Chung-Ju Wu <jasonwucj@gmail.com>
1275 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
1276 * config/nds32/nds32-n9-2r1w.md: New file.
1277 * config/nds32/nds32-n9-3r2w.md: New file.
1278 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
1279 nds32_register_ports): New or modify for cpu n9.
1280 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
1282 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
1283 * config/nds32/nds32-utils.c: New file.
1284 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
1285 TARGET_MUL_SLOW): Define.
1286 * config/nds32/nds32.md (pipeline_model): New attribute.
1287 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
1288 New options that support cpu n9.
1289 * config/nds32/pipelines.md: Include n9 settings.
1290 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
1292 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1294 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
1295 information if necessary.
1296 (output_cond_branch_compare_zero): Likewise.
1297 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
1298 (nds32_target_alignment): Refine for alignment.
1299 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
1300 (FUNCTION_BOUNDARY): Modify.
1301 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
1303 * config/nds32/nds32.opt (malways-align, malign-functions): New.
1305 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1307 * config/nds32/constants.md (unspec_volatile_element): Add values for
1308 TLB operation and data prefetch.
1309 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1310 functions for TLB operation and data prefetch.
1311 * config/nds32/nds32-intrinsic.md: Likewise.
1312 * config/nds32/nds32_intrinsic.h: Likewise.
1313 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
1314 (nds32_print_operand): Likewise.
1315 * config/nds32/nds32.h (nds32_builtins): Likewise.
1317 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
1318 Andrew Pinski <pinsika@gcc.gnu.org>
1321 * match.pd: Use constant_boolean_node of correct type instead of
1322 boolean_true_node or boolean_false_node for simplifying
1323 pointer comparisons to zero.
1325 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1327 PR tree-optimization/80021
1328 * tree.c (verify_type_variant): Make error call in verify_variant_match
1329 translatable and remove final full stop.
1331 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1333 * config/nds32/constants.md (unspec_volatile_element): Add
1334 UNSPEC_VOLATILE_EH_RETURN.
1335 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
1336 nds32_output_stack_pop): Support dwarf exception handling process.
1337 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
1338 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
1339 exception handling process.
1340 (nds32_compute_stack_frame): Likewise.
1341 (nds32_return_addr_rtx): Likewise.
1342 (nds32_initial_elimination_offset): Likewise.
1343 (nds32_expand_prologue): Likewise.
1344 (nds32_expand_epilogue): Likewise.
1345 (nds32_dynamic_chain_address): New function.
1346 * config/nds32/nds32.h (machine_function): Add fields for dwarf
1348 (DYNAMIC_CHAIN_ADDRESS): Define.
1349 (EH_RETURN_DATA_REGNO): Define.
1350 (EH_RETURN_STACKADJ_RTX): Define.
1351 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
1352 patterns for dwarf exception handling.
1354 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1356 * config/nds32/nds32.h: Clean up obsolete macros.
1358 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1360 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1361 Add enum values for particular instructions.
1362 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1363 particular intrinsic functions.
1364 * config/nds32/nds32-intrinsic.md: Likewise.
1365 * config/nds32/nds32_intrinsic.h: Likewise.
1366 * config/nds32/nds32.h (nds32_builtins): Likewise.
1367 * config/nds32/nds32.md (type): Add pbsad and pbsada.
1368 (btst, ave): New patterns for particular instructions.
1370 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1372 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1373 Add enum values for atomic load/store and memory sync.
1374 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
1376 * config/nds32/nds32-intrinsic.md: Likewise.
1377 * config/nds32/nds32_intrinsic.h: Likewise.
1378 * config/nds32/nds32.h (nds32_builtins): Likewise.
1380 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1382 PR tree-optimization/85257
1383 * fold-const.c (native_encode_vector): If not all elts could fit
1384 and off is -1, return 0 rather than offset.
1385 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
1386 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
1387 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
1388 adjust buffer in native_interpret_expr call.
1390 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1392 * config/nds32/constants.md (unspec_volatile_element): Add cache
1393 control enum values.
1394 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
1395 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
1396 * config/nds32/nds32.c (nds32_cctl_names): New.
1397 (nds32_print_operand): Handle cache control register names.
1398 * config/nds32/nds32.h (nds32_builtins): New enum values.
1399 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
1401 * config/nds32/nds32.md (type): Add mmu.
1402 * config/nds32/pipelines.md (simple_insn): Add mmu.
1404 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1406 * config/nds32/nds32.md (type): Remove call.
1407 * config/nds32/pipelines.md (simple_insn): Likewise.
1409 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1411 * config/nds32/constants.md (unspec_volatile_element): Add
1412 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
1413 UNSPEC_VOLATILE_FMFCFG.
1414 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
1415 description for fmfcfg and fmfcsr.
1416 (bdesc_1arg): Add fmtcsr.
1417 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
1418 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
1419 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
1420 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
1421 unspec_fmfcfg): New patterns.
1422 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
1423 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
1424 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
1425 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
1426 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
1427 __nds32__fmfcfg): Define.
1429 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1431 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
1432 intrinsic register names.
1433 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
1434 intrinsic register enum values and macros.
1436 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1438 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
1439 for load/store addressing form.
1440 (nds32_print_operand_address): Likewise.
1442 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
1445 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
1446 based on LABEL_REF. Remove useless assertion.
1447 (pic_address_needs_scratch): Fix formatting.
1448 (sparc_legitimize_pic_address): Minor tweaks.
1449 (sparc_delegitimize_address): Adjust assertion accordingly.
1450 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
1451 into symbolic_operand.
1452 (movsi_high_pic_label_ref): Likewise.
1453 (movsi_lo_sum_pic_label_ref): Likewise.
1454 (movdi_pic_label_ref): Likewise.
1455 (movdi_high_pic_label_ref): Likewise.
1456 (movdi_lo_sum_pic_label_ref): Likewise.
1458 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
1460 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
1461 custom LIB_SPEC setup.
1463 2018-04-06 Ruslan Bukin <br@bsdpad.com>
1464 Kito Cheng <kito.cheng@gmail.com>
1466 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
1467 * config/riscv/freebsd.h: New.
1469 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1471 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
1472 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
1475 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1476 Kito Cheng <kito.cheng@gmail.com>
1478 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
1479 nds32_output_call, nds32_symbol_binds_local_p): New functions.
1480 * config/nds32/nds32-protos.h (nds32_output_call,
1481 nds32_output_return): Declare.
1482 * config/nds32/nds32.md: Refine all the call and return patterns.
1484 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1487 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
1488 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
1490 PR rtl-optimization/84872
1491 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
1492 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
1495 2018-04-06 Tamar Christina <tamar.christina@arm.com>
1497 * expr.c (copy_blkmode_to_reg): Revert 254862.
1498 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
1500 2018-04-06 Richard Biener <rguenther@suse.de>
1503 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
1504 after seeing a component reference with an adjacent field. Treat
1505 refs to arrays at struct end of external decls similar to
1506 refs to unconstrained commons.
1508 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1511 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
1512 look through SAVE_EXPRs with non-side-effects argument. Adjust
1514 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
1517 2018-04-06 Richard Biener <rguenther@suse.de>
1520 * alias.c (find_base_term): New wrapper around find_base_term
1521 unwinding CSELIB_VAL_PTR changes.
1522 (find_base_term): Do not restore CSELIB_VAL_PTR during the
1525 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1527 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
1529 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
1530 constant definitions.
1531 ("nop"): lr 0,0 -> nopr r0
1532 ("nop_lr0", "nop_lr1"): New insn definitions.
1534 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1536 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
1537 NDS32_V3PUSH_AVAILABLE_P macro.
1539 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
1540 Chung-Ju Wu <jasonwucj@gmail.com>
1542 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
1543 (nds32*-*-*): Add float and fpu_config into supported_defaults.
1544 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
1545 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
1546 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
1547 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
1548 * config/nds32/constraints.md: New constraints and checking for hard
1549 float configuration.
1550 * config/nds32/iterators.md: New mode iterator and attribute for hard
1551 float configuration.
1552 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
1554 * config/nds32/nds32-fpu.md: New file.
1555 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
1556 deal with hard float code generation.
1557 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
1559 (abi_type, float_reg_number): New enum type.
1560 * config/nds32/nds32-predicates.c: New predicates for hard float.
1561 * config/nds32/nds32-protos.h: Declare functions for hard float.
1562 * config/nds32/nds32.c: Implementation for hard float configuration.
1563 * config/nds32/nds32.h: Definitions for hard float configuration.
1564 * config/nds32/nds32.md: Include hard float machine description and
1565 modify patterns for hard float configuration.
1566 * config/nds32/nds32.opt: New options for hard float configuration.
1567 * config/nds32/predicates.md: New predicates for hard float
1570 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1572 * common/config/nds32/nds32-common.c
1573 (nds32_option_optimization_table): Enable -mreleax-hint by default.
1575 2018-04-05 Jakub Jelinek <jakub@redhat.com>
1578 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
1579 CONSTRUCTOR_ELT (ctor, ...)->value.
1581 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
1584 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
1586 2018-04-05 Tom de Vries <tom@codesourcery.com>
1589 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
1592 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1593 Kito Cheng <kito.cheng@gmail.com>
1595 * config/nds32/constraints.md (U33): Fine-tune checking condition.
1596 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
1597 * config/nds32/nds32.h (nds32_16bit_address_type): Add
1598 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
1600 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1601 Kito Cheng <kito.cheng@gmail.com>
1603 * config/nds32/constraints.md (Ufe): New memory constraint.
1604 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
1605 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
1606 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
1608 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
1609 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
1611 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1613 * config/nds32/nds32.md: Use optimize_size in the condition for
1614 alu-shift instructions.
1616 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1618 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
1620 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1622 * config/nds32/nds32.md (negsi2): Refine pattern.
1624 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1625 Chung-Ju Wu <jasonwucj@gmail.com>
1627 * config/nds32/iterators.md (shift_rotate): New code iterator.
1628 (shift): New code attribute.
1629 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
1630 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
1631 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
1632 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
1633 bit-wise operations.
1634 (andsi3, *andsi3): Ditto.
1635 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
1636 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
1637 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
1638 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
1639 nds32_ior_operand, nds32_xor_operand): New predicates.
1641 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1643 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
1644 (addsi3, subsi3): ... this.
1646 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1648 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
1650 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1652 * config/nds32/nds32.md: Adjust indention.
1654 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1656 * config/nds32/nds32.md (feature): New attribute.
1658 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1660 * config/nds32/nds32.md (subtype): New attribute.
1662 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
1665 * config/arm/arm-builtins.c (arm_expand_builtin): Change
1666 expansion to perform a bitwise AND of the argument followed by a
1667 boolean negation of the result.
1669 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
1671 PR rtl-optimization/84878
1672 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
1673 the basic block. Assert the use reference is not artificial and that
1674 it has an associated insn.
1676 2018-04-04 Michael Matz <matz@suse.de>
1678 * builtins.c (compute_objsize): Pass correct operand
1679 to array_at_struct_end_p.
1681 2018-04-04 Richard Biener <rguenther@suse.de>
1684 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
1685 from contexts for DINFO_LEVEL_TERSE and below.
1687 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1689 * config/nds32/nds32-doubleword.md (move_<mode>): Require
1690 resiter_operand condition.
1691 * config/nds32/nds32.md (*move<mode>): Ditto.
1693 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1694 Monk Chiang <sh.chiang04@gmail.com>
1696 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
1698 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1700 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
1702 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1703 Kito Cheng <kito.cheng@gmail.com>
1705 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
1706 nds32_cond_code_str, output_cond_branch,
1707 output_cond_branch_compare_zero, nds32_expand_cbranch,
1708 nds32_expand_cstore, nds32_expand_movcc,
1709 nds32_output_cbranchsi4_equality_zero,
1710 nds32_output_cbranchsi4_equality_reg,
1711 nds32_output_cbranchsi4_equality_reg_or_const_int,
1712 nds32_output_cbranchsi4_greater_less_zero: New functions.
1713 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
1714 nds32_expand_cstore, nds32_expand_movcc,
1715 nds32_output_cbranchsi4_equality_zero,
1716 nds32_output_cbranchsi4_equality_reg,
1717 nds32_output_cbranchsi4_equality_reg_or_const_int,
1718 nds32_output_cbranchsi4_greater_less_zero): Declare.
1719 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
1720 nds32_rimm11s_operand): New predicates.
1721 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
1722 * config/nds32/nds32.md: Rewrite all the branch and conditional move
1725 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1727 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
1728 * config/nds32/nds32.md: Ditto.
1729 * config/nds32/pipelines.md: Ditto.
1731 2018-04-04 Richard Biener <rguenther@suse.de>
1733 PR tree-optimization/85168
1734 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
1735 propagating abnormals.
1737 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1739 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
1741 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1742 Kito Cheng <kito.cheng@gmail.com>
1744 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
1745 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
1746 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
1747 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
1748 * config/nds32/nds32.md (sibcall_internal): New.
1749 (sibcall_register): Remove.
1750 (sibcall_immediate): Remove.
1751 (sibcall_value_internal): New.
1752 (sibcall_value_register): Remove.
1753 (sibcall_value_immediate): Remove.
1754 * config/nds32/predicates.md (nds32_general_register_operand): New.
1755 (nds32_call_address_operand): New.
1757 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1759 PR rtl-optimization/85167
1760 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
1761 bb_defs if *split_p, instead preinitialize it to NULL.
1763 PR tree-optimization/85156
1764 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
1765 evaluating the argument multiple times.
1767 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
1769 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
1771 (_mm_cvtpd_ps): Likewise.
1772 (_mm_cvttpd_epi32): Likewise.
1773 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
1774 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
1775 vector, pixel, and bool following altivec.h include.
1777 2018-04-03 Martin Sebor <msebor@redhat.com>
1779 * doc/extend.texi (Common Function Attributes): Clarify.
1780 (const attribute): Likewise.
1781 (pure attribute): Likewise.
1783 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1786 * config/i386/i386.c (ix86_expand_vector_set): Use
1787 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
1789 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
1791 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
1792 instructions when changing rounding bits to preserve precision bits
1793 in the x87 control word.
1795 2018-04-03 Martin Liska <mliska@suse.cz>
1797 PR tree-optimization/82491
1798 * rtl.h (strip_offset_and_add): Replace += suboffset with
1799 poly_uint64 () + suboffset.
1801 2018-03-29 Martin Liska <mliska@suse.cz>
1802 Martin Jambor <mjambor@suse.cz>
1805 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
1806 param_type is not an integral or pointer type.
1808 2018-04-03 Richard Biener <rguenther@suse.de>
1810 * sese.h (recompute_all_dominators): Remove.
1812 2018-04-02 Martin Sebor <msebor@redhat.com>
1814 * doc/invoke.texi (-Wrestrict): Fix typos.
1816 2018-04-02 Jim Wilson <jimw@sifive.com>
1818 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
1819 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
1820 (<optab>di3, <optab>si3_extend): Likewise.
1821 (<optab>si3_mask, <optab>si3_mask_1): New.
1822 (<optab>di3_mask, <optab>di3_mask_1): New.
1823 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
1824 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
1825 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
1827 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
1829 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
1832 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
1834 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
1835 (nds32_canonicalize_comparison): New function.
1837 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1838 Kito Cheng <kito.cheng@gmail.com>
1839 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1841 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
1842 * config/nds32/constants.md (unspec_volatile_element): Add
1843 UNSPEC_VOLATILE_RELAX_GROUP.
1844 * config/nds32/nds32-relax-opt.c: New file.
1845 * config/nds32/nds32-predicates.c
1846 (nds32_symbol_load_store_p): New function.
1847 * config/nds32/nds32-protos.h
1848 (nds32_symbol_load_store_p): Declare function.
1849 (make_pass_nds32_relax_opt): Declare new rtl pass function.
1850 * config/nds32/nds32.c
1851 (nds32_register_pass): New function to register pass.
1852 (nds32_register_passes): New function to register passes.
1853 * config/nds32/nds32.md (relax_group): New pattern.
1854 * config/nds32/nds32.opt (mrelax-hint): New option.
1855 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
1857 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
1859 * config/nds32/t-nds32: Modify files dependency.
1861 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1863 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
1864 (PROFILE_HOOK): Define its implementation.
1866 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1868 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
1869 type and 32-bit size.
1871 2018-04-01 Jakub Jelinek <jakub@redhat.com>
1874 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
1875 (V_128_256): New mode iterator.
1876 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
1877 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
1878 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
1880 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
1881 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
1883 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
1886 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
1887 NaN inputs correctly.
1889 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
1892 * config/rs6000/vsx.md (??r): New mode attribute.
1893 (*vsx_mov<mode>_64bit): Use it.
1894 (*vsx_mov<mode>_32bit): Likewise.
1896 2018-03-30 Martin Sebor <msebor@redhat.com>
1898 PR tree-optimization/84818
1899 * builtins.c (check_access): Use warning_n.
1901 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1904 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
1906 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
1909 2018-03-30 Julia Koval <julia.koval@intel.com>
1912 * x86-tune.def (movx, partial_reg_dependency): Enable for
1915 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
1918 * lra-constraints.c (process_alt_operands): Move setting
1919 this_alternative_matches below.
1921 2018-03-29 Martin Liska <mliska@suse.cz>
1924 * doc/invoke.texi: Document how LTO works with debug info.
1925 Describe auto-load support of binutils. Mention 'x86-64'
1926 as valid option value of -march option.
1928 2018-03-29 Jakub Jelinek <jakub@redhat.com>
1930 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
1933 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
1934 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
1935 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
1938 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1941 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
1942 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
1943 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
1944 for RS6000_BTM_POWERPC64.
1945 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
1946 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
1947 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
1952 2018-03-28 Carl Love <cel@us.ibm.com>
1955 2017-09-27 Carl Love <cel@us.ibm.com>
1957 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
1958 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
1959 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
1962 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1964 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
1965 instead of __vector bool.
1966 (_mm_max_pu8): Likewise.
1967 (_mm_min_pi16): Likewise.
1969 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1972 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
1973 (DIVWEUO): Likewise.
1975 (DIVDEUO): Likewise.
1976 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
1977 DIVWEUO and DIVDEUO.
1978 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
1979 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
1980 (div_extend): Likewise.
1981 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
1983 (__builtin_divweuo): Likewise.
1984 (__builtin_divdeo): Likewise.
1985 (__builtin_divdeuo): Likewise.
1987 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1990 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
1991 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
1993 PR tree-optimization/82004
1994 * gimple-match-head.c (optimize_pow_to_exp): New function.
1995 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
1996 Don't fold to exp if optimize_pow_to_exp is false.
1998 2018-03-28 Martin Liska <mliska@suse.cz>
2001 * calls.c (initialize_argument_information): Fix trailing space.
2002 * common.opt: Fix typo and provide better explanation for
2003 -fsanitize-coverage option.
2004 * config/i386/i386.opt: Fix typo.
2006 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2007 Martin Liska <mliska@suse.cz>
2010 * gimplify.c (asan_poison_variable): Don't do the check for
2011 gimplify_omp_ctxp here.
2012 (gimplify_decl_expr): Do it here.
2013 (gimplify_target_expr): Likewise.
2015 2018-03-28 Martin Liska <mliska@suse.cz>
2018 * config/i386/i386.c (ix86_function_arg_advance): Do not call
2019 chkp_type_bounds_count if MPX is not enabled.
2021 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
2023 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
2025 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
2028 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
2029 function to create the function decl for complex long double
2030 multiply and divide for -mabi=ieeelongdouble.
2031 (init_float128_ieee): Call it.
2033 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
2036 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
2037 -fcf-protection=branch -mibt.
2038 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
2040 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2043 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
2045 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
2048 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
2049 extern array declarations.
2051 2018-03-27 Richard Biener <rguenther@suse.de>
2054 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
2055 explicit single_use checks.
2057 2018-03-27 Richard Biener <rguenther@suse.de>
2059 PR tree-optimization/85082
2060 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
2063 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2065 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
2066 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
2067 Turn on fasynchronous-unwind-tables and funwind-tables.
2069 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
2072 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
2073 (*bmi_blsr_<mode>_ccz): Ditto.
2075 2018-03-26 Tom de Vries <tom@codesourcery.com>
2077 PR tree-optimization/85063
2078 * omp-general.c (offloading_function_p): New function. Factor out
2080 * omp-offload.c (pass_omp_target_link::gate): ... here.
2081 * omp-general.h (offloading_function_p): Declare.
2082 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
2083 with attribute omp declare target for offloading functions.
2085 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
2087 PR tree-optimization/84005
2088 * tree-data-ref.h (get_base_for_alignment): Declare.
2089 * tree-data-ref.c (get_base_for_alignment_1): New function.
2090 (get_base_for_alignment): Likewise.
2091 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
2092 get_base_for_alignment to find a suitable base object, instead
2093 of always using drb->base_address.
2095 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2098 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
2099 known size by default.
2101 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
2104 * lra-constraints.c (process_alt_operands): Don't match BLKmode
2105 and non BLKmode operands.
2107 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2110 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
2111 Clean up attributes.
2113 2018-03-23 Richard Biener <rguenther@suse.de>
2116 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
2117 we are going to emit early debug for LTO.
2119 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2122 * function.c (match_asm_constraints_1): Don't optimize if input
2123 doesn't satisfy general_operand predicate for output's mode.
2126 * alias.c (write_dependence_p): Don't require for x_canonicalized
2127 non-VOIDmode if x has VOIDmode.
2130 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
2131 just don't try to optimize it rather than assert it never happens.
2133 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2135 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
2136 macro expansions for definition of ST_INTERNAL_<mode> and
2137 LD_INTERNAL_<mode> builtins.
2138 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
2140 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
2142 (altivec_expand_st_builtin): Likewise.
2143 (altivec_expand_builtin): Remove calls to deleted functions.
2144 (rs6000_address_for_altivec): Delete this function.
2145 * config/rs6000/vector.md: Remove expands for
2146 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
2148 2018-03-22 Sudakshina Das <sudi.das@arm.com>
2151 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
2152 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
2153 re-computing once computed.
2154 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
2155 (arm_init_machine_status): Initialize
2156 machine->static_chain_stack_bytes.
2158 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2161 * doc/extend.texi: Add four new prototypes for vec_ld.
2162 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
2163 definitions for more logical presentation.
2164 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
2165 entries for V1TI variants of __builtin_altivec_ld builtin.
2166 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
2167 handling of V1TI variant of LVX icode pattern.
2168 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
2169 (rs6000_gimple_fold_builtin): Likewise.
2170 (altivec_init_builtins): Add code to define
2171 __builtin_altivec_lvx_v1ti function.
2173 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2176 * function.c (match_asm_constraints_1): Don't do the optimization
2177 if input isn't a REG, SUBREG, MEM or constant.
2179 2018-03-22 Tom de Vries <tom@codesourcery.com>
2181 PR tree-optimization/84956
2182 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
2183 bb_has_abnormal_pred.
2185 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2188 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
2189 DECL_INITIAL (decl) to decl at the end.
2190 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
2193 2018-03-21 Joseph Myers <joseph@codesourcery.com>
2195 * doc/extend.texi (__builtin_tgmath): Document when complex
2196 integer types are treated as _Complex _Float64.
2198 2018-03-21 Tom de Vries <tom@codesourcery.com>
2200 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
2202 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2204 PR tree-optimization/84960
2205 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
2206 if it is ENTRY block, move them into single succ of ENTRY in that case.
2208 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
2210 PR tree-optimization/84811
2211 * poly-int.h (poly_span_traits): Remove the T3 parameter and
2212 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
2213 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
2214 (known_subrange_p): Update accordingly. Cast each value involved
2215 in the size comparison, rather than casting the result of the
2218 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2220 PR tree-optimization/84982
2221 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
2222 by flipping the least significant bit rather than all bits from
2223 bitpos to bitpos + bitsize - 1.
2225 2018-03-21 Nathan Sidwell <nathan@acm.org>
2227 * doc/extend.texi (Deprecated Features): Remove mention of
2228 long-deleted deprecations.
2230 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2233 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
2234 * configure: Regenerate.
2236 2018-03-21 Tom de Vries <tom@codesourcery.com>
2238 PR tree-optimization/83126
2239 * tree-parloops.c (num_phis): New function.
2240 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
2242 2018-03-21 Nathan Sidwell <nathan@acm.org>
2244 * doc/extend.texi (Deprecated Features): Update deprecated flags,
2245 mention anon-struct/union members and trailing attributes.
2247 2018-03-21 Bin Cheng <bin.cheng@arm.com>
2249 PR tree-optimization/84969
2250 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
2251 builtin memset partitions if they set different rhs values.
2253 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2255 PR rtl-optimization/84989
2256 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
2257 VEC_DUPLICATE with scalar result mode.
2259 2018-03-21 Martin Liska <mliska@suse.cz>
2262 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
2263 not intended return statement.
2265 2018-03-21 Martin Liska <mliska@suse.cz>
2268 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
2269 (chkp_find_bound_slots_1): Limit number of iterations.
2271 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
2274 * Minor grammar fixes for x86 options.
2276 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2279 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
2280 holding REG_CFA_RESTORE notes, instead turn them into a USE.
2282 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
2285 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
2286 (altivec_lvx_<mode>_1op): Likewise.
2287 (altivec_stvx_<mode>_2op): Likewise.
2288 (altivec_stvx_<mode>_1op): Likewise.
2289 (altivec_lvx_<VM2:mode>): New define_expand.
2290 (altivec_stvx_<VM2:mode>): Likewise.
2291 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
2292 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2293 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
2294 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2295 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
2296 (rs6000_gen_lvx): Likewise.
2297 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
2298 (altivec_expand_stv_builtin): Likewise.
2299 (altivec_expand_builtin): Likewise.
2300 * config/rs6000/vector.md: Likewise.
2302 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2305 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
2308 2018-03-20 Richard Biener <rguenther@suse.de>
2311 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
2312 sign-conversions as zero, fall back to standard scalar_stmt
2315 2018-03-20 Martin Liska <mliska@suse.cz>
2318 * predict.c (rebuild_frequencies): Handle case when we have
2319 PROFILE_ABSENT, but flag_guess_branch_prob is false.
2321 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2324 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
2325 flag_section_anchors.
2326 * varasm.c (use_blocks_for_decl_p): Remove hack for
2327 dw2_force_const_mem.
2330 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
2332 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
2333 be created, use lowpart_subreg of operands[0] rather than operands[0]
2335 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
2336 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
2337 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
2338 and n constraint instead of aarch64_shift_imm_di and Usd.
2339 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
2340 (*aarch64_<optab>_reg_minus<mode>3): ... this.
2342 2018-03-20 Sudakshina Das <sudi.das@arm.com>
2345 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
2346 to favor GPR over NEON registers.
2347 (<shift>di3_neon): Likewise.
2349 2018-03-20 Tom de Vries <tom@codesourcery.com>
2352 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
2353 (nvptx_process_pars): Emit bar.sync asap and alap.
2355 2018-03-20 Tom de Vries <tom@codesourcery.com>
2358 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
2359 seen_label if seen_label is already set.
2361 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2364 * config/i386/i386.c (fold_builtin_cpu): For features above 31
2365 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
2366 Use 1U instead of 1. Formatting fixes.
2369 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
2370 instead of TREE_TYPE (s1) for the return value.
2372 2018-03-19 Jakub Jelinek <jakub@redhat.com>
2374 PR tree-optimization/84946
2375 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
2376 bitsize + bitsize in poly_uint64 rather than poly_int64.
2379 * dwarf2asm.c: Include fold-const.c.
2380 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
2381 of decl rather than decl itself.
2383 PR rtl-optimization/84643
2384 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
2386 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
2389 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
2390 calling assemble_variable.
2392 2018-03-19 Sudakshina Das <sudi.das@arm.com>
2395 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
2396 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
2398 2018-03-19 Jim Wilson <jimw@sifive.com>
2401 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
2402 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
2403 (riscv_first_stack_step): Likewise.
2404 (riscv_option_override): Use STACK_BOUNDARY instead of
2406 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
2408 (BIGGEST_ALIGNMENT): Set to 128.
2409 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
2410 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
2413 2018-03-19 Richard Biener <rguenther@suse.de>
2415 PR tree-optimization/84933
2416 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
2417 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
2419 2018-03-19 Richard Biener <rguenther@suse.de>
2421 PR tree-optimization/84859
2422 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
2423 (cond_if_else_store_replacement): Perform sinking operation on
2424 single-store BBs regardless of MAX_STORES_TO_SINK setting.
2425 Generalize what a BB with a single eligible store is.
2427 2018-03-19 Richard Biener <rguenther@suse.de>
2429 PR tree-optimization/84929
2430 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
2431 chrec_is_positive against non-chrec arg.
2433 2018-03-19 Tamar Christina <tamar.christina@arm.com>
2436 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
2438 2018-03-18 Martin Liska <mliska@suse.cz>
2440 PR rtl-optimization/84635
2441 * regrename.c (build_def_use): Use matches_mode only when
2444 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
2446 PR tree-optimization/84913
2447 * tree-vect-loop.c (vectorizable_reduction): Don't try to
2448 vectorize chains of COND_EXPRs.
2450 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2452 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
2454 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2456 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
2458 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2460 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
2462 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2463 Kito Cheng <kito.cheng@gmail.com>
2465 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
2466 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
2467 (nds32_adjust_reg_alloc_order): New function.
2468 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
2470 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
2472 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
2473 nds32_print_operand, nds32_print_operand_address): Use
2474 HOST_WIDE_INT_PRINT_DEC instead.
2476 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2478 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
2480 2018-03-17 Jakub Jelinek <jakub@redhat.com>
2483 * config/i386/i386.c (initial_ix86_tune_features,
2484 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
2486 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
2487 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
2488 rather than 1u << ix86_tune. Formatting fix.
2489 (ix86_option_override_internal): Change ix86_arch_mask from
2490 unsigned int to unsigned HOST_WIDE_INT, initialize to
2491 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
2492 (ix86_function_specific_restore): Likewise.
2494 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2497 * postreload.c (reload_combine_recognize_pattern): Perform
2498 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
2499 truncate_int_for_mode the result for the destination's mode.
2502 * hsa-gen.c (mem_type_for_type): Fix comment typo.
2503 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
2505 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2508 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
2511 * lra-assigns.c (lra_split_hard_reg_for): Don't use
2512 regno_allocno_class_array and sorted_pseudos.
2513 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
2514 insns where regno is used.
2516 2018-03-16 Martin Liska <mliska@suse.cz>
2519 * multiple_target.c (create_dispatcher_calls): Redirect
2520 reference in the symbol table.
2522 2018-03-16 Martin Liska <mliska@suse.cz>
2525 * multiple_target.c (create_dispatcher_calls): Redirect also
2528 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2532 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
2533 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
2534 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
2536 2018-03-16 Julia Koval <julia.koval@intel.com>
2538 * doc/invoke.texi (Skylake Server): Add CLWB.
2539 Cannonlake): Remove CLWB.
2541 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2543 PR tree-optimization/84841
2544 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
2546 (FLOAT_ONE_CONST_TYPE): Define.
2547 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
2548 (sort_by_operand_rank): Put entries with higher constant_type last
2549 rather than first to match comments.
2551 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
2553 * config/nios2/nios2.md (movsi_internal): Fix thinko in
2556 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2559 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
2560 check and use address of target always.
2562 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
2565 * config/i386/i386.c (indirect_thunk_needed): Update comments.
2566 (indirect_thunk_bnd_needed): Likewise.
2567 (indirect_thunks_used): Likewise.
2568 (indirect_thunks_bnd_used): Likewise.
2569 (indirect_return_needed): New.
2570 (indirect_return_bnd_needed): Likewise.
2571 (output_indirect_thunk_function): Add a bool argument for
2573 (output_indirect_thunk_function): Don't generate alias for
2574 function return thunk.
2575 (ix86_code_end): Call output_indirect_thunk_function to generate
2576 function return thunks.
2577 (ix86_output_function_return): Set indirect_return_bnd_needed
2578 and indirect_return_needed instead of indirect_thunk_bnd_needed
2579 and indirect_thunk_needed.
2581 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
2583 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
2584 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
2585 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
2587 2018-03-15 David Malcolm <dmalcolm@redhat.com>
2588 Paul Hua <paul.hua.gm@gmail.com>
2591 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
2593 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
2595 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
2596 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
2599 2018-03-15 Tamar Christina <tamar.christina@arm.com>
2602 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
2603 instead of GET_MODE_SIZE when comparing Units.
2605 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2608 * varasm.c (hash_section): Return an unchangeble hash value
2609 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
2610 Return !aarch64_can_use_per_function_literal_pools_p ().
2612 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2615 * optabs.c (emit_conditional_move): Pass address of cmode's copy
2616 rather than address of cmode as last argument to prepare_cmp_insn.
2618 2018-03-15 Julia Koval <julia.koval@intel.com>
2620 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
2621 F_AVX512VNNI, F_AVX512BITALG): New.
2623 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
2626 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
2627 insn for floating-point loads and stores.
2629 2018-03-14 Carl Love <cel@us.ibm.com>
2631 * config/rs6000/rs6000-c.c: Add macro definitions for
2632 ALTIVEC_BUILTIN_VEC_PERMXOR.
2633 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
2634 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
2635 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
2636 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
2638 * config/doc/extend.texi: Add prototypes for vec_permxor.
2640 2018-03-14 David Malcolm <dmalcolm@redhat.com>
2643 * diagnostic-show-locus.c (class layout_point): Convert m_line
2644 from int to linenum_type.
2645 (line_span::comparator): Use linenum "compare" function when
2646 comparing line numbers.
2647 (test_line_span): New function.
2648 (layout_range::contains_point): Convert param "row" from int to
2650 (layout_range::intersects_line_p): Likewise.
2651 (layout::will_show_line_p): Likewise.
2652 (layout::print_source_line): Likewise.
2653 (layout::should_print_annotation_line_p): Likewise.
2654 (layout::print_annotation_line): Likewise.
2655 (layout::print_leading_fixits): Likewise.
2656 (layout::annotation_line_showed_range_p): Likewise.
2657 (struct line_corrections): Likewise for field m_row.
2658 (line_corrections::line_corrections): Likewise for param "row".
2659 (layout::print_trailing_fixits): Likewise.
2660 (layout::get_state_at_point): Likewise.
2661 (layout::get_x_bound_for_row): Likewise.
2662 (layout::print_line): Likewise.
2663 (diagnostic_show_locus): Likewise for locals "last_line" and
2665 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
2666 * input.c (selftest::test_linenum_comparisons): New function.
2667 (selftest::input_c_tests): Call it.
2668 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
2669 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
2670 * selftest.h (ASSERT_GT): New macro.
2671 (ASSERT_GT_AT): New macro.
2672 (ASSERT_LT): New macro.
2673 (ASSERT_LT_AT): New macro.
2675 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
2677 PR rtl-optimization/84780
2678 * combine.c (distribute_links): Don't make a link based on pc_rtx.
2680 2018-03-14 Martin Liska <mliska@suse.cz>
2682 * tree.c (record_node_allocation_statistics): Use
2683 get_stats_node_kind.
2684 (get_stats_node_kind): New function extracted from
2685 record_node_allocation_statistics.
2686 (free_node): Use get_stats_node_kind.
2688 2018-03-14 Richard Biener <rguenther@suse.de>
2690 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
2691 that the value-set of ANTIC_IN doesn't grow.
2694 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2696 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2697 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2699 2018-03-14 Julia Koval <julia.koval@intel.com>
2701 * config.gcc (icelake-client, icelake-server): New.
2703 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
2704 (initial_ix86_arch_features): Ditto.
2705 (PTA_SKYLAKE): Add SGX.
2706 (PTA_ICELAKE): Remove.
2707 (PTA_ICELAKE_CLIENT): New.
2708 (PTA_ICELAKE_SERVER): New.
2709 (ix86_option_override_internal): Split up icelake on icelake client and
2711 (get_builtin_code_for_version): Ditto.
2712 (fold_builtin_cpu): Ditto.
2713 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
2714 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
2715 * config/i386/i386.h (processor_type): Ditto.
2716 * doc/invoke.texi: Ditto.
2718 2018-03-14 Jakub Jelinek <jakub@redhat.com>
2721 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
2722 INTEGER_CST offset, add it together with bitpos / 8 and
2723 sign extend based on POINTER_SIZE.
2727 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
2730 * config/i386/constraints.md (Yc): New register constraint.
2731 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
2732 Use Yc constraint for alternative 2 of operand 0. Remove
2733 preferred_for_speed attribute.
2735 2018-03-14 Richard Biener <rguenther@suse.de>
2737 PR tree-optimization/84830
2738 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
2739 with the old one to avoid oscillations.
2741 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
2744 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
2746 (assign_by_spills): Return a flag of reload assignment failure.
2747 Do not process the reload assignment failures. Do not spill other
2748 reload pseudos if they has the same reg class. Update n if
2750 (lra_assign): Add a return arg. Set up from the result of
2751 assign_by_spills call.
2752 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2753 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2754 usage_insns if it is not NULL.
2755 (spill_hard_reg_in_range): New function.
2756 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2757 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2758 function prototypes.
2759 (lra_assign): Change prototype.
2760 * lra.c (lra): Add code to deal with fails by splitting hard reg
2763 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
2765 * config/riscv/riscv.opt (mrelax): New option.
2766 * config/riscv/riscv.c (riscv_file_start): Emit ".option
2767 "norelax" when riscv_mrelax is disabled.
2768 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
2770 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2773 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
2774 reassociation for int modes.
2776 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2778 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
2779 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
2781 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
2782 * config/aarch64/aarch64-sve.md
2783 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
2784 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
2785 (*extend<mode><Vwide>2): Rename to...
