1 2018-05-11 Martin Jambor <mjambor@suse.cz>
4 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
7 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
10 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
12 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
15 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
17 (cortex-m0): Use armv6s-m isa.
18 (cortex-m0plus): Likewise.
19 (cortex-m1): Likewise.
20 (cortex-m0.small-multiply): Likewise.
21 (cortex-m0plus.small-multiply): Likewise.
22 (cortex-m1.small-multiply): Likewise.
24 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
26 PR rtl-optimization/85638
27 * bb-reorder.c: Include common/common-target.h.
28 (create_forwarder_block): New function extracted from...
29 (fix_up_crossing_landing_pad): ...here. Rename into...
30 (dw2_fix_up_crossing_landing_pad): ...this.
31 (sjlj_fix_up_crossing_landing_pad): New function.
32 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
33 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
34 from both partitions and exit the loop after one iteration.
36 2018-05-08 Richard Biener <rguenther@suse.de>
38 Backport from mainline
39 2018-05-04 Richard Biener <rguenther@suse.de>
42 * fold-const.c (negate_expr_p): Restrict negation of operand
43 zero of a division to when we know that can happen without
45 (fold_negate_expr_1): Likewise.
47 2018-05-03 Richard Biener <rguenther@suse.de>
49 PR tree-optimization/85615
50 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
51 to loops not nested in BBs loop father to avoid creating multi-entry
54 2018-05-02 Richard Biener <rguenther@suse.de>
57 * gimplify.c (gimplify_save_expr): When in SSA form allow
58 SAVE_EXPRs to compute to SSA vars.
60 2018-05-02 Richard Biener <rguenther@suse.de>
62 PR tree-optimization/85597
63 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
64 do not use split vect_get_vec_defs call but call vect_get_slp_defs
67 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
70 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
71 (check_arch): Likewise.
72 (check_fpu): Return the result rather than printing it.
73 (end arch): Fix operator precedence.
75 (END): Print the result from check_fpu.
77 2018-05-08 Richard Biener <rguenther@suse.de>
79 Backport from mainline
80 2018-05-04 Richard Biener <rguenther@suse.de>
82 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
83 * gimple-ssa-store-merging.c
84 (imm_store_chain_info::output_merged_store): Remove redundant create,
85 release split_store vector contents on failure.
86 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
87 scalar stmt vector on cache hit.
89 2018-05-04 Richard Biener <rguenther@suse.de>
92 * tree-complex.c (tree_lower_complex): Handle removed blocks.
94 2018-05-07 Chung-Ju Wu <jasonwucj@gmail.com>
96 Backport from mainline.
97 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
99 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
100 GET_MODE_MASK before any checking.
101 (nds32_can_use_bset_p): Likewise.
102 (nds32_can_use_btgl_p): Likewise.
104 2018-05-07 Chung-Ju Wu <jasonwucj@gmail.com>
106 Backport from mainline.
107 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
109 * config/nds32/nds32-doubleword.md: New define_split pattern for
110 illegal register number.
112 2018-05-07 Chung-Ju Wu <jasonwucj@gmail.com>
114 Backport from mainline.
115 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
117 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
119 2018-05-07 Chung-Ju Wu <jasonwucj@gmail.com>
121 Backport from mainline.
122 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
124 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
126 2018-05-06 Jakub Jelinek <jakub@redhat.com>
129 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
130 the type is addressable. Don't force op into register if it has
133 2018-05-04 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
136 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
139 * config/aarch64/constraints.md (Usg): Limit to 31.
142 2018-05-02 Jakub Jelinek <jakub@redhat.com>
144 * BASE-VER: Set to 8.1.1.
146 2018-05-02 Release Manager
148 * GCC 8.1.0 released.
150 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
152 PR tree-optimization/85586
153 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
154 exit early for statements in the same group if the accesses are
157 2018-05-01 Jakub Jelinek <jakub@redhat.com>
160 * doc/install.texi2html: Replace _002d with - and _002a with * in
161 generated html files using sed.
163 2018-04-27 Martin Jambor <mjambor@suse.cz>
166 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
167 the jump function allows for passing through aggregate values.
169 2018-04-27 Jeff Law <law@redhat.com>
171 * config.gcc: Mark tile* targets as deprecated/obsolete.
173 2018-04-27 Jakub Jelinek <jakub@redhat.com>
175 PR tree-optimization/85529
176 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
177 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
178 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
179 zero extension or masking of the MSB bit.
180 (optimize_range_tests): Add FIRST_BB argument, pass it through
181 to optimize_range_tests_var_bound.
182 (maybe_optimize_range_tests, reassociate_bb): Adjust
183 optimize_range_tests callers.
185 2018-04-26 Richard Biener <rguenther@suse.de>
186 Jakub Jelinek <jakub@redhat.com>
188 * cgraph.h (symbol_table): Just declare debug method here.
189 * symtab.c (symbol_table::debug): Define.
191 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
193 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
195 2018-04-25 H.J. Lu <hongjiu.lu@intel.com>
197 Backport from mainline
198 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
201 * config/i386/i386.c (ix86_expand_builtin): Change memory
202 operand to XI, extend p0 to Pmode.
203 * config/i386/i386.md: Change unspec volatile and operand
204 1 mode to XI, change operand 0 mode to P.
206 2018-04-25 Jakub Jelinek <jakub@redhat.com>
208 * DEV-PHASE: Set to prerelease.
210 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
212 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
213 in __abskf2 and __powikf2.
215 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
218 * config/aarch64/constraints.md (Usg, Usj): New constraints.
219 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
220 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
221 Use the above on operand 2. Reindent.
222 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
224 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
227 * common/config/i386/i386-common.c (ix86_handle_option): Don't
229 * config/i386/i386.opt (mcet): Removed.
230 * doc/install.texi: Remove -mcet documentation.
231 * doc/invoke.texi: Likewise.
233 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
236 * doc/install.texi: Remove -mcet from bootstrap-cet.
238 2018-04-24 Jakub Jelinek <jakub@redhat.com>
241 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
242 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
246 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
247 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
248 containing a CONST_VECTOR.
250 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
252 * doc/install.texi: Update newlib dependency for nvptx.
254 2018-04-24 Jakub Jelinek <jakub@redhat.com>
257 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
258 instead of INTVAL when shifting x left.
260 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
262 PR tree-optimization/85478
263 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
264 vect_grouped_store_supported for single element vectors.
266 2018-04-24 Richard Biener <rguenther@suse.de>
269 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
270 load cost increase to the case of non-constant step.
272 2018-04-24 Jakub Jelinek <jakub@redhat.com>
275 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
276 destination if any_malformed_asm.
278 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
281 * expr.c (store_field): In the bitfield case, if the value comes from
282 a function call and is returned in registers by means of a PARALLEL,
283 do not change the mode of the temporary unless BLKmode and VOIDmode.
285 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
287 PR rtl-optimization/85423
288 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
289 dependencies to debug insns when the previous insn is non-debug.
291 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
293 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
294 enums into a single definition.
295 (fls): Fix predicates and printing.
298 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
300 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
301 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
302 and short u6 immediate.
303 (check_if_valid_sleep_operand): Remove.
304 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
306 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
308 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
309 flag_always_save_lp condition.
310 * config/nds32/nds32.opt (malways-save-lp): New option.
312 2018-04-22 Shiva Chen <shiva0217@gmail.com>
314 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
315 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
316 * config/nds32/nds32.h
317 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
318 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
320 2018-04-22 Shiva Chen <shiva0217@gmail.com>
322 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
323 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
325 2018-04-22 Shiva Chen <shiva0217@gmail.com>
326 Chung-Ju Wu <jasonwucj@gmail.com>
328 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
330 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
331 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
333 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
335 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
337 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
339 * config/nds32/nds32-protos.h (nds32_data_alignment,
340 nds32_local_alignment): Declare.
341 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
342 nds32_local_alignment): New functions.
343 (TARGET_CONSTANT_ALIGNMENT): Define.
344 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
346 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
348 * config/nds32/nds32.c
349 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
350 (TARGET_MODES_TIEABLE_P): Likewise.
352 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
354 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
357 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
358 Chung-Ju Wu <jasonwucj@gmail.com>
360 * config/nds32/constants.md (unspec_volatile_element): Add enum values
361 for unaligned access.
362 * config/nds32/nds32-intrinsic.c: Implementation of expanding
364 * config/nds32/nds32-intrinsic.md: Likewise.
365 * config/nds32/nds32_intrinsic.h: Likewise.
366 * config/nds32/nds32.h (nds32_builtins): Likewise.
367 * config/nds32/nds32.opt (munaligned-access): New option.
368 * config/nds32/nds32.c (nds32_asm_file_start): Display
369 flag_unaligned_access status.
371 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
373 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
374 -mno-relax is present.
375 * config/riscv/linux.h (LINK_SPEC): Ditto.
377 2018-04-20 Martin Sebor <msebor@redhat.com>
380 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
382 (gimple_fold_builtin_stxcpy_chk): Same.
383 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
385 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
388 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
389 __powikf2 when long double is IEEE 128-bit.