2786 (aarch64_sve_extend<mode><Vwide>2): ...this.
2787 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
2788 renaming the old pattern to...
2789 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
2791 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
2792 define_expand, renaming the old pattern to...
2793 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
2794 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
2795 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
2796 account when deciding which SVE instruction the optab should use.
2797 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
2799 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2801 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
2802 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
2803 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
2804 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
2805 (tlsdesc_small_<mode>): Turn a define_expand and use
2806 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
2807 (tlsdesc_small_advsimd_<mode>): ...this.
2808 (tlsdesc_small_sve_<mode>): New pattern.
2810 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2812 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
2813 (UNSPEC_UMUL_HIGHPART): New constants.
2814 (MUL_HIGHPART): New int iteraor.
2815 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
2816 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
2818 (*<su>mul<mode>3_highpart): New define_insn.
2820 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2823 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2826 2018-03-13 Martin Liska <mliska@suse.cz>
2829 * (sem_item_optimizer::sem_item_optimizer): Initialize new
2831 (sem_item_optimizer::~sem_item_optimizer): Release it.
2832 (sem_item_optimizer::merge_classes): Register variable aliases.
2833 (sem_item_optimizer::fixup_pt_set): New function.
2834 (sem_item_optimizer::fixup_points_to_sets): Likewise.
2835 * ipa-icf.h: Declare new variables and functions.
2837 2018-03-13 Jakub Jelinek <jakub@redhat.com>
2840 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
2841 integer_pow2p@2 and test integer_pow2p in condition.
2842 (A < 0 ? C : 0): Similarly for @1.
2845 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
2846 characters starting at p contain '\0' character, don't look beyond
2850 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
2851 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
2854 * reg-stack.c (change_stack): Change update_end var from int to
2855 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
2856 also call set_block_for_insn on the newly added insns and rescan.
2859 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
2860 on the last operand.
2863 * tree.c (stabilize_reference_1): Return save_expr (e) for
2864 STATEMENT_LIST even if it doesn't have side-effects.
2866 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
2868 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
2870 2018-03-12 Renlin Li <renlin.li@arm.com>
2872 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
2873 aarch64_output_scalar_simd_mov_immediate.
2875 2018-03-12 Martin Sebor <msebor@redhat.com>
2877 PR tree-optimization/83456
2878 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
2879 for perfectly overlapping calls to memcpy.
2880 (gimple_fold_builtin_memory_chk): Same.
2881 (gimple_fold_builtin_strcpy): Handle no-warning.
2882 (gimple_fold_builtin_stxcpy_chk): Same.
2883 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
2885 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2887 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
2888 parameter. Use it for SFmode.
2889 (rs6000_function_arg_advance_1): Adjust.
2890 (rs6000_function_arg): Adjust.
2891 (rs6000_gimplify_va_arg): Pass false for that new parameter.
2893 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2895 PR rtl-optimization/84169
2896 PR rtl-optimization/84780
2897 * combine.c (can_combine_p): Check for a 2-insn combination whether
2898 the destination register is used between the two insns, too.
2900 2018-03-12 Richard Biener <rguenther@suse.de>
2902 PR tree-optimization/84803
2903 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
2904 for refs DR analysis didn't process.
2906 2018-03-12 Richard Biener <rguenther@suse.de>
2908 PR tree-optimization/84777
2909 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
2910 force-vectorize loops ignore whether we are optimizing for size.
2912 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
2914 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
2915 (TARGET_MD_ASM_ADJUST): Define.
2917 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
2918 Kito Cheng <kito.cheng@gmail.com>
2919 Chung-Ju Wu <jasonwucj@gmail.com>
2921 * config/nds32/nds32.c (nds32_compute_stack_frame,
2922 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
2923 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
2924 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
2925 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
2926 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
2927 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
2928 * config/nds32/nds32.md (prologue, epilogue): Use macro
2929 NDS32_V3PUSH_AVAILABLE_P to do checking.
2931 2018-03-11 Jakub Jelinek <jakub@redhat.com>
2934 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
2935 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
2936 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
2937 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
2938 addition of most attributes on !orig_type_die or the attribute not
2939 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
2941 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2942 Chung-Ju Wu <jasonwucj@gmail.com>
2944 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
2946 * config/nds32/nds32.opt (mvh): New option.
2948 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2949 Chung-Ju Wu <jasonwucj@gmail.com>
2951 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
2953 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
2954 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
2957 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2958 Chung-Ju Wu <jasonwucj@gmail.com>
2960 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
2962 * config/nds32/nds32-multiple.md (strlensi): New pattern.
2963 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
2965 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
2966 Kito Cheng <kito.cheng@gmail.com>
2967 Chung-Ju Wu <jasonwucj@gmail.com>
2969 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
2970 UNSPEC_FFMISM and UNSPEC_FLMISM.
2971 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
2972 for ffb, ffmism and flmism.
2973 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
2974 (unspec_ffmism): Ditto.
2975 (unspec_flmism): Ditto.
2976 (nds32_expand_builtin_impl): Check if string extension is available.
2977 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
2978 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
2980 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
2983 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2986 * lra-assigns.c (assign_by_spills): Return a flag of reload
2987 assignment failure. Do not process the reload assignment
2988 failures. Do not spill other reload pseudos if they has the same
2990 (lra_assign): Add a return arg. Set up from the result of
2991 assign_by_spills call.
2992 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2993 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2994 usage_insns if it is not NULL.
2995 (spill_hard_reg_in_range): New function.
2996 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2997 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2998 function prototypes.
2999 (lra_assign): Change prototype.
3000 * lra.c (lra): Add code to deal with fails by splitting hard reg
3003 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
3006 * config/i386/i386.opt: Replace Enforcment with Enforcement.
3008 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
3011 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
3012 (dw_val_node): Add val_symbolic_view.
3013 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
3014 (symview_upper_bound): New.
3015 (new_line_info_table): Initialize symviews_since_reset.
3016 (dwarf2out_source_line): Count symviews_since_reset and set
3017 symview_upper_bound.
3018 (dw_val_equal_p): Handle symview.
3019 (add_AT_symview): New.
3020 (print_dw_val): Handle symview.
3021 (attr_checksum, attr_checksum_ordered): Likewise.
3022 (same_dw_val_p, size_of_die): Likewise.
3023 (value_format, output_die): Likewise.
3024 (add_high_low_attributes): Use add_AT_symview for entry_view.
3025 (dwarf2out_finish): Reset symview_upper_bound, clear
3028 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
3031 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
3032 Add strict argument and use it.
3033 (rs6000_split_multireg_move): Update for new strict argument.
3034 (mem_operand_gpr): Disallow all non-offsettable addresses.
3035 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
3037 2018-03-09 Jakub Jelinek <jakub@redhat.com>
3040 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
3041 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
3042 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
3045 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
3046 decl, use remap_type if we want to use the type.
3048 2018-03-09 Martin Sebor <msebor@redhat.com>
3050 PR tree-optimization/84526
3051 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3053 (builtin_access::generic_overlap): Be prepared to handle non-array
3056 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
3058 PR rtl-optimization/84682
3059 * lra-constraints.c (process_address_1): Check is_address flag
3060 for address constraints.
3061 (process_alt_operands): Likewise.
3062 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
3063 preprocess_constraints.
3064 * recog.h (preprocess_constraints): Add oploc parameter.
3066 * recog.c (preprocess_constraints): Test address_operand for
3067 CT_ADDRESS constraints.
3069 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3072 * lra-assigns.c (assign_by_spills): Return a flag of reload
3073 assignment failure. Do not process the reload assignment
3074 failures. Do not spill other reload pseudos if they has the same
3076 (lra_assign): Add a return arg. Set up from the result of
3077 assign_by_spills call.
3078 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3079 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3080 usage_insns if it is not NULL.
3081 (spill_hard_reg_in_range): New function.
3082 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3083 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3084 function prototypes.
3085 (lra_assign): Change prototype.
3086 * lra.c (lra): Add code to deal with fails by splitting hard reg
3089 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3092 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
3093 Accept complain bool parameter. Only emit errors if it is true.
3094 (arm_parse_cpu_option_name): Likewise.
3095 (arm_target_thumb_only): Adjust callers of the above.
3096 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
3097 prototype to take a default true bool parameter.
3098 (arm_parse_arch_option_name): Likewise.
3100 2018-03-09 David Malcolm <dmalcolm@redhat.com>
3101 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
3105 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
3106 * configure: Regenerate.
3107 * configure.ac ("linker --version-script option"): New.
3108 ("linker soname option"): New.
3110 2018-03-09 Richard Biener <rguenther@suse.de>
3112 PR tree-optimization/84775
3113 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
3114 immediate uses of predicate stmts and mark them modified.
3117 PR tree-optimization/84178
3118 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3120 (version_loop_for_if_conversion): Delay update_ssa call.
3121 (tree_if_conversion): Delay update_ssa until after predicate
3124 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
3127 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
3128 when the function accesses prior frames.
3130 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3133 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
3134 gen_llsym, otherwise call maybe_gen_llsym.
3137 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
3138 has ',' character inside of it.
3140 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3143 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
3144 as clobbering CC_REGNUM.
3146 2018-03-08 Richard Biener <rguenther@suse.de>
3149 * tree-scalar-evolution.c: Include tree-into-ssa.h.
3150 (follow_copies_to_constant): Do not follow SSA names registered
3153 2018-03-08 Richard Biener <rguenther@suse.de>
3155 PR tree-optimization/84178
3156 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3158 (version_loop_for_if_conversion): Delay update_ssa call.
3159 (tree_if_conversion): Delay update_ssa until after predicate
3162 2018-03-08 David Malcolm <dmalcolm@redhat.com>
3164 PR tree-optimization/84178
3165 * tree-if-conv.c (release_bb_predicate): Remove the
3166 the assertion that the stmts have NULL use_ops.
3167 Discard the statements, asserting that they haven't
3168 yet been added to a BB.
3170 2018-03-08 Richard Biener <rguenther@suse.de>
3172 PR tree-optimization/84746
3173 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
3174 (phi_translate): Pass in destination ANTIC_OUT set.
3175 (phi_translate_1): Likewise. For a simplified result lookup
3176 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
3177 (phi_translate_set): Adjust.
3178 (do_pre_regular_insertion): Likewise.
3179 (do_pre_partial_partial_insertion): Likewise.
3181 2018-03-08 Martin Liska <mliska@suse.cz>
3183 PR gcov-profile/84735
3184 * doc/gcov.texi: Document usage of profile files.
3185 * gcov-io.h: Document changes in the format.
3187 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
3191 * dwarf2out.c (struct dw_line_info_table): Update comments for
3193 (FORCE_RESET_NEXT_VIEW): New.
3194 (FORCE_RESETTING_VIEW_P): New.
3195 (RESETTING_VIEW_P): Check for -1 too.
3196 (ZERO_VIEW_P): Likewise.
3197 (new_line_info_table): Force-reset next view.
3198 (dwarf2out_begin_function): Likewise.
3199 (dwarf2out_source_line): Simplify zero_view_p initialization.
3200 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
3201 view directly. Omit view when omitting .loc at line 0.
3203 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3205 PR tree-optimization/84740
3206 * tree-switch-conversion.c (process_switch): Call build_constructors
3207 only if info.phi_count is non-zero.
3209 PR tree-optimization/84739
3210 * tree-tailcall.c (find_tail_calls): Check call arguments against
3211 DECL_ARGUMENTS (current_function_decl) rather than
3212 DECL_ARGUMENTS (func) when checking for tail recursion.
3214 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3216 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
3217 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
3218 Volker Reichelt's entry and add entries for people that perform
3219 GCC fuzzy testing and report numerous bugs.
3221 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
3224 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
3225 readonly data in sdata, if that is disabled.
3226 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
3227 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3228 -mreadonly-in-sdata option.
3230 2018-03-07 Martin Sebor <msebor@redhat.com>
3232 PR tree-optimization/84468
3233 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
3234 basic block when looking for nul assignment.
3236 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
3239 * except.h (output_function_exception_table): Adjust prototype.
3240 * except.c (output_function_exception_table): Remove FNNAME parameter
3241 and add SECTION parameter. Ouput one part of the table at a time.
3242 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
3243 the first part of the exception table and emit unwind directives.
3244 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
3245 (i386_pe_seh_cold_init): Likewise.
3246 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
3247 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
3248 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
3249 (ix86_output_call_insn): Emit a nop in one more case for SEH.
3250 * config/i386/winnt.c: Include except.h.
3251 (struct seh_frame_state): Add reg_offset, after_prologue and
3252 in_cold_section fields.
3253 (i386_pe_seh_end_prologue): Set seh->after_prologue.
3254 (i386_pe_seh_cold_init): New function.
3255 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
3256 to seh->in_cold_section.
3257 (seh_emit_push): Record the offset of the push.
3258 (seh_emit_save): Record the offet of the save.
3259 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
3260 Test seh->after_prologue to disregard the epilogue.
3261 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
3262 (i386_pe_end_cold_function): New function.
3264 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3267 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
3268 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
3271 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
3273 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
3274 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
3277 * multiple_target.c: Include tree-inline.h and intl.h.
3278 (expand_target_clones): Diagnose and fail if node->definition and
3279 !tree_versionable_function_p (node->decl).
3281 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
3283 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
3285 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
3286 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
3287 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
3289 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3292 * combine.c (try_combine): Use reg_or_subregno instead of handling
3293 just paradoxical SUBREGs and REGs.
3295 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
3297 * config/arc/arc.c (arc_finalize_pic): Remove function.
3298 (arc_must_save_register): We use single base PIC register, remove
3299 checks to save/restore the PIC register.
3300 (arc_expand_prologue): Likewise.
3301 * config/arc/arc-protos.h (arc_set_default_type_attributes):
3303 (arc_verify_short): Likewise.
3304 (arc_attr_type): Likewise.
3305 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
3306 (walk_stores): Likewise.
3307 (arc_address_cost): Make it static.
3308 (arc_verify_short): Likewise.
3309 (branch_dest): Likewise.
3310 (arc_attr_type): Likewise.
3311 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
3312 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
3313 (arc_final_prescan_insn): Remove inserting the nops due to
3314 hardware hazards. It is done in reorg step.
3315 (insn_length_variant_t): Remove.
3316 (insn_length_parameters_t): Likewise.
3317 (arc_insn_length_parameters): Likewise.
3318 (arc_get_insn_variants): Likewise.
3319 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
3321 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3324 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
3327 PR tree-optimization/84687
3328 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
3330 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
3332 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3334 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
3335 Rename to ppc_speculation_barrier.
3336 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
3337 __builtin_ppc_speculation_barrier.
3339 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3342 * combine.c (combine_simplify_rtx): Don't try to simplify if
3343 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
3346 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
3348 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
3349 to 32 bytes when compiling for POWER9.
3351 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3354 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
3355 regparm >= 3 with no arg reg available also for calls with
3356 flag_force_indirect_call. Pass decl to ix86_function_regparm.
3359 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
3361 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
3363 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
3366 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
3368 2018-03-05 Richard Biener <rguenther@suse.de>
3370 PR tree-optimization/84486
3371 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
3372 When inserting a __builtin_assume_aligned call set the LHS
3373 SSA name alignment info accordingly.
3375 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
3377 PR tree-optimization/84114
3378 * config/aarch64/aarch64.c (aarch64_reassociation_width)
3379 Avoid reassociation of FLOAT_MODE addition.
3381 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
3383 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
3384 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
3385 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
3386 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
3387 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
3388 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
3389 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
3391 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
3392 __builtin_ia32_wbinvd): New builtins.
3393 (SPECIAL_ARGS2): New.
3394 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
3395 (SPECIAL_ARGS2): New.
3396 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
3397 (ix86_valid_target_attribute_inner_p): Ditto.
3398 (ix86_init_mmx_sse_builtins): Add special_args2.
3399 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
3400 TARGET_WBNOINVD_P): New.
3401 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
3402 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
3403 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
3404 * config/i386/immintrin.h (_wbinvd): New intrinsic.
3405 * config/i386/pconfigintrin.h: New file.
3406 * config/i386/wbnoinvdintrin.h: Ditto.
3407 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
3408 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
3410 2018-03-05 Richard Biener <rguenther@suse.de>
3412 PR tree-optimization/84670
3413 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3415 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3416 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3417 (compute_antic_aux): Only assert the number of values in ANTIC_IN
3418 doesn't grow if all successors (recursively) were visited at least
3421 2018-03-05 Richard Biener <rguenther@suse.de>
3423 PR tree-optimization/84650
3424 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
3425 if executed in the loop pipeline.
3427 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
3429 * doc/configfiles.texi (Configuration Files): Move info about
3430 conditionalizing $target-protos.h to...
3431 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
3432 differs from $target-protos.h.
3434 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
3435 Chung-Ju Wu <jasonwucj@gmail.com>
3437 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
3438 * config/nds32/nds32-multiple.md (setmemsi): Define.
3439 * config/nds32/nds32-memory-manipulation.c
3440 (nds32_gen_dup_4_byte_to_word_value): New.
3441 (emit_setmem_word_loop): New.
3442 (emit_setmem_byte_loop): New.
3443 (nds32_expand_setmem_loop): New.
3444 (nds32_expand_setmem_loop_v3m): New.
3445 (nds32_expand_setmem_unroll): New.
3446 (nds32_expand_setmem): New.
3448 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3449 Chung-Ju Wu <jasonwucj@gmail.com>
3451 * config/nds32/nds32-memory-manipulation.c
3452 (nds32_emit_load_store): New.
3453 (nds32_emit_post_inc_load_store): New.
3454 (nds32_emit_mem_move): New.
3455 (nds32_emit_mem_move_block): New.
3456 (nds32_expand_movmemsi_loop_unknown_size): New.
3457 (nds32_expand_movmemsi_loop_known_size): New.
3458 (nds32_expand_movmemsi_loop): New.
3459 (nds32_expand_movmemsi_unroll): New.
3460 (nds32_expand_movmemqi): Rename ...
3461 (nds32_expand_movmemsi): ... to this.
3462 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
3463 (movmemsi): ... to this.
3464 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
3465 (nds32_expand_movmemsi): ... to this.
3467 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3468 Monk Chiang <sh.chiang04@gmail.com>
3469 Chung-Ju Wu <jasonwucj@gmail.com>
3471 * config/nds32/nds32-protos.h
3472 (nds32_expand_load_multiple): New arguments.
3473 (nds32_expand_store_multiple): Ditto.
3474 (nds32_valid_multiple_load_store): Rename ...
3475 (nds32_valid_multiple_load_store_p): ... to this.
3476 * config/nds32/nds32-memory-manipulation.c
3477 (nds32_expand_load_multiple): Refine implementation.
3478 (nds32_expand_store_multiple): Ditto.
3479 * config/nds32/nds32-multiple.md
3480 (load_multiple): Update nds32_expand_load_multiple interface.
3481 (store_multiple): Update nds32_expand_store_multiple interface.
3482 * config/nds32/nds32-predicates.c
3483 (nds32_valid_multiple_load_store): Rename ...
3484 (nds32_valid_multiple_load_store_p): ... to this and refine
3486 * config/nds32/predicates.md
3487 (nds32_load_multiple_and_update_address_operation): New predicate.
3488 (nds32_store_multiple_and_update_address_operation): New predicate.
3490 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3491 Chung-Ju Wu <jasonwucj@gmail.com>
3493 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
3494 (combo): New attribute.
3495 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
3497 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
3499 * config/nds32/nds32.opt: Change -mcmodel= default value.
3501 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
3502 Monk Chiang <sh.chiang04@gmail.com>
3503 Chung-Ju Wu <jasonwucj@gmail.com>
3505 * config/nds32/constants.md (unspec_element): New enum.
3506 * config/nds32/constraints.md (Umw): New constraint.
3507 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
3508 * config/nds32/nds32-intrinsic.md: Likewise.
3509 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
3510 (nds32_valid_smw_lwm_base_p): New.
3511 (nds32_output_smw_single_word): New.
3512 (nds32_output_lmw_single_word): New.
3513 (nds32_expand_unaligned_load): New.
3514 (nds32_expand_unaligned_store): New.
3515 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
3516 (nds32_output_smw_single_word): Declare.
3517 (nds32_output_lmw_single_word): Declare.
3518 (nds32_expand_unaligned_load): Declare.
3519 (nds32_expand_unaligned_store): Declare.
3520 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
3521 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
3522 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
3523 NDS32_BUILTIN_UASTORE_DW.
3524 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
3527 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
3528 Kito Cheng <kito.cheng@gmail.com>
3529 Chung-Ju Wu <jasonwucj@gmail.com>
3531 * config/nds32/nds32-intrinsic.c
3532 (nds32_expand_builtin_null_ftype_reg): Delete.
3533 (nds32_expand_builtin_reg_ftype_imm): Ditto.
3534 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
3535 (nds32_read_argument): New.
3536 (nds32_legitimize_target): Ditto.
3537 (nds32_legitimize_argument): Ditto.
3538 (nds32_check_constant_argument): Ditto.
3539 (nds32_expand_unop_builtin): Ditto.
3540 (nds32_expand_unopimm_builtin): Ditto.
3541 (nds32_expand_binop_builtin): Ditto.
3542 (nds32_builtin_decl_impl): Ditto.
3543 (builtin_description): Ditto.
3544 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
3545 (nds32_init_builtins_impl): Ditto.
3546 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
3547 (nds32_builtin_decl): New.
3548 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
3549 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
3551 2018-03-02 Jeff Law <law@redhat.com>
3553 * reorg.c (stop_search_p): Handle DEBUG_INSN.
3554 (redundant_insn, fill_simple_delay_slots): Likewise.
3555 (fill_slots_from_thread): Likewise.
3556 * resource.c (mark_referenced_resources): Likewise.
3557 (mark_set_resources, find_dead_or_set_registers): Likewise.
3559 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3561 * substring-locations.h (format_warning_va): Formatting fix for
3563 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
3565 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
3566 * substring-locations.c: Include intl.h.
3567 (format_warning_va): Turned into small wrapper around
3568 format_warning_n_va, renamed to ...
3569 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
3570 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
3572 (format_warning_at_substring_n): New function.
3573 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
3574 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
3575 format_warning_at_substring with just a shorter name instead of
3576 const function pointer.
3577 (fmtwarn_n): New function.
3578 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
3579 appropriate, get rid of all the fmtstr temporaries, move conditionals
3580 with G_() wrapped string literals directly into fmtwarn arguments,
3581 cast dir.len to (int), formatting fixes.
3583 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
3585 * doc/invoke.texi: Remove "Cilk Plus" references.
3587 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3588 Richard Biener <rguenther@suse.de>
3591 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
3592 for error or warning attributes if CALL_FROM_THUNK_P is set.
3595 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3598 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
3599 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
3602 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
3603 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
3605 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
3606 instead of -1U in last predictors element's probability member.
3608 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3611 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3612 arguments if they are comparable.
3614 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3616 PR tree-optimization/84634
3617 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
3618 masks and masked_loop_p with a single loop_masks, making sure it's
3619 null for bb vectorization.
3621 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3623 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3624 (vect_analyze_data_ref_access): Use loop->safe_len rather than
3625 loop->force_vectorize to check whether there is no alias.
3627 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3630 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
3632 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
3634 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
3635 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
3636 instead of a loop around prev_real_insn.
3637 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
3641 * config/i386/i386.c (ix86_print_operand): Use conditional
3642 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
3645 2018-03-02 Richard Biener <rguenther@suse.de>
3647 PR tree-optimization/84427
3648 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
3649 (bitmap_set_subtract_values): Rewrite to handle multiple
3652 (prune_clobbered_mems): Likewise.
3653 (phi_translate): Take edge instead of pred/phiblock.
3654 (phi_translate_1): Likewise.
3655 (phi_translate_set): Likewise. Insert all translated
3656 exprs for a value into the set, keeping possibly multiple
3657 expressions per value.
3658 (compute_antic_aux): Adjust for phi_translate changes.
3659 When intersecting union the expressions and prune those
3660 not in the final value set, keeping possibly multiple
3661 expressions per value. Do not use value-insertion
3662 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
3663 all expressions. Add verification that the value-sets
3664 only shrink during iteration.
3665 (compute_partial_antic_aux): Adjust for the phi_translate changes.
3666 (do_pre_regular_insertion): Likewise.
3667 (do_pre_partial_partial_insertion): Likewise.
3669 2018-03-02 Richard Biener <rguenther@suse.de>
3672 * config/darwin.c (saved_debug_info_level): New static global.
3673 (darwin_asm_lto_start): Disable debug info generation for LTO out.
3674 (darwin_asm_lto_end): Restore debug info generation settings.
3676 2018-03-01 Martin Liska <mliska@suse.cz>
3679 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
3682 2018-03-01 Richard Biener <rguenther@suse.de>
3685 * dwarf2out.c (gen_variable_die): Properly handle late VLA
3686 type annotation with LTO when debug was disabled at compile-time.
3688 2018-03-01 Matthew Fortune <mfortune@gmail.com>
3690 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
3692 (mips_final_postscan_insn): Likewise.
3694 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
3696 PR rtl-optimization/84528
3697 * alias.c (init_alias_target): Add commentary.
3698 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
3699 a unique base value if the frame pointer is not eliminated
3700 to the stack pointer.
3702 2018-03-01 Tom de Vries <tom@codesourcery.com>
3704 PR rtl-optimization/83327
3705 * lra-int.h (hard_regs_spilled_into): Declare.
3706 * lra.c (hard_regs_spilled_into): Define.
3707 (init_reg_info): Init hard_regs_spilled_into.
3708 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
3709 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
3710 (process_bb_lives): Handle hard_regs_spilled_into.
3711 (lra_create_live_ranges_1): Before doing liveness propagation, clear
3712 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
3714 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
3716 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
3717 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
3718 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
3719 * config/rs6000/aix72.h: New file.
3721 2018-02-28 Jakub Jelinek <jakub@redhat.com>
3723 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
3724 instead of warning_at with conditional singular and plural messages
3728 * stor-layout.c (update_alignment_for_field): For
3729 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
3730 && !DECL_PACKED (field), do the alignment update, just use
3731 only desired_align instead of MAX (type_align, desired_align)
3733 (place_field): Don't do known_align < desired_align handling
3734 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
3735 is non-NULL, instead do it after rli->prev_field handling and
3736 only if not within a bitfield word. For DECL_PACKED (field)
3737 use type_align of BITS_PER_UNIT.
3739 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
3741 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
3742 superfluous parentheses and trailing spaces.
3744 2018-02-28 Richard Biener <rguenther@suse.de>
3746 PR tree-optimization/84584
3747 * graphite-scop-detection.c (scop_detection::add_scop): Discard
3748 SCoPs with fake exit edge.
3750 2018-02-28 Martin Liska <mliska@suse.cz>
3753 * timevar.c (timer::print): Fix format to properly print 100%
3756 2018-02-28 Richard Biener <rguenther@suse.de>
3759 * genmatch.c (capture_info::walk_match): Do not mark
3760 captured expressions without operands as expr_p given
3761 they act more like predicates and should be subject to
3762 "lost tail" side-effect preserving.
3764 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
3766 PR rtl-optimization/81611
3767 * auto-inc-dec.c (attempt_change): Move dead note from
3768 mem_insn if it's the next use of regno
3769 (find_address): Take address use of reg holding
3770 non-incremented value. Add parm to limit search to the named
3772 (merge_in_block): Attempt to use a mem insn that is the next
3773 use of the original regno.
3775 2018-02-27 Martin Sebor <msebor@redhat.com>
3778 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
3779 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
3781 2018-02-27 Martin Sebor <msebor@redhat.com>
3783 PR translation/84207
3784 * diagnostic-core.h (warning_n, error_n, inform_n): Change
3785 n argument to unsigned HOST_WIDE_INT.
3786 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
3787 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
3788 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
3789 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
3791 2018-02-27 Richard Biener <rguenther@suse.de>
3793 PR tree-optimization/84512
3794 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
3795 Do not use the estimate returned from record_stmt_cost for
3796 the scalar iteration cost but sum properly using add_stmt_cost.
3798 2018-02-27 Richard Biener <rguenther@suse.de>
3800 PR tree-optimization/84466
3801 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3802 Adjust last change to less strictly validate use operands.
3804 2018-02-27 Martin Liska <mliska@suse.cz>
3806 PR gcov-profile/84548
3807 * gcov.c (process_file): Allow partial overlap and consider it
3808 also as group functions.
3809 (output_lines): Properly calculate range of lines for a group.
3811 2018-02-27 Martin Liska <mliska@suse.cz>
3813 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
3814 'ggc' suffixes. Change first column width.
3815 (timer::print): Fix formatting of the column.
3817 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
3819 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
3820 preserve inline entry blocks for the sake of debug inline
3821 entry point markers alone.
3822 (remove_unused_locals): Suggest in comments a better place to
3823 force the preservation of inline entry blocks that are
3824 otherwise unused, but do not preserve them.
3826 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3828 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
3830 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3833 * config/i386/constraints.md (Bs): Replace
3834 ix86_indirect_branch_register with
3835 TARGET_INDIRECT_BRANCH_REGISTER.
3837 * config/i386/i386.md (indirect_jump): Likewise.
3838 (tablejump): Likewise.
3839 (*sibcall_memory): Likewise.
3840 (*sibcall_value_memory): Likewise.
3841 Peepholes of indirect call and jump via memory: Likewise.
3842 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
3843 (*sibcall_value_GOT_32): Likewise.
3844 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3845 (GOT_memory_operand): Likewise.
3846 (call_insn_operand): Likewise.
3847 (sibcall_insn_operand): Likewise.
3848 (GOT32_symbol_operand): Likewise.
3849 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
3851 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3853 PR rtl-optimization/83496
3854 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
3855 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
3856 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
3857 redundant insn, if any.
3858 (relax_delay_slots): Likewise.
3859 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
3861 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
3863 PR tree-optimization/83965
3864 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
3865 that grouped statements are part of a reduction chain. Return
3866 true if the statement is not marked as a reduction itself but
3868 (vect_recog_dot_prod_pattern): Don't check whether the statement
3869 is part of a group here.
3870 (vect_recog_sad_pattern): Likewise.
3871 (vect_recog_widen_sum_pattern): Likewise.
3873 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3876 * final.c (rest_of_clean_state): Also look for calls inside sequences.
3878 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3881 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
3883 (ix86_output_indirect_function_return): New prototype.
3884 (ix86_split_simple_return_pop_internal): Likewise.
3885 * config/i386/i386.c (indirect_return_via_cx): New.
3886 (indirect_return_via_cx_bnd): Likewise.
3887 (indirect_thunk_name): Handle return va CX_REG.
3888 (output_indirect_thunk_function): Create alias for
3889 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
3890 (ix86_output_indirect_jmp): Remove the bool argument.
3891 (ix86_output_indirect_function_return): New function.
3892 (ix86_split_simple_return_pop_internal): Likewise.
3893 * config/i386/i386.md (*indirect_jump): Don't pass false
3894 to ix86_output_indirect_jmp.
3895 (*tablejump_1): Likewise.
3896 (simple_return_pop_internal): Change it to define_insn_and_split.
3897 Call ix86_split_simple_return_pop_internal to split it for
3899 (simple_return_indirect_internal): Call
3900 ix86_output_indirect_function_return instead of
3901 ix86_output_indirect_jmp.
3903 2018-02-26 Jakub Jelinek <jakub@redhat.com>
3906 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
3907 memset and value initialization afterwards.
3909 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
3911 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
3913 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3916 * common/config/aarch64/aarch64-common.c
3917 (aarch_option_optimization_table[]): Switch
3918 off fomit-frame-pointer
3920 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3921 Chung-Ju Wu <jasonwucj@gmail.com>
3923 * config/nds32/nds32-multiple.md (load_multiple): Disallow
3925 (store_multiple): Ditto.
3927 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3929 * config.gcc: Add --with-cpu support for nds32 target.
3930 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
3931 * config/nds32/nds32.opt: Add -mcpu= option.
3933 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
3935 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
3936 isel=yes): Warn for these deprecated options.
3938 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
3940 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
3941 ISA_2_5_MASKS_EMBEDDED.
3943 2018-02-23 Jakub Jelinek <jakub@redhat.com>
3945 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
3946 p->max as pointers rather than using iterative_hash_expr.
3948 2018-02-23 Carl Love <cel@us.ibm.com>
3950 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
3951 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
3953 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
3954 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
3955 P8V_BUILTIN_VEC_VUNSIGNED2.
3957 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
3960 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
3961 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
3963 * lra-constraints.c (curr_insn_transform): Set up
3964 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
3966 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
3967 * lra-eliminations.c (spill_pseudos): Ditto.
3968 (process_insn_for_elimination): Ditto.
3969 * lra-lives.c (reg_early_clobber_p): Use the new macros.
3970 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
3971 LRA_NON_CLOBBERED_ALT.
3973 2018-02-22 Martin Sebor <msebor@redhat.com>
3975 PR tree-optimization/84480
3976 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
3977 to maybe_diag_stxncpy_trunc. Call it.
3978 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
3979 from gimple_fold_builtin_strcpy. Print inlining stack.
3980 (handle_builtin_stxncpy): Print inlining stack.
3981 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
3983 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
3986 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
3987 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
3988 and -fcheck-pointer-bounds are used together.
3989 (indirect_thunk_prefix): New enum.
3990 (indirect_thunk_need_prefix): New function.
3991 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
3992 "_nt" instead of "_bnd" for NOTRACK prefix.
3993 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
3994 (output_indirect_thunk_function): Likewise.
3996 (ix86_code_end): Update output_indirect_thunk_function calls.
3997 (ix86_output_indirect_branch_via_reg): Replace
3998 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
3999 (ix86_output_indirect_branch_via_push): Likewise.
4000 (ix86_output_function_return): Likewise.
4001 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
4002 incompatible with -fcf-protection=branch and
4003 -fcheck-pointer-bounds.
4005 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4008 * config/aarch64/aarch64.c (aarch64_print_address_internal):
4009 Change gcc_assert call to output_operand_lossage.
4011 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4013 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
4015 2018-02-22 DJ Delorie <dj@redhat.com>
4016 Sebastian Perta <sebastian.perta@renesas.com>
4017 Oleg Endo <olegendo@gcc.gnu.org>
4019 * config/rx/rx.c (rx_rtx_costs): New function.
4020 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
4022 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
4024 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
4026 2018-02-22 Martin Liska <mliska@suse.cz>
4029 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
4030 Add "native" as a possible value.
4032 2018-02-22 Martin Liska <mliska@suse.cz>
4035 * config/i386/i386.c (ix86_option_override_internal):
4036 Add "native" as a possible value for -march and -mtune.
4038 2018-02-22 Jakub Jelinek <jakub@redhat.com>
4041 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
4042 to all type variants.
4044 PR tree-optimization/84503
4045 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
4046 width as info->bitpos + info->bitsize - start.
4047 (merged_store_group::merge_overlapping): Simplify width computation.
4048 (check_no_overlap): New function.
4049 (imm_store_chain_info::try_coalesce_bswap): Compute expected
4050 start + width and last_order of the group, fail if check_no_overlap
4052 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
4053 to group if check_no_overlap fails.
4055 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4057 * config/rs6000/altivec.md: Delete contraint arguments to
4058 define_expand, define_split, and define_peephole2, and in
4059 define_insn_and_split if always unused.
4060 * config/rs6000/darwin.md: Ditto.
4061 * config/rs6000/dfp.md: Ditto.
4062 * config/rs6000/rs6000.md: Ditto.
4063 * config/rs6000/sync.md: Ditto.
4064 * config/rs6000/vector.md: Ditto.
4065 * config/rs6000/vsx.md: Ditto.
4067 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4069 * config/rs6000/altivec.md: Write output control strings as braced
4070 blocks instead of double-quoted strings.
4071 * config/rs6000/darwin.md: Ditto.
4072 * config/rs6000/rs6000.md: Ditto.
4073 * config/rs6000/vector.md: Ditto.
4074 * config/rs6000/vsx.md: Ditto.
4076 2018-02-21 Jason Merrill <jason@redhat.com>
4078 PR c++/84314 - ICE with templates and fastcall attribute.
4079 * attribs.c (build_type_attribute_qual_variant): Remove assert.
4081 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4083 * ipa-cp.c (determine_versionability): Fix comment typos.
4085 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4088 * ipa-cp.c (determine_versionability): Do not version functions caling
4091 2018-02-21 Martin Liska <mliska@suse.cz>
4094 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
4095 Add "native" as a possible value.
4096 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
4097 the macro when native cpu detection is available.
4099 2018-02-21 Martin Liska <mliska@suse.cz>
4102 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
4103 Add "native" as a possible value.
4104 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
4105 when native cpu detection is available.
4107 2018-02-21 Jakub Jelinek <jakub@redhat.com>
4108 Martin Sebor <msebor@redhat.com>
4110 PR tree-optimization/84478
4111 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
4113 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
4114 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
4115 support which is conservatively correct, for 2 only stay conservative
4116 for maxlen. Formatting and comment capitalization fixes. Add STRICT
4117 argument to the 2 argument get_range_strlen, adjust 6 arg
4118 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
4120 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
4121 (gimple_fold_builtin_strlen): Pass true as last argument to
4124 2018-02-20 Martin Sebor <msebor@redhat.com>
4127 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
4128 (builtin_memref::set_base_and_offset): Same. Handle inner references.
4129 (builtin_memref::builtin_memref): Factor out parts into
4130 set_base_and_offset and call it.
4132 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4135 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
4136 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
4137 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
4138 search at the associated MODE_INT.
4140 2018-02-20 Jeff Law <law@redhat.com>
4143 PR tree-optimization/81592
4145 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
4146 for range data rather than using global data.
4147 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
4148 range data rather than using global data.
4149 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
4150 pass it to children as needed.
4151 (struct directive::fmtresult): Similarly.
4152 (struct directive::set_width): Similarly.
4153 (struct directive::set_precision): Similarly.
4154 (format_integer, format_directive, parse_directive): Similarly.
4155 (format_none): Accept unnamed vr_values parameter.
4156 (format_percent, format_floating, format_character): Similarly.
4157 (format_string, format_plain): Similarly.
4158 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
4159 the EVRP range analyzer for range data rather than using global data.
4160 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
4161 gimple-ssa-evrp-analyze.h
4162 (class sprintf_dom_walker): Add after_dom_children member function.
4163 Add evrp_range_analyzer member.
4164 (sprintf_dom_walker::before_dom_children): Call into the EVRP
4165 range analyzer as needed.
4166 (sprintf_dom_walker::after_dom_children): New member function.
4167 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
4169 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
4170 (evrp_range_analyzer::pop_to_marker): Likewise.
4172 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4174 PR tree-optimization/84419
4175 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
4176 with the required type if its current type is compatible but
4179 2018-02-20 Jakub Jelinek <jakub@redhat.com>
4182 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
4183 after vectorization.
4185 2018-02-20 Martin Liska <mliska@suse.cz>
4188 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
4189 possible values if we don't have a hint.
4191 2018-02-20 Martin Liska <mliska@suse.cz>
4195 * final.c (shorten_branches): Build align_tab array with one
4197 * opts.c (finish_options): Add alignment option limit check.
4198 (MAX_CODE_ALIGN): Likewise.
4199 (MAX_CODE_ALIGN_VALUE): Likewise.
4200 * doc/invoke.texi: Document maximum allowed option value for
4201 all -falign-* options.
4203 2018-02-19 Jakub Jelinek <jakub@redhat.com>
4206 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
4207 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
4208 * var-tracking.c (emit_note_insn_var_location): Remove all references
4209 to NOTE_INSN_CALL_ARG_LOCATION.
4210 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
4211 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
4212 Use copy_rtx_if_shared.
4213 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
4214 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
4215 (dwarf2out_var_location): Remove handling of
4216 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
4218 * final.c (final_scan_insn): Remove all references to
4219 NOTE_INSN_CALL_ARG_LOCATION.
4220 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
4221 before dumping final insns.
4222 * except.c (emit_note_eh_region_end): Remove all references to
4223 NOTE_INSN_CALL_ARG_LOCATION.
4224 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
4225 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
4226 * config/arc/arc.c (hwloop_optimize): Likewise.
4227 * config/arm/arm.c (create_fix_barrier): Likewise.
4228 * config/s390/s390.c (s390_chunkify_start): Likewise.
4229 * config/sh/sh.c (find_barrier): Likewise.
4230 * config/i386/i386.c (rest_of_insert_endbranch,
4231 ix86_seh_fixup_eh_fallthru): Likewise.
4232 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
4233 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
4234 * config/frv/frv.c (frv_function_prologue): Likewise.
4235 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
4237 (note_outside_basic_block_p): Remove all references to
4238 NOTE_INSN_CALL_ARG_LOCATION.
4239 * gengtype.c (adjust_field_rtx_def): Likewise.
4240 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
4242 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
4243 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
4246 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
4249 PR tree-optimization/84452
4250 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
4251 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
4254 2018-02-19 Martin Liska <mliska@suse.cz>
4257 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
4259 2018-02-19 Martin Liska <mliska@suse.cz>
4260 Richard Sandiford <richard.sandiford@linaro.org>
4262 PR tree-optimization/82491
4263 * gimple-fold.c (get_base_constructor): Make earlier bail out
4266 2018-02-19 Carl Love <cel@us.ibm.com>
4268 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
4269 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
4271 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
4272 P8V_BUILTIN_VEC_NEG.
4274 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
4276 * config/rl78/rl78.md (movdf): New define expand.
4278 2018-02-19 Martin Liska <mliska@suse.cz>
4281 * doc/invoke.texi: Fix typo.
4282 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
4284 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
4286 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
4287 handle rs6000_single_float and rs6000_double_float specially for
4290 2018-02-16 Jeff Law <law@redhat.com>
4292 * config/rx/rx.c (add_pop_cfi_notes): New function.;
4295 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4298 * ipa-inline.c (inline_small_functions): Fix a typo.
4300 2018-02-16 Nathan Sidwell <nathan@acm.org>
4302 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
4304 2018-02-16 Carl Love <cel@us.ibm.com>
4306 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
4307 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
4308 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
4309 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
4310 expansion to P8V_BUILTIN_VEC_FLOAT2.
4312 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
4314 PR rtl-optimization/70023
4315 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
4316 src_regno into account.
4318 2018-02-16 Carl Love <cel@us.ibm.com>
4320 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
4321 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
4322 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
4323 * config/rs6000/rs6000.c: Remove case statements for
4324 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
4325 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
4326 and P9V_BUILTIN_VEC_VINSERT4B.
4327 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
4328 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
4329 * config/rs6000/vsx.md:
4330 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
4333 2018-02-16 Carl Love <cel@us.ibm.com>
4335 * config/rs6000/altivec.h: Add builtin names vec_extract4b
4337 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
4339 * config/rs6000/rs6000-c.c: Add the definitions for
4340 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
4341 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
4342 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
4343 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
4344 definition for insert4b and define insn *insert3b_internal.
4345 * doc/extend.texi: Add documentation for vec_extract4b.
4347 2018-02-16 Nathan Sidwell <nathan@acm.org>
4349 * doc/extend.texi (Backwards Compatibility): Mention friend
4350 injection. Note for-scope is deprecated.
4351 * doc/invoke.texi (-ffriend-injection): Deprecate.
4353 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
4355 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
4356 that moved to I2, also allow destinations that are a paradoxical
4357 subreg (instead of a normal reg).
4359 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
4362 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
4365 2018-02-16 Richard Biener <rguenther@suse.de>
4367 PR tree-optimization/84037
4368 PR tree-optimization/84016
4370 * config/i386/i386.c (ix86_builtin_vectorization_cost):
4371 Adjust vec_construct for the fact we need additional higher latency
4372 128bit inserts for AVX256 and AVX512 vector builds.
4373 (ix86_add_stmt_cost): Scale vector construction cost for
4376 2018-02-16 Richard Biener <rguenther@suse.de>
4378 PR tree-optimization/84417
4379 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
4380 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
4381 (non_rewritable_lvalue_p): Likewise, use poly-ints.
4383 2018-02-16 Martin Liska <mliska@suse.cz>
4386 * internal-fn.def (ASAN_CHECK): Set proper flags.
4387 (ASAN_MARK): Likewise.
4389 2018-02-16 Julia Koval <julia.koval@intel.com>
4391 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
4392 from PTA_CANNONLAKE.
4394 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4397 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
4398 Use ++iter rather than iter++ for std::list iterators.
4399 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
4400 defer deleting them until all nodes in the forest are processed. Do
4401 free even leaf nodes. Change to_process into auto_vec.
4404 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
4405 * vec.h (vec_default_construct): Use memset instead of placement new
4406 if BROKEN_VALUE_INITIALIZATION is defined.
4407 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
4408 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
4411 PR rtl-optimization/83723
4412 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
4413 * lra.c (lra_substitute_pseudo): Likewise. If true, use
4414 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
4416 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
4418 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
4420 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
4422 PR rtl-optimization/81443
4423 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
4424 from inner REGs to paradoxical SUBREGs.
4426 2018-02-16 Richard Biener <rguenther@suse.de>
4428 PR tree-optimization/84399
4429 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4430 For operands we can analyze at their definition make sure we can
4431 analyze them at each use as well.
4433 2018-02-16 Richard Biener <rguenther@suse.de>
4435 PR tree-optimization/84190
4436 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
4437 volatile accesses if the decl isn't volatile.
4439 2018-02-15 Jason Merrill <jason@redhat.com>
4441 PR c++/84314 - ICE with templates and fastcall attribute.
4442 * attribs.c (build_type_attribute_qual_variant): Don't clobber
4443 TYPE_CANONICAL on an existing type.
4445 2018-02-15 Jakub Jelinek <jakub@redhat.com>
4447 PR tree-optimization/84383
4448 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
4449 dstoff nor call operand_equal_p if dstbase is NULL.
4451 PR tree-optimization/84334
4452 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
4453 also a CONSTANT_CLASS_P, punt.
4455 2018-02-14 Jim Wilson <jimw@sifive.com>
4457 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
4458 first SMALL_OPERAND check. New local min_second_step. Move assert
4459 to where locals are set. Add TARGET_RVC support.
4460 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
4462 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
4464 * doc/invoke.texi: Correct -Wformat-overflow code sample.
4466 2018-02-14 Martin Sebor <msebor@redhat.com>
4468 PR tree-optimization/83698
4469 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
4470 arrays constrain the offset range to their bounds.
4471 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
4472 (builtin_access::overlap): Avoid setting the size of overlap if it's
4474 (maybe_diag_overlap): Also consider arrays when deciding what values
4475 of offsets to include in diagnostics.
4477 2018-02-14 Martin Sebor <msebor@redhat.com>
4480 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
4481 that correspond to the kind of a declaration.
4483 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
4486 * config/pa/pa.md: Load address of PIC label using the linkage table
4487 if the label is nonlocal.
4489 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
4491 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
4492 warning message if user requests -maltivec=be.
4493 * doc/invoke.texi: Document deprecation of -maltivec=be.
4495 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
4498 * config/rs6000/rs6000-c.c: Update definitions for
4499 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
4500 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
4502 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4505 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
4506 add _get_ssp intrinsics. Remove argument from
4507 __builtin_ia32_rdssp[d|q].
4508 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
4509 * config/i386/i386-builtin.def: Remove argument from
4510 __builtin_ia32_rdssp[d|q].
4511 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
4512 ix86_expand_special_args_builtin for _rdssp[d|q].
4513 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
4514 Clear register before usage.
4515 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
4516 Add documentation for new _get_ssp and _inc_ssp intrinsics.
4518 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
4520 PR tree-optimization/84357
4521 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
4522 operand 1 of an ARRAY_REF too.
4524 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
4527 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
4528 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4530 (set_of_reg): New struct.
4531 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
4532 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
4533 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4535 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
4536 Split into bitclr, bitset, bitinvert patterns if appropriate.
4537 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
4538 use rx_fuse_in_memory_bitop.
4539 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
4540 to named insn, correct maximum insn length.
4542 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
4545 * machmode.def: Define a complex mode for PARTIAL_INT.
4546 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
4548 * doc/rtl.texi: Document CSPImode.
4549 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
4551 (msp430_hard_regno_nregs_with_padding): Likewise.
4553 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
4556 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
4558 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
4560 PR rtl-optimization/84169
4561 * combine.c (try_combine): New variable split_i2i3. Set it to true if
4562 we generated a parallel as new i3 and we split that to new i2 and i3
4563 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
4564 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
4565 those to i2, not i1. Partially rewrite this scan code.
4567 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4570 * stor-layout.c (place_field): For variable length fields, adjust
4571 offset_align afterwards not just based on the field's alignment,
4572 but also on the size.
4575 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
4576 of exps and logs in the use_exp2 case.
4578 2018-02-13 Jeff Law <law@redhat.com>
4580 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
4583 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
4586 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
4590 * common.opt (gas-loc-support, gas-locview-support): New.
4591 (ginline-points, ginternal-reset-location-views): New.
4592 * doc/invoke.texi: Document them. Use @itemx where intended.
4593 (gvariable-location-views): Adjust.
4594 * target.def (reset_location_view): New.
4595 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
4596 (TARGET_RESET_LOCATION_VIEW): New.
4597 * doc/tm.texi: Rebuilt.
4598 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
4599 (dwarf2out_default_as_locview_support): New.
4600 (output_asm_line_debug_info): Use option variables.
4601 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
4602 (output_loc_list): Likewise.
4603 (add_high_low_attributes): Check option variables.
4604 Don't output entry view attribute in strict mode.
4605 (gen_inlined_subroutine_die): Check option variables.
4606 (dwarf2out_inline_entry): Likewise.
4607 (init_sections_and_labels): Likewise.
4608 (dwarf2out_early_finish): Likewise.
4609 (maybe_reset_location_view): New, from...
4610 (dwarf2out_var_location): ... here. Call it.
4611 * debug.h (dwarf2out_default_as_loc_support): Declare.
4612 (dwarf2out_default_as_locview_support): Declare.
4613 * hooks.c (hook_int_rtx_insn_0): New.
4614 * hooks.h (hook_int_rtx_insn_0): Declare.
4615 * toplev.c (process_options): Take -gas-loc-support and
4616 -gas-locview-support from dwarf2out. Enable
4617 -gvariable-location-views by default only with locview
4618 assembler support. Enable -ginternal-reset-location-views by
4619 default only if the target defines the corresponding hook.
4620 Enable -ginline-points by default if location views are
4621 enabled; force it disabled if statement frontiers are
4623 * tree-inline.c (expand_call_inline): Check option variables.
4624 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
4626 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
4628 PR tree-optimization/84321
4629 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
4630 handling. Also check whether the anti-range contains any values
4631 that satisfy the mask; switch to a VR_RANGE if not.
4633 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
4636 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
4638 2018-02-13 Martin Jambor <mjambor@suse.cz>
4641 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
4642 of call statements, also set location of a load to a temporary.
4644 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
4646 * config/rl78/rl78.c (add_vector_labels): New function.
4647 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
4648 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
4649 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
4650 which checks that no arguments are passed.
4651 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
4652 * doc/extend.texi: Documentation for the new attribute.
4654 2018-02-13 Andreas Schwab <schwab@suse.de>
4656 * config/riscv/linux.h (CPP_SPEC): Define.
4658 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4661 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
4662 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
4663 OPTION_MASK_ISA_AES as first argument to def_builtin_const
4664 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
4665 instead of OPTION_MASK_ISA_PCLMUL as first argument to
4666 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
4667 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
4668 temporarily for AES and PCLMUL builtins.
4670 PR tree-optimization/84339
4671 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
4672 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
4676 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
4677 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
4678 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
4680 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
4682 * omp-simd-clone.h: New file.
4683 * omp-simd-clone.c: Include omp-simd-clone.h.
4684 (expand_simd_clones): No longer static.
4685 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
4686 cgraph.h and omp-simd-clone.h.
4687 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
4688 (vect_recog_widen_shift_pattern): Formatting fix.
4689 (vect_pattern_recog_1): Don't check optab for calls.
4692 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
4693 operands[2] into a REG before using gen_lowpart on it.
4695 2018-02-12 Jeff Law <law@redhat.com>
4698 * config/sh/sh.c (find_barrier): Consider a sibling call
4701 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
4702 successfully back substituting a reg.
4704 2018-02-12 Richard Biener <rguenther@suse.de>
4706 PR tree-optimization/84037
4707 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
4708 parameter, move visited init to caller.
4709 (vect_slp_analyze_operations): Separate cost from validity
4710 check, initialize visited once for all instances.
4711 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
4713 * tree-vect-stmts.c (vect_model_simple_cost): Make early
4715 (vect_model_promotion_demotion_cost): Likewise.
4716 (vectorizable_bswap): Guard cost modeling with !slp_node
4717 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
4719 (vectorizable_call): Likewise.
4720 (vectorizable_conversion): Likewise.
4721 (vectorizable_assignment): Likewise.
4722 (vectorizable_shift): Likewise.
4723 (vectorizable_operation): Likewise.
4724 (vectorizable_store): Likewise.
4725 (vectorizable_load): Likewise.
4726 (vectorizable_condition): Likewise.
4727 (vectorizable_comparison): Likewise.
4729 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
4732 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
4733 (ASAN_MARK): Fix fnspec to account for return value, change pointer
4734 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
4736 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4739 * params.def (inline-min-speedup): Increase from 8 to 15.
4740 (max-inline-insns-auto): Decrease from 40 to 30.
4741 * ipa-split.c (consider_split): Add some buffer for function to
4742 be considered inlining candidate.
4743 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
4746 2018-02-12 Richard Biener <rguenther@suse.de>
4748 PR tree-optimization/84037
4749 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
4750 matched stmts if we cannot swap the non-matched ones.
4752 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4754 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
4755 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
4756 _mm_maskz_scalef_round_ss): New intrinsics.
4757 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
4758 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
4759 __builtin_ia32_scalefss_round): Remove.
4760 (__builtin_ia32_scalefsd_mask_round,
4761 __builtin_ia32_scalefss_mask_round): New intrinsics.
4762 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
4763 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
4764 ((match_operand:VF_128 2 "<round_nimm_predicate>"
4765 "<round_constraint>")): Changed to ...
4766 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
4767 "<round_scalar_constraint>")): ... this.
4768 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
4769 %0, %1, %2<round_op3>}"): Changed to ...
4770 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
4771 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
4772 %2<round_scalar_mask_op3>}"): ... this.
4773 * config/i386/subst.md (round_scalar_nimm_predicate): New.
4775 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4777 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
4778 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
4779 (_mm_maskz_sqrt_round_ss): New intrinsics.
4780 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
4781 (__builtin_ia32_sqrtsd_mask_round)
4782 (__builtin_ia32_sqrtss_mask_round): New builtins.
4783 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
4784 (__builtin_ia32_sqrtss_round): Remove.
4785 (__builtin_ia32_sqrtsd_mask_round)
4786 (__builtin_ia32_sqrtss_mask_round): New builtins.
4787 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
4788 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
4789 ((match_operand:VF_128 1 "vector_operand"
4790 "xBm,<round_constraint>")): Changed to ...
4791 ((match_operand:VF_128 1 "vector_operand"
4792 "xBm,<round_scalar_constraint>")): ... this.
4793 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
4794 %0, %2, %<iptr>1<round_op3>}): Changed to ...
4795 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
4796 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
4797 %<iptr>1<round_scalar_mask_op3>}): ... this.
4798 ((set_attr "prefix" "<round_prefix>")): Changed to ...
4799 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
4801 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
4804 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
4805 Cast vec_cmpeq result to correct type.
4806 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
4807 Cast vec_cmpgt result to correct type.
4809 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
4811 * final.c (final_scan_insn_1): Renamed from...
4812 (final_scan_insn): ... this. New wrapper, to recover
4813 seen from the outermost call in recursive ones.
4814 * config/sparc/sparc.c (output_return): Drop seen from call.
4815 (output_sibcall): Likewise.
4816 * config/visium/visium.c (output_branch): Likewise.
4818 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
4820 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
4823 2018-02-10 Alan Modra <amodra@gmail.com>
4826 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
4827 Specify LR as an input.
4829 2018-02-10 Jakub Jelinek <jakub@redhat.com>
4832 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
4833 remove_member_access_dummy_vars): New functions.
4834 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
4835 lower_omp_1, execute_lower_omp): Use them.
4837 PR rtl-optimization/84308
4838 * shrink-wrap.c (spread_components): Release todo vector.
4840 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
4842 PR rtl-optimization/57193
4843 * ira-color.c (struct allocno_color_data): Add member
4844 conflict_allocno_hard_prefs.
4845 (update_conflict_allocno_hard_prefs): New.
4846 (bucket_allocno_compare_func): Add a preference based on
4847 conflict_allocno_hard_prefs.
4848 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
4849 (color_allocnos): Remove a dead code. Initiate
4850 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
4852 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4855 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
4856 constraint from =wa to wa. Avoid a subreg on the output operand,
4857 instead use a pseudo and subreg it in a move.
4858 (p9_xxbrd_<mode>): Changed to ...
4859 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
4860 (p9_xxbrd_v2df): New expander.
4861 (p9_xxbrw_<mode>): Changed to ...
4862 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
4863 (p9_xxbrw_v4sf): New expander.
4865 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4867 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
4869 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
4872 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
4873 multiply in 32-bit mode.
4874 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
4875 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
4878 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4880 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
4881 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
4882 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
4883 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
4885 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
4888 * dwarf2out.c (is_trivial_indirect_ref): New function.
4889 (dwarf2out_late_global_decl): Do not generate a location
4890 attribute for variables that have a non-trivial DECL_VALUE_EXPR
4891 and that are not defined in the current unit.
4893 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4895 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
4896 instead of a libcall for UNORDERED.
4898 2018-02-09 Tamar Christina <tamar.christina@arm.com>
4901 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
4902 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
4904 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4907 * config/s390/s390.c (s390_set_current_function): Invoke
4908 s390_indirect_branch_settings also if fndecl didn't change.
4910 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4912 * config/rs6000/rs6000.md (blockage): Set length to zero.
4914 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4916 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
4918 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4921 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
4922 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
4926 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
4927 PARALLEL incoming that failed vt_get_decl_and_offset check.
4930 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
4931 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
4933 (get_variable_section): For decls in named .bss* sections pass true as
4934 second argument to bss_initializer_p.
4936 2018-02-09 Marek Polacek <polacek@redhat.com>
4937 Jakub Jelinek <jakub@redhat.com>
4940 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
4941 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
4942 Sync some changes from cxx_fold_indirect_ref.
4944 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4946 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
4948 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
4949 (BLOCK_INLINE_ENTRY_LABEL): New.
4950 (dwarf2out_var_location): Disregard inline entry markers.
4951 (inline_entry_data): New struct.
4952 (inline_entry_data_hasher): New hashtable type.
4953 (inline_entry_data_hasher::hash): New.
4954 (inline_entry_data_hasher::equal): New.
4955 (inline_entry_data_table): New variable.
4956 (add_high_low_attributes): Add DW_AT_entry_pc and
4957 DW_AT_GNU_entry_view attributes if a pending entry is found
4958 in inline_entry_data_table. Add old entry_pc attribute only
4959 if debug nonbinding markers are disabled.
4960 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
4961 markers are enabled.
4962 (block_within_block_p, dwarf2out_inline_entry): New.
4963 (dwarf2out_finish): Check that no entries remained in
4964 inline_entry_data_table.
4965 * final.c (reemit_insn_block_notes): Handle inline entry notes.
4966 (final_scan_insn, notice_source_line): Likewise.
4967 (rest_of_clean_state): Skip inline entry markers.
4968 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
4970 * gimple.c (gimple_build_debug_inline_entry): New.
4971 * gimple.h (enum gimple_debug_subcode): Add
4972 GIMPLE_DEBUG_INLINE_ENTRY.
4973 (gimple_build_debug_inline_entry): Declare.
4974 (gimple_debug_inline_entry_p): New.
4975 (gimple_debug_nonbind_marker_p): Adjust.
4976 * insn-notes.def (INLINE_ENTRY): New.
4977 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
4978 inline entry marker notes.
4979 (print_insn): Likewise.
4980 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
4981 (INSN_DEBUG_MARKER_KIND): Likewise.
4982 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
4983 * tree-inline.c (expand_call_inline): Build and insert
4984 debug_inline_entry stmt.
4985 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
4986 inline entry blocks early, if nonbind markers are enabled.
4987 (dump_scope_block): Dump fragment info.
4988 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
4989 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
4990 (gimple_build_debug_inline_entry): New.
4991 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
4992 Enable/disable inline entry points too.
4993 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
4994 (DEBUG_INSN): Describe inline entry markers.
4996 * common.opt (gvariable-location-views): New.
4997 (gvariable-location-views=incompat5): New.
4998 * config.in: Rebuilt.
4999 * configure: Rebuilt.
5000 * configure.ac: Test assembler for view support.
5001 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
5002 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
5003 * dwarf2out.c (var_loc_view): New typedef.
5004 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
5005 (dwarf2out_locviews_in_attribute): New.
5006 (dwarf2out_locviews_in_loclist): New.
5007 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
5008 (enum dw_line_info_opcode): Add LI_adv_address.
5009 (struct dw_line_info_table): Add view.
5010 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
5011 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
5012 (zero_view_p): New variable.
5013 (ZERO_VIEW_P): New macro.
5014 (output_asm_line_debug_info): New.
5015 (struct var_loc_node): Add view.
5016 (add_AT_view_list, AT_loc_list): New.
5017 (add_var_loc_to_decl): Add view param. Test it against last.
5018 (new_loc_list): Add view params. Record them.
5019 (AT_loc_list_ptr): Handle loc and view lists.
5020 (view_list_to_loc_list_val_node): New.
5021 (print_dw_val): Handle dw_val_class_view_list.
5022 (size_of_die): Likewise.
5023 (value_format): Likewise.
5024 (loc_list_has_views): New.
5025 (gen_llsym): Set vl_symbol too.
5026 (maybe_gen_llsym, skip_loc_list_entry): New.
5027 (dwarf2out_maybe_output_loclist_view_pair): New.
5028 (output_loc_list): Output view list or entries too.
5029 (output_view_list_offset): New.
5030 (output_die): Handle dw_val_class_view_list.
5031 (output_dwarf_version): New.
5032 (output_compilation_unit_header): Use it.
5033 (output_skeleton_debug_sections): Likewise.
5034 (output_rnglists, output_line_info): Likewise.
5035 (output_pubnames, output_aranges): Update version comments.
5036 (output_one_line_info_table): Output view numbers in asm comments.
5037 (dw_loc_list): Determine current endview, pass it to new_loc_list.
5038 Call maybe_gen_llsym.
5039 (loc_list_from_tree_1): Adjust.
5040 (add_AT_location_description): Create view list attribute if
5041 needed, check it's absent otherwise.
5042 (convert_cfa_to_fb_loc_list): Adjust.
5043 (maybe_emit_file): Call output_asm_line_debug_info for test.
5044 (dwarf2out_var_location): Reset views as needed. Precompute
5045 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
5046 attribute. Set view.
5047 (new_line_info_table): Reset next view.
5048 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
5049 (dwarf2out_source_line): Likewise. Output view resets and labels to
5050 the assembler, or select appropriate line info opcodes.
5051 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
5052 (optimize_string_length): Catch it. Adjust.
5053 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
5054 dw_val_class_view_list, and remove it if no longer needed.
5055 (hash_loc_list): Hash view numbers.
5056 (loc_list_hasher::equal): Compare them.
5057 (optimize_location_lists): Check whether a view list symbol is
5058 needed, and whether the locview attribute is present, and
5059 whether they match. Remove the locview attribute if no longer
5061 (index_location_lists): Call skip_loc_list_entry for test.
5062 (dwarf2out_finish): Call output_asm_line_debug_info for test.
5063 Use output_dwarf_version.
5064 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
5065 (struct dw_val_node): Add val_view_list.
5066 * final.c (SEEN_NEXT_VIEW): New.
5067 (set_next_view_needed): New.
5068 (clear_next_view_needed): New.
5069 (maybe_output_next_view): New.
5070 (final_start_function): Rename to...
5071 (final_start_function_1): ... this. Take pointer to FIRST,
5072 add SEEN parameter. Emit param bindings in the initial view.
5073 (final_start_function): Reintroduce SEEN-less interface.
5074 (final): Rename to...
5075 (final_1): ... this. Take SEEN parameter. Output final pending
5076 next view at the end.
5077 (final): Reintroduce seen-less interface.
5078 (final_scan_insn): Output pending next view before switching
5079 sections or ending a block. Mark the next view as needed when
5080 outputting variable locations. Notify debug backend of section
5081 changes, and of location view changes.
5082 (rest_of_handle_final): Adjust.
5083 * toplev.c (process_options): Autodetect value for debug variable
5084 location views option. Warn on incompat5 without -gdwarf-5.
5085 * doc/invoke.texi (gvariable-location-views): New.
5086 (gvariable-location-views=incompat5): New.
5087 (gno-variable-location-views): New.
5089 2018-02-08 David Malcolm <dmalcolm@redhat.com>
5091 PR tree-optimization/84136
5092 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
5093 that the result of find_edge is non-NULL.
5095 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5098 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
5099 storing integer register in SImode. Fix cost of 256 and 512
5100 byte aligned SSE register store.
5102 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5104 * config/i386/i386.c (ix86_multiplication_cost): Fix
5105 multiplication cost for TARGET_AVX512DQ.
5107 2018-02-08 Marek Polacek <polacek@redhat.com>
5109 PR tree-optimization/84238
5110 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
5113 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5115 PR tree-optimization/84265
5116 * tree-vect-stmts.c (vectorizable_store): Don't treat
5117 VMAT_CONTIGUOUS accesses as grouped.
5118 (vectorizable_load): Likewise.
5120 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5122 PR tree-optimization/81635
5123 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
5124 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
5125 (test_round_for_mask): New functions.
5126 (wide_int_cc_tests): Call test_round_for_mask.
5127 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
5128 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
5129 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
5130 range returned by get_range_info.
5132 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5135 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
5136 * symtab.c: Include builtins.h
5137 (symtab_node::output_to_lto_symbol_table_p): Move here
5138 from lto-streamer-out.c:output_symbol_p.
5139 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
5140 (output_symbol_p): Move all logic to symtab.c
5141 (produce_symtab): Update.
5143 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5145 * config/s390/s390-opts.h (enum indirect_branch): Define.
5146 * config/s390/s390-protos.h (s390_return_addr_from_memory)
5147 (s390_indirect_branch_via_thunk)
5148 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
5149 (enum s390_indirect_branch_type): Define.
5150 * config/s390/s390.c (struct s390_frame_layout, struct
5151 machine_function): Remove.
5152 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
5153 (indirect_branch_table_label_no, indirect_branch_table_name):
5155 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
5156 (enum s390_indirect_branch_option): Define.
5157 (s390_return_addr_from_memory): New function.
5158 (s390_handle_string_attribute): New function.
5159 (s390_attribute_table): Add new attribute handler.
5160 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
5161 (s390_indirect_branch_via_thunk): New function.
5162 (s390_indirect_branch_via_inline_thunk): New function.
5163 (s390_function_ok_for_sibcall): When jumping via thunk disallow
5164 sibling call optimization for non z10 compiles.
5165 (s390_emit_call): Force indirect branch target to be a single
5166 register. Add r1 clobber for non-z10 compiles.
5167 (s390_emit_epilogue): Emit return jump via return_use expander.
5168 (s390_reorg): Handle JUMP_INSNs as execute targets.
5169 (s390_option_override_internal): Perform validity checks for the
5170 new command line options.
5171 (s390_indirect_branch_attrvalue): New function.
5172 (s390_indirect_branch_settings): New function.
5173 (s390_set_current_function): Invoke s390_indirect_branch_settings.
5174 (s390_output_indirect_thunk_function): New function.
5175 (s390_code_end): Implement target hook.
5176 (s390_case_values_threshold): Implement target hook.
5177 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
5179 * config/s390/s390.h (struct s390_frame_layout)
5180 (struct machine_function): Move here from s390.c.
5181 (TARGET_INDIRECT_BRANCH_NOBP_RET)
5182 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
5183 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
5184 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
5185 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
5186 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
5187 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
5188 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
5189 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
5190 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
5191 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
5192 (mnemonic attribute): Add values which aren't recognized
5194 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
5195 pattern for branch conversion. Fix mnemonic attribute.
5196 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
5197 indirect branch via thunk if requested.
5198 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
5199 ("*indirect_jump"): Disable for branch conversion using out of
5201 ("indirect_jump_via_thunk<mode>_z10")
5202 ("indirect_jump_via_thunk<mode>")
5203 ("indirect_jump_via_inlinethunk<mode>_z10")
5204 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
5205 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
5206 ("casesi_jump_via_inlinethunk<mode>_z10")
5207 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
5208 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
5209 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
5210 ("*indirect2_jump"): Disable for branch conversion.
5211 ("casesi_jump"): Turn into expander and expand patterns for branch
5213 ("return_use"): New expander.
5214 ("*return"): Emit return via thunk and rename it to ...
5215 ("*return<mode>"): ... this one.
5216 * config/s390/s390.opt: Add new options and and enum for the
5219 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5221 * lra-constraints.c (match_reload): Unconditionally use
5222 gen_lowpart_SUBREG, rather than selecting between that
5223 and equivalent gen_rtx_SUBREG code.
5225 2018-02-08 Richard Biener <rguenther@suse.de>
5227 PR tree-optimization/84233
5228 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
5229 changed flag instead of boguously re-using phi_inserted.
5231 2018-02-08 Martin Jambor <mjambor@suse.cz>
5233 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
5234 static local variables.
5236 2018-02-08 Richard Biener <rguenther@suse.de>
5238 PR tree-optimization/84278
5239 * tree-vect-stmts.c (vectorizable_store): When looking for
5240 smaller vector types to perform grouped strided loads/stores
5241 make sure the mode is supported by the target.
5242 (vectorizable_load): Likewise.
5244 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5246 * config/aarch64/aarch64.c (aarch64_components_for_bb):
5247 Increase LDP/STP opportunities by adding adjacent callee-saves.
5249 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5251 PR rtl-optimization/84068
5252 PR rtl-optimization/83459
5253 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
5255 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
5257 PR tree-optimization/84224
5258 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
5259 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
5262 2018-02-07 Iain Sandoe <iain@codesourcery.com>
5265 * config/rs6000/altivec.md (*restore_world): Remove LR use.
5266 * config/rs6000/predicates.md (restore_world_operation): Adjust op
5267 count, remove one USE.
5269 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
5271 * doc/install.texi (Configuration): Document the
5272 --with-long-double-format={ibm,ieee} PowerPC configuration
5276 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
5277 Convert from define_expand to be define_insn_and_split. Rework
5278 float/double/_Float128 conversions to QI/HI/SImode to work with
5279 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
5280 conversions to QI/HImode types did a store and then a load to
5281 truncate the value. For conversions to VSX registers, don't split
5282 the insn, instead emit the code directly. Use the code iterator
5283 any_fix to combine signed and unsigned conversions.
5284 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
5285 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
5286 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5287 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
5288 (fix_<mode>di2_hw): Likewise.
5289 (fixuns_<mode>di2_hw): Likewise.
5290 (fix_<mode>si2_hw): Likewise.
5291 (fixuns_<mode>si2_hw): Likewise.
5292 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
5293 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5294 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
5295 fix<uns>_trunc<SFDF:mode>si2_p8.
5296 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
5298 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
5299 (fix<uns>_<mode>_mem): Likewise.
5300 (fctiw<u>z_<mode>_mem): Likewise.
5301 (fix<uns>_<mode>_mem): Likewise.
5302 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
5303 the register allocator from doing a direct move to the GPRs to do
5304 a store, and instead use the ISA 3.0 store byte/half-word from
5305 vector register instruction. For IEEE 128-bit floating point,
5306 also optimize stores of 32-bit ints.
5307 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
5309 2018-02-07 Alan Hayward <alan.hayward@arm.com>
5311 * genextract.c (push_pathstr_operand): New function to support
5313 (walk_rtx): Call push_pathstr_operand.
5314 (print_path): Support [a-zA-Z].
5316 2018-02-07 Richard Biener <rguenther@suse.de>
5318 PR tree-optimization/84037
5319 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
5320 (cse_and_gimplify_to_preheader): Declare.
5321 (vect_get_place_in_interleaving_chain): Likewise.
5322 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5324 (_loop_vec_info::~_loop_vec_info): Delete it.
5325 (cse_and_gimplify_to_preheader): New function.
5326 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
5327 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
5328 (vectorizable_load): Likewise. For grouped stores always base
5329 the IV on the first element.
5330 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
5331 condition before gimplifying.
5333 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5335 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
5336 *DIV_EXPR and *MOD_EXPR.
5338 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
5341 * config/i386/i386.c (ix86_option_override_internal): Mask out
5342 the CF_SET bit when checking -fcf-protection.
5344 2018-02-07 Tom de Vries <tom@codesourcery.com>
5347 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
5350 2018-02-07 Richard Biener <rguenther@suse.de>
5352 PR tree-optimization/84204
5353 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
5356 PR tree-optimization/84205
5357 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
5358 special-case isl_ast_op_zdiv_r.
5360 PR tree-optimization/84223
5361 * graphite-scop-detection.c (gather_bbs::before_dom_children):
5362 Only add conditions from within the region.
5363 (gather_bbs::after_dom_children): Adjust.
5365 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
5368 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
5369 * config/avr/avr.md: Only post-reload split REG-REG moves if
5370 either register is GENERAL_REG_P.
5372 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5374 PR tree-optimization/84235
5375 * tree-ssa-scopedtables.c
5376 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
5377 if the subtraction is performed in floating point type where NaNs are
5378 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
5379 build 1. Formatting fix.
5381 2018-02-06 Jakub Jelinek <jakub@redhat.com>
5384 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
5385 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
5386 and skip it regardless of bb boundaries. Use CALL_P macro,
5387 don't test INSN_P (insn) together with CALL_P or JUMP_P check
5388 unnecessarily, formatting fix.
5390 2018-02-06 Michael Collison <michael.collison@arm.com>
5392 * config/arm/thumb2.md:
5393 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
5394 (*thumb_mov_notscc): Ditto.
5396 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
5399 * config/rs6000/rs6000.md (su code attribute): Use "u" for
5400 unsigned_fix, not "s".
5402 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5404 * configure.ac (gcc_fn_eh_frame_ro): New function.
5405 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
5406 correct .eh_frame permissions.
5407 * configure: Regenerate.
5409 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
5411 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
5414 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5416 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5417 Display warning message for -mno-speculate-indirect-jumps.
5419 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
5421 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
5423 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
5425 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
5427 PR tree-optimization/84225
5428 * tree-eh.c (find_trapping_overflow): Only call
5429 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
5431 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5434 * config/i386/i386.c: Reimplement the check of possible options
5435 -mibt/-mshstk conbination. Change error messages.