391 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
393 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
394 step to make sure stack always aligned.
396 2018-04-20 Carl Love <cel@us.ibm.com>
399 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
402 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
403 Tom de Vries <tom@codesourcery.com>
406 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
407 Emit insns for calls too.
408 (nvptx_find_par): Always look for worker-level predecessor insn.
409 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
411 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
412 (nvptx_process_pars): Propagate frames for calls.
414 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
417 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
419 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
420 (ix86_handle_option): Don't handle OPT_mibt.
421 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
423 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
425 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
426 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
427 (ix86_target_macros): Define __CET__ with flag_cf_protection
429 * config/i386/i386.c (isa2_opts): Remove -mibt.
430 * config/i386/i386.h (TARGET_IBT): Removed.
431 (TARGET_IBT_P): Likewise.
432 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
433 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
434 * config/i386/i386.opt (mcet): Update help message.
437 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
438 -mcet as an alias for -mshstk.
440 2018-04-20 Richard Biener <rguenther@suse.de>
443 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
444 complexity by forcing a single use of the multiply operand.
446 2018-04-20 Martin Jambor <mjambor@suse.cz>
449 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
450 recursion dependency to only apply to non-clones.
452 2018-04-20 Martin Jambor <mjambor@suse.cz>
455 * ipa-cp.c (create_specialized_node): Check that clones of
456 self-recursive edges exist during IPA-CP.
458 2018-04-19 Toon Moene <toon@moene.org>
460 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
463 2018-04-19 Jakub Jelinek <jakub@redhat.com>
465 PR tree-optimization/85467
466 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
467 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
468 VECTOR_CST element to type.
470 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
473 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
474 * config/i386/i386.md (builtin_setjmp_setup): Removed.
475 (builtin_longjmp): Likewise.
476 (save_stack_nonlocal): New pattern.
477 (restore_stack_nonlocal): Likewise.
479 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
482 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
483 Replace ASM_OUTPUT_LABEL with fprintf.
485 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
488 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
489 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
490 * config/i386/i386-c.c (ix86_target_macros_internal): Also
491 define __IBT__ and __SHSTK__ for -fcf-protection.
492 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
494 (ix86_trampoline_init): Likewise.
495 (x86_output_mi_thunk): Likewise.
496 (ix86_notrack_prefixed_insn_p): Likewise.
497 (ix86_option_override_internal): Don't disallow -fcf-protection.
498 * config/i386/i386.md (rdssp<mode>): Also enable for
500 (incssp<mode>): Likewise.
501 (nop_endbr): Likewise.
502 * config/i386/i386.opt (mcet): Change help message to built-in
506 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
507 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
508 enable CET built-in functions.
510 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
512 * common/config/i386/i386-common.c
513 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
514 OPTION_MASK_ISA_MOVDIRI_UNSET,
515 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
516 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
517 * config.gcc (movdirintrin.h): New header.
518 * config/i386/cpuid.h (bit_MOVDIRI,
519 bit_MOVDIR64B): New bits.
520 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
522 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
523 (VOID, PVOID, PCVOID)): New function types.
524 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
525 __builtin_ia32_directstoreu_u64,
526 __builtin_ia32_movdir64b): New builtins.
527 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
528 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
530 (ix86_valid_target_attribute_inner_p): Ditto.
531 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
532 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
533 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
534 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
535 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
536 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
537 (movdiri<mode>, movdir64b_<mode>): New.
538 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
539 * config/i386/immintrin.h: Include movdirintrin.h.
540 * config/i386/movdirintrin.h: New file.
541 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
543 2018-04-19 Richard Biener <rguenther@suse.de>
546 * cfg.c (clear_bb_flags): When loop state says we have
547 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
549 2018-04-19 Richard Biener <rguenther@suse.de>
551 PR tree-optimization/84737
552 * tree-vect-data-refs.c (vect_copy_ref_info): New function
553 copying restrict info.
554 (vect_setup_realignment): Use it.
555 * tree-vectorizer.h (vect_copy_ref_info): Declare.
556 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
557 the first DR to all generated stores.
558 (vectorizable_load): Likewise for loads.
560 2018-04-19 Jakub Jelinek <jakub@redhat.com>
562 PR tree-optimization/85446
563 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
564 the integral and pointer types to have the same precision.
566 * doc/install.texi: Document --disable-cet being the default and
569 2018-04-18 Martin Liska <mliska@suse.cz>
571 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
574 2018-04-18 Martin Liska <mliska@suse.cz>
577 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
580 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
581 arguments if they are comparable.