5436 * doc/invoke.texi: Fix a typo: remove extra '='.
5438 2018-02-06 Marek Polacek <polacek@redhat.com>
5440 PR tree-optimization/84228
5441 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
5443 2018-02-06 Tamar Christina <tamar.christina@arm.com>
5446 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
5447 emitted arch directives.
5448 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
5449 __ARM_FEATURE_COPROC before changing architectures.
5451 2018-02-06 Richard Biener <rguenther@suse.de>
5453 * config/i386/i386.c (print_reg): Fix typo.
5454 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
5456 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
5458 * configure: Regenerate.
5460 2018-02-05 Martin Sebor <msebor@redhat.com>
5462 PR tree-optimization/83369
5463 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
5466 2018-02-05 Martin Liska <mliska@suse.cz>
5468 * doc/invoke.texi: Cherry-pick upstream r323995.
5470 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
5472 * ira.c (ira_init_register_move_cost): Adjust comment.
5474 2018-02-05 Martin Liska <mliska@suse.cz>
5476 PR gcov-profile/84137
5477 * doc/gcov.texi: Fix typo in documentation.
5479 2018-02-05 Martin Liska <mliska@suse.cz>
5481 PR gcov-profile/83879
5482 * doc/gcov.texi: Document necessity of --dynamic-list-data when
5483 using dlopen functionality.
5485 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
5487 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
5488 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
5489 _mm_maskz_range_ss, _mm_mask_range_round_ss,
5490 _mm_maskz_range_round_ss): New intrinsics.
5491 (__builtin_ia32_rangesd128_round)
5492 (__builtin_ia32_rangess128_round): Remove.
5493 (__builtin_ia32_rangesd128_mask_round,
5494 __builtin_ia32_rangess128_mask_round): New builtins.
5495 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
5496 __builtin_ia32_rangess128_round): Remove.
5497 (__builtin_ia32_rangesd128_mask_round,
5498 __builtin_ia32_rangess128_mask_round): New builtins.
5499 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
5500 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
5501 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
5502 "<round_saeonly_constraint>")): Changed to ...
5503 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
5504 "<round_saeonly_scalar_constraint>")): ... this.
5505 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
5506 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
5507 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
5508 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
5509 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
5511 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
5513 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
5515 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
5516 Remove all values except native, 8540 and 8548.
5518 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
5520 * config/i386/i386.c (ix86_output_function_return): Pass
5521 INVALID_REGNUM, instead of -1, as invalid register number to
5522 indirect_thunk_name and output_indirect_thunk.
5524 2018-02-02 Julia Koval <julia.koval@intel.com>
5526 * config.gcc: Add -march=icelake.
5527 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
5528 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
5529 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
5530 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
5531 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
5532 (processor_target_table): Add icelake.
5533 (ix86_option_override_internal): Handle new PTAs.
5534 (get_builtin_code_for_version): Handle icelake.
5535 (M_INTEL_COREI7_ICELAKE): New.
5536 (fold_builtin_cpu): Handle icelake.
5537 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
5538 * doc/invoke.texi: Add -march=icelake.
5540 2018-02-02 Julia Koval <julia.koval@intel.com>
5542 * config/i386/i386.c (ix86_option_override_internal): Change flags type
5543 to wide_int_bitmask.
5544 * wide-int-bitmask.h: New.
5546 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5549 * config/i386/i386.md: Replace Pmode with word_mode in
5550 builtin_setjmp_setup and builtin_longjmp to support x32.
5552 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
5556 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
5558 (rs6000_supported_cpu_names): New static variable.
5559 (linux_cpu_translation_table): Likewise.
5560 (elf_platform) <cpu>: Define new static variable and use it.
5561 Translate kernel AT_PLATFORM name to canonical name if needed.
5562 Error if platform name is unknown.
5564 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
5567 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
5569 2018-02-01 Jeff Law <law@redhat.com>
5572 * config/i386/i386.c (release_scratch_register_on_entry): Add new
5573 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
5574 the scratch if RELEASE_VIA_POP is false.
5575 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
5576 If we have to save a temporary register, decrement SIZE appropriately.
5577 Pass new arguments to release_scratch_register_on_entry.
5578 (ix86_adjust_stack_and_probe): Likewise.
5579 (ix86_emit_probe_stack_range): Pass new arguments to
5580 release_scratch_register_on_entry.
5582 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
5584 PR rtl-optimization/84157
5585 * combine.c (change_zero_ext): Use REG_P predicate in
5586 front of HARD_REGISTER_P predicate.
5588 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
5590 * config/avr/avr.c (avr_option_override): Move disabling of
5591 -fdelete-null-pointer-checks to...
5592 * common/config/avr/avr-common.c (avr_option_optimization_table):
5595 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5597 PR tree-optimization/81635
5598 * tree-data-ref.c (split_constant_offset_1): For types that
5599 wrap on overflow, try to use range info to prove that wrapping
5602 2018-02-01 Renlin Li <renlin.li@arm.com>
5605 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
5607 (aarch64_register_move_cost): Likewise.
5608 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
5610 (REG_CLASS_NAMES): Likewise.
5611 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
5612 TAILCALL_ADDR_REGS. Remove IP registers.
5613 * config/aarch64/aarch64.md (Ucs): Update register constraint.
5615 2018-02-01 Richard Biener <rguenther@suse.de>
5617 * domwalk.h (dom_walker::dom_walker): Add additional constructor
5618 for specifying RPO order and allow NULL for that.
5619 * domwalk.c (dom_walker::dom_walker): Likewise.
5620 (dom_walker::walk): Handle NULL RPO order.
5621 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
5623 (rewrite_update_dom_walker): Likewise.
5624 (mark_def_dom_walker): Likewise.
5626 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5628 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
5629 (aarch64_maybe_expand_sve_subreg_move): Declare.
5630 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
5631 * config/aarch64/predicates.md (aarch64_any_register_operand): New
5633 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
5634 that are semantically a reverse operation.
5635 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
5636 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
5637 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
5639 (aarch64_can_change_mode_class): For big-endian, forbid changes
5640 between two SVE modes if they have different element sizes.
5642 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5644 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
5645 the TImode handling for big-endian targets.
5647 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5649 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
5650 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
5652 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
5653 Remove BSWAP handing for big-endian targets and use the form of
5654 LD1RQ appropariate for the mode.
5656 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5658 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
5659 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
5662 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5665 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
5666 check for operands that need to go through aarch64_sve_reload_be.
5668 2018-02-01 Jakub Jelinek <jakub@redhat.com>
5670 PR tree-optimization/81661
5671 PR tree-optimization/84117
5672 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
5673 * tree-eh.c: Include gimplify.h.
5674 (find_trapping_overflow, replace_trapping_overflow,
5675 rewrite_to_non_trapping_overflow): New functions.
5676 * tree-vect-loop.c: Include tree-eh.h.
5677 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
5678 * tree-data-ref.c: Include tree-eh.h.
5679 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
5681 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
5683 PR rtl-optimization/84123
5684 * combine.c (change_zero_ext): Check if hard register satisfies
5685 can_change_dest_mode before calling gen_lowpart_SUBREG.
5687 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
5690 * ira.c (ira_init_register_move_cost): Remove assert.
5692 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5694 PR rtl-optimization/84071
5695 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
5696 * doc/tm.texi: Regenerate.
5698 2018-01-31 Richard Biener <rguenther@suse.de>
5700 PR tree-optimization/84132
5701 * tree-data-ref.c (analyze_miv_subscript): Properly
5702 check whether evolution_function_is_affine_multivariate_p
5703 before calling gcd_of_steps_may_divide_p.
5705 2018-01-31 Julia Koval <julia.koval@intel.com>
5708 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
5709 * config/i386/i386.md (rdpid_rex64) New.
5710 (rdpid): Make 32bit only.
5712 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
5715 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
5716 an IDENTIFIER_NODE for FUNCTION_TYPE's.
5718 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5721 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5723 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5725 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5727 PR rtl-optimization/84071
5728 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
5729 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
5731 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5733 * config/arc/arc.c (arc_handle_aux_attribute): New function.
5734 (arc_attribute_table): Add 'aux' attribute.
5735 (arc_in_small_data_p): Consider aux like variables.
5736 (arc_is_aux_reg_p): New function.
5737 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
5738 (arc_get_aux_arg): New function.
5739 (prepare_move_operands): Handle aux-register access.
5740 (arc_handle_aux_attribute): New function.
5741 * doc/extend.texi (ARC Variable attributes): Add subsection.
5743 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5745 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
5746 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
5747 (arc_attribute_table): Add 'uncached' attribute.
5748 (arc_print_operand): Print '.di' flag for uncached memory
5750 (arc_in_small_data_p): Do not consider for small data the uncached
5752 (arc_is_uncached_mem_p): New function.
5753 * config/arc/predicates.md (compact_store_memory_operand): Check
5754 for uncached memory accesses.
5755 (nonvol_nonimm_operand): Likewise.
5756 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
5758 2018-01-31 Jakub Jelinek <jakub@redhat.com>
5761 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
5762 falign-loops=): Add Optimization flag.
5764 2018-01-30 Jeff Law <law@redhat.com>
5767 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
5768 INT_REGISTERS_SAVED. Check it prior to calling
5769 get_scratch_register_on_entry.
5770 (ix86_adjust_stack_and_probe): Similarly.
5771 (ix86_emit_probe_stack_range): Similarly.
5772 (ix86_expand_prologue): Corresponding changes.
5774 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5777 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
5778 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
5780 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
5783 * lra-constraints.c (curr_insn_transform): Process AND in the
5786 2018-01-30 Jakub Jelinek <jakub@redhat.com>
5788 PR rtl-optimization/83986
5789 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
5790 dependence against last_pending_memory_flush in addition to
5793 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
5795 PR tree-optimization/81611
5796 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
5799 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5802 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
5805 2018-01-30 Richard Biener <rguenther@suse.de>
5806 Jakub Jelinek <jakub@redhat.com>
5808 PR tree-optimization/84111
5809 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
5810 inner loops added during recursion, as they don't have up-to-date
5813 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5816 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
5817 (can_inline_edge_by_limits_p): ... here.
5818 (can_early_inline_edge_p, check_callers,
5819 update_caller_keys, update_callee_keys, recursive_inlining,
5820 add_new_edges_to_heap, speculation_useful_p,
5821 inline_small_functions,
5822 inline_small_functions, flatten_function,
5823 inline_to_all_callers_1): Update.
5825 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5827 * profile-count.c (profile_count::combine_with_ipa_count): Handle
5830 2018-01-30 Richard Biener <rguenther@suse.de>
5832 PR tree-optimization/83008
5833 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
5834 invariant and constant vector uses in stmts when they need
5837 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5840 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
5841 * configure: Regenerate.
5843 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5845 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
5847 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
5848 Use gen_rtx_REG rather than gen_lowpart.
5850 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5852 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
5853 rather than 0 when creating partial subregs.
5855 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5857 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
5860 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
5863 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
5864 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
5865 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
5866 flags. This restores the settings used before the 2017-07-24.
5867 Turning off pre increment/decrement/modify allows IVOPTS to
5868 optimize DF/SF loops where the index is an int.
5870 2018-01-29 Richard Biener <rguenther@suse.de>
5871 Kelvin Nilsen <kelvin@gcc.gnu.org>
5874 * tree-vect-stmts.c (vectorizable_call): Don't call
5875 targetm.vectorize_builtin_md_vectorized_function if callee is
5878 2018-01-22 Carl Love <cel@us.ibm.com>
5880 * doc/extend.tex: Fix typo in second arg in
5881 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
5883 2018-01-29 Richard Biener <rguenther@suse.de>
5885 PR tree-optimization/84086
5886 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
5887 (flush_ssaname_freelist): When SSA names were released reset
5888 the SCEV hash table.
5890 2018-01-29 Richard Biener <rguenther@suse.de>
5892 PR tree-optimization/84057
5893 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
5894 removed paths when removing edges.
5896 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
5898 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
5899 -mfunction-return=@var{choice}.
5901 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
5904 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
5906 (layout::print_source_line): Likewise.
5907 (test_get_line_width_without_trailing_whitespace): Add test cases.
5909 2018-01-27 Jakub Jelinek <jakub@redhat.com>
5912 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
5915 2018-01-26 Jim Wilson <jimw@sifive.com>
5917 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
5919 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
5922 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5924 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
5925 and CMP + SUB-immediate -> SUBS.
5927 2018-01-26 Martin Sebor <msebor@redhat.com>
5929 PR tree-optimization/83896
5930 * tree-ssa-strlen.c (get_string_len): Rename...
5931 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
5932 Avoid assuming length is constant.
5933 (handle_char_store): Use HOST_WIDE_INT for string length.
5935 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
5938 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
5939 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
5941 2018-01-26 Richard Biener <rguenther@suse.de>
5943 PR rtl-optimization/84003
5944 * dse.c (record_store): Only record redundant stores when
5945 the earlier store aliases at least all accesses the later one does.
5947 2018-01-26 Jakub Jelinek <jakub@redhat.com>
5949 PR rtl-optimization/83985
5950 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
5951 REG_CFA_RESTORE insns.
5952 (delete_unmarked_insns): Don't ignore separate shrink wrapping
5953 REG_CFA_RESTORE insns here.
5956 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
5957 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
5959 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5961 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
5962 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
5963 (arc_init): Likewise.
5964 (arc_override_options): Likewise.
5965 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
5967 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
5969 * config/arc/arc.h (TARGET_DBNZ): Define.
5970 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
5971 properly set the tune attribute.
5972 (dbnz): Use TARGET_DBNZ guard.
5973 * config/arc/arc.opt (mtune): Add core3 option.
5975 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5977 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
5978 recognize new pic like addresses.
5979 (arc_delegitimize_address): Clean up.
5981 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5983 * config/arc/arc-arches.def: Option mrf16 valid for all
5985 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
5986 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
5987 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
5988 * config/arc/arc-tables.opt: Regenerate.
5989 * config/arc/arc.c (arc_conditional_register_usage): Handle
5990 reduced register file case.
5991 (arc_file_start): Set must have build attributes.
5992 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
5994 * config/arc/arc.opt (mrf16): Add new option.
5995 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
5996 * config/arc/genmultilib.awk: Handle new mrf16 option.
5997 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
5998 * config/arc/t-multilib: Regenerate.
5999 * doc/invoke.texi (ARC Options): Document mrf16 option.
6001 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6003 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
6004 * config/arc/arc.c (arc_handle_secure_attribute): New function.
6005 (arc_attribute_table): Add 'secure_call' attribute.
6006 (arc_print_operand): Print secure call operand.
6007 (arc_function_ok_for_sibcall): Don't optimize tail calls when
6009 (arc_is_secure_call_p): New function. * config/arc/arc.md
6010 (call_i): Add support for sjli instruction.
6011 (call_value_i): Likewise.
6012 * config/arc/constraints.md (Csc): New constraint.
6014 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6015 John Eric Martin <John.Martin@emmicro-us.com>
6017 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
6018 * config/arc/arc.c (_arc_jli_section): New struct.
6019 (arc_jli_section): New type.
6020 (rc_jli_sections): New static variable.
6021 (arc_handle_jli_attribute): New function.
6022 (arc_attribute_table): Add jli_always and jli_fixed attribute.
6023 (arc_file_end): New function.
6024 (TARGET_ASM_FILE_END): Define.
6025 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
6026 (arc_add_jli_section): New function.
6027 (jli_call_scan): Likewise.
6028 (arc_reorg): Call jli_call_scan.
6029 (arc_output_addsi): Remove 'S' from printing asm operand.
6030 (arc_is_jli_call_p): New function.
6031 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
6033 (movhi_insn): Likewise.
6034 (movsi_insn): Likewise.
6035 (movsi_set_cc_insn): Likewise.
6036 (loadqi_update): Likewise.
6037 (load_zeroextendqisi_update): Likewise.
6038 (load_signextendqisi_update): Likewise.
6039 (loadhi_update): Likewise.
6040 (load_zeroextendhisi_update): Likewise.
6041 (load_signextendhisi_update): Likewise.
6042 (loadsi_update): Likewise.
6043 (loadsf_update): Likewise.
6044 (movsicc_insn): Likewise.
6045 (bset_insn): Likewise.
6046 (bxor_insn): Likewise.
6047 (bclr_insn): Likewise.
6048 (bmsk_insn): Likewise.
6049 (bicsi3_insn): Likewise.
6050 (cmpsi_cc_c_insn): Likewise.
6051 (movsi_ne): Likewise.
6052 (movsi_cond_exec): Likewise.
6053 (clrsbsi2): Likewise.
6064 (call_i): Remove 'S' asm letter, add jli instruction.
6065 (call_value_i): Likewise.
6066 * config/arc/arc.op (mjli-always): New option.
6067 * config/arc/constraints.md (Cji): New constraint.
6068 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
6070 (subsf3_fpx): Likewise.
6071 (mulsf3_fpx): Likewise.
6072 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
6074 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
6076 * doc/invoke.texi (ARC): Document mjli-always option.
6078 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
6080 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
6081 avoid addition with 0 and use incw and decw where possible.
6083 2018-01-26 Richard Biener <rguenther@suse.de>
6085 PR tree-optimization/81082
6086 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
6087 association if it requires casting to unsigned.
6088 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
6089 from fold_plusminus_mult_expr to catch important cases late when
6090 range info is available.
6092 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6094 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
6095 * configure.ac (hidden_linkonce): New test.
6096 * configure: Regenerate.
6097 * config.in: Regenerate.
6099 2018-01-26 Julia Koval <julia.koval@intel.com>
6101 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
6102 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
6103 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
6104 _mm_mask_bitshuffle_epi64_mask): Fix type.
6105 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
6106 USI_FTYPE_V4DI_V4DI_USI): Remove.
6107 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
6108 __builtin_ia32_vpshufbitqmb256_mask,
6109 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
6110 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
6111 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
6113 2018-01-26 Alan Modra <amodra@gmail.com>
6116 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
6117 UNSPEC_VBPERMQ. Sort other unspecs.
6119 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
6121 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
6123 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
6126 * predict.c (drop_profile): Do not push/pop cfun; update also
6128 (handle_missing_profiles): Fix logic looking for zero profiles.
6130 2018-01-25 Jakub Jelinek <jakub@redhat.com>
6133 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
6134 on functions with #pragma omp declare simd or functions with simd
6136 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
6137 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
6138 Remove trailing \n from warning_at calls.
6140 2018-01-25 Tom de Vries <tom@codesourcery.com>
6143 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6144 for neutered workers.
6146 2018-01-24 Joseph Myers <joseph@codesourcery.com>
6149 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
6150 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
6152 2018-01-24 Jeff Law <law@redhat.com>
6155 * i386.c (get_probe_interval): Move to earlier point.
6156 (ix86_compute_frame_layout): If -fstack-clash-protection and
6157 the frame is larger than the probe interval, then use pushes
6158 to save registers rather than reg->mem moves.
6159 (ix86_expand_prologue): Remove conditional for int_registers_saved
6162 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
6165 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
6166 min/max for never referenced object.
6168 2018-01-24 Jakub Jelinek <jakub@redhat.com>
6171 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
6173 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
6174 attributes from DECL_ATTRIBUTES (decl) without affecting
6175 DECL_ATTRIBUTES (current_function_decl).
6176 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
6177 functions with non-NULL DECL_ABSTRACT_ORIGIN.
6179 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
6181 PR tree-optimization/83979
6182 * fold-const.c (fold_comparison): Use constant_boolean_node
6183 instead of boolean_{true,false}_node.
6185 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
6187 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
6190 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6192 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6193 Simplify the clause that sets the length attribute.
6194 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6195 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
6196 clause that sets the length attribute.
6197 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6199 2018-01-24 Tom de Vries <tom@codesourcery.com>
6202 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
6203 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
6204 Add strict parameter.
6205 (prevent_branch_around_nothing): Insert dummy insn between branch to
6206 label and label with no ptx insn inbetween.
6207 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
6209 2018-01-24 Tom de Vries <tom@codesourcery.com>
6212 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6213 for neutered threads in warp.
6214 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
6216 2018-01-24 Richard Biener <rguenther@suse.de>
6218 PR tree-optimization/83176
6219 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
6222 2018-01-24 Richard Biener <rguenther@suse.de>
6224 PR tree-optimization/82819
6225 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
6226 code generating pluses that are no-ops in the target precision.
6228 2018-01-24 Richard Biener <rguenther@suse.de>
6231 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
6233 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6235 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
6236 to merge probabilities.
6237 * predict.c (probably_never_executed): Also mark as cold functions
6238 with global 0 profile and guessed local profile.
6239 * profile-count.c (profile_probability::combine_with_count): New
6241 * profile-count.h (profile_probability::operator*,
6242 profile_probability::operator*=, profile_probability::operator/,
6243 profile_probability::operator/=): Reduce precision to adjusted
6244 and set value to guessed on contradictory divisions.
6245 (profile_probability::combine_with_freq): Remove.
6246 (profile_probability::combine_wiht_count): Declare.
6247 (profile_count::force_nonzero):: Set to adjusted.
6248 (profile_count::probability_in):: Set quality to adjusted.
6249 * tree-ssa-tail-merge.c (replace_block_by): Use
6252 2018-01-23 Andrew Waterman <andrew@sifive.com>
6253 Jim Wilson <jimw@sifive.com>
6255 * config/riscv/riscv.c (riscv_stack_boundary): New.
6256 (riscv_option_override): Set riscv_stack_boundary. Handle
6257 riscv_preferred_stack_boundary_arg.
6258 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
6259 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
6260 (STACK_BOUNDARY): Set to riscv_stack_boundary.
6261 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
6262 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
6263 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
6265 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
6268 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
6269 of struct ix86_frame.
6270 (ix86_expand_epilogue): Likewise. Add a local variable for
6271 the reg_save_offset field in struct ix86_frame.
6273 2018-01-23 Bin Cheng <bin.cheng@arm.com>
6275 PR tree-optimization/82604
6276 * tree-loop-distribution.c (enum partition_kind): New enum item
6277 PKIND_PARTIAL_MEMSET.
6278 (partition_builtin_p): Support above new enum item.
6279 (generate_code_for_partition): Ditto.
6280 (compute_access_range): Differentiate cases that equality can be
6281 proven at all loops, the innermost loops or no loops.
6282 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
6283 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
6284 (finalize_partitions, distribute_loop): Don't fuse partition of
6285 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
6286 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
6289 2018-01-23 Martin Liska <mliska@suse.cz>
6291 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
6292 order to ignore the predictor.
6293 (PRED_POLYMORPHIC_CALL): Likewise.
6294 (PRED_RECURSIVE_CALL): Likewise.
6296 2018-01-23 Martin Liska <mliska@suse.cz>
6298 * tree-profile.c (tree_profiling): Print function header to
6299 aware reader which function we are working on.
6300 * value-prof.c (gimple_find_values_to_profile): Do not print
6301 not interesting value histograms.
6303 2018-01-23 Martin Liska <mliska@suse.cz>
6305 * profile-count.h (enum profile_quality): Add
6306 profile_uninitialized as the first value. Do not number values
6307 as they are zero based.
6308 (profile_count::verify): Update sanity check.
6309 (profile_probability::verify): Likewise.
6311 2018-01-23 Nathan Sidwell <nathan@acm.org>
6313 * doc/invoke.texi (ffor-scope): Deprecate.
6315 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6317 PR tree-optimization/83510
6318 * domwalk.c (set_all_edges_as_executable): New function.
6319 (dom_walker::dom_walker): Convert bool param
6320 "skip_unreachable_blocks" to enum reachability. Move setup of
6321 edge flags to set_all_edges_as_executable and only do it when
6322 reachability is REACHABLE_BLOCKS.
6323 * domwalk.h (enum dom_walker::reachability): New enum.
6324 (dom_walker::dom_walker): Convert bool param
6325 "skip_unreachable_blocks" to enum reachability.
6326 (set_all_edges_as_executable): New decl.
6327 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
6328 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
6330 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
6331 but converting true to REACHABLE_BLOCKS.
6332 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
6334 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
6335 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
6336 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
6338 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
6339 if check_all_array_refs will be called.
6341 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6343 * tree.c (selftest::test_location_wrappers): Add more test
6346 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6348 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
6349 (selftest::test_bit_in_range): Likewise.
6351 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6354 * doc/sourcebuild.texi (vect_float): Say that the selector
6355 only describes the situation when -funsafe-math-optimizations is on.
6356 (vect_float_strict): Document.
6358 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6360 PR tree-optimization/83965
6361 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
6362 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
6363 instead of checking only for a reduction.
6364 (vect_recog_widen_sum_pattern): Likewise.
6366 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6368 * predict.c (probably_never_executed): Only use precise profile info.
6369 (compute_function_frequency): Skip after inlining hack since we now
6370 have quality checking.
6372 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6374 * profile-count.h (profile_probability::very_unlikely,
6375 profile_probability::unlikely, profile_probability::even): Set
6376 precision to guessed.
6378 2018-01-23 Richard Biener <rguenther@suse.de>
6380 PR tree-optimization/83963
6381 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
6382 Properly terminate dominator walk when crossing the exit edge not
6383 when visiting its source block.
6385 2018-01-23 Jakub Jelinek <jakub@redhat.com>
6388 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
6389 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
6391 2018-01-22 Jakub Jelinek <jakub@redhat.com>
6393 PR tree-optimization/83957
6394 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
6395 semicolon after for body surrounded by braces.
6397 PR tree-optimization/83081
6398 * profile-count.h (profile_probability::split): New method.
6399 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
6400 Use profile_probability::split.
6401 (do_compare_rtx_and_jump): Fix adjustment of probabilities
6402 when splitting a single conditional jump into 2.
6404 2018-01-22 David Malcolm <dmalcolm@redhat.com>
6406 PR tree-optimization/69452
6407 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
6410 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6412 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
6413 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
6414 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
6416 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6418 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
6419 * config/rl78/rl78.md (movdi): New define_expand.
6420 * config/rl78/rl78.c (rl78_split_movdi): New function.
6422 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
6425 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
6427 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
6428 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
6429 128-bit to produce an UNSPEC move to get the double word with the
6430 signbit and then a shift directly to do signbit.
6431 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
6432 implementation with a new version that just does either a direct
6433 move or a regular move. Move memory interface to separate insns.
6434 Move insns so they are next to the expander.
6435 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
6436 with signbit move. Split big and little endian case.
6437 (signbit<mode>2_dm_mem_le): Likewise.
6438 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
6439 (signbit<mode>2_dm2): Likewise.
6441 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6443 * config/rl78/rl78.md (anddi3): New define_expand.
6445 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6447 * config/rl78/rl78.md (umindi3): New define_expand.
6449 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6451 * config/rl78/rl78.md (smindi3): New define_expand.
6453 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6455 * config/rl78/rl78.md (smaxdi3): New define_expand.
6457 2018-01-22 Carl Love <cel@us.ibm.com>
6459 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
6460 LVX_V1TI): Add macro expansion.
6461 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
6462 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
6463 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
6464 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
6465 Change check to determine if the instruction is a byte reversing
6466 entry. Fix typo in comment.
6467 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
6468 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
6469 Add def_builtin calls for new builtins.
6470 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
6471 Add define_insn expansion.
6473 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6475 * config/rl78/rl78.md (umaxdi3): New define_expand.
6477 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6479 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
6480 for non-QImode registers.
6482 2018-01-22 Richard Biener <rguenther@suse.de>
6484 PR tree-optimization/83963
6485 * graphite-scop-detection.c (scop_detection::get_sese): Delay
6486 including the loop exit block.
6487 (scop_detection::merge_sese): Likewise.
6488 (scop_detection::add_scop): Do it here instead.
6490 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6492 * doc/sourcebuild.texi (arm_softfloat): Document.
6494 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
6497 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
6498 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
6499 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
6501 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6502 David Edelsohn <dje.gcc@gmail.com>
6505 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6506 Change "crset eq" to "crset 2".
6507 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6508 (*call_indirect_aix<mode>_nospec): Likewise.
6509 (*call_value_indirect_aix<mode>_nospec): Likewise.
6510 (*call_indirect_elfv2<mode>_nospec): Likewise.
6511 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
6512 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
6513 change assembly output from . to $.
6514 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6515 (indirect_jump<mode>_nospec): Change assembly output from . to $.
6516 (*tablejump<mode>_internal1_nospec): Likewise.
6518 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
6521 * config/sh/sh_optimize_sett_clrt.cc:
6522 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
6524 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
6526 PR tree-optimization/83940
6527 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
6528 offset_dt to vect_constant_def rather than vect_unknown_def_type.
6529 (vect_check_load_store_mask): Add a mask_dt_out parameter and
6530 use it to pass back the definition type.
6531 (vect_check_store_rhs): Likewise rhs_dt_out.
6532 (vect_build_gather_load_calls): Add a mask_dt argument and use
6533 it instead of a call to vect_is_simple_use.
6534 (vectorizable_store): Update calls to vect_check_load_store_mask
6535 and vect_check_store_rhs. Use the dt returned by the latter instead
6536 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
6537 instead of calls to vect_is_simple_use. Pass the scalar rather
6538 than the vector operand to vect_is_simple_use when handling
6539 second and subsequent copies of an rhs value.
6540 (vectorizable_load): Update calls to vect_check_load_store_mask
6541 and vect_build_gather_load_calls. Use the cached mask_dt and
6542 gs_info.offset_dt instead of calls to vect_is_simple_use.
6544 2018-01-20 Jakub Jelinek <jakub@redhat.com>
6547 * tree-emutls.c: Include gimplify.h.
6548 (lower_emutls_2): New function.
6549 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
6550 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
6551 it before further processing.
6554 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
6555 UINTVAL (trueop1) instead of INTVAL (op1).
6557 2018-01-19 Jakub Jelinek <jakub@redhat.com>
6561 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
6562 INCOMING_FRAME_SP_OFFSET if not defined.
6563 (scan_trace): Add ENTRY argument. If true and
6564 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
6565 emit a note to adjust the CFA offset.
6566 (create_cfi_notes): Adjust scan_trace callers.
6567 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
6568 INCOMING_FRAME_SP_OFFSET in the CIE.
6569 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
6570 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
6572 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
6573 * doc/tm.texi: Regenerated.
6575 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6577 PR rtl-optimization/83147
6578 * lra-constraints.c (remove_inheritance_pseudos): Use
6579 lra_substitute_pseudo_within_insn.
6581 2018-01-19 Tom de Vries <tom@codesourcery.com>
6582 Cesar Philippidis <cesar@codesourcery.com>
6585 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
6587 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
6590 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
6591 spaces for function labels.
6593 2018-01-19 Martin Liska <mliska@suse.cz>
6595 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
6596 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
6597 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
6598 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
6599 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
6600 (PRED_CONST_RETURN): Change from 69 to 65.
6601 (PRED_NULL_RETURN): Change from 91 to 71.
6602 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
6603 (PRED_LOOP_GUARD): Change from 66 to 73.
6605 2018-01-19 Martin Liska <mliska@suse.cz>
6607 * predict.c (predict_insn_def): Add new assert.
6608 (struct branch_predictor): Change type to signed integer.
6609 (test_prediction_value_range): Amend test to cover
6611 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
6612 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
6613 (PRED_LOOP_ITERATIONS_MAX): Likewise.
6614 (PRED_LOOP_IV_COMPARE): Likewise.
6615 * predict.h (PROB_UNINITIALIZED): Define new constant.
6617 2018-01-19 Martin Liska <mliska@suse.cz>
6619 * predict.c (dump_prediction): Add new format for
6620 analyze_brprob.py script which is enabled with -details
6622 * profile-count.h (precise_p): New function.
6624 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6626 PR tree-optimization/83922
6627 * tree-vect-loop.c (vect_verify_full_masking): Return false if
6628 there are no statements that need masking.
6629 (vect_active_double_reduction_p): New function.
6630 (vect_analyze_loop_operations): Use it when handling phis that
6631 are not in the loop header.
6633 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6635 PR tree-optimization/83914
6636 * tree-vect-loop.c (vectorizable_induction): Don't convert
6637 init_expr or apply the peeling adjustment for inductions
6638 that are nested within the vectorized loop.
6640 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6642 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
6645 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6649 * function.h (gimplify_parameters): Add gimple_seq * argument.
6650 * function.c: Include gimple.h and options.h.
6651 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
6652 for the added local temporaries if needed.
6653 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
6654 if there are any parameter cleanups, wrap whole body into a
6655 try/finally with the cleanups.
6657 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
6660 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
6661 Use GET_MODE_CLASS for scalar floating point.
6663 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6667 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
6668 Fix call of call_cgraph_insertion_hooks.
6670 2018-01-18 Martin Sebor <msebor@redhat.com>
6672 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
6674 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6677 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
6680 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
6683 * common.opt: (-ffile-prefix-map): New option.
6684 * opts.c (common_handle_option): Defer it.
6685 * opts-global.c (handle_common_deferred_options): Handle it.
6686 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
6687 * file-prefix-map.h: New file.
6688 (remap_debug_filename, add_debug_prefix_map): ...here.
6689 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
6690 * final.c (debug_prefix_map, add_debug_prefix_map
6691 remap_debug_filename): Move to...
6692 * file-prefix-map.c: New file.
6693 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
6694 generalize, get rid of alloca(), use strrchr() instead of strchr().
6695 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
6696 Implement in terms of add_prefix_map().
6697 (remap_macro_filename, remap_debug_filename): Implement in term of
6699 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
6700 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
6701 * dbxout.c: Include file-prefix-map.h.
6702 * varasm.c: Likewise.
6703 * vmsdbgout.c: Likewise.
6704 * xcoffout.c: Likewise.
6705 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
6706 * doc/cppopts.texi (-fmacro-prefix-map): Document.
6707 * doc/invoke.texi (-ffile-prefix-map): Document.
6708 (-fdebug-prefix-map): Update description.
6710 2018-01-18 Martin Liska <mliska@suse.cz>
6712 * config/i386/i386.c (indirect_thunk_name): Document that also
6714 (output_indirect_thunk): Document why both instructions
6715 (pause and lfence) are generated.
6717 2018-01-18 Richard Biener <rguenther@suse.de>
6719 PR tree-optimization/83887
6720 * graphite-scop-detection.c
6721 (scop_detection::get_nearest_dom_with_single_entry): Remove.
6722 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
6723 (scop_detection::merge_sese): Re-implement with a flood-fill
6724 algorithm that properly finds a SESE region if it exists.
6726 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6729 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
6730 pointer_diff optimizations use view_convert instead of convert.
6732 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6734 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6735 Generate different code for -mno-speculate-indirect-jumps.
6736 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6737 (*call_indirect_aix<mode>): Disable for
6738 -mno-speculate-indirect-jumps.
6739 (*call_indirect_aix<mode>_nospec): New define_insn.
6740 (*call_value_indirect_aix<mode>): Disable for
6741 -mno-speculate-indirect-jumps.
6742 (*call_value_indirect_aix<mode>_nospec): New define_insn.
6743 (*sibcall_nonlocal_sysv<mode>): Generate different code for
6744 -mno-speculate-indirect-jumps.
6745 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6747 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6749 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
6750 long double type, set the flags for noting the default long double
6751 type, even if we don't pass or return a long double type.
6753 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
6756 * ipa-inline.c (flatten_function): Do not overwrite final inlining
6759 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
6761 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
6762 support for merge[hl].
6763 (fold_mergehl_helper): New helper function.
6764 (tree-vector-builder.h): New #include for tree_vector_builder usage.
6765 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
6766 (altivec_vmrglw_direct): Add xxmrglw insn.
6768 2018-01-17 Andrew Waterman <andrew@sifive.com>
6770 * config/riscv/riscv.c (riscv_conditional_register_usage): If
6771 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
6773 2018-01-17 David Malcolm <dmalcolm@redhat.com>
6776 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
6777 call the lto_location_cache before reading the
6778 DECL_SOURCE_LOCATION of the types.
6780 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
6781 Richard Sandiford <richard.sandiford@linaro.org>
6783 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
6784 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
6785 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
6786 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
6787 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
6789 * config/aarch64/constraints.md (aarch64_movti_operand):
6791 * config/aarch64/predicates.md (Uti): Add new constraint.
6793 2018-01-17 Carl Love <cel@us.ibm.com>
6795 * config/rs6000/vsx.md (define_expand xl_len_r,
6796 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
6797 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
6799 (define_expand, define_insn): Move the shift left from the
6800 define_insn to the define_expand for lxvl and stxvl instructions.
6801 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
6802 and XL_LEN_R definitions to PURE.
6804 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
6806 * config/i386/i386.c (indirect_thunk_name): Declare regno
6807 as unsigned int. Compare regno with INVALID_REGNUM.
6808 (output_indirect_thunk): Ditto.
6809 (output_indirect_thunk_function): Ditto.
6810 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
6811 in the call to output_indirect_thunk_function.
6813 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
6816 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
6817 rather than the size of inner_type to determine the stack slot size
6818 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
6820 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
6823 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
6826 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
6828 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
6829 endian Linux systems to optionally enable multilibs for selecting
6830 the long double type if the user configured an explicit type.
6831 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
6832 have no long double multilibs if not defined.
6833 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
6834 warn if the user used -mabi={ieee,ibm}longdouble and we built
6835 multilibs for long double.
6836 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
6837 appropriate multilib option.
6838 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
6840 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
6841 for building long double multilibs.
6842 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
6844 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
6846 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
6849 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
6851 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
6854 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
6857 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
6860 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6862 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
6863 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
6865 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
6867 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
6868 different rtl trees depending on TARGET_64BIT.
6869 (rs6000_gen_lvx): Likewise.
6871 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6873 * config/visium/visium.md (nop): Tweak comment.
6874 (hazard_nop): Likewise.
6876 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6878 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
6879 -mspeculate-indirect-jumps.
6880 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
6881 for -mno-speculate-indirect-jumps.
6882 (*call_indirect_elfv2<mode>_nospec): New define_insn.
6883 (*call_value_indirect_elfv2<mode>): Disable for
6884 -mno-speculate-indirect-jumps.
6885 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
6886 (indirect_jump): Emit different RTL for
6887 -mno-speculate-indirect-jumps.
6888 (*indirect_jump<mode>): Disable for
6889 -mno-speculate-indirect-jumps.
6890 (*indirect_jump<mode>_nospec): New define_insn.
6891 (tablejump): Emit different RTL for
6892 -mno-speculate-indirect-jumps.
6893 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
6894 (tablejumpsi_nospec): New define_expand.
6895 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
6896 (tablejumpdi_nospec): New define_expand.
6897 (*tablejump<mode>_internal1): Disable for
6898 -mno-speculate-indirect-jumps.
6899 (*tablejump<mode>_internal1_nospec): New define_insn.
6900 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
6903 2018-01-16 Artyom Skrobov tyomitch@gmail.com
6905 * caller-save.c (insert_save): Drop unnecessary parameter. All
6908 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6909 Richard Biener <rguenth@suse.de>
6912 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
6913 return early, inline manually is_gimple_sizepos. Make sure if we
6914 call gimplify_expr we don't end up with a gimple constant.
6915 * tree.c (variably_modified_type_p): Don't return true for
6916 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
6917 * gimplify.h (is_gimple_sizepos): Remove.
6919 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6921 PR tree-optimization/83857
6922 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
6923 vectorizable_live_operation for pure SLP statements.
6924 (vectorizable_live_operation): Handle PHIs.
6926 2018-01-16 Richard Biener <rguenther@suse.de>
6928 PR tree-optimization/83867
6929 * tree-vect-stmts.c (vect_transform_stmt): Precompute
6930 nested_in_vect_loop_p since the scalar stmt may get invalidated.
6932 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6935 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
6936 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
6937 If off is not INTEGER_CST, issue a may not be aligned warning
6938 rather than isn't aligned. Use isn%'t rather than isn't.
6939 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
6941 <case MULT_EXPR>: Improve the case when bottom and one of the
6942 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
6943 operand, in that case check if the other operand is multiple of
6944 bottom divided by the INTEGER_CST operand.
6946 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6949 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
6950 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
6951 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
6952 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
6953 * config/pa/pa.c (pa_function_arg_advance): Likewise.
6954 (pa_function_arg, pa_arg_partial_bytes): Likewise.
6955 (pa_function_arg_size): New function.
6957 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6959 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
6960 in a separate statement.
6962 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6964 PR tree-optimization/83847
6965 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
6966 group gathers and scatters.
6968 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6970 PR rtl-optimization/86620
6971 * params.def (max-sched-ready-insns): Bump minimum value to 1.
6973 PR rtl-optimization/83213
6974 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
6975 to last if both are JUMP_INSNs.
6977 PR tree-optimization/83843
6978 * gimple-ssa-store-merging.c
6979 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
6980 store_immediate_info for bswap/nop orig_stores.
6982 2018-01-15 Andrew Waterman <andrew@sifive.com>
6984 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
6986 <UDIV>: Increase cost if !TARGET_DIV.
6988 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
6990 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
6991 (define_attr "cr_logical_3op"): New.
6992 (cceq_ior_compare): Adjust.
6993 (cceq_ior_compare_complement): Adjust.
6994 (*cceq_rev_compare): Adjust.
6995 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
6996 (is_cracked_insn): Adjust.
6997 (insn_must_be_first_in_group): Adjust.
6998 * config/rs6000/40x.md: Adjust.
6999 * config/rs6000/440.md: Adjust.
7000 * config/rs6000/476.md: Adjust.
7001 * config/rs6000/601.md: Adjust.
7002 * config/rs6000/603.md: Adjust.
7003 * config/rs6000/6xx.md: Adjust.
7004 * config/rs6000/7450.md: Adjust.
7005 * config/rs6000/7xx.md: Adjust.
7006 * config/rs6000/8540.md: Adjust.
7007 * config/rs6000/cell.md: Adjust.
7008 * config/rs6000/e300c2c3.md: Adjust.
7009 * config/rs6000/e500mc.md: Adjust.
7010 * config/rs6000/e500mc64.md: Adjust.
7011 * config/rs6000/e5500.md: Adjust.
7012 * config/rs6000/e6500.md: Adjust.
7013 * config/rs6000/mpc.md: Adjust.
7014 * config/rs6000/power4.md: Adjust.
7015 * config/rs6000/power5.md: Adjust.
7016 * config/rs6000/power6.md: Adjust.
7017 * config/rs6000/power7.md: Adjust.
7018 * config/rs6000/power8.md: Adjust.
7019 * config/rs6000/power9.md: Adjust.
7020 * config/rs6000/rs64.md: Adjust.
7021 * config/rs6000/titan.md: Adjust.
7023 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7025 * config/i386/predicates.md (indirect_branch_operand): Rewrite
7026 ix86_indirect_branch_register logic.
7028 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7030 * config/i386/constraints.md (Bs): Update
7031 ix86_indirect_branch_register check. Don't check
7032 ix86_indirect_branch_register with GOT_memory_operand.
7034 * config/i386/predicates.md (GOT_memory_operand): Don't check
7035 ix86_indirect_branch_register here.
7036 (GOT32_symbol_operand): Likewise.
7038 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7040 * config/i386/predicates.md (constant_call_address_operand):
7041 Rewrite ix86_indirect_branch_register logic.
7042 (sibcall_insn_operand): Likewise.
7044 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7046 * config/i386/constraints.md (Bs): Replace
7047 ix86_indirect_branch_thunk_register with
7048 ix86_indirect_branch_register.
7050 * config/i386/i386.md (indirect_jump): Likewise.
7051 (tablejump): Likewise.
7052 (*sibcall_memory): Likewise.
7053 (*sibcall_value_memory): Likewise.
7054 Peepholes of indirect call and jump via memory: Likewise.
7055 * config/i386/i386.opt: Likewise.
7056 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7057 (GOT_memory_operand): Likewise.
7058 (call_insn_operand): Likewise.
7059 (sibcall_insn_operand): Likewise.
7060 (GOT32_symbol_operand): Likewise.
7062 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7065 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
7066 type rather than type addr's type points to.
7067 (expand_omp_atomic_mutex): Likewise.
7068 (expand_omp_atomic): Likewise.
7070 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7073 * config/i386/i386.c (output_indirect_thunk_function): Use
7074 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
7075 for __x86_return_thunk.
7077 2018-01-15 Richard Biener <rguenther@suse.de>
7080 * expmed.c (extract_bit_field_1): Fix typo.
7082 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7085 * config/arm/iterators.md (VF): New mode iterator.
7086 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
7087 Remove integer-related logic from pattern.
7088 (neon_vabd<mode>_3): Likewise.
7090 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7093 * common.opt (fstrict-overflow): No longer an alias.
7094 (fwrapv-pointer): New option.
7095 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
7096 also for pointer types based on flag_wrapv_pointer.
7097 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
7098 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
7099 opts->x_flag_wrapv got set.
7100 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
7101 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
7102 POINTER_TYPE_OVERFLOW_UNDEFINED.
7103 * match.pd: Likewise in address comparison pattern.
7104 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
7106 2018-01-15 Richard Biener <rguenther@suse.de>
7109 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
7110 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
7111 Reset type names to their identifier if their TYPE_DECL doesn't
7112 have linkage (and thus is used for ODR and devirt).
7113 (save_debug_info_for_decl): Remove.
7114 (save_debug_info_for_type): Likewise.
7115 (add_tree_to_fld_list): Adjust.
7116 * tree-pretty-print.c (dump_generic_node): Make dumping of
7117 type names more robust.
7119 2018-01-15 Richard Biener <rguenther@suse.de>
7121 * BASE-VER: Bump to 8.0.1.
7123 2018-01-14 Martin Sebor <msebor@redhat.com>
7126 * builtins.c (check_access): Avoid warning when the no-warning bit
7129 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
7131 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
7132 * ira-color (allocno_hard_regs_compare): Likewise.
7134 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
7137 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
7138 Use .pushsection/.popsection.
7140 2018-01-14 Martin Sebor <msebor@redhat.com>
7143 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
7145 2018-01-14 Jakub Jelinek <jakub@redhat.com>
7147 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
7148 entry from extra_headers.
7149 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
7150 extra_headers, make the list bitwise identical to the i?86-*-* one.
7152 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7154 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
7155 -mcmodel=large with -mindirect-branch=thunk,
7156 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
7157 -mfunction-return=thunk-extern.
7158 * doc/invoke.texi: Document -mcmodel=large is incompatible with
7159 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
7160 -mfunction-return=thunk and -mfunction-return=thunk-extern.
7162 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7164 * config/i386/i386.c (print_reg): Print the name of the full
7165 integer register without '%'.
7166 (ix86_print_operand): Handle 'V'.
7167 * doc/extend.texi: Document 'V' modifier.
7169 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7171 * config/i386/constraints.md (Bs): Disallow memory operand for
7172 -mindirect-branch-register.
7174 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7175 (GOT_memory_operand): Likewise.
7176 (call_insn_operand): Likewise.
7177 (sibcall_insn_operand): Likewise.
7178 (GOT32_symbol_operand): Likewise.
7179 * config/i386/i386.md (indirect_jump): Call convert_memory_address
7180 for -mindirect-branch-register.
7181 (tablejump): Likewise.
7182 (*sibcall_memory): Likewise.
7183 (*sibcall_value_memory): Likewise.
7184 Disallow peepholes of indirect call and jump via memory for
7185 -mindirect-branch-register.
7186 (*call_pop): Replace m with Bw.
7187 (*call_value_pop): Likewise.
7188 (*sibcall_pop_memory): Replace m with Bs.
7189 * config/i386/i386.opt (mindirect-branch-register): New option.
7190 * doc/invoke.texi: Document -mindirect-branch-register option.
7192 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7194 * config/i386/i386-protos.h (ix86_output_function_return): New.
7195 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
7196 set function_return_type.
7197 (indirect_thunk_name): Add ret_p to indicate thunk for function
7199 (output_indirect_thunk_function): Pass false to
7200 indirect_thunk_name.
7201 (ix86_output_indirect_branch_via_reg): Likewise.
7202 (ix86_output_indirect_branch_via_push): Likewise.
7203 (output_indirect_thunk_function): Create alias for function
7204 return thunk if regno < 0.
7205 (ix86_output_function_return): New function.
7206 (ix86_handle_fndecl_attribute): Handle function_return.
7207 (ix86_attribute_table): Add function_return.
7208 * config/i386/i386.h (machine_function): Add
7209 function_return_type.
7210 * config/i386/i386.md (simple_return_internal): Use
7211 ix86_output_function_return.
7212 (simple_return_internal_long): Likewise.
7213 * config/i386/i386.opt (mfunction-return=): New option.
7214 (indirect_branch): Mention -mfunction-return=.
7215 * doc/extend.texi: Document function_return function attribute.
7216 * doc/invoke.texi: Document -mfunction-return= option.
7218 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7220 * config/i386/i386-opts.h (indirect_branch): New.
7221 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
7222 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
7223 with local indirect jump when converting indirect call and jump.
7224 (ix86_set_indirect_branch_type): New.
7225 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
7226 (indirectlabelno): New.
7227 (indirect_thunk_needed): Likewise.
7228 (indirect_thunk_bnd_needed): Likewise.
7229 (indirect_thunks_used): Likewise.
7230 (indirect_thunks_bnd_used): Likewise.
7231 (INDIRECT_LABEL): Likewise.
7232 (indirect_thunk_name): Likewise.
7233 (output_indirect_thunk): Likewise.
7234 (output_indirect_thunk_function): Likewise.
7235 (ix86_output_indirect_branch_via_reg): Likewise.
7236 (ix86_output_indirect_branch_via_push): Likewise.
7237 (ix86_output_indirect_branch): Likewise.
7238 (ix86_output_indirect_jmp): Likewise.
7239 (ix86_code_end): Call output_indirect_thunk_function if needed.
7240 (ix86_output_call_insn): Call ix86_output_indirect_branch if
7242 (ix86_handle_fndecl_attribute): Handle indirect_branch.
7243 (ix86_attribute_table): Add indirect_branch.
7244 * config/i386/i386.h (machine_function): Add indirect_branch_type
7245 and has_local_indirect_jump.
7246 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
7248 (tablejump): Likewise.
7249 (*indirect_jump): Use ix86_output_indirect_jmp.
7250 (*tablejump_1): Likewise.
7251 (simple_return_indirect_internal): Likewise.
7252 * config/i386/i386.opt (mindirect-branch=): New option.
7253 (indirect_branch): New.
7256 (thunk-inline): Likewise.
7257 (thunk-extern): Likewise.
7258 * doc/extend.texi: Document indirect_branch function attribute.
7259 * doc/invoke.texi: Document -mindirect-branch= option.
7261 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
7264 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
7266 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
7268 * ipa-inline.c (want_inline_small_function_p): Return false if
7269 inlining has already failed with CIF_FINAL_ERROR.
7270 (update_caller_keys): Call want_inline_small_function_p before
7272 (update_callee_keys): Likewise.
7274 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7276 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
7278 (rs6000_quadword_masked_address_p): Likewise.
7279 (quad_aligned_load_p): Likewise.
7280 (quad_aligned_store_p): Likewise.
7281 (const_load_sequence_p): Add comment to describe the outer-most loop.
7282 (mimic_memory_attributes_and_flags): New function.
7283 (rs6000_gen_stvx): Likewise.
7284 (replace_swapped_aligned_store): Likewise.
7285 (rs6000_gen_lvx): Likewise.
7286 (replace_swapped_aligned_load): Likewise.
7287 (replace_swapped_load_constant): Capitalize argument name in
7288 comment describing this function.
7289 (rs6000_analyze_swaps): Add a third pass to search for vector loads
7290 and stores that access quad-word aligned addresses and replace
7291 with stvx or lvx instructions when appropriate.
7292 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
7293 New function prototype.
7294 (rs6000_quadword_masked_address_p): Likewise.
7295 (rs6000_gen_lvx): Likewise.
7296 (rs6000_gen_stvx): Likewise.
7297 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
7298 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
7299 when memory address is aligned.
7300 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
7301 this split to select lvx instruction when memory address is aligned.
7302 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
7303 instruction when memory address is aligned.
7304 (*vsx_le_perm_load_v16qi): Likewise.
7305 (four unnamed splitters): Modify to select the stvx instruction
7306 when memory is aligned.
7308 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
7310 * predict.c (determine_unlikely_bbs): Handle correctly BBs
7311 which appears in the queue multiple times.
7313 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7314 Alan Hayward <alan.hayward@arm.com>
7315 David Sherwood <david.sherwood@arm.com>
7317 * tree-vectorizer.h (vec_lower_bound): New structure.
7318 (_loop_vec_info): Add check_nonzero and lower_bounds.
7319 (LOOP_VINFO_CHECK_NONZERO): New macro.
7320 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
7321 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
7322 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
7323 fields. Make seg_len the distance travelled, not including the
7325 (dr_direction_indicator): Declare.
7326 (dr_zero_step_indicator): Likewise.
7327 (dr_known_forward_stride_p): Likewise.
7328 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
7330 (runtime_alias_check_p): Allow runtime alias checks with
7332 (operator ==): Compare access_size and align.
7333 (prune_runtime_alias_test_list): Rework for new distinction between
7334 the access_size and seg_len.
7335 (create_intersect_range_checks_index): Likewise. Cope with polynomial
7337 (get_segment_min_max): New function.
7338 (create_intersect_range_checks): Use it.
7339 (dr_step_indicator): New function.
7340 (dr_direction_indicator): Likewise.
7341 (dr_zero_step_indicator): Likewise.
7342 (dr_known_forward_stride_p): Likewise.
7343 * tree-loop-distribution.c (data_ref_segment_size): Return
7344 DR_STEP * (niters - 1).
7345 (compute_alias_check_pairs): Update call to the dr_with_seg_len
7347 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
7348 (vect_preserves_scalar_order_p): New function, split out from...
7349 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
7350 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
7351 (vect_vfa_access_size): New function.
7352 (vect_vfa_align): Likewise.
7353 (vect_compile_time_alias): Take access_size_a and access_b arguments.
7354 (dump_lower_bound): New function.
7355 (vect_check_lower_bound): Likewise.
7356 (vect_small_gap_p): Likewise.
7357 (vectorizable_with_step_bound_p): Likewise.
7358 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
7359 depencies if the vectorization factor is 1. Convert the checks
7360 for nonzero steps into checks on the bounds of DR_STEP. Try using
7361 a bunds check for variable steps if the minimum required step is
7362 relatively small. Update calls to the dr_with_seg_len
7363 constructor and to vect_compile_time_alias.
7364 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
7366 (vect_loop_versioning): Call it.
7367 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
7369 (vect_estimate_min_profitable_iters): Account for any bounds checks.
7371 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7372 Alan Hayward <alan.hayward@arm.com>
7373 David Sherwood <david.sherwood@arm.com>
7375 * doc/sourcebuild.texi (vect_scatter_store): Document.
7376 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
7378 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
7380 * genopinit.c (main): Add supports_vec_scatter_store and
7381 supports_vec_scatter_store_cached to target_optabs.
7382 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
7383 IFN_MASK_SCATTER_STORE.
7384 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
7386 * internal-fn.h (internal_store_fn_p): Declare.
7387 (internal_fn_stored_value_index): Likewise.
7388 * internal-fn.c (scatter_store_direct): New macro.
7389 (expand_scatter_store_optab_fn): New function.
7390 (direct_scatter_store_optab_supported_p): New macro.
7391 (internal_store_fn_p): New function.
7392 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
7393 IFN_MASK_SCATTER_STORE.
7394 (internal_fn_mask_index): Likewise.
7395 (internal_fn_stored_value_index): New function.
7396 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
7398 * optabs-query.h (supports_vec_scatter_store_p): Declare.
7399 * optabs-query.c (supports_vec_scatter_store_p): New function.
7400 * tree-vectorizer.h (vect_get_store_rhs): Declare.
7401 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
7402 true for scatter stores.
7403 (vect_gather_scatter_fn_p): Handle scatter stores too.
7404 (vect_check_gather_scatter): Consider using scatter stores if
7405 supports_vec_scatter_store_p.
7406 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
7408 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7409 internal_fn_stored_value_index.
7410 (check_load_store_masking): Handle scatter stores too.
7411 (vect_get_store_rhs): Make public.
7412 (vectorizable_call): Use internal_store_fn_p.
7413 (vectorizable_store): Handle scatter store internal functions.
7414 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
7415 when deciding whether the end of the group has been reached.
7416 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
7417 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
7418 (mask_scatter_store<mode>): New insns.
7420 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7421 Alan Hayward <alan.hayward@arm.com>
7422 David Sherwood <david.sherwood@arm.com>
7424 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
7425 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
7426 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
7428 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
7429 Use vect_truncate_gather_scatter_offset if we can't treat the
7430 operation as a normal gather load or scatter store.
7431 (get_group_load_store_type): Take the gather_scatter_info
7432 as argument. Try using a gather load or scatter store for
7433 single-element groups.
7434 (get_load_store_type): Update calls to get_group_load_store_type
7435 and vect_use_strided_gather_scatters_p.
7437 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7438 Alan Hayward <alan.hayward@arm.com>
7439 David Sherwood <david.sherwood@arm.com>
7441 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
7442 optional tree argument.
7443 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
7445 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
7446 but continue to use the current value as a fallback.
7447 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
7448 to compare the updates.
7449 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
7450 (get_load_store_type): Use it when handling a strided access.
7451 (vect_get_strided_load_store_ops): New function.
7452 (vect_get_data_ptr_increment): Likewise.
7453 (vectorizable_load): Handle strided gather loads. Always pass
7454 a step to vect_create_data_ref_ptr and bump_vector_ptr.
7456 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7457 Alan Hayward <alan.hayward@arm.com>
7458 David Sherwood <david.sherwood@arm.com>
7460 * doc/md.texi (gather_load@var{m}): Document.
7461 (mask_gather_load@var{m}): Likewise.
7462 * genopinit.c (main): Add supports_vec_gather_load and
7463 supports_vec_gather_load_cached to target_optabs.
7464 * optabs-tree.c (init_tree_optimization_optabs): Use
7465 ggc_cleared_alloc to allocate target_optabs.
7466 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
7467 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
7469 * internal-fn.h (internal_load_fn_p): Declare.
7470 (internal_gather_scatter_fn_p): Likewise.
7471 (internal_fn_mask_index): Likewise.
7472 (internal_gather_scatter_fn_supported_p): Likewise.
7473 * internal-fn.c (gather_load_direct): New macro.
7474 (expand_gather_load_optab_fn): New function.
7475 (direct_gather_load_optab_supported_p): New macro.
7476 (direct_internal_fn_optab): New function.
7477 (internal_load_fn_p): Likewise.
7478 (internal_gather_scatter_fn_p): Likewise.
7479 (internal_fn_mask_index): Likewise.
7480 (internal_gather_scatter_fn_supported_p): Likewise.
7481 * optabs-query.c (supports_at_least_one_mode_p): New function.
7482 (supports_vec_gather_load_p): Likewise.
7483 * optabs-query.h (supports_vec_gather_load_p): Declare.
7484 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
7485 and memory_type field.
7486 (NUM_PATTERNS): Bump to 15.
7487 * tree-vect-data-refs.c: Include internal-fn.h.
7488 (vect_gather_scatter_fn_p): New function.
7489 (vect_describe_gather_scatter_call): Likewise.
7490 (vect_check_gather_scatter): Try using internal functions for
7491 gather loads. Recognize existing calls to a gather load function.
7492 (vect_analyze_data_refs): Consider using gather loads if
7493 supports_vec_gather_load_p.
7494 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
7495 (vect_get_gather_scatter_offset_type): Likewise.
7496 (vect_convert_mask_for_vectype): Likewise.
7497 (vect_add_conversion_to_patterm): Likewise.
7498 (vect_try_gather_scatter_pattern): Likewise.
7499 (vect_recog_gather_scatter_pattern): New pattern recognizer.
7500 (vect_vect_recog_func_ptrs): Add it.
7501 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7502 internal_fn_mask_index and internal_gather_scatter_fn_p.
7503 (check_load_store_masking): Take the gather_scatter_info as an
7504 argument and handle gather loads.
7505 (vect_get_gather_scatter_ops): New function.
7506 (vectorizable_call): Check internal_load_fn_p.
7507 (vectorizable_load): Likewise. Handle gather load internal
7509 (vectorizable_store): Update call to check_load_store_masking.
7510 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
7511 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
7512 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
7513 (aarch64_gather_scale_operand_d): New predicates.
7514 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
7515 (mask_gather_load<mode>): New insns.
7517 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7518 Alan Hayward <alan.hayward@arm.com>
7519 David Sherwood <david.sherwood@arm.com>
7521 * optabs.def (fold_left_plus_optab): New optab.
7522 * doc/md.texi (fold_left_plus_@var{m}): Document.
7523 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
7524 * internal-fn.c (fold_left_direct): Define.
7525 (expand_fold_left_optab_fn): Likewise.
7526 (direct_fold_left_optab_supported_p): Likewise.
7527 * fold-const-call.c (fold_const_fold_left): New function.
7528 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
7529 * tree-parloops.c (valid_reduction_p): New function.
7530 (gather_scalar_reductions): Use it.
7531 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
7532 (vect_finish_replace_stmt): Declare.
7533 * tree-vect-loop.c (fold_left_reduction_fn): New function.
7534 (needs_fold_left_reduction_p): New function, split out from...
7535 (vect_is_simple_reduction): ...here. Accept reductions that
7536 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
7537 (vect_force_simple_reduction): Also store the reduction type in
7538 the assignment's STMT_VINFO_REDUC_TYPE.
7539 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
7540 (merge_with_identity): New function.
7541 (vect_expand_fold_left): Likewise.
7542 (vectorize_fold_left_reduction): Likewise.
7543 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
7544 scalar phi in place for it. Check for target support and reject
7545 cases that would reassociate the operation. Defer the transform
7546 phase to vectorize_fold_left_reduction.
7547 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
7548 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
7549 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
7551 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7553 * tree-if-conv.c (predicate_mem_writes): Remove redundant
7554 call to ifc_temp_var.
7556 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7557 Alan Hayward <alan.hayward@arm.com>
7558 David Sherwood <david.sherwood@arm.com>
7560 * target.def (legitimize_address_displacement): Take the original
7561 offset as a poly_int.
7562 * targhooks.h (default_legitimize_address_displacement): Update
7564 * targhooks.c (default_legitimize_address_displacement): Likewise.
7565 * doc/tm.texi: Regenerate.
7566 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
7567 as an argument, moving assert of ad->disp == ad->disp_term to...
7568 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
7569 Try calling targetm.legitimize_address_displacement before expanding
7570 the address rather than afterwards, and adjust for the new interface.
7571 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
7572 Match the new hook interface. Handle SVE addresses.
7573 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
7576 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7578 * Makefile.in (OBJS): Add early-remat.o.
7579 * target.def (select_early_remat_modes): New hook.
7580 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
7581 * doc/tm.texi: Regenerate.
7582 * targhooks.h (default_select_early_remat_modes): Declare.
7583 * targhooks.c (default_select_early_remat_modes): New function.
7584 * timevar.def (TV_EARLY_REMAT): New timevar.
7585 * passes.def (pass_early_remat): New pass.
7586 * tree-pass.h (make_pass_early_remat): Declare.
7587 * early-remat.c: New file.
7588 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
7590 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
7592 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7593 Alan Hayward <alan.hayward@arm.com>
7594 David Sherwood <david.sherwood@arm.com>
7596 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
7597 vfm1 with a bound_epilog parameter.
7598 (vect_do_peeling): Update calls accordingly, and move the prologue
7599 call earlier in the function. Treat the base bound_epilog as 0 for
7600 fully-masked loops and retain vf - 1 for other loops. Add 1 to
7601 this base when peeling for gaps.
7602 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
7603 with fully-masked loops.
7604 (vect_estimate_min_profitable_iters): Handle the single peeled
7605 iteration in that case.
7607 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7608 Alan Hayward <alan.hayward@arm.com>
7609 David Sherwood <david.sherwood@arm.com>
7611 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
7612 single-element interleaving even if the size is not a power of 2.
7613 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
7614 accesses for single-element interleaving if the group size is
7617 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7618 Alan Hayward <alan.hayward@arm.com>
7619 David Sherwood <david.sherwood@arm.com>
7621 * doc/md.texi (fold_extract_last_@var{m}): Document.
7622 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
7623 * optabs.def (fold_extract_last_optab): New optab.
7624 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
7625 * internal-fn.c (fold_extract_direct): New macro.
7626 (expand_fold_extract_optab_fn): Likewise.
7627 (direct_fold_extract_optab_supported_p): Likewise.
7628 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
7629 * tree-vect-loop.c (vect_model_reduction_cost): Handle
7630 EXTRACT_LAST_REDUCTION.
7631 (get_initial_def_for_reduction): Do not create an initial vector
7632 for EXTRACT_LAST_REDUCTION reductions.
7633 (vectorizable_reduction): Leave the scalar phi in place for
7634 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
7635 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
7636 epilogue code for EXTRACT_LAST_REDUCTION and defer the
7637 transform phase to vectorizable_condition.
7638 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
7640 (vect_finish_stmt_generation): ...here.
7641 (vect_finish_replace_stmt): New function.
7642 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
7643 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
7645 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
7647 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7648 Alan Hayward <alan.hayward@arm.com>
7649 David Sherwood <david.sherwood@arm.com>
7651 * doc/md.texi (extract_last_@var{m}): Document.
7652 * optabs.def (extract_last_optab): New optab.
7653 * internal-fn.def (EXTRACT_LAST): New internal function.
7654 * internal-fn.c (cond_unary_direct): New macro.
7655 (expand_cond_unary_optab_fn): Likewise.
7656 (direct_cond_unary_optab_supported_p): Likewise.
7657 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
7658 loops using EXTRACT_LAST.
7659 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
7660 (extract_last_<mode>): ...this optab.
7661 (vec_extract<mode><Vel>): Update accordingly.
7663 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7664 Alan Hayward <alan.hayward@arm.com>
7665 David Sherwood <david.sherwood@arm.com>
7667 * target.def (empty_mask_is_expensive): New hook.
7668 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
7669 * doc/tm.texi: Regenerate.
7670 * targhooks.h (default_empty_mask_is_expensive): Declare.
7671 * targhooks.c (default_empty_mask_is_expensive): New function.
7672 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
7673 if the target says that empty masks are expensive.
7674 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
7676 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
7678 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7679 Alan Hayward <alan.hayward@arm.com>
7680 David Sherwood <david.sherwood@arm.com>
7682 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
7683 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
7684 (vect_use_loop_mask_for_alignment_p): New function.
7685 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
7686 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
7687 niters_skip argument. Make sure that the first niters_skip elements
7688 of the first iteration are inactive.
7689 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
7690 Update call to vect_set_loop_masks_directly.
7691 (get_misalign_in_elems): New function, split out from...
7692 (vect_gen_prolog_loop_niters): ...here.
7693 (vect_update_init_of_dr): Take a code argument that specifies whether
7694 the adjustment should be added or subtracted.
7695 (vect_update_init_of_drs): Likewise.
7696 (vect_prepare_for_masked_peels): New function.
7697 (vect_do_peeling): Skip prologue peeling if we're using a mask
7698 instead. Update call to vect_update_inits_of_drs.
7699 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7701 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
7702 alignment. Do not include the number of peeled iterations in
7703 the minimum threshold in that case.
7704 (vectorizable_induction): Adjust the start value down by
7705 LOOP_VINFO_MASK_SKIP_NITERS iterations.
7706 (vect_transform_loop): Call vect_prepare_for_masked_peels.
7707 Take the number of skipped iterations into account when calculating
7709 * tree-vect-stmts.c (vect_gen_while_not): New function.
7711 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7712 Alan Hayward <alan.hayward@arm.com>
7713 David Sherwood <david.sherwood@arm.com>
7715 * doc/sourcebuild.texi (vect_fully_masked): Document.
7716 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
7718 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
7720 (vect_analyze_loop_2): ...here. Don't check the vectorization
7721 factor against the number of loop iterations if the loop is
7724 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7725 Alan Hayward <alan.hayward@arm.com>
7726 David Sherwood <david.sherwood@arm.com>
7728 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
7729 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
7730 (dump_groups): Update accordingly.
7731 (iv_use::mem_type): New member variable.
7732 (address_p): New function.
7733 (record_use): Add a mem_type argument and initialize the new
7735 (record_group_use): Add a mem_type argument. Use address_p.
7736 Remove obsolete null checks of base_object. Update call to record_use.
7737 (find_interesting_uses_op): Update call to record_group_use.
7738 (find_interesting_uses_cond): Likewise.
7739 (find_interesting_uses_address): Likewise.
7740 (get_mem_type_for_internal_fn): New function.
7741 (find_address_like_use): Likewise.
7742 (find_interesting_uses_stmt): Try find_address_like_use before
7743 calling find_interesting_uses_op.
7744 (addr_offset_valid_p): Use the iv mem_type field as the type
7745 of the addressed memory.
7746 (add_autoinc_candidates): Likewise.
7747 (get_address_cost): Likewise.
7748 (split_small_address_groups_p): Use address_p.
7749 (split_address_groups): Likewise.
7750 (add_iv_candidate_for_use): Likewise.
7751 (autoinc_possible_for_pair): Likewise.
7752 (rewrite_groups): Likewise.
7753 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
7754 (determine_group_iv_cost): Update after split of USE_ADDRESS.
7755 (get_alias_ptr_type_for_ptr_address): New function.
7756 (rewrite_use_address): Rewrite address uses in calls that were
7757 identified by find_address_like_use.
7759 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7760 Alan Hayward <alan.hayward@arm.com>
7761 David Sherwood <david.sherwood@arm.com>
7763 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
7765 * gimple-expr.h (is_gimple_addressable: Likewise.
7766 * gimple-expr.c (is_gimple_address): Likewise.
7767 * internal-fn.c (expand_call_mem_ref): New function.
7768 (expand_mask_load_optab_fn): Use it.
7769 (expand_mask_store_optab_fn): Likewise.
7771 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7772 Alan Hayward <alan.hayward@arm.com>
7773 David Sherwood <david.sherwood@arm.com>
7775 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
7776 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
7777 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
7778 (cond_umax@var{mode}): Document.
7779 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
7780 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
7781 (cond_umin_optab, cond_umax_optab): New optabs.
7782 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
7783 (COND_IOR, COND_XOR): New internal functions.
7784 * internal-fn.h (get_conditional_internal_fn): Declare.
7785 * internal-fn.c (cond_binary_direct): New macro.
7786 (expand_cond_binary_optab_fn): Likewise.
7787 (direct_cond_binary_optab_supported_p): Likewise.
7788 (get_conditional_internal_fn): New function.
7789 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
7790 Cope with reduction statements that are vectorized as calls rather
7792 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
7793 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
7794 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
7795 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
7796 (UNSPEC_COND_EOR): New unspecs.
7797 (optab): Add mappings for them.
7798 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
7799 (sve_int_op, sve_fp_op): New int attributes.
7801 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7802 Alan Hayward <alan.hayward@arm.com>
7803 David Sherwood <david.sherwood@arm.com>
7805 * optabs.def (while_ult_optab): New optab.
7806 * doc/md.texi (while_ult@var{m}@var{n}): Document.
7807 * internal-fn.def (WHILE_ULT): New internal function.
7808 * internal-fn.h (direct_internal_fn_supported_p): New override
7809 that takes two types as argument.
7810 * internal-fn.c (while_direct): New macro.
7811 (expand_while_optab_fn): New function.
7812 (convert_optab_supported_p): Likewise.
7813 (direct_while_optab_supported_p): New macro.
7814 * wide-int.h (wi::udiv_ceil): New function.
7815 * tree-vectorizer.h (rgroup_masks): New structure.
7816 (vec_loop_masks): New typedef.
7817 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
7819 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
7820 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
7821 (vect_max_vf): New function.
7822 (slpeel_make_loop_iterate_ntimes): Delete.
7823 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
7824 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
7825 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
7826 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
7827 internal-fn.h, stor-layout.h and optabs-query.h.
7828 (vect_set_loop_mask): New function.
7829 (add_preheader_seq): Likewise.
7830 (add_header_seq): Likewise.
7831 (interleave_supported_p): Likewise.
7832 (vect_maybe_permute_loop_masks): Likewise.
7833 (vect_set_loop_masks_directly): Likewise.
7834 (vect_set_loop_condition_masked): Likewise.
7835 (vect_set_loop_condition_unmasked): New function, split out from
7836 slpeel_make_loop_iterate_ntimes.
7837 (slpeel_make_loop_iterate_ntimes): Rename to..
7838 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
7839 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
7840 (vect_do_peeling): Update call accordingly.
7841 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
7843 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7844 mask_compare_type, can_fully_mask_p and fully_masked_p.
7845 (release_vec_loop_masks): New function.
7846 (_loop_vec_info): Use it to free the loop masks.
7847 (can_produce_all_loop_masks_p): New function.
7848 (vect_get_max_nscalars_per_iter): Likewise.
7849 (vect_verify_full_masking): Likewise.
7850 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
7851 retries, and free the mask rgroups before retrying. Check loop-wide
7852 reasons for disallowing fully-masked loops. Make the final decision
7853 about whether use a fully-masked loop or not.
7854 (vect_estimate_min_profitable_iters): Do not assume that peeling
7855 for the number of iterations will be needed for fully-masked loops.
7856 (vectorizable_reduction): Disable fully-masked loops.
7857 (vectorizable_live_operation): Likewise.
7858 (vect_halve_mask_nunits): New function.
7859 (vect_double_mask_nunits): Likewise.
7860 (vect_record_loop_mask): Likewise.
7861 (vect_get_loop_mask): Likewise.
7862 (vect_transform_loop): Handle the case in which the final loop
7863 iteration might handle a partial vector. Call vect_set_loop_condition
7864 instead of slpeel_make_loop_iterate_ntimes.
7865 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
7866 (check_load_store_masking): New function.
7867 (prepare_load_store_mask): Likewise.
7868 (vectorizable_store): Handle fully-masked loops.
7869 (vectorizable_load): Likewise.
7870 (supportable_widening_operation): Use vect_halve_mask_nunits for
7872 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
7873 (vect_gen_while): New function.
7874 * config/aarch64/aarch64.md (umax<mode>3): New expander.
7875 (aarch64_uqdec<mode>): New insn.
7877 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7878 Alan Hayward <alan.hayward@arm.com>
7879 David Sherwood <david.sherwood@arm.com>
7881 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
7882 (reduc_xor_scal_optab): New optabs.
7883 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
7884 (reduc_xor_scal_@var{m}): Document.
7885 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
7886 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
7888 * fold-const-call.c (fold_const_call): Handle them.
7889 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
7890 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
7891 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
7892 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
7893 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
7894 (UNSPEC_XORV): New unspecs.
7895 (optab): Add entries for them.
7896 (BITWISEV): New int iterator.
7897 (bit_reduc_op): New int attributes.
7899 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7900 Alan Hayward <alan.hayward@arm.com>
7901 David Sherwood <david.sherwood@arm.com>
7903 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
7904 * internal-fn.def (VEC_SHL_INSERT): New internal function.
7905 * optabs.def (vec_shl_insert_optab): New optab.
7906 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
7907 (duplicate_and_interleave): Likewise.
7908 * tree-vect-loop.c: Include internal-fn.h.
7909 (neutral_op_for_slp_reduction): New function, split out from
7910 get_initial_defs_for_reduction.
7911 (get_initial_def_for_reduction): Handle option 2 for variable-length
7912 vectors by loading the neutral value into a vector and then shifting
7913 the initial value into element 0.
7914 (get_initial_defs_for_reduction): Replace the code argument with
7915 the neutral value calculated by neutral_op_for_slp_reduction.
7916 Use gimple_build_vector for constant-length vectors.
7917 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
7918 but the first group_size elements have a neutral value.
7919 Use duplicate_and_interleave otherwise.
7920 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
7921 Update call to get_initial_defs_for_reduction. Handle SLP
7922 reductions for variable-length vectors by creating one vector
7923 result for each scalar result, with the elements associated
7924 with other scalar results stubbed out with the neutral value.
7925 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
7926 Require IFN_VEC_SHL_INSERT for double reductions on
7927 variable-length vectors, or SLP reductions that have
7928 a neutral value. Require can_duplicate_and_interleave_p
7929 support for variable-length unchained SLP reductions if there
7930 is no neutral value, such as for MIN/MAX reductions. Also require
7931 the number of vector elements to be a multiple of the number of
7932 SLP statements when doing variable-length unchained SLP reductions.
7933 Update call to vect_create_epilog_for_reduction.
7934 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
7935 and remove initial values.
7936 (duplicate_and_interleave): Make public.
7937 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
7938 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
7940 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7941 Alan Hayward <alan.hayward@arm.com>
7942 David Sherwood <david.sherwood@arm.com>
7944 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
7945 (can_duplicate_and_interleave_p): New function.
7946 (vect_get_and_check_slp_defs): Take the vector of statements
7947 rather than just the current one. Remove excess parentheses.
7948 Restriction rejectinon of vect_constant_def and vect_external_def
7949 for variable-length vectors to boolean types, or types for which
7950 can_duplicate_and_interleave_p is false.
7951 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
7952 (duplicate_and_interleave): New function.
7953 (vect_get_constant_vectors): Use gimple_build_vector for
7954 constant-length vectors and suitable variable-length constant
7955 vectors. Use duplicate_and_interleave for other variable-length
7956 vectors. Don't defer the update when inserting new statements.
7958 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7959 Alan Hayward <alan.hayward@arm.com>
7960 David Sherwood <david.sherwood@arm.com>
7962 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
7963 min_profitable_iters doesn't go negative.
7965 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7966 Alan Hayward <alan.hayward@arm.com>
7967 David Sherwood <david.sherwood@arm.com>
7969 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
7970 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
7971 * optabs.def (vec_mask_load_lanes_optab): New optab.
7972 (vec_mask_store_lanes_optab): Likewise.
7973 * internal-fn.def (MASK_LOAD_LANES): New internal function.
7974 (MASK_STORE_LANES): Likewise.
7975 * internal-fn.c (mask_load_lanes_direct): New macro.
7976 (mask_store_lanes_direct): Likewise.
7977 (expand_mask_load_optab_fn): Handle masked operations.
7978 (expand_mask_load_lanes_optab_fn): New macro.
7979 (expand_mask_store_optab_fn): Handle masked operations.
7980 (expand_mask_store_lanes_optab_fn): New macro.
7981 (direct_mask_load_lanes_optab_supported_p): Likewise.
7982 (direct_mask_store_lanes_optab_supported_p): Likewise.
7983 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
7985 (vect_load_lanes_supported): Likewise.
7986 * tree-vect-data-refs.c (strip_conversion): New function.
7987 (can_group_stmts_p): Likewise.
7988 (vect_analyze_data_ref_accesses): Use it instead of checking
7989 for a pair of assignments.
7990 (vect_store_lanes_supported): Take a masked_p parameter.
7991 (vect_load_lanes_supported): Likewise.
7992 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
7993 vect_store_lanes_supported and vect_load_lanes_supported.
7994 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7995 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
7996 parameter. Don't allow gaps for masked accesses.
7997 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
7998 and vect_load_lanes_supported.
7999 (get_load_store_type): Take a masked_p parameter and update
8000 call to get_group_load_store_type.
8001 (vectorizable_store): Update call to get_load_store_type.
8002 Handle IFN_MASK_STORE_LANES.
8003 (vectorizable_load): Update call to get_load_store_type.
8004 Handle IFN_MASK_LOAD_LANES.
8006 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8007 Alan Hayward <alan.hayward@arm.com>
8008 David Sherwood <david.sherwood@arm.com>
8010 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
8012 * config/aarch64/aarch64-protos.h
8013 (aarch64_sve_struct_memory_operand_p): Declare.
8014 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
8015 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
8016 (VPRED, vpred): Handle SVE structure modes.
8017 * config/aarch64/constraints.md (Utx): New constraint.
8018 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
8019 (aarch64_sve_struct_nonimmediate_operand): New predicates.
8020 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
8021 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
8022 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
8023 structure modes. Split into pieces after RA.
8024 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
8025 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
8027 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
8028 SVE structure modes.
8029 (aarch64_classify_address): Likewise.
8030 (sizetochar): Move earlier in file.
8031 (aarch64_print_operand): Handle SVE register lists.
8032 (aarch64_array_mode): New function.
8033 (aarch64_sve_struct_memory_operand_p): Likewise.
8034 (TARGET_ARRAY_MODE): Redefine.
8036 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8037 Alan Hayward <alan.hayward@arm.com>
8038 David Sherwood <david.sherwood@arm.com>
8040 * target.def (array_mode): New target hook.
8041 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
8042 * doc/tm.texi: Regenerate.
8043 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
8044 * hooks.c (hook_optmode_mode_uhwi_none): New function.
8045 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
8047 * stor-layout.c (mode_for_array): Likewise. Support polynomial
8050 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8051 Alan Hayward <alan.hayward@arm.com>
8052 David Sherwood <david.sherwood@arm.com>
8054 * fold-const.c (fold_binary_loc): Check the argument types
8055 rather than the result type when testing for a vector operation.
8057 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8059 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
8060 * doc/tm.texi: Regenerate.
8062 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8063 Alan Hayward <alan.hayward@arm.com>
8064 David Sherwood <david.sherwood@arm.com>
8066 * doc/invoke.texi (-msve-vector-bits=): Document new option.
8067 (sve): Document new AArch64 extension.
8068 * doc/md.texi (w): Extend the description of the AArch64
8069 constraint to include SVE vectors.
8070 (Upl, Upa): Document new AArch64 predicate constraints.
8071 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
8073 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
8074 (msve-vector-bits=): New option.
8075 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
8076 SVE when these are disabled.
8077 (sve): New extension.
8078 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
8079 modes. Adjust their number of units based on aarch64_sve_vg.
8080 (MAX_BITSIZE_MODE_ANY_MODE): Define.
8081 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
8082 aarch64_addr_query_type.
8083 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
8084 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
8085 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
8086 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
8087 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
8088 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
8089 (aarch64_simd_imm_zero_p): Delete.
8090 (aarch64_check_zero_based_sve_index_immediate): Declare.
8091 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8092 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8093 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8094 (aarch64_sve_float_mul_immediate_p): Likewise.
8095 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8097 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
8098 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
8099 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
8100 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
8101 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
8102 (aarch64_regmode_natural_size): Likewise.
8103 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
8104 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
8106 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
8107 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
8108 for VG and the SVE predicate registers.
8109 (V_ALIASES): Add a "z"-prefixed alias.
8110 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
8111 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
8112 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
8113 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
8114 (REG_CLASS_NAMES): Add entries for them.
8115 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
8116 and the predicate registers.
8117 (aarch64_sve_vg): Declare.
8118 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
8119 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
8120 (REGMODE_NATURAL_SIZE): Define.
8121 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
8123 * config/aarch64/aarch64.c: Include cfgrtl.h.
8124 (simd_immediate_info): Add a constructor for series vectors,
8125 and an associated step field.
8126 (aarch64_sve_vg): New variable.
8127 (aarch64_dbx_register_number): Handle VG and the predicate registers.
8128 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
8129 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
8130 (VEC_ANY_DATA, VEC_STRUCT): New constants.
8131 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
8132 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
8133 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
8134 (aarch64_get_mask_mode): New functions.
8135 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
8136 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
8137 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
8138 predicate modes and predicate registers. Explicitly restrict
8139 GPRs to modes of 16 bytes or smaller. Only allow FP registers
8140 to store a vector mode if it is recognized by
8141 aarch64_classify_vector_mode.
8142 (aarch64_regmode_natural_size): New function.
8143 (aarch64_hard_regno_caller_save_mode): Return the original mode
8145 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
8146 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
8147 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
8148 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
8150 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
8151 does not overlap dest if the function is frame-related. Handle
8153 (aarch64_split_add_offset): New function.
8154 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
8155 them aarch64_add_offset.
8156 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
8157 and update call to aarch64_sub_sp.
8158 (aarch64_add_cfa_expression): New function.
8159 (aarch64_expand_prologue): Pass extra temporary registers to the
8160 functions above. Handle the case in which we need to emit new
8161 DW_CFA_expressions for registers that were originally saved
8162 relative to the stack pointer, but now have to be expressed
8163 relative to the frame pointer.
8164 (aarch64_output_mi_thunk): Pass extra temporary registers to the
8166 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
8167 IP0 and IP1 values for SVE frames.
8168 (aarch64_expand_vec_series): New function.
8169 (aarch64_expand_sve_widened_duplicate): Likewise.
8170 (aarch64_expand_sve_const_vector): Likewise.
8171 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
8172 Handle SVE constants. Use emit_move_insn to move a force_const_mem
8173 into the register, rather than emitting a SET directly.
8174 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
8175 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
8176 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
8177 (offset_9bit_signed_scaled_p): New functions.
8178 (aarch64_replicate_bitmask_imm): New function.
8179 (aarch64_bitmask_imm): Use it.
8180 (aarch64_cannot_force_const_mem): Reject expressions involving
8181 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
8182 (aarch64_classify_index): Handle SVE indices, by requiring
8183 a plain register index with a scale that matches the element size.
8184 (aarch64_classify_address): Handle SVE addresses. Assert that
8185 the mode of the address is VOIDmode or an integer mode.
8186 Update call to aarch64_classify_symbol.
8187 (aarch64_classify_symbolic_expression): Update call to
8188 aarch64_classify_symbol.
8189 (aarch64_const_vec_all_in_range_p): New function.
8190 (aarch64_print_vector_float_operand): Likewise.
8191 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
8192 "vN" for FP registers with SVE modes. Handle (const ...) vectors
8193 and the FP immediates 1.0 and 0.5.
8194 (aarch64_print_address_internal): Handle SVE addresses.
8195 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
8196 (aarch64_regno_regclass): Handle predicate registers.
8197 (aarch64_secondary_reload): Handle big-endian reloads of SVE
8199 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
8200 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
8201 (aarch64_convert_sve_vector_bits): New function.
8202 (aarch64_override_options): Use it to handle -msve-vector-bits=.
8203 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8205 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
8206 Handle SVE vector and predicate modes. Accept VL-based constants
8207 that need only one temporary register, and VL offsets that require
8208 no temporary registers.
8209 (aarch64_conditional_register_usage): Mark the predicate registers
8210 as fixed if SVE isn't available.
8211 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
8212 Return true for SVE vector and predicate modes.
8213 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
8214 rather than an unsigned int. Handle SVE modes.
8215 (aarch64_preferred_simd_mode): Update call accordingly. Handle
8217 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
8219 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8220 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8221 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8222 (aarch64_sve_float_mul_immediate_p): New functions.
8223 (aarch64_sve_valid_immediate): New function.
8224 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
8225 Explicitly reject structure modes. Check for INDEX constants.
8226 Handle PTRUE and PFALSE constants.
8227 (aarch64_check_zero_based_sve_index_immediate): New function.
8228 (aarch64_simd_imm_zero_p): Delete.
8229 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
8230 vector modes. Accept constants in the range of CNT[BHWD].
8231 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
8232 ask for an Advanced SIMD mode.
8233 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
8234 (aarch64_simd_vector_alignment): Handle SVE predicates.
8235 (aarch64_vectorize_preferred_vector_alignment): New function.
8236 (aarch64_simd_vector_alignment_reachable): Use it instead of
8238 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
8239 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
8241 (MAX_VECT_LEN): Delete.
8242 (expand_vec_perm_d): Add a vec_flags field.
8243 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
8244 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
8245 (aarch64_evpc_ext): Don't apply a big-endian lane correction
8247 (aarch64_evpc_rev): Rename to...
8248 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
8249 (aarch64_evpc_rev_global): New function.
8250 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
8251 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
8253 (aarch64_evpc_sve_tbl): New function.
8254 (aarch64_expand_vec_perm_const_1): Update after rename of
8255 aarch64_evpc_rev. Handle SVE permutes too, trying
8256 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
8257 than aarch64_evpc_tbl.
8258 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
8259 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
8260 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
8261 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
8262 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
8263 (aarch64_expand_sve_vcond): New functions.
8264 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
8265 of aarch64_vector_mode_p.
8266 (aarch64_dwarf_poly_indeterminate_value): New function.
8267 (aarch64_compute_pressure_classes): Likewise.
8268 (aarch64_can_change_mode_class): Likewise.
8269 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
8270 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
8271 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
8272 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
8273 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
8274 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
8275 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
8276 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
8278 (Dn, Dl, Dr): Accept const as well as const_vector.
8279 (Dz): Likewise. Compare against CONST0_RTX.
8280 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
8281 of "vector" where appropriate.
8282 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
8283 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
8284 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
8285 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
8286 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
8287 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
8288 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
8289 (v_int_equiv): Extend to SVE modes.
8290 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
8292 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
8293 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
8294 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
8295 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
8296 (SVE_COND_FP_CMP): New int iterators.
8297 (perm_hilo): Handle the new unpack unspecs.
8298 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
8300 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
8301 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
8302 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
8303 (aarch64_equality_operator, aarch64_constant_vector_operand)
8304 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
8305 (aarch64_sve_nonimmediate_operand): Likewise.
8306 (aarch64_sve_general_operand): Likewise.
8307 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
8308 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
8309 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
8310 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
8311 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
8312 (aarch64_sve_float_arith_immediate): Likewise.
8313 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
8314 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
8315 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
8316 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
8317 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
8318 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
8319 (aarch64_sve_float_arith_operand): Likewise.
8320 (aarch64_sve_float_arith_with_sub_operand): Likewise.
8321 (aarch64_sve_float_mul_operand): Likewise.
8322 (aarch64_sve_vec_perm_operand): Likewise.
8323 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
8324 (aarch64_mov_operand): Accept const_poly_int and const_vector.
8325 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
8326 as well as const_vector.
8327 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
8328 in file. Use CONST0_RTX and CONSTM1_RTX.
8329 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
8330 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
8331 Use aarch64_simd_imm_zero.
8332 * config/aarch64/aarch64-sve.md: New file.
8333 * config/aarch64/aarch64.md: Include it.
8334 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
8335 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
8336 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
8337 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
8338 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
8339 (sve): New attribute.
8340 (enabled): Disable instructions with the sve attribute unless
8342 (movqi, movhi): Pass CONST_POLY_INT operaneds through
8343 aarch64_expand_mov_immediate.
8344 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
8345 CNT[BHSD] immediates.
8346 (movti): Split CONST_POLY_INT moves into two halves.
8347 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
8348 Split additions that need a temporary here if the destination
8349 is the stack pointer.
8350 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
8351 (*add<mode>3_poly_1): New instruction.
8352 (set_clobber_cc): New expander.
8354 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8356 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
8357 parameter and use it instead of GET_MODE_SIZE (innermode). Use
8358 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
8359 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
8360 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
8361 Change innermode from fixed_mode_size to machine_mode.
8362 (simplify_subreg): Update call accordingly. Handle a constant-sized
8363 subreg of a variable-length CONST_VECTOR.
8365 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8366 Alan Hayward <alan.hayward@arm.com>
8367 David Sherwood <david.sherwood@arm.com>
8369 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
8370 (add_offset_to_base): New function, split out from...
8371 (create_mem_ref): ...here. When handling a scale other than 1,
8372 check first whether the address is valid without the offset.
8373 Add it into the base if so, leaving the index and scale as-is.
8375 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8378 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
8379 fold_for_warn before checking if arg2 is INTEGER_CST.
8381 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
8383 * config/rs6000/predicates.md (load_multiple_operation): Delete.
8384 (store_multiple_operation): Delete.
8385 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
8386 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
8387 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
8388 guarded by TARGET_STRING.
8389 (rs6000_output_load_multiple): Delete.
8390 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
8391 OPTION_MASK_STRING / TARGET_STRING handling.
8392 (print_operand) <'N', 'O'>: Add comment that these are unused now.
8393 (const rs6000_opt_masks) <"string">: Change mask to 0.
8394 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
8395 (MASK_STRING): Delete.
8396 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
8398 (load_multiple): Delete.
8405 (store_multiple): Delete.
8412 (movmemsi_8reg): Delete.
8413 (corresponding unnamed define_insn): Delete.
8414 (movmemsi_6reg): Delete.
8415 (corresponding unnamed define_insn): Delete.
8416 (movmemsi_4reg): Delete.
8417 (corresponding unnamed define_insn): Delete.
8418 (movmemsi_2reg): Delete.
8419 (corresponding unnamed define_insn): Delete.
8420 (movmemsi_1reg): Delete.
8421 (corresponding unnamed define_insn): Delete.
8422 * config/rs6000/rs6000.opt (mno-string): New.
8423 (mstring): Replace by deprecation warning stub.
8424 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
8426 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8428 * regrename.c (regrename_do_replace): If replacing the same
8429 reg multiple times, try to reuse last created gen_raw_REG.
8432 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
8433 main to workaround a bug in GDB.
8435 2018-01-12 Tom de Vries <tom@codesourcery.com>
8438 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
8440 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
8442 PR rtl-optimization/80481
8443 * ira-color.c (get_cap_member): New function.
8444 (allocnos_conflict_by_live_ranges_p): Use it.
8445 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
8446 (setup_slot_coalesced_allocno_live_ranges): Ditto.
8448 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
8451 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
8452 (*saddl_se_1): Ditto.
8454 (*ssubl_se_1): Ditto.
8456 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8458 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
8459 rather than wi::to_widest for DR_INITs.
8460 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
8461 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
8462 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
8464 (vect_analyze_group_access_1): Note that here.
8466 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8468 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
8469 polynomial type sizes.
8471 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8473 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
8474 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
8475 (gimple_add_tmp_var): Likewise.
8477 2018-01-12 Martin Liska <mliska@suse.cz>
8479 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
8480 (gimple_alloc_sizes): Likewise.
8481 (dump_gimple_statistics): Use PRIu64 in printf format.
8482 * gimple.h: Change uint64_t to int.
8484 2018-01-12 Martin Liska <mliska@suse.cz>
8486 * tree-core.h: Use uint64_t instead of int.
8487 * tree.c (tree_node_counts): Likewise.
8488 (tree_node_sizes): Likewise.
8489 (dump_tree_statistics): Use PRIu64 in printf format.
8491 2018-01-12 Martin Liska <mliska@suse.cz>
8493 * Makefile.in: As qsort_chk is implemented in vec.c, add
8494 vec.o to linkage of gencfn-macros.
8495 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
8496 passing the info to record_node_allocation_statistics.
8497 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
8499 * ggc-common.c (struct ggc_usage): Add operator== and use
8500 it in operator< and compare function.
8501 * mem-stats.h (struct mem_usage): Likewise.
8502 * vec.c (struct vec_usage): Remove operator< and compare
8503 function. Can be simply inherited.
8505 2018-01-12 Martin Jambor <mjambor@suse.cz>
8508 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
8509 * tree-ssa-math-opts.c: Include domwalk.h.
8510 (convert_mult_to_fma_1): New function.
8511 (fma_transformation_info): New type.
8512 (fma_deferring_state): Likewise.
8513 (cancel_fma_deferring): New function.
8514 (result_of_phi): Likewise.
8515 (last_fma_candidate_feeds_initial_phi): Likewise.
8516 (convert_mult_to_fma): Added deferring logic, split actual
8517 transformation to convert_mult_to_fma_1.
8518 (math_opts_dom_walker): New type.
8519 (math_opts_dom_walker::after_dom_children): New method, body moved
8520 here from pass_optimize_widening_mul::execute, added deferring logic
8522 (pass_optimize_widening_mul::execute): Moved most of code to
8523 math_opts_dom_walker::after_dom_children.
8524 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
8525 * config/i386/i386.c (ix86_option_override_internal): Added
8526 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
8528 2018-01-12 Richard Biener <rguenther@suse.de>
8531 * dwarf2out.c (gen_variable_die): Do not reset old_die for
8532 inline instance vars.
8534 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
8537 * config/rx/rx.c (rx_is_restricted_memory_address):
8540 2018-01-12 Richard Biener <rguenther@suse.de>
8542 PR tree-optimization/80846
8543 * target.def (split_reduction): New target hook.
8544 * targhooks.c (default_split_reduction): New function.
8545 * targhooks.h (default_split_reduction): Declare.
8546 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
8547 target requests first reduce vectors by combining low and high
8549 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
8550 (get_vectype_for_scalar_type_and_size): Export.
8551 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
8552 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
8553 * doc/tm.texi: Regenerate.
8554 * config/i386/i386.c (ix86_split_reduction): Implement
8555 TARGET_VECTORIZE_SPLIT_REDUCTION.
8557 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8560 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
8561 in PIC mode except for TARGET_VXWORKS_RTP.
8562 * config/sparc/sparc.c: Include cfgrtl.h.
8563 (TARGET_INIT_PIC_REG): Define.
8564 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
8565 (sparc_pic_register_p): New predicate.
8566 (sparc_legitimate_address_p): Use it.
8567 (sparc_legitimize_pic_address): Likewise.
8568 (sparc_delegitimize_address): Likewise.
8569 (sparc_mode_dependent_address_p): Likewise.
8570 (gen_load_pcrel_sym): Remove 4th parameter.
8571 (load_got_register): Adjust call to above. Remove obsolete stuff.
8572 (sparc_expand_prologue): Do not call load_got_register here.
8573 (sparc_flat_expand_prologue): Likewise.
8574 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
8575 (sparc_use_pseudo_pic_reg): New function.
8576 (sparc_init_pic_reg): Likewise.
8577 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8578 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
8580 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
8582 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
8583 Add item for branch_cost.
8585 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8587 PR rtl-optimization/83565
8588 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
8589 not extend the result to a larger mode for rotate operations.
8590 (num_sign_bit_copies1): Likewise.
8592 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8595 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
8597 Use values-Xc.o for -pedantic.
8598 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
8600 2018-01-12 Martin Liska <mliska@suse.cz>
8603 * ipa-devirt.c (final_warning_record::grow_type_warnings):
8605 (possible_polymorphic_call_targets): Use it.
8606 (ipa_devirt): Likewise.
8608 2018-01-12 Martin Liska <mliska@suse.cz>
8610 * profile-count.h (enum profile_quality): Use 0 as invalid
8611 enum value of profile_quality.
8613 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
8615 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
8616 -mext-string options.
8618 2018-01-12 Richard Biener <rguenther@suse.de>
8620 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
8621 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
8622 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
8624 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
8626 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
8628 * configure.ac (--with-long-double-format): Add support for the
8629 configuration option to change the default long double format on
8631 * config.gcc (powerpc*-linux*-*): Likewise.
8632 * configure: Regenerate.
8633 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
8634 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
8635 used without modification.
8637 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8639 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
8640 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
8641 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
8642 MISC_BUILTIN_SPEC_BARRIER.
8643 (rs6000_init_builtins): Likewise.
8644 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
8646 (speculation_barrier): New define_insn.
8647 * doc/extend.texi: Document __builtin_speculation_barrier.
8649 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8652 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
8653 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
8654 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
8656 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
8657 integral modes instead of "ss" and "sd".
8658 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
8659 vectors with 32-bit and 64-bit elements.
8660 (vecdupssescalarmodesuffix): New mode attribute.
8661 (vec_dup<mode>): Use it.
8663 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
8666 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
8667 frame if argument is passed on stack.
8669 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8672 * ree.c (combine_reaching_defs): Optimize also
8673 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
8674 reg2=any_extend(exp); reg1=reg2;, formatting fix.
8676 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8679 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
8681 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8684 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
8685 after they are computed.
8687 2018-01-11 Bin Cheng <bin.cheng@arm.com>
8689 PR tree-optimization/83695
8690 * gimple-loop-linterchange.cc
8691 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
8692 reset cached scev information after interchange.
8693 (pass_linterchange::execute): Remove call to scev_reset_htab.
8695 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8697 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
8698 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
8699 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
8700 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
8701 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
8702 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
8703 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
8704 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
8705 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
8706 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
8707 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
8708 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
8709 (V_lane_reg): Likewise.
8710 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
8712 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
8713 (vfmal_lane_low<mode>_intrinsic,
8714 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
8715 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
8716 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
8717 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
8718 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
8719 vfmsl_lane_high<mode>_intrinsic): New define_insns.
8721 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8723 * config/arm/arm-cpus.in (fp16fml): New feature.
8724 (ALL_SIMD): Add fp16fml.
8725 (armv8.2-a): Add fp16fml as an option.
8726 (armv8.3-a): Likewise.
8727 (armv8.4-a): Add fp16fml as part of fp16.
8728 * config/arm/arm.h (TARGET_FP16FML): Define.
8729 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
8731 * config/arm/arm-modes.def (V2HF): Define.
8732 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
8733 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
8734 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
8735 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
8736 vfmsl_low, vfmsl_high): New set of builtins.
8737 * config/arm/iterators.md (PLUSMINUS): New code iterator.
8738 (vfml_op): New code attribute.
8739 (VFMLHALVES): New int iterator.
8740 (VFML, VFMLSEL): New mode attributes.
8741 (V_reg): Define mapping for V2HF.
8742 (V_hi, V_lo): New mode attributes.
8743 (VF_constraint): Likewise.
8744 (vfml_half, vfml_half_selector): New int attributes.
8745 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
8747 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
8748 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
8750 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
8751 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
8752 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
8753 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
8755 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
8756 Document new effective target and option set.
8758 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8760 * config/arm/arm-cpus.in (armv8_4): New feature.
8761 (ARMv8_4a): New fgroup.
8762 (armv8.4-a): New arch.
8763 * config/arm/arm-tables.opt: Regenerate.
8764 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
8765 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
8766 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
8767 Add matching rules for -march=armv8.4-a and extensions.
8768 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
8770 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
8773 * config/rx/rx.md (BW): New mode attribute.
8774 (sync_lock_test_and_setsi): Add mode suffix to insn output.
8776 2018-01-11 Richard Biener <rguenther@suse.de>
8778 PR tree-optimization/83435
8779 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
8780 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
8781 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
8783 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8784 Alan Hayward <alan.hayward@arm.com>
8785 David Sherwood <david.sherwood@arm.com>
8787 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
8789 (aarch64_classify_address): Initialize it. Track polynomial offsets.
8790 (aarch64_print_address_internal): Use it to check for a zero offset.
8792 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8793 Alan Hayward <alan.hayward@arm.com>
8794 David Sherwood <david.sherwood@arm.com>
8796 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
8797 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
8798 Return a poly_int64 rather than a HOST_WIDE_INT.
8799 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
8800 rather than a HOST_WIDE_INT.
8801 * config/aarch64/aarch64.h (aarch64_frame): Protect with
8802 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
8803 hard_fp_offset, frame_size, initial_adjust, callee_offset and
8804 final_offset from HOST_WIDE_INT to poly_int64.
8805 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
8806 to_constant when getting the number of units in an Advanced SIMD
8808 (aarch64_builtin_vectorized_function): Check for a constant number
8810 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
8812 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
8813 attribute instead of GET_MODE_NUNITS.
8814 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
8815 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
8816 GET_MODE_SIZE for fixed-size registers.
8817 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
8818 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
8819 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
8820 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
8821 (aarch64_print_operand, aarch64_print_address_internal)
8822 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
8823 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
8824 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
8825 Handle polynomial GET_MODE_SIZE.
8826 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
8827 wider than SImode without modification.
8828 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
8829 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
8830 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
8831 passing and returning SVE modes.
8832 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
8833 rather than GEN_INT.
8834 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
8835 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
8836 (aarch64_allocate_and_probe_stack_space): Likewise.
8837 (aarch64_layout_frame): Cope with polynomial offsets.
8838 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
8839 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
8841 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
8842 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
8843 poly_int64 rather than a HOST_WIDE_INT.
8844 (aarch64_get_separate_components, aarch64_process_components)
8845 (aarch64_expand_prologue, aarch64_expand_epilogue)
8846 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
8847 (aarch64_anchor_offset): New function, split out from...
8848 (aarch64_legitimize_address): ...here.
8849 (aarch64_builtin_vectorization_cost): Handle polynomial
8850 TYPE_VECTOR_SUBPARTS.
8851 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
8853 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
8854 number of elements from the PARALLEL rather than the mode.
8855 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
8856 rather than GET_MODE_BITSIZE.
8857 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
8858 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
8859 (aarch64_expand_vec_perm_const_1): Handle polynomial
8860 d->perm.length () and d->perm elements.
8861 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
8862 Apply to_constant to d->perm elements.
8863 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
8864 polynomial CONST_VECTOR_NUNITS.
8865 (aarch64_move_pointer): Take amount as a poly_int64 rather
8867 (aarch64_progress_pointer): Avoid temporary variable.
8868 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
8869 the mode attribute instead of GET_MODE.
8871 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8872 Alan Hayward <alan.hayward@arm.com>
8873 David Sherwood <david.sherwood@arm.com>
8875 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
8876 x exists before using it.
8877 (aarch64_add_constant_internal): Rename to...
8878 (aarch64_add_offset_1): ...this. Replace regnum with separate
8879 src and dest rtxes. Handle the case in which they're different,
8880 including when the offset is zero. Replace scratchreg with an rtx.
8881 Use 2 additions if there is no spare register into which we can
8882 move a 16-bit constant.
8883 (aarch64_add_constant): Delete.
8884 (aarch64_add_offset): Replace reg with separate src and dest
8885 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
8886 Use aarch64_add_offset_1.
8887 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
8888 an rtx rather than an int. Take the delta as a poly_int64
8889 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
8890 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
8891 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
8892 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
8893 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
8895 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
8896 aarch64_add_constant.
8898 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8900 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
8901 Use scalar_float_mode.
8903 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8905 * config/aarch64/aarch64-simd.md
8906 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
8907 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
8908 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
8909 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
8910 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
8911 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
8912 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
8913 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
8914 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
8915 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
8917 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8920 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
8921 targ_options->x_arm_arch_string is non NULL.
8923 2018-01-11 Tamar Christina <tamar.christina@arm.com>
8925 * config/aarch64/aarch64.h
8926 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
8928 2018-01-11 Sudakshina Das <sudi.das@arm.com>
8931 * expmed.c (emit_store_flag_force): Swap if const op0
8932 and change VOIDmode to mode of op0.
8934 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8936 PR rtl-optimization/83761
8937 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
8938 than bytes to mode_for_size.
8940 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8943 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
8944 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
8947 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8950 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
8951 when in layout mode.
8952 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
8953 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
8956 2018-01-10 Michael Collison <michael.collison@arm.com>
8958 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
8959 * config/aarch64/aarch64-option-extension.def: Add
8960 AARCH64_OPT_EXTENSION of 'fp16fml'.
8961 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8962 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
8963 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
8964 * config/aarch64/constraints.md (Ui7): New constraint.
8965 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
8966 (VFMLA_SEL_W): Ditto.
8969 (VFMLA16_LOW): New int iterator.
8970 (VFMLA16_HIGH): Ditto.
8971 (UNSPEC_FMLAL): New unspec.
8972 (UNSPEC_FMLSL): Ditto.
8973 (UNSPEC_FMLAL2): Ditto.
8974 (UNSPEC_FMLSL2): Ditto.
8975 (f16mac): New code attribute.
8976 * config/aarch64/aarch64-simd-builtins.def
8977 (aarch64_fmlal_lowv2sf): Ditto.
8978 (aarch64_fmlsl_lowv2sf): Ditto.
8979 (aarch64_fmlalq_lowv4sf): Ditto.
8980 (aarch64_fmlslq_lowv4sf): Ditto.
8981 (aarch64_fmlal_highv2sf): Ditto.
8982 (aarch64_fmlsl_highv2sf): Ditto.
8983 (aarch64_fmlalq_highv4sf): Ditto.
8984 (aarch64_fmlslq_highv4sf): Ditto.
8985 (aarch64_fmlal_lane_lowv2sf): Ditto.
8986 (aarch64_fmlsl_lane_lowv2sf): Ditto.
8987 (aarch64_fmlal_laneq_lowv2sf): Ditto.
8988 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
8989 (aarch64_fmlalq_lane_lowv4sf): Ditto.
8990 (aarch64_fmlsl_lane_lowv4sf): Ditto.
8991 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
8992 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
8993 (aarch64_fmlal_lane_highv2sf): Ditto.
8994 (aarch64_fmlsl_lane_highv2sf): Ditto.
8995 (aarch64_fmlal_laneq_highv2sf): Ditto.
8996 (aarch64_fmlsl_laneq_highv2sf): Ditto.
8997 (aarch64_fmlalq_lane_highv4sf): Ditto.
8998 (aarch64_fmlsl_lane_highv4sf): Ditto.
8999 (aarch64_fmlalq_laneq_highv4sf): Ditto.
9000 (aarch64_fmlsl_laneq_highv4sf): Ditto.
9001 * config/aarch64/aarch64-simd.md:
9002 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
9003 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9004 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
9005 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9006 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
9007 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
9008 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
9009 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
9010 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
9011 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
9012 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
9013 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
9014 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
9015 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
9016 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
9017 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
9018 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
9019 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
9020 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
9021 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
9022 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
9023 (vfmlsl_low_u32): Ditto.
9024 (vfmlalq_low_u32): Ditto.
9025 (vfmlslq_low_u32): Ditto.
9026 (vfmlal_high_u32): Ditto.
9027 (vfmlsl_high_u32): Ditto.
9028 (vfmlalq_high_u32): Ditto.
9029 (vfmlslq_high_u32): Ditto.
9030 (vfmlal_lane_low_u32): Ditto.
9031 (vfmlsl_lane_low_u32): Ditto.
9032 (vfmlal_laneq_low_u32): Ditto.
9033 (vfmlsl_laneq_low_u32): Ditto.
9034 (vfmlalq_lane_low_u32): Ditto.
9035 (vfmlslq_lane_low_u32): Ditto.
9036 (vfmlalq_laneq_low_u32): Ditto.
9037 (vfmlslq_laneq_low_u32): Ditto.
9038 (vfmlal_lane_high_u32): Ditto.
9039 (vfmlsl_lane_high_u32): Ditto.
9040 (vfmlal_laneq_high_u32): Ditto.
9041 (vfmlsl_laneq_high_u32): Ditto.
9042 (vfmlalq_lane_high_u32): Ditto.
9043 (vfmlslq_lane_high_u32): Ditto.
9044 (vfmlalq_laneq_high_u32): Ditto.
9045 (vfmlslq_laneq_high_u32): Ditto.
9046 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
9047 (AARCH64_FL_FOR_ARCH8_4): New.
9048 (AARCH64_ISA_F16FML): New ISA flag.
9049 (TARGET_F16FML): New feature flag for fp16fml.
9050 (doc/invoke.texi): Document new fp16fml option.
9052 2018-01-10 Michael Collison <michael.collison@arm.com>
9054 * config/aarch64/aarch64-builtins.c:
9055 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
9056 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9057 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
9058 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
9059 (AARCH64_ISA_SHA3): New ISA flag.
9060 (TARGET_SHA3): New feature flag for sha3.
9061 * config/aarch64/iterators.md (sha512_op): New int attribute.
9062 (CRYPTO_SHA512): New int iterator.
9063 (UNSPEC_SHA512H): New unspec.
9064 (UNSPEC_SHA512H2): Ditto.
9065 (UNSPEC_SHA512SU0): Ditto.
9066 (UNSPEC_SHA512SU1): Ditto.
9067 * config/aarch64/aarch64-simd-builtins.def
9068 (aarch64_crypto_sha512hqv2di): New builtin.
9069 (aarch64_crypto_sha512h2qv2di): Ditto.
9070 (aarch64_crypto_sha512su0qv2di): Ditto.
9071 (aarch64_crypto_sha512su1qv2di): Ditto.
9072 (aarch64_eor3qv8hi): Ditto.
9073 (aarch64_rax1qv2di): Ditto.
9074 (aarch64_xarqv2di): Ditto.
9075 (aarch64_bcaxqv8hi): Ditto.
9076 * config/aarch64/aarch64-simd.md:
9077 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
9078 (aarch64_crypto_sha512su0qv2di): Ditto.
9079 (aarch64_crypto_sha512su1qv2di): Ditto.
9080 (aarch64_eor3qv8hi): Ditto.
9081 (aarch64_rax1qv2di): Ditto.
9082 (aarch64_xarqv2di): Ditto.
9083 (aarch64_bcaxqv8hi): Ditto.
9084 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
9085 (vsha512h2q_u64): Ditto.
9086 (vsha512su0q_u64): Ditto.
9087 (vsha512su1q_u64): Ditto.
9088 (veor3q_u16): Ditto.
9089 (vrax1q_u64): Ditto.
9091 (vbcaxq_u16): Ditto.
9092 * config/arm/types.md (crypto_sha512): New type attribute.
9093 (crypto_sha3): Ditto.
9094 (doc/invoke.texi): Document new sha3 option.
9096 2018-01-10 Michael Collison <michael.collison@arm.com>
9098 * config/aarch64/aarch64-builtins.c:
9099 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
9100 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9101 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
9102 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
9103 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
9104 (AARCH64_ISA_SM4): New ISA flag.
9105 (TARGET_SM4): New feature flag for sm4.
9106 * config/aarch64/aarch64-simd-builtins.def
9107 (aarch64_sm3ss1qv4si): Ditto.
9108 (aarch64_sm3tt1aq4si): Ditto.
9109 (aarch64_sm3tt1bq4si): Ditto.
9110 (aarch64_sm3tt2aq4si): Ditto.
9111 (aarch64_sm3tt2bq4si): Ditto.
9112 (aarch64_sm3partw1qv4si): Ditto.
9113 (aarch64_sm3partw2qv4si): Ditto.
9114 (aarch64_sm4eqv4si): Ditto.
9115 (aarch64_sm4ekeyqv4si): Ditto.
9116 * config/aarch64/aarch64-simd.md:
9117 (aarch64_sm3ss1qv4si): Ditto.
9118 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
9119 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
9120 (aarch64_sm4eqv4si): Ditto.
9121 (aarch64_sm4ekeyqv4si): Ditto.
9122 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
9123 (sm3part_op): Ditto.
9124 (CRYPTO_SM3TT): Ditto.
9125 (CRYPTO_SM3PART): Ditto.
9126 (UNSPEC_SM3SS1): New unspec.
9127 (UNSPEC_SM3TT1A): Ditto.
9128 (UNSPEC_SM3TT1B): Ditto.
9129 (UNSPEC_SM3TT2A): Ditto.
9130 (UNSPEC_SM3TT2B): Ditto.
9131 (UNSPEC_SM3PARTW1): Ditto.
9132 (UNSPEC_SM3PARTW2): Ditto.
9133 (UNSPEC_SM4E): Ditto.
9134 (UNSPEC_SM4EKEY): Ditto.
9135 * config/aarch64/constraints.md (Ui2): New constraint.
9136 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
9137 * config/arm/types.md (crypto_sm3): New type attribute.
9138 (crypto_sm4): Ditto.
9139 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
9140 (vsm3tt1aq_u32): Ditto.
9141 (vsm3tt1bq_u32): Ditto.
9142 (vsm3tt2aq_u32): Ditto.
9143 (vsm3tt2bq_u32): Ditto.
9144 (vsm3partw1q_u32): Ditto.
9145 (vsm3partw2q_u32): Ditto.
9146 (vsm4eq_u32): Ditto.
9147 (vsm4ekeyq_u32): Ditto.
9148 (doc/invoke.texi): Document new sm4 option.
9150 2018-01-10 Michael Collison <michael.collison@arm.com>
9152 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
9153 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
9154 (AARCH64_FL_FOR_ARCH8_4): New.
9155 (AARCH64_FL_V8_4): New flag.
9156 (doc/invoke.texi): Document new armv8.4-a option.
9158 2018-01-10 Michael Collison <michael.collison@arm.com>
9160 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9161 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
9162 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
9163 * config/aarch64/aarch64-option-extension.def: Add
9164 AARCH64_OPT_EXTENSION of 'sha2'.
9165 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
9166 (crypto): Disable sha2 and aes if crypto disabled.
9167 (crypto): Enable aes and sha2 if enabled.
9168 (simd): Disable sha2 and aes if simd disabled.
9169 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
9171 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
9172 (TARGET_SHA2): New feature flag for sha2.
9173 (TARGET_AES): New feature flag for aes.
9174 * config/aarch64/aarch64-simd.md:
9175 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
9176 conditional on TARGET_AES.
9177 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
9178 (aarch64_crypto_sha1hsi): Make pattern conditional
9180 (aarch64_crypto_sha1hv4si): Ditto.
9181 (aarch64_be_crypto_sha1hv4si): Ditto.
9182 (aarch64_crypto_sha1su1v4si): Ditto.
9183 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
9184 (aarch64_crypto_sha1su0v4si): Ditto.
9185 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
9186 (aarch64_crypto_sha256su0v4si): Ditto.
9187 (aarch64_crypto_sha256su1v4si): Ditto.
9188 (doc/invoke.texi): Document new aes and sha2 options.
9190 2018-01-10 Martin Sebor <msebor@redhat.com>
9192 PR tree-optimization/83781
9193 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
9196 2018-01-11 Martin Sebor <msebor@gmail.com>
9197 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9199 PR tree-optimization/83501
9200 PR tree-optimization/81703
9202 * tree-ssa-strlen.c (get_string_cst): Rename...
9203 (get_string_len): ...to this. Handle global constants.
9204 (handle_char_store): Adjust.
9206 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
9207 Jim Wilson <jimw@sifive.com>
9209 * config/riscv/riscv-protos.h (riscv_output_return): New.
9210 * config/riscv/riscv.c (struct machine_function): New naked_p field.
9211 (riscv_attribute_table, riscv_output_return),
9212 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
9213 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
9214 (riscv_compute_frame_info): Only compute frame->mask if not a naked
9216 (riscv_expand_prologue): Add early return for naked function.
9217 (riscv_expand_epilogue): Likewise.
9218 (riscv_function_ok_for_sibcall): Return false for naked function.
9219 (riscv_set_current_function): New.
9220 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
9221 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
9222 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
9223 * doc/extend.texi (RISC-V Function Attributes): New.
9225 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
9227 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
9228 check for 128-bit long double before checking TCmode.
9229 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
9230 128-bit long doubles before checking TFmode or TCmode.
9231 (FLOAT128_IBM_P): Likewise.
9233 2018-01-10 Martin Sebor <msebor@redhat.com>
9235 PR tree-optimization/83671
9236 * builtins.c (c_strlen): Unconditionally return zero for the empty
9238 Use -Warray-bounds for warnings.
9239 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
9240 for non-constant array indices with COMPONENT_REF, arrays of
9241 arrays, and pointers to arrays.
9242 (gimple_fold_builtin_strlen): Determine and set length range for
9243 non-constant character arrays.
9245 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
9248 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
9251 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
9253 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
9255 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9258 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
9259 VECTOR_MEM_ALTIVEC_OR_VSX_P.
9260 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
9261 indexed_or_indirect_operand predicate.
9262 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
9263 (*vsx_le_perm_load_v8hi): Likewise.
9264 (*vsx_le_perm_load_v16qi): Likewise.
9265 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
9266 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
9267 (*vsx_le_perm_store_v8hi): Likewise.
9268 (*vsx_le_perm_store_v16qi): Likewise.
9269 (eight unnamed splitters): Likewise.
9271 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9273 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
9274 * config/rs6000/emmintrin.h: Likewise.
9275 * config/rs6000/mmintrin.h: Likewise.
9276 * config/rs6000/xmmintrin.h: Likewise.
9278 2018-01-10 David Malcolm <dmalcolm@redhat.com>
9281 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
9283 * tree.c (tree_nop_conversion): Return true for location wrapper
9285 (maybe_wrap_with_location): New function.
9286 (selftest::check_strip_nops): New function.
9287 (selftest::test_location_wrappers): New function.
9288 (selftest::tree_c_tests): Call it.
9289 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
9290 (maybe_wrap_with_location): New decl.
9291 (EXPR_LOCATION_WRAPPER_P): New macro.
9292 (location_wrapper_p): New inline function.
9293 (tree_strip_any_location_wrapper): New inline function.
9295 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
9298 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
9299 stack_realign_offset for the largest alignment of stack slot
9301 (ix86_find_max_used_stack_alignment): New function.
9302 (ix86_finalize_stack_frame_flags): Use it. Set
9303 max_used_stack_alignment if we don't realign stack.
9304 * config/i386/i386.h (machine_function): Add
9305 max_used_stack_alignment.
9307 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
9309 * config/arm/arm.opt (-mbranch-cost): New option.
9310 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
9313 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
9316 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
9317 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
9319 2018-01-10 Richard Biener <rguenther@suse.de>
9322 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
9323 early out so it also covers the case where we have a non-NULL
9326 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9328 PR tree-optimization/83753
9329 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
9330 for non-strided grouped accesses if the number of elements is 1.
9332 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9335 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
9336 * i386.h (TARGET_USE_GATHER): Define.
9337 * x86-tune.def (X86_TUNE_USE_GATHER): New.
9339 2018-01-10 Martin Liska <mliska@suse.cz>
9342 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
9343 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
9345 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
9346 CLEANUP_NO_PARTITIONING is not set.
9348 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9350 * doc/rtl.texi: Remove documentation of (const ...) wrappers
9351 for vectors, as a partial revert of r254296.
9352 * rtl.h (const_vec_p): Delete.
9353 (const_vec_duplicate_p): Don't test for vector CONSTs.
9354 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
9355 * expmed.c (make_tree): Likewise.
9358 * common.md (E, F): Use CONSTANT_P instead of checking for
9360 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9361 checking for CONST_VECTOR.
9363 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9366 * predict.c (force_edge_cold): Handle in more sane way edges
9369 2018-01-09 Carl Love <cel@us.ibm.com>
9371 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
9373 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
9374 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
9375 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
9376 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
9377 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
9378 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
9379 * config/rs6000/rs6000-protos.h: Add extern defition for
9380 rs6000_generate_float2_double_code.
9381 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
9383 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
9384 (float2_v2df): Add define_expand.
9386 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
9389 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
9390 op_mode in the force_to_mode call.
9392 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9394 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
9395 instead of checking each element individually.
9396 (aarch64_evpc_uzp): Likewise.
9397 (aarch64_evpc_zip): Likewise.
9398 (aarch64_evpc_ext): Likewise.
9399 (aarch64_evpc_rev): Likewise.
9400 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
9401 instead of checking each element individually. Return true without
9403 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
9404 whether all selected elements come from the same input, instead of
9405 checking each element individually. Remove calls to gen_rtx_REG,
9406 start_sequence and end_sequence and instead assert that no rtl is
9409 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9411 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
9412 order of HIGH and CONST checks.
9414 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9416 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
9417 if the destination isn't an SSA_NAME.
9419 2018-01-09 Richard Biener <rguenther@suse.de>
9421 PR tree-optimization/83668
9422 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
9424 (canonicalize_loop_form): ... here, renamed from ...
9425 (canonicalize_loop_closed_ssa_form): ... this and amended to
9426 swap successor edges for loop exit blocks to make us use
9427 the RPO order we need for initial schedule generation.
9429 2018-01-09 Joseph Myers <joseph@codesourcery.com>
9431 PR tree-optimization/64811
9432 * match.pd: When optimizing comparisons with Inf, avoid
9433 introducing or losing exceptions from comparisons with NaN.
9435 2018-01-09 Martin Liska <mliska@suse.cz>
9438 * asan.c (shadow_mem_size): Add gcc_assert.
9440 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
9442 Don't save registers in main().
9445 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
9446 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
9447 * config/avr/avr.c (avr_set_current_function): Don't error if
9448 naked, OS_task or OS_main are specified at the same time.
9449 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
9451 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
9453 * common/config/avr/avr-common.c (avr_option_optimization_table):
9454 Switch on -mmain-is-OS_task for optimizing compilations.
9456 2018-01-09 Richard Biener <rguenther@suse.de>
9458 PR tree-optimization/83572
9459 * graphite.c: Include cfganal.h.
9460 (graphite_transform_loops): Connect infinite loops to exit
9461 and remove fake edges at the end.
9463 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9465 * ipa-inline.c (edge_badness): Revert accidental checkin.
9467 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9470 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
9471 symbols; not inline clones.
9473 2018-01-09 Jakub Jelinek <jakub@redhat.com>
9476 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
9477 hard registers. Formatting fixes.
9479 PR preprocessor/83722
9480 * gcc.c (try_generate_repro): Pass
9481 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
9482 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
9485 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
9486 Kito Cheng <kito.cheng@gmail.com>
9488 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
9489 (riscv_leaf_function_p): Delete.
9490 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
9492 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9494 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
9496 (do_ifelse): New function.
9497 (do_isel): New function.
9498 (do_sub3): New function.
9499 (do_add3): New function.
9500 (do_load_mask_compare): New function.
9501 (do_overlap_load_compare): New function.
9502 (expand_compare_loop): New function.
9503 (expand_block_compare): Call expand_compare_loop() when appropriate.
9504 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
9506 (-mblock-compare-inline-loop-limit): New option.
9508 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9511 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
9512 Reverse order of second and third operands in first alternative.
9513 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
9514 of first and second elements in UNSPEC_VPERMR vector.
9515 (altivec_expand_vec_perm_le): Likewise.
9517 2018-01-08 Jeff Law <law@redhat.com>
9519 PR rtl-optimizatin/81308
9520 * tree-switch-conversion.c (cfg_altered): New file scoped static.
9521 (process_switch): If group_case_labels makes a change, then set
9523 (pass_convert_switch::execute): If a switch is converted, then
9524 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
9526 PR rtl-optimization/81308
9527 * recog.c (split_all_insns): Conditionally cleanup the CFG after
9530 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
9532 PR target/83663 - Revert r255946
9533 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
9534 generation for cases where splatting a value is not useful.
9535 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
9536 across a vec_duplicate and a paradoxical subreg forming a vector
9537 mode to a vec_concat.
9539 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9541 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
9542 -march=armv8.3-a variants.
9543 * config/arm/t-multilib: Likewise.
9544 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
9546 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9548 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
9550 (cceq_ior_compare_complement): Give it a name so I can use it, and
9551 change boolean_or_operator predicate to boolean_operator so it can
9552 be used to generate a crand.
9553 (eqne): New code iterator.
9554 (bd/bd_neg): New code_attrs.
9555 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
9556 a single define_insn.
9557 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
9558 decrement (bdnzt/bdnzf/bdzt/bdzf).
9559 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
9560 with the new names of the branch decrement patterns, and added the
9561 names of the branch decrement conditional patterns.
9563 2018-01-08 Richard Biener <rguenther@suse.de>
9565 PR tree-optimization/83563
9566 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
9569 2018-01-08 Richard Biener <rguenther@suse.de>
9572 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
9574 2018-01-08 Richard Biener <rguenther@suse.de>
9576 PR tree-optimization/83685
9577 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
9578 references to abnormals.
9580 2018-01-08 Richard Biener <rguenther@suse.de>
9583 * dwarf2out.c (output_indirect_strings): Handle empty
9584 skeleton_debug_str_hash.
9585 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
9587 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9589 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
9590 (emit_store_direct): Likewise.
9591 (arc_trampoline_adjust_address): Likewise.
9592 (arc_asm_trampoline_template): New function.
9593 (arc_initialize_trampoline): Use asm_trampoline_template.
9594 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
9595 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
9596 * config/arc/arc.md (flush_icache): Delete pattern.
9598 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9600 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
9601 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
9604 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9607 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
9608 by not USED_FOR_TARGET.
9609 (make_pass_resolve_sw_modes): Likewise.
9611 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9613 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
9616 2018-01-08 Richard Biener <rguenther@suse.de>
9619 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
9621 2018-01-08 Richard Biener <rguenther@suse.de>
9624 * match.pd ((t * 2) / 2) -> t): Add missing :c.
9626 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
9629 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
9630 basic blocks with a small number of successors.
9631 (convert_control_dep_chain_into_preds): Improve handling of
9633 (dump_predicates): Split apart into...
9634 (dump_pred_chain): ...here...
9635 (dump_pred_info): ...and here.
9636 (can_one_predicate_be_invalidated_p): Add debugging printfs.
9637 (can_chain_union_be_invalidated_p): Improve check for invalidation
9639 (uninit_uses_cannot_happen): Avoid unnecessary if
9640 convert_control_dep_chain_into_preds yielded nothing.
9642 2018-01-06 Martin Sebor <msebor@redhat.com>
9644 PR tree-optimization/83640
9645 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
9646 subtracting negative offset from size.
9647 (builtin_access::overlap): Adjust offset bounds of the access to fall
9648 within the size of the object if possible.
9650 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
9652 PR rtl-optimization/83699
9653 * expmed.c (extract_bit_field_1): Restrict the vector usage of
9654 extract_bit_field_as_subreg to cases in which the extracted
9655 value is also a vector.
9657 * lra-constraints.c (process_alt_operands): Test for the equivalence
9658 substitutions when detecting a possible reload cycle.
9660 2018-01-06 Jakub Jelinek <jakub@redhat.com>
9663 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
9664 by default if flag_selective_schedling{,2}. Formatting fixes.
9666 PR rtl-optimization/83682
9667 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
9668 if it has non-VECTOR_MODE element mode.
9669 (vec_duplicate_p): Likewise.
9672 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
9673 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
9675 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9678 * config/i386/i386-builtin.def
9679 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
9680 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
9681 Require also OPTION_MASK_ISA_AVX512F in addition to
9682 OPTION_MASK_ISA_GFNI.
9683 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
9684 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
9685 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
9686 to OPTION_MASK_ISA_GFNI.
9687 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
9688 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
9689 OPTION_MASK_ISA_AVX512BW.
9690 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
9691 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
9692 addition to OPTION_MASK_ISA_GFNI.
9693 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
9694 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
9695 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
9696 to OPTION_MASK_ISA_GFNI.
9697 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
9698 a requirement for all ISAs rather than any of them with a few
9700 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
9702 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
9703 bitmasks to be enabled with 3 exceptions, instead of requiring any
9704 enabled ISA with lots of exceptions.
9705 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
9706 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
9707 Change avx512bw in isa attribute to avx512f.
9708 * config/i386/sgxintrin.h: Add license boilerplate.
9709 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
9710 to __AVX512F__ and __AVX512VL to __AVX512VL__.
9711 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
9712 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
9714 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
9715 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
9716 temporarily sse2 rather than sse if not enabled already.
9719 * config/i386/sse.md (VI248_VLBW): Rename to ...
9720 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
9721 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
9722 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
9723 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
9724 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
9725 mode iterator instead of VI248_VLBW.
9727 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
9729 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
9730 (record_modified): Skip clobbers; add debug output.
9731 (param_change_prob): Use sreal frequencies.
9733 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9735 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
9736 punt for user-aligned variables.
9738 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9740 * tree-chrec.c (chrec_contains_symbols): Return true for
9743 2018-01-05 Sudakshina Das <sudi.das@arm.com>
9746 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
9747 of (x|y) == x for BICS pattern.
9749 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9751 PR tree-optimization/83605
9752 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
9753 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
9756 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
9758 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
9759 * config/epiphany/rtems.h: New file.
9761 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9762 Uros Bizjak <ubizjak@gmail.com>
9765 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
9766 QIreg_operand instead of register_operand predicate.
9767 * config/i386/i386.c (ix86_rop_should_change_byte_p,
9768 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
9769 comments instead of -fmitigate[-_]rop.
9771 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9774 * cgraphunit.c (symbol_table::compile): Switch to text_section
9775 before calling assembly_start debug hook.
9776 * run-rtl-passes.c (run_rtl_passes): Likewise.
9779 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9781 * tree-vrp.c (extract_range_from_binary_expr_1): Check
9782 range_int_cst_p rather than !symbolic_range_p before calling
9783 extract_range_from_multiplicative_op_1.
9785 2018-01-04 Jeff Law <law@redhat.com>
9787 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
9788 redundant test in assertion.
9790 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9792 * doc/rtl.texi: Document machine_mode wrapper classes.
9794 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9796 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
9799 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9801 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
9802 the VEC_PERM_EXPR fold to fail.
9804 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9807 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
9808 to switched_sections.
9810 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9813 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
9816 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
9819 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
9820 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
9822 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9825 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
9826 is BLKmode and bitpos not zero or mode change is needed.
9828 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9831 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
9834 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
9837 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
9838 instead of MULT rtx. Update all corresponding splitters.
9840 (*ssub<modesuffix>): Ditto.
9842 (*cmp_sadd_di): Update split patterns.
9843 (*cmp_sadd_si): Ditto.
9844 (*cmp_sadd_sidi): Ditto.
9845 (*cmp_ssub_di): Ditto.
9846 (*cmp_ssub_si): Ditto.
9847 (*cmp_ssub_sidi): Ditto.
9848 * config/alpha/predicates.md (const23_operand): New predicate.
9849 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
9850 Look for ASHIFT, not MULT inner operand.
9851 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
9853 2018-01-04 Martin Liska <mliska@suse.cz>
9855 PR gcov-profile/83669
9856 * gcov.c (output_intermediate_file): Add version to intermediate
9858 * doc/gcov.texi: Document new field 'version' in intermediate
9859 file format. Fix location of '-k' option of gcov command.
9861 2018-01-04 Martin Liska <mliska@suse.cz>
9864 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
9866 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9868 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
9870 2018-01-03 Martin Sebor <msebor@redhat.com>
9872 PR tree-optimization/83655
9873 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
9874 checking calls with invalid arguments.
9876 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9878 * tree-vect-stmts.c (vect_get_store_rhs): New function.
9879 (vectorizable_mask_load_store): Delete.
9880 (vectorizable_call): Return false for masked loads and stores.
9881 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
9882 instead of gimple_assign_rhs1.
9883 (vectorizable_load): Handle IFN_MASK_LOAD.
9884 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
9886 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9888 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
9890 (vectorizable_mask_load_store): ...here.
9891 (vectorizable_load): ...and here.
9893 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9895 * tree-vect-stmts.c (vect_build_all_ones_mask)
9896 (vect_build_zero_merge_argument): New functions, split out from...
9897 (vectorizable_load): ...here.
9899 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9901 * tree-vect-stmts.c (vect_check_store_rhs): New function,
9903 (vectorizable_mask_load_store): ...here.
9904 (vectorizable_store): ...and here.
9906 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9908 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
9910 (vectorizable_mask_load_store): ...here.
9912 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9914 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
9915 (vect_model_store_cost): Take a vec_load_store_type instead of a
9917 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
9918 (vect_model_store_cost): Take a vec_load_store_type instead of a
9920 (vectorizable_mask_load_store): Update accordingly.
9921 (vectorizable_store): Likewise.
9922 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
9924 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9926 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
9927 IFN_MASK_LOAD calls here rather than...
9928 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
9930 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9931 Alan Hayward <alan.hayward@arm.com>
9932 David Sherwood <david.sherwood@arm.com>
9934 * expmed.c (extract_bit_field_1): For vector extracts,
9935 fall back to extract_bit_field_as_subreg if vec_extract
9938 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9939 Alan Hayward <alan.hayward@arm.com>
9940 David Sherwood <david.sherwood@arm.com>
9942 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
9943 they are variable or constant sized.
9944 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
9945 slots for constant-sized data.
9947 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9948 Alan Hayward <alan.hayward@arm.com>
9949 David Sherwood <david.sherwood@arm.com>
9951 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
9952 handling COND_EXPRs with boolean comparisons, try to find a better
9953 basis for the mask type than the boolean itself.
9955 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9957 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
9958 is calculated and how it can be overridden.
9959 * genmodes.c (max_bitsize_mode_any_mode): New variable.
9960 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
9962 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
9965 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9966 Alan Hayward <alan.hayward@arm.com>
9967 David Sherwood <david.sherwood@arm.com>
9969 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
9970 Remove the mode argument.
9971 (aarch64_simd_valid_immediate): Remove the mode and inverse
9973 * config/aarch64/iterators.md (bitsize): New iterator.
9974 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
9975 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
9976 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
9977 aarch64_simd_valid_immediate.
9978 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
9979 (aarch64_reg_or_bic_imm): Likewise.
9980 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
9981 with an insn_type enum and msl with a modifier_type enum.
9982 Replace element_width with a scalar_mode. Change the shift
9983 to unsigned int. Add constructors for scalar_float_mode and
9984 scalar_int_mode elements.
9985 (aarch64_vect_float_const_representable_p): Delete.
9986 (aarch64_can_const_movi_rtx_p)
9987 (aarch64_simd_scalar_immediate_valid_for_move)
9988 (aarch64_simd_make_constant): Update call to
9989 aarch64_simd_valid_immediate.
9990 (aarch64_advsimd_valid_immediate_hs): New function.
9991 (aarch64_advsimd_valid_immediate): Likewise.
9992 (aarch64_simd_valid_immediate): Remove mode and inverse
9993 arguments. Rewrite to use the above. Use const_vec_duplicate_p
9994 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
9995 and aarch64_float_const_representable_p on the result.
9996 (aarch64_output_simd_mov_immediate): Remove mode argument.
9997 Update call to aarch64_simd_valid_immediate and use of
9998 simd_immediate_info.
9999 (aarch64_output_scalar_simd_mov_immediate): Update call
10002 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10003 Alan Hayward <alan.hayward@arm.com>
10004 David Sherwood <david.sherwood@arm.com>
10006 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
10007 (mode_nunits): Likewise CONST_MODE_NUNITS.
10008 * machmode.def (ADJUST_NUNITS): Document.
10009 * genmodes.c (mode_data::need_nunits_adj): New field.
10010 (blank_mode): Update accordingly.
10011 (adj_nunits): New variable.
10012 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
10014 (emit_mode_size_inline): Set need_bytesize_adj for all modes
10015 listed in adj_nunits.
10016 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
10017 listed in adj_nunits. Don't emit case statements for such modes.
10018 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
10019 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
10020 nothing if adj_nunits is nonnull.
10021 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
10022 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
10023 (emit_mode_fbit): Update use of print_maybe_const_decl.
10024 (emit_move_size): Likewise. Treat the array as non-const
10026 (emit_mode_adjustments): Handle adj_nunits.
10028 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10030 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
10031 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
10032 (VECTOR_MODES): Use it.
10033 (make_vector_modes): Take the prefix as an argument.
10035 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10036 Alan Hayward <alan.hayward@arm.com>
10037 David Sherwood <david.sherwood@arm.com>
10039 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
10040 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
10041 for MODE_VECTOR_BOOL.
10042 * machmode.def (VECTOR_BOOL_MODE): Document.
10043 * genmodes.c (VECTOR_BOOL_MODE): New macro.
10044 (make_vector_bool_mode): New function.
10045 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
10047 * lto-streamer-in.c (lto_input_mode_table): Likewise.
10048 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
10050 * stor-layout.c (int_mode_for_mode): Likewise.
10051 * tree.c (build_vector_type_for_mode): Likewise.
10052 * varasm.c (output_constant_pool_2): Likewise.
10053 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
10054 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
10055 for MODE_VECTOR_BOOL.
10056 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
10057 of mode class checks.
10058 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
10059 instead of a list of mode class checks.
10060 (expand_vector_scalar_condition): Likewise.
10061 (type_for_widest_vector_mode): Handle BImode as an inner mode.
10063 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10064 Alan Hayward <alan.hayward@arm.com>
10065 David Sherwood <david.sherwood@arm.com>
10067 * machmode.h (mode_size): Change from unsigned short to
10069 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
10070 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10071 or if measurement_type is not polynomial.
10072 (fixed_size_mode::includes_p): Check for constant-sized modes.
10073 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
10074 return a poly_uint16 rather than an unsigned short.
10075 (emit_mode_size): Change the type of mode_size from unsigned short
10076 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
10077 (emit_mode_adjustments): Cope with polynomial vector sizes.
10078 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10080 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10082 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
10083 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
10084 * caller-save.c (setup_save_areas): Likewise.
10085 (replace_reg_with_saved_mem): Likewise.
10086 * calls.c (emit_library_call_value_1): Likewise.
10087 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
10088 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
10089 (gen_lowpart_for_combine): Likewise.
10090 * convert.c (convert_to_integer_1): Likewise.
10091 * cse.c (equiv_constant, cse_insn): Likewise.
10092 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
10093 (cselib_subst_to_values): Likewise.
10094 * dce.c (word_dce_process_block): Likewise.
10095 * df-problems.c (df_word_lr_mark_ref): Likewise.
10096 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
10097 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
10098 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
10099 (rtl_for_decl_location): Likewise.
10100 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
10101 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
10102 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
10103 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
10104 (expand_expr_real_1): Likewise.
10105 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
10106 (pad_below): Likewise.
10107 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10108 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
10109 * ira.c (get_subreg_tracking_sizes): Likewise.
10110 * ira-build.c (ira_create_allocno_objects): Likewise.
10111 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
10112 (ira_sort_regnos_for_alter_reg): Likewise.
10113 * ira-costs.c (record_operand_costs): Likewise.
10114 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
10115 (resolve_simple_move): Likewise.
10116 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
10117 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
10118 (lra_constraints): Likewise.
10119 (CONST_POOL_OK_P): Reject variable-sized modes.
10120 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
10121 (add_pseudo_to_slot, lra_spill): Likewise.
10122 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10123 * optabs-query.c (get_best_extraction_insn): Likewise.
10124 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10125 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
10126 (expand_mult_highpart, valid_multiword_target_p): Likewise.
10127 * recog.c (offsettable_address_addr_space_p): Likewise.
10128 * regcprop.c (maybe_mode_change): Likewise.
10129 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
10130 * regrename.c (build_def_use): Likewise.
10131 * regstat.c (dump_reg_info): Likewise.
10132 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
10133 (find_reloads, find_reloads_subreg_address): Likewise.
10134 * reload1.c (eliminate_regs_1): Likewise.
10135 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
10136 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
10137 (simplify_binary_operation_1, simplify_subreg): Likewise.
10138 * targhooks.c (default_function_arg_padding): Likewise.
10139 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
10140 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
10141 (verify_gimple_assign_ternary): Likewise.
10142 * tree-inline.c (estimate_move_cost): Likewise.
10143 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10144 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
10145 (get_address_cost_ainc): Likewise.
10146 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
10147 (vect_supportable_dr_alignment): Likewise.
10148 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10149 (vectorizable_reduction): Likewise.
10150 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
10151 (vectorizable_operation, vectorizable_load): Likewise.
10152 * tree.c (build_same_sized_truth_vector_type): Likewise.
10153 * valtrack.c (cleanup_auto_inc_dec): Likewise.
10154 * var-tracking.c (emit_note_insn_var_location): Likewise.
10155 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
10156 (ADDR_VEC_ALIGN): Likewise.
10158 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10159 Alan Hayward <alan.hayward@arm.com>
10160 David Sherwood <david.sherwood@arm.com>
10162 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
10164 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10165 or if measurement_type is polynomial.
10166 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
10167 * combine.c (make_extraction): Likewise.
10168 * dse.c (find_shift_sequence): Likewise.
10169 * dwarf2out.c (mem_loc_descriptor): Likewise.
10170 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
10171 (extract_bit_field, extract_low_bits): Likewise.
10172 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
10173 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
10174 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
10175 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
10176 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10177 * reload.c (find_reloads): Likewise.
10178 * reload1.c (alter_reg): Likewise.
10179 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
10180 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
10181 * tree-if-conv.c (predicate_mem_writes): Likewise.
10182 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
10183 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
10184 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
10185 * valtrack.c (dead_debug_insert_temp): Likewise.
10186 * varasm.c (mergeable_constant_section): Likewise.
10187 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
10189 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10190 Alan Hayward <alan.hayward@arm.com>
10191 David Sherwood <david.sherwood@arm.com>
10193 * expr.c (expand_assignment): Cope with polynomial mode sizes
10194 when assigning to a CONCAT.
10196 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10197 Alan Hayward <alan.hayward@arm.com>
10198 David Sherwood <david.sherwood@arm.com>
10200 * machmode.h (mode_precision): Change from unsigned short to
10202 (mode_to_precision): Return a poly_uint16 rather than an unsigned
10204 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
10205 or if measurement_type is not polynomial.
10206 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
10207 in which the mode is already known to be a scalar_int_mode.
10208 * genmodes.c (emit_mode_precision): Change the type of mode_precision
10209 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
10211 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10212 for GET_MODE_PRECISION.
10213 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10214 for GET_MODE_PRECISION.
10215 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
10217 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
10218 (expand_field_assignment, make_extraction): Likewise.
10219 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
10220 (get_last_value): Likewise.
10221 * convert.c (convert_to_integer_1): Likewise.
10222 * cse.c (cse_insn): Likewise.
10223 * expr.c (expand_expr_real_1): Likewise.
10224 * lra-constraints.c (simplify_operand_subreg): Likewise.
10225 * optabs-query.c (can_atomic_load_p): Likewise.
10226 * optabs.c (expand_atomic_load): Likewise.
10227 (expand_atomic_store): Likewise.
10228 * ree.c (combine_reaching_defs): Likewise.
10229 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
10230 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
10231 * tree.h (type_has_mode_precision_p): Likewise.
10232 * ubsan.c (instrument_si_overflow): Likewise.
10234 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10235 Alan Hayward <alan.hayward@arm.com>
10236 David Sherwood <david.sherwood@arm.com>
10238 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
10239 polynomial numbers of units.
10240 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
10241 (valid_vector_subparts_p): New function.
10242 (build_vector_type): Remove temporary shim and take the number
10243 of units as a poly_uint64 rather than an int.
10244 (build_opaque_vector_type): Take the number of units as a
10245 poly_uint64 rather than an int.
10246 * tree.c (build_vector_from_ctor): Handle polynomial
10247 TYPE_VECTOR_SUBPARTS.
10248 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
10249 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
10250 (build_vector_from_val): If the number of units is variable,
10251 use build_vec_duplicate_cst for constant operands and
10252 VEC_DUPLICATE_EXPR otherwise.
10253 (make_vector_type): Remove temporary is_constant ().
10254 (build_vector_type, build_opaque_vector_type): Take the number of
10255 units as a poly_uint64 rather than an int.
10256 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
10258 * cfgexpand.c (expand_debug_expr): Likewise.
10259 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
10260 (store_constructor, expand_expr_real_1): Likewise.
10261 (const_scalar_mask_from_tree): Likewise.
10262 * fold-const-call.c (fold_const_reduction): Likewise.
10263 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
10264 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
10265 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
10266 (fold_relational_const): Likewise.
10267 (native_interpret_vector): Likewise. Change the size from an
10268 int to an unsigned int.
10269 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
10270 TYPE_VECTOR_SUBPARTS.
10271 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
10272 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
10273 duplicating a non-constant operand into a variable-length vector.
10274 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
10275 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
10276 * ipa-icf.c (sem_variable::equals): Likewise.
10277 * match.pd: Likewise.
10278 * omp-simd-clone.c (simd_clone_subparts): Likewise.
10279 * print-tree.c (print_node): Likewise.
10280 * stor-layout.c (layout_type): Likewise.
10281 * targhooks.c (default_builtin_vectorization_cost): Likewise.
10282 * tree-cfg.c (verify_gimple_comparison): Likewise.
10283 (verify_gimple_assign_binary): Likewise.
10284 (verify_gimple_assign_ternary): Likewise.
10285 (verify_gimple_assign_single): Likewise.
10286 * tree-pretty-print.c (dump_generic_node): Likewise.
10287 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10288 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
10289 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
10290 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
10291 (vect_shift_permute_load_chain): Likewise.
10292 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
10293 (expand_vector_condition, optimize_vector_constructor): Likewise.
10294 (lower_vec_perm, get_compute_type): Likewise.
10295 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10296 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
10297 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
10298 (vect_recog_mask_conversion_pattern): Likewise.
10299 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
10300 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
10301 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10302 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
10303 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
10304 (vectorizable_shift, vectorizable_operation, vectorizable_store)
10305 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
10306 (supportable_widening_operation): Likewise.
10307 (supportable_narrowing_operation): Likewise.
10308 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
10310 * varasm.c (output_constant): Likewise.
10312 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10313 Alan Hayward <alan.hayward@arm.com>
10314 David Sherwood <david.sherwood@arm.com>
10316 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
10317 so that both the length == 3 and length != 3 cases set up their
10318 own permute vectors. Add comments explaining why we know the
10319 number of elements is constant.
10320 (vect_permute_load_chain): Likewise.
10322 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10323 Alan Hayward <alan.hayward@arm.com>
10324 David Sherwood <david.sherwood@arm.com>
10326 * machmode.h (mode_nunits): Change from unsigned char to
10328 (ONLY_FIXED_SIZE_MODES): New macro.
10329 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
10330 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
10331 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
10333 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
10334 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
10335 or if measurement_type is not polynomial.
10336 * genmodes.c (ZERO_COEFFS): New macro.
10337 (emit_mode_nunits_inline): Make mode_nunits_inline return a
10339 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
10340 Use ZERO_COEFFS when emitting initializers.
10341 * data-streamer.h (bp_pack_poly_value): New function.
10342 (bp_unpack_poly_value): Likewise.
10343 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10344 for GET_MODE_NUNITS.
10345 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10346 for GET_MODE_NUNITS.
10347 * tree.c (make_vector_type): Remove temporary shim and make
10348 the real function take the number of units as a poly_uint64
10349 rather than an int.
10350 (build_vector_type_for_mode): Handle polynomial nunits.
10351 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
10352 * emit-rtl.c (const_vec_series_p_1): Likewise.
10353 (gen_rtx_CONST_VECTOR): Likewise.
10354 * fold-const.c (test_vec_duplicate_folding): Likewise.
10355 * genrecog.c (validate_pattern): Likewise.
10356 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
10357 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10358 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
10359 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
10360 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
10361 * rtlanal.c (subreg_get_info): Likewise.
10362 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10363 (vect_grouped_load_supported): Likewise.
10364 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
10365 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
10366 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
10367 (simplify_const_unary_operation, simplify_binary_operation_1)
10368 (simplify_const_binary_operation, simplify_ternary_operation)
10369 (test_vector_ops_duplicate, test_vector_ops): Likewise.
10370 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
10371 instead of CONST_VECTOR_NUNITS.
10372 * varasm.c (output_constant_pool_2): Likewise.
10373 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
10374 explicit-encoded elements in the XVEC for variable-length vectors.
10376 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10378 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
10380 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10381 Alan Hayward <alan.hayward@arm.com>
10382 David Sherwood <david.sherwood@arm.com>
10384 * coretypes.h (fixed_size_mode): Declare.
10385 (fixed_size_mode_pod): New typedef.
10386 * builtins.h (target_builtins::x_apply_args_mode)
10387 (target_builtins::x_apply_result_mode): Change type to
10388 fixed_size_mode_pod.
10389 * builtins.c (apply_args_size, apply_result_size, result_vector)
10390 (expand_builtin_apply_args_1, expand_builtin_apply)
10391 (expand_builtin_return): Update accordingly.
10393 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10395 * cse.c (hash_rtx_cb): Hash only the encoded elements.
10396 * cselib.c (cselib_hash_rtx): Likewise.
10397 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
10398 CONST_VECTOR encoding.
10400 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10401 Jeff Law <law@redhat.com>
10404 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
10405 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
10406 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
10407 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
10410 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
10411 explicitly probe *sp in a noreturn function if there were any callee
10412 register saves or frame pointer is needed.
10414 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10417 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
10418 BLKmode for ternary, binary or unary expressions.
10421 * var-tracking.c (delete_vta_debug_insn): New inline function.
10422 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
10423 insns from get_insns () to NULL instead of each bb separately.
10424 Use delete_vta_debug_insn. No longer static.
10425 (vt_debug_insns_local, variable_tracking_main_1): Adjust
10426 delete_vta_debug_insns callers.
10427 * rtl.h (delete_vta_debug_insns): Declare.
10428 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
10429 instead of variable_tracking_main.
10431 2018-01-03 Martin Sebor <msebor@redhat.com>
10433 PR tree-optimization/83603
10434 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
10435 arguments past the endof the argument list in functions declared
10436 without a prototype.
10437 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
10438 Avoid checking when arguments are null.
10440 2018-01-03 Martin Sebor <msebor@redhat.com>
10443 * doc/extend.texi (attribute const): Fix a typo.
10444 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
10445 issuing -Wsuggest-attribute for void functions.
10447 2018-01-03 Martin Sebor <msebor@redhat.com>
10449 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
10450 offset_int::from instead of wide_int::to_shwi.
10451 (maybe_diag_overlap): Remove assertion.
10452 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
10453 * gimple-ssa-sprintf.c (format_directive): Same.
10454 (parse_directive): Same.
10455 (sprintf_dom_walker::compute_format_length): Same.
10456 (try_substitute_return_value): Same.
10458 2018-01-03 Jeff Law <law@redhat.com>
10460 PR middle-end/83654
10461 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
10462 non-constant residual for zero at runtime and avoid probing in
10463 that case. Reorganize code for trailing problem to mirror handling
10466 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10468 PR tree-optimization/83501
10469 * tree-ssa-strlen.c (get_string_cst): New.
10470 (handle_char_store): Call get_string_cst.
10472 2018-01-03 Martin Liska <mliska@suse.cz>
10474 PR tree-optimization/83593
10475 * tree-ssa-strlen.c: Include tree-cfg.h.
10476 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
10477 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
10478 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
10480 (strlen_dom_walker::before_dom_children): Call
10481 gimple_purge_dead_eh_edges. Dump tranformation with details
10483 (strlen_dom_walker::before_dom_children): Update call by adding
10484 new argument cleanup_eh.
10485 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
10487 2018-01-03 Martin Liska <mliska@suse.cz>
10490 * cif-code.def (VARIADIC_THUNK): New enum value.
10491 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
10494 2018-01-03 Jan Beulich <jbeulich@suse.com>
10496 * sse.md (mov<mode>_internal): Tighten condition for when to use
10497 vmovdqu<ssescalarsize> for TI and OI modes.
10499 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10501 Update copyright years.
10503 2018-01-03 Martin Liska <mliska@suse.cz>
10506 * ipa-visibility.c (function_and_variable_visibility): Skip
10507 functions with noipa attribure.
10509 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10511 * gcc.c (process_command): Update copyright notice dates.
10512 * gcov-dump.c (print_version): Ditto.
10513 * gcov.c (print_version): Ditto.
10514 * gcov-tool.c (print_version): Ditto.
10515 * gengtype.c (create_file): Ditto.
10516 * doc/cpp.texi: Bump @copying's copyright year.
10517 * doc/cppinternals.texi: Ditto.
10518 * doc/gcc.texi: Ditto.
10519 * doc/gccint.texi: Ditto.
10520 * doc/gcov.texi: Ditto.
10521 * doc/install.texi: Ditto.
10522 * doc/invoke.texi: Ditto.
10524 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10526 * vector-builder.h (vector_builder::m_full_nelts): Change from
10527 unsigned int to poly_uint64.
10528 (vector_builder::full_nelts): Update prototype accordingly.
10529 (vector_builder::new_vector): Likewise.
10530 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
10531 (vector_builder::operator ==): Likewise.
10532 (vector_builder::finalize): Likewise.
10533 * int-vector-builder.h (int_vector_builder::int_vector_builder):
10534 Take the number of elements as a poly_uint64 rather than an
10536 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
10537 from unsigned int to poly_uint64.
10538 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
10539 (vec_perm_indices::new_vector): Likewise.
10540 (vec_perm_indices::length): Likewise.
10541 (vec_perm_indices::nelts_per_input): Likewise.
10542 (vec_perm_indices::input_nelts): Likewise.
10543 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
10544 number of elements per input as a poly_uint64 rather than an
10545 unsigned int. Use the original encoding for variable-length
10546 vectors, rather than clamping each individual element.
10547 For the second and subsequent elements in each pattern,
10548 clamp the step and base before clamping their sum.
10549 (vec_perm_indices::series_p): Handle polynomial element counts.
10550 (vec_perm_indices::all_in_range_p): Likewise.
10551 (vec_perm_indices_to_tree): Likewise.
10552 (vec_perm_indices_to_rtx): Likewise.
10553 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
10554 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
10555 (tree_vector_builder::new_binary_operation): Handle polynomial
10556 element counts. Return false if we need to know the number
10557 of elements at compile time.
10558 * fold-const.c (fold_vec_perm): Punt if the number of elements
10559 isn't known at compile time.
10561 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10563 * vec-perm-indices.h (vec_perm_builder): Change element type
10564 from HOST_WIDE_INT to poly_int64.
10565 (vec_perm_indices::element_type): Update accordingly.
10566 (vec_perm_indices::clamp): Handle polynomial element_types.
10567 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10568 (vec_perm_indices::all_in_range_p): Likewise.
10569 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
10571 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
10572 polynomial vec_perm_indices element types.
10573 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
10574 * fold-const.c (fold_vec_perm): Likewise.
10575 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
10576 * tree-vect-generic.c (lower_vec_perm): Likewise.
10577 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10578 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
10579 element type to HOST_WIDE_INT.
10581 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10582 Alan Hayward <alan.hayward@arm.com>
10583 David Sherwood <david.sherwood@arm.com>
10585 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
10586 rather than an int. Use plus_constant.
10587 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
10588 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
10590 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10591 Alan Hayward <alan.hayward@arm.com>
10592 David Sherwood <david.sherwood@arm.com>
10594 * calls.c (emit_call_1, expand_call): Change struct_value_size from
10595 a HOST_WIDE_INT to a poly_int64.
10597 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10598 Alan Hayward <alan.hayward@arm.com>
10599 David Sherwood <david.sherwood@arm.com>
10601 * calls.c (load_register_parameters): Cope with polynomial
10602 mode sizes. Require a constant size for BLKmode parameters
10603 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
10604 forces a parameter to be padded at the lsb end in order to
10605 fill a complete number of words, require the parameter size
10606 to be ordered wrt UNITS_PER_WORD.
10608 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10609 Alan Hayward <alan.hayward@arm.com>
10610 David Sherwood <david.sherwood@arm.com>
10612 * reload1.c (spill_stack_slot_width): Change element type
10613 from unsigned int to poly_uint64_pod.
10614 (alter_reg): Treat mode sizes as polynomial.
10616 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10617 Alan Hayward <alan.hayward@arm.com>
10618 David Sherwood <david.sherwood@arm.com>
10620 * reload.c (complex_word_subreg_p): New function.
10621 (reload_inner_reg_of_subreg, push_reload): Use it.
10623 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10624 Alan Hayward <alan.hayward@arm.com>
10625 David Sherwood <david.sherwood@arm.com>
10627 * lra-constraints.c (process_alt_operands): Reject matched
10628 operands whose sizes aren't ordered.
10629 (match_reload): Refer to this check here.
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 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
10636 that the mode size is in the set {1, 2, 4, 8, 16}.
10638 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10639 Alan Hayward <alan.hayward@arm.com>
10640 David Sherwood <david.sherwood@arm.com>
10642 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
10643 Use plus_constant instead of gen_rtx_PLUS.
10645 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10646 Alan Hayward <alan.hayward@arm.com>
10647 David Sherwood <david.sherwood@arm.com>
10649 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
10650 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
10651 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
10652 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
10653 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
10654 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
10655 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
10656 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
10657 * config/i386/i386.c (ix86_push_rounding): ...this new function.
10658 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
10660 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
10661 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
10662 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
10663 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
10664 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
10665 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
10666 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
10667 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
10668 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
10669 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
10671 * expr.c (emit_move_resolve_push): Treat the input and result
10672 of PUSH_ROUNDING as a poly_int64.
10673 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
10674 (emit_push_insn): Likewise.
10675 * lra-eliminations.c (mark_not_eliminable): Likewise.
10676 * recog.c (push_operand): Likewise.
10677 * reload1.c (elimination_effects): Likewise.
10678 * rtlanal.c (nonzero_bits1): Likewise.
10679 * calls.c (store_one_arg): Likewise. Require the padding to be
10680 known at compile time.
10682 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10683 Alan Hayward <alan.hayward@arm.com>
10684 David Sherwood <david.sherwood@arm.com>
10686 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
10687 Use plus_constant instead of gen_rtx_PLUS.
10689 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10690 Alan Hayward <alan.hayward@arm.com>
10691 David Sherwood <david.sherwood@arm.com>
10693 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
10694 rather than an int.
10696 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10697 Alan Hayward <alan.hayward@arm.com>
10698 David Sherwood <david.sherwood@arm.com>
10700 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
10701 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
10702 via stack temporaries. Treat the mode size as polynomial too.
10704 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10705 Alan Hayward <alan.hayward@arm.com>
10706 David Sherwood <david.sherwood@arm.com>
10708 * expr.c (expand_expr_real_2): When handling conversions involving
10709 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
10710 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
10711 as a poly_uint64 too.
10713 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10714 Alan Hayward <alan.hayward@arm.com>
10715 David Sherwood <david.sherwood@arm.com>
10717 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
10719 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10720 Alan Hayward <alan.hayward@arm.com>
10721 David Sherwood <david.sherwood@arm.com>
10723 * combine.c (can_change_dest_mode): Handle polynomial
10724 REGMODE_NATURAL_SIZE.
10725 * expmed.c (store_bit_field_1): Likewise.
10726 * expr.c (store_constructor): Likewise.
10727 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
10728 and polynomial REGMODE_NATURAL_SIZE.
10729 (gen_lowpart_common): Likewise.
10730 * reginfo.c (record_subregs_of_mode): Likewise.
10731 * rtlanal.c (read_modify_subreg_p): Likewise.
10733 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10734 Alan Hayward <alan.hayward@arm.com>
10735 David Sherwood <david.sherwood@arm.com>
10737 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
10738 numbers of elements.
10740 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10741 Alan Hayward <alan.hayward@arm.com>
10742 David Sherwood <david.sherwood@arm.com>
10744 * match.pd: Cope with polynomial numbers of vector elements.
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 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
10751 in a POINTER_PLUS_EXPR.
10753 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10754 Alan Hayward <alan.hayward@arm.com>
10755 David Sherwood <david.sherwood@arm.com>
10757 * omp-simd-clone.c (simd_clone_subparts): New function.
10758 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
10759 (ipa_simd_modify_function_body): Likewise.
10761 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10762 Alan Hayward <alan.hayward@arm.com>
10763 David Sherwood <david.sherwood@arm.com>
10765 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
10766 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
10767 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
10768 (expand_vector_condition, vector_element): Likewise.
10769 (subparts_gt): New function.
10770 (get_compute_type): Use subparts_gt.
10771 (count_type_subparts): Delete.
10772 (expand_vector_operations_1): Use subparts_gt instead of
10773 count_type_subparts.
10775 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10776 Alan Hayward <alan.hayward@arm.com>
10777 David Sherwood <david.sherwood@arm.com>
10779 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
10780 (vect_compile_time_alias): ...this new function. Do the calculation
10781 on poly_ints rather than trees.
10782 (vect_prune_runtime_alias_test_list): Update call accordingly.
10784 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10785 Alan Hayward <alan.hayward@arm.com>
10786 David Sherwood <david.sherwood@arm.com>
10788 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
10790 (vect_schedule_slp_instance): Likewise.
10792 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10793 Alan Hayward <alan.hayward@arm.com>
10794 David Sherwood <david.sherwood@arm.com>
10796 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
10797 constant and extern definitions for variable-length vectors.
10798 (vect_get_constant_vectors): Note that the number of units
10799 is known to be constant.
10801 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10802 Alan Hayward <alan.hayward@arm.com>
10803 David Sherwood <david.sherwood@arm.com>
10805 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
10806 of units as polynomial. Choose between WIDE and NARROW based
10809 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10810 Alan Hayward <alan.hayward@arm.com>
10811 David Sherwood <david.sherwood@arm.com>
10813 * tree-vect-stmts.c (simd_clone_subparts): New function.
10814 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
10816 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10817 Alan Hayward <alan.hayward@arm.com>
10818 David Sherwood <david.sherwood@arm.com>
10820 * tree-vect-stmts.c (vectorizable_call): Treat the number of
10821 vectors as polynomial. Use build_index_vector for
10822 IFN_GOMP_SIMD_LANE.
10824 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10825 Alan Hayward <alan.hayward@arm.com>
10826 David Sherwood <david.sherwood@arm.com>
10828 * tree-vect-stmts.c (get_load_store_type): Treat the number of
10829 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
10830 for variable-length vectors.
10831 (vectorizable_mask_load_store): Treat the number of units as
10832 polynomial, asserting that it is constant if the condition has
10833 already been enforced.
10834 (vectorizable_store, vectorizable_load): Likewise.
10836 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10837 Alan Hayward <alan.hayward@arm.com>
10838 David Sherwood <david.sherwood@arm.com>
10840 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
10841 of units as polynomial. Punt if we can't tell at compile time
10842 which vector contains the final result.
10844 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10845 Alan Hayward <alan.hayward@arm.com>
10846 David Sherwood <david.sherwood@arm.com>
10848 * tree-vect-loop.c (vectorizable_induction): Treat the number
10849 of units as polynomial. Punt on SLP inductions. Use an integer
10850 VEC_SERIES_EXPR for variable-length integer reductions. Use a
10851 cast of such a series for variable-length floating-point
10854 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10855 Alan Hayward <alan.hayward@arm.com>
10856 David Sherwood <david.sherwood@arm.com>
10858 * tree.h (build_index_vector): Declare.
10859 * tree.c (build_index_vector): New function.
10860 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
10861 of units as polynomial, forcibly converting it to a constant if
10862 vectorizable_reduction has already enforced the condition.
10863 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
10864 to create a {1,2,3,...} vector.
10865 (vectorizable_reduction): Treat the number of units as polynomial.
10866 Choose vectype_in based on the largest scalar element size rather
10867 than the smallest number of units. Enforce the restrictions
10870 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10871 Alan Hayward <alan.hayward@arm.com>
10872 David Sherwood <david.sherwood@arm.com>
10874 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
10875 number of units as polynomial.
10877 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10878 Alan Hayward <alan.hayward@arm.com>
10879 David Sherwood <david.sherwood@arm.com>
10881 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
10882 * target.def (autovectorize_vector_sizes): Return the vector sizes
10883 by pointer, using vector_sizes rather than a bitmask.
10884 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
10885 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
10886 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
10888 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
10889 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
10890 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
10891 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
10892 * omp-general.c (omp_max_vf): Likewise.
10893 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10894 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
10895 * tree-vect-loop.c (vect_analyze_loop): Likewise.
10896 * tree-vect-slp.c (vect_slp_bb): Likewise.
10897 * doc/tm.texi: Regenerate.
10898 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
10900 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
10901 the vector size as a poly_uint64 rather than an unsigned int.
10902 (current_vector_size): Change from an unsigned int to a poly_uint64.
10903 (get_vectype_for_scalar_type): Update accordingly.
10904 * tree.h (build_truth_vector_type): Take the size and number of
10905 units as a poly_uint64 rather than an unsigned int.
10906 (build_vector_type): Add a temporary overload that takes
10907 the number of units as a poly_uint64 rather than an unsigned int.
10908 * tree.c (make_vector_type): Likewise.
10909 (build_truth_vector_type): Take the number of units as a poly_uint64
10910 rather than an unsigned int.
10912 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10913 Alan Hayward <alan.hayward@arm.com>
10914 David Sherwood <david.sherwood@arm.com>
10916 * target.def (get_mask_mode): Take the number of units and length
10917 as poly_uint64s rather than unsigned ints.
10918 * targhooks.h (default_get_mask_mode): Update accordingly.
10919 * targhooks.c (default_get_mask_mode): Likewise.
10920 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
10921 * doc/tm.texi: Regenerate.
10923 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10924 Alan Hayward <alan.hayward@arm.com>
10925 David Sherwood <david.sherwood@arm.com>
10927 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
10928 * omp-general.c (omp_max_vf): Likewise.
10929 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
10930 (expand_omp_simd): Handle polynomial safelen.
10931 * omp-low.c (omplow_simd_context): Add a default constructor.
10932 (omplow_simd_context::max_vf): Change from int to poly_uint64.
10933 (lower_rec_simd_input_clauses): Update accordingly.
10934 (lower_rec_input_clauses): Likewise.
10936 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10937 Alan Hayward <alan.hayward@arm.com>
10938 David Sherwood <david.sherwood@arm.com>
10940 * tree-vectorizer.h (vect_nunits_for_cost): New function.
10941 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
10942 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
10943 (vect_analyze_slp_cost): Likewise.
10944 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
10945 (vect_model_load_cost): Likewise.
10947 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10948 Alan Hayward <alan.hayward@arm.com>
10949 David Sherwood <david.sherwood@arm.com>
10951 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
10952 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
10953 from an unsigned int * to a poly_uint64_pod *.
10954 (calculate_unrolling_factor): New function.
10955 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
10957 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10958 Alan Hayward <alan.hayward@arm.com>
10959 David Sherwood <david.sherwood@arm.com>
10961 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
10962 from an unsigned int to a poly_uint64.
10963 (_loop_vec_info::slp_unrolling_factor): Likewise.
10964 (_loop_vec_info::vectorization_factor): Change from an int
10966 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
10967 (vect_get_num_vectors): New function.
10968 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
10969 (vect_get_num_copies): Use vect_get_num_vectors.
10970 (vect_analyze_data_ref_dependences): Change max_vf from an int *
10971 to an unsigned int *.
10972 (vect_analyze_data_refs): Change min_vf from an int * to a
10974 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10975 than an unsigned HOST_WIDE_INT.
10976 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
10977 (vect_analyze_data_ref_dependence): Change max_vf from an int *
10978 to an unsigned int *.
10979 (vect_analyze_data_ref_dependences): Likewise.
10980 (vect_compute_data_ref_alignment): Handle polynomial vf.
10981 (vect_enhance_data_refs_alignment): Likewise.
10982 (vect_prune_runtime_alias_test_list): Likewise.
10983 (vect_shift_permute_load_chain): Likewise.
10984 (vect_supportable_dr_alignment): Likewise.
10985 (dependence_distance_ge_vf): Take the vectorization factor as a
10986 poly_uint64 rather than an unsigned HOST_WIDE_INT.
10987 (vect_analyze_data_refs): Change min_vf from an int * to a
10989 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
10990 vfm1 as a poly_uint64 rather than an int. Make the same change
10991 for the returned bound_scalar.
10992 (vect_gen_vector_loop_niters): Handle polynomial vf.
10993 (vect_do_peeling): Likewise. Update call to
10994 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
10995 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
10997 * tree-vect-loop.c (vect_determine_vectorization_factor)
10998 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
10999 (vect_get_known_peeling_cost): Likewise.
11000 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
11001 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
11002 (vect_transform_loop): Likewise. Use the lowest possible VF when
11003 updating the upper bounds of the loop.
11004 (vect_min_worthwhile_factor): Make static. Return an unsigned int
11005 rather than an int.
11006 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
11007 polynomial unroll factors.
11008 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
11009 (vect_make_slp_decision): Likewise.
11010 (vect_supported_load_permutation_p): Likewise, and polynomial
11012 (vect_analyze_slp_cost): Handle polynomial vf.
11013 (vect_slp_analyze_node_operations): Likewise.
11014 (vect_slp_analyze_bb_1): Likewise.
11015 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11016 than an unsigned HOST_WIDE_INT.
11017 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
11018 (vectorizable_load): Handle polynomial vf.
11019 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
11021 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
11023 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11024 Alan Hayward <alan.hayward@arm.com>
11025 David Sherwood <david.sherwood@arm.com>
11027 * match.pd: Handle bit operations involving three constants
11028 and try to fold one pair.
11030 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11032 * tree-vect-loop-manip.c: Include gimple-fold.h.
11033 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
11034 niters_maybe_zero parameters. Handle other cases besides a step of 1.
11035 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
11036 Add a path that uses a step of VF instead of 1, but disable it
11038 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
11039 and niters_no_overflow parameters. Update calls to
11040 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
11041 Create a new SSA name if the latter choses to use a ste other
11042 than zero, and return it via niters_vector_mult_vf_var.
11043 * tree-vect-loop.c (vect_transform_loop): Update calls to
11044 vect_do_peeling, vect_gen_vector_loop_niters and
11045 slpeel_make_loop_iterate_ntimes.
11046 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
11047 (vect_gen_vector_loop_niters): Update declarations after above changes.
11049 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
11051 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
11052 128-bit round to integer instructions.
11053 (ceil<mode>2): Likewise.
11054 (btrunc<mode>2): Likewise.
11055 (round<mode>2): Likewise.
11057 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11059 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
11060 unaligned VSX load/store on P8/P9.
11061 (expand_block_clear): Allow the use of unaligned VSX
11062 load/store on P8/P9.
11064 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11066 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
11068 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
11069 swap associated with both a load and a store.
11071 2018-01-02 Andrew Waterman <andrew@sifive.com>
11073 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
11074 * config/riscv/riscv.md (clear_cache): Use it.
11076 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
11078 * web.c: Remove out-of-date comment.
11080 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11082 * expr.c (fixup_args_size_notes): Check that any existing
11083 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
11084 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
11085 (emit_single_push_insn): ...here.
11087 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11089 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
11090 (const_vector_encoded_nelts): New function.
11091 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
11092 (const_vector_int_elt, const_vector_elt): Declare.
11093 * emit-rtl.c (const_vector_int_elt_1): New function.
11094 (const_vector_elt): Likewise.
11095 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
11096 of CONST_VECTOR_ELT.
11098 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11100 * expr.c: Include rtx-vector-builder.h.
11101 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
11102 directly on the tree encoding.
11103 (const_vector_from_tree): Likewise.
11104 * optabs.c: Include rtx-vector-builder.h.
11105 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
11106 sequence of "u" values.
11107 * vec-perm-indices.c: Include rtx-vector-builder.h.
11108 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
11109 directly on the vec_perm_indices encoding.
11111 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11113 * doc/rtl.texi (const_vector): Describe new encoding scheme.
11114 * Makefile.in (OBJS): Add rtx-vector-builder.o.
11115 * rtx-vector-builder.h: New file.
11116 * rtx-vector-builder.c: Likewise.
11117 * rtl.h (rtx_def::u2): Add a const_vector field.
11118 (CONST_VECTOR_NPATTERNS): New macro.
11119 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
11120 (CONST_VECTOR_DUPLICATE_P): Likewise.
11121 (CONST_VECTOR_STEPPED_P): Likewise.
11122 (CONST_VECTOR_ENCODED_ELT): Likewise.
11123 (const_vec_duplicate_p): Check for a duplicated vector encoding.
11124 (unwrap_const_vec_duplicate): Likewise.
11125 (const_vec_series_p): Check for a non-duplicated vector encoding.
11126 Say that the function only returns true for integer vectors.
11127 * emit-rtl.c: Include rtx-vector-builder.h.
11128 (gen_const_vec_duplicate_1): Delete.
11129 (gen_const_vector): Call gen_const_vec_duplicate instead of
11130 gen_const_vec_duplicate_1.
11131 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
11132 (gen_const_vec_duplicate): Use rtx_vector_builder.
11133 (gen_const_vec_series): Likewise.
11134 (gen_rtx_CONST_VECTOR): Likewise.
11135 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
11136 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11137 Build a new vector rather than modifying a CONST_VECTOR in-place.
11138 (handle_special_swappables): Update call accordingly.
11139 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
11140 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11141 Build a new vector rather than modifying a CONST_VECTOR in-place.
11142 (handle_special_swappables): Update call accordingly.
11144 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11146 * simplify-rtx.c (simplify_const_binary_operation): Use
11147 CONST_VECTOR_ELT instead of XVECEXP.
11149 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11151 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
11152 the selector elements to be different from the data elements
11153 if the selector is a VECTOR_CST.
11154 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
11155 ssizetype for the selector.
11157 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11159 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
11160 before testing each element individually.
11161 * tree-vect-generic.c (lower_vec_perm): Likewise.
11163 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11165 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
11166 * selftest-run-tests.c (selftest::run_tests): Call it.
11167 * vector-builder.h (vector_builder::operator ==): New function.
11168 (vector_builder::operator !=): Likewise.
11169 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
11170 (vec_perm_indices::all_from_input_p): New function.
11171 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11172 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
11173 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
11174 instead of reading the VECTOR_CST directly. Detect whether both
11175 vector inputs are the same before constructing the vec_perm_indices,
11176 and update the number of inputs argument accordingly. Use the
11177 utility functions added above. Only construct sel2 if we need to.
11179 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11181 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
11182 the broadcast of the low byte.
11183 (expand_mult_highpart): Use an explicit encoding for the permutes.
11184 * optabs-query.c (can_mult_highpart_p): Likewise.
11185 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
11186 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11187 (vectorizable_bswap): Likewise.
11188 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
11189 explicit encoding for the power-of-2 permutes.
11190 (vect_permute_store_chain): Likewise.
11191 (vect_grouped_load_supported): Likewise.
11192 (vect_permute_load_chain): Likewise.
11194 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11196 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
11197 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
11198 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
11199 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11200 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
11201 (vect_gen_perm_mask_any): Likewise.
11203 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11205 * int-vector-builder.h: New file.
11206 * vec-perm-indices.h: Include int-vector-builder.h.
11207 (vec_perm_indices): Redefine as an int_vector_builder.
11208 (auto_vec_perm_indices): Delete.
11209 (vec_perm_builder): Redefine as a stand-alone class.
11210 (vec_perm_indices::vec_perm_indices): New function.
11211 (vec_perm_indices::clamp): Likewise.
11212 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
11213 (vec_perm_indices::new_vector): New function.
11214 (vec_perm_indices::new_expanded_vector): Update for new
11215 vec_perm_indices class.
11216 (vec_perm_indices::rotate_inputs): New function.
11217 (vec_perm_indices::all_in_range_p): Operate directly on the
11218 encoded form, without computing elided elements.
11219 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
11220 encoding. Update for new vec_perm_indices class.
11221 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
11222 the given vec_perm_builder.
11223 (expand_vec_perm_var): Update vec_perm_builder constructor.
11224 (expand_mult_highpart): Use vec_perm_builder instead of
11225 auto_vec_perm_indices.
11226 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
11227 vec_perm_indices instead of auto_vec_perm_indices. Use a single
11228 or double series encoding as appropriate.
11229 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
11230 vec_perm_indices instead of auto_vec_perm_indices.
11231 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11232 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11233 (vect_permute_store_chain): Likewise.
11234 (vect_grouped_load_supported): Likewise.
11235 (vect_permute_load_chain): Likewise.
11236 (vect_shift_permute_load_chain): Likewise.
11237 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11238 (vect_transform_slp_perm_load): Likewise.
11239 (vect_schedule_slp_instance): Likewise.
11240 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11241 (vectorizable_mask_load_store): Likewise.
11242 (vectorizable_bswap): Likewise.
11243 (vectorizable_store): Likewise.
11244 (vectorizable_load): Likewise.
11245 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
11246 vec_perm_indices instead of auto_vec_perm_indices. Use
11247 tree_to_vec_perm_builder to read the vector from a tree.
11248 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
11249 vec_perm_builder instead of a vec_perm_indices.
11250 (have_whole_vector_shift): Use vec_perm_builder and
11251 vec_perm_indices instead of auto_vec_perm_indices. Leave the
11252 truncation to calc_vec_perm_mask_for_shift.
11253 (vect_create_epilog_for_reduction): Likewise.
11254 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
11255 from auto_vec_perm_indices to vec_perm_indices.
11256 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11257 instead of changing individual elements.
11258 (aarch64_vectorize_vec_perm_const): Use new_vector to install
11259 the vector in d.perm.
11260 * config/arm/arm.c (expand_vec_perm_d::perm): Change
11261 from auto_vec_perm_indices to vec_perm_indices.
11262 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11263 instead of changing individual elements.
11264 (arm_vectorize_vec_perm_const): Use new_vector to install
11265 the vector in d.perm.
11266 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
11267 Update vec_perm_builder constructor.
11268 (rs6000_expand_interleave): Likewise.
11269 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
11270 (rs6000_expand_interleave): Likewise.
11272 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11274 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
11275 to qimode could truncate the indices.
11276 * optabs.c (expand_vec_perm_var): Likewise.
11278 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11280 * Makefile.in (OBJS): Add vec-perm-indices.o.
11281 * vec-perm-indices.h: New file.
11282 * vec-perm-indices.c: Likewise.
11283 * target.h (vec_perm_indices): Replace with a forward class
11285 (auto_vec_perm_indices): Move to vec-perm-indices.h.
11286 * optabs.h: Include vec-perm-indices.h.
11287 (expand_vec_perm): Delete.
11288 (selector_fits_mode_p, expand_vec_perm_var): Declare.
11289 (expand_vec_perm_const): Declare.
11290 * target.def (vec_perm_const_ok): Replace with...
11291 (vec_perm_const): ...this new hook.
11292 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
11293 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
11294 * doc/tm.texi: Regenerate.
11295 * optabs.def (vec_perm_const): Delete.
11296 * doc/md.texi (vec_perm_const): Likewise.
11297 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
11298 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
11299 expand_vec_perm for constant permutation vectors. Assert that
11300 the mode of variable permutation vectors is the integer equivalent
11301 of the mode that is being permuted.
11302 * optabs-query.h (selector_fits_mode_p): Declare.
11303 * optabs-query.c: Include vec-perm-indices.h.
11304 (selector_fits_mode_p): New function.
11305 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
11306 is defined, instead of checking whether the vec_perm_const_optab
11307 exists. Use targetm.vectorize.vec_perm_const instead of
11308 targetm.vectorize.vec_perm_const_ok. Check whether the indices
11309 fit in the vector mode before using a variable permute.
11310 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
11311 vec_perm_indices instead of an rtx.
11312 (expand_vec_perm): Replace with...
11313 (expand_vec_perm_const): ...this new function. Take the selector
11314 as a vec_perm_indices rather than an rtx. Also take the mode of
11315 the selector. Update call to shift_amt_for_vec_perm_mask.
11316 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
11317 Use vec_perm_indices::new_expanded_vector to expand the original
11318 selector into bytes. Check whether the indices fit in the vector
11319 mode before using a variable permute.
11320 (expand_vec_perm_var): Make global.
11321 (expand_mult_highpart): Use expand_vec_perm_const.
11322 * fold-const.c: Includes vec-perm-indices.h.
11323 * tree-ssa-forwprop.c: Likewise.
11324 * tree-vect-data-refs.c: Likewise.
11325 * tree-vect-generic.c: Likewise.
11326 * tree-vect-loop.c: Likewise.
11327 * tree-vect-slp.c: Likewise.
11328 * tree-vect-stmts.c: Likewise.
11329 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
11331 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
11332 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
11333 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
11334 (aarch64_vectorize_vec_perm_const): ...this new function.
11335 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11336 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11337 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
11338 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
11339 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11340 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11341 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
11343 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
11344 check for NEON modes.
11345 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
11346 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
11347 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
11348 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
11350 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
11351 the old VEC_PERM_CONST conditions.
11352 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
11353 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
11354 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
11355 (ia64_vectorize_vec_perm_const_ok): Merge into...
11356 (ia64_vectorize_vec_perm_const): ...this new function.
11357 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
11358 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
11359 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
11360 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
11361 * config/mips/mips.c (mips_expand_vec_perm_const)
11362 (mips_vectorize_vec_perm_const_ok): Merge into...
11363 (mips_vectorize_vec_perm_const): ...this new function.
11364 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
11365 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
11366 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
11367 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
11368 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
11369 (rs6000_expand_vec_perm_const): Delete.
11370 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
11372 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11373 (altivec_expand_vec_perm_const_le): Take each operand individually.
11374 Operate on constant selectors rather than rtxes.
11375 (altivec_expand_vec_perm_const): Likewise. Update call to
11376 altivec_expand_vec_perm_const_le.
11377 (rs6000_expand_vec_perm_const): Delete.
11378 (rs6000_vectorize_vec_perm_const_ok): Delete.
11379 (rs6000_vectorize_vec_perm_const): New function.
11380 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11381 an element count and rtx array.
11382 (rs6000_expand_extract_even): Update call accordingly.
11383 (rs6000_expand_interleave): Likewise.
11384 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
11385 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
11386 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
11387 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
11388 (rs6000_expand_vec_perm_const): Delete.
11389 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11390 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11391 (altivec_expand_vec_perm_const_le): Take each operand individually.
11392 Operate on constant selectors rather than rtxes.
11393 (altivec_expand_vec_perm_const): Likewise. Update call to
11394 altivec_expand_vec_perm_const_le.
11395 (rs6000_expand_vec_perm_const): Delete.
11396 (rs6000_vectorize_vec_perm_const_ok): Delete.
11397 (rs6000_vectorize_vec_perm_const): New function. Remove stray
11398 reference to the SPE evmerge intructions.
11399 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11400 an element count and rtx array.
11401 (rs6000_expand_extract_even): Update call accordingly.
11402 (rs6000_expand_interleave): Likewise.
11403 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
11404 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
11406 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11408 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11410 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
11411 vector mode and that that mode matches the mode of the data
11413 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
11414 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
11415 directly using expand_vec_perm_1 when forcing selectors into
11417 (expand_vec_perm_var): New function, split out from expand_vec_perm.
11419 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11421 * optabs-query.h (can_vec_perm_p): Delete.
11422 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
11423 * optabs-query.c (can_vec_perm_p): Split into...
11424 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
11425 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
11426 particular selector is valid.
11427 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11428 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11429 (vect_grouped_load_supported): Likewise.
11430 (vect_shift_permute_load_chain): Likewise.
11431 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11432 (vect_transform_slp_perm_load): Likewise.
11433 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11434 (vectorizable_bswap): Likewise.
11435 (vect_gen_perm_mask_checked): Likewise.
11436 * fold-const.c (fold_ternary_loc): Likewise. Don't take
11437 implementations of variable permutation vectors into account
11438 when deciding which selector to use.
11439 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
11440 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
11441 with a false third argument.
11442 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
11443 to test whether the constant selector is valid and can_vec_perm_var_p
11444 to test whether a variable selector is valid.
11446 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11448 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
11449 * optabs-query.c (can_vec_perm_p): Likewise.
11450 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
11451 instead of vec_perm_indices.
11452 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
11453 (vect_gen_perm_mask_checked): Likewise,
11454 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
11455 (vect_gen_perm_mask_checked): Likewise,
11457 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11459 * optabs-query.h (qimode_for_vec_perm): Declare.
11460 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
11461 (qimode_for_vec_perm): ...this new function.
11462 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
11464 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11466 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
11467 does not have a conditional at the top.
11469 2018-01-02 Richard Biener <rguenther@suse.de>
11471 * ipa-inline.c (big_speedup_p): Fix expression.
11473 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11476 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
11479 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11483 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
11484 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
11485 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
11486 cond_taken_branch_cost 3->4.
11488 2018-01-01 Jakub Jelinek <jakub@redhat.com>
11490 PR tree-optimization/83581
11491 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
11492 TODO_cleanup_cfg if any changes have been made.
11494 PR middle-end/83608
11495 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
11496 convert_modes if target mode has the right side, but different mode
11499 PR middle-end/83609
11500 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
11501 last argument when extracting from CONCAT. If either from_real or
11502 from_imag is NULL, use expansion through memory. If result is not
11503 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
11504 the parts directly to inner mode, if even that fails, use expansion
11507 PR middle-end/83623
11508 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
11509 check for bswap in mode rather than HImode and use that in expand_unop
11512 Copyright (C) 2018 Free Software Foundation, Inc.
11514 Copying and distribution of this file, with or without modification,
11515 are permitted in any medium without royalty provided the copyright
11516 notice and this notice are preserved.