583 2018-04-18 Martin Liska <mliska@suse.cz>
586 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
589 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
592 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
595 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
596 ENDBR after calling __morestack.
598 2018-04-18 David Malcolm <dmalcolm@redhat.com>
601 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
602 by using gcc_base_ver to generate a gcc_driver_version, and use
603 it when generating GCC_DRIVER_NAME.
604 * configure: Regenerate.
606 2018-04-18 Jakub Jelinek <jakub@redhat.com>
609 * config.gcc: Obsolete powerpc*-*-*spe*.
611 2018-04-17 Jakub Jelinek <jakub@redhat.com>
614 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
615 (stabstr_D): Change type of unum from unsigned int to
616 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
619 2018-04-17 Jim Wilson <jimw@sifive.com>
622 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
623 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
624 Set arg_pointer_offset after using pretend_args_size.
626 2018-04-17 Jakub Jelinek <jakub@redhat.com>
628 PR rtl-optimization/85431
629 * dse.c (record_store): Ignore zero width stores.
632 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
633 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
634 __builtin_stack_restore rather than after it.
635 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
636 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
637 argument instead of virtual_dynamic_stack_rtx.
639 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
641 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
643 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
644 Add note to error message to explain internal mapping of overloaded
645 built-in function name to non-overloaded built-in function name.
646 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
649 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
652 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
653 where the inputs overlap with the output.
655 2018-04-17 Jakub Jelinek <jakub@redhat.com>
657 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
658 (=v, v) alternative and explicit "memory" attribute.
659 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
660 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
662 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
663 "sselog1" type instead of "sselog".
664 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
665 "sselog". Remove explicit "memory" attribute.
666 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
667 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
669 (vec_extract_hi_v32hi): Merge all alternatives into one, use
670 "sselog1" type instead of "sselog". Remove explicit "memory"
672 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
673 use "sselog1" type instead of "sselog". Remove explicit "memory"
675 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
676 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
678 (vec_extract_hi_v64qi): Merge all alternatives into one, use
679 "sselog1" type instead of "sselog". Remove explicit "memory"
681 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
682 use "sselog1" type instead of "sselog". Remove explicit "memory"
686 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
689 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
692 2018-04-17 Martin Jambor <mjambor@suse.cz>
695 * ipa-cp.c (create_specialized_node): Call
696 expand_all_artificial_thunks if necessary.
698 2018-04-17 Martin Liska <mliska@suse.cz>
701 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
702 in message, remote space in between '_G' and '('.
704 2018-04-17 Jakub Jelinek <jakub@redhat.com>
707 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
708 avx512f_vmcmp<mode>3<round_saeonly_name>,
709 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
710 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
711 avx512f_rndscale<mode><round_saeonly_name>,
712 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
713 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
714 Use %<iptr>2 instead of %2 for -masm=intel.
715 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
716 avx512f_vcvttss2usi<round_saeonly_name>,
717 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
719 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
720 avx512f_vcvttsd2usi<round_saeonly_name>,
721 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
722 Use %q1 instead of %1 for -masm=intel.
723 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
724 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
725 of %3 for -masm=intel.
726 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
728 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
730 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
732 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
733 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
735 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
737 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
738 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
739 %g1 and one with %0 and %1.
740 (avx512er_vmrcp28<mode><round_saeonly_name>,
741 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
743 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
744 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
745 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
746 of %0 and %{%4%} for -masm=intel.
747 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
748 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
749 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
750 order of %0 and %{%5%}%{z%} for -masm=intel.
752 2018-04-17 Jan Hubicka <jh@suse.cz>
755 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
757 2018-04-17 Martin Liska <mliska@suse.cz>
760 * multiple_target.c (create_dispatcher_calls): Set apostrophes
761 for target_clone error message. Make default implementation
762 clone to be a local declaration.
763 (separate_attrs): Add new argument and check for an empty
765 (expand_target_clones): Handle it.
766 (ipa_target_clone): Make redirection just for target_clones
769 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
770 Tom de Vries <tom@codesourcery.com>
773 * omp-expand.c (expand_oacc_for): Add dummy false branch for
774 tiled basic blocks without omp continue statements.
776 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
779 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
780 vec_extract expression as having side effects to make sure it gets
783 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
786 * config/i386/i386.c (get_builtin_code_for_version): Check
789 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
792 * gcc/config.gcc: Support "skylake".
793 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
795 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
796 (processor_target_table): Add "skylake".
797 (ix86_option_override_internal): Add "skylake".
798 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
799 PROCESSOR_CANNONLAKE.
800 (get_builtin_code_for_version): Fix priority for
801 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
802 PROCESSOR_SKYLAKE-AVX512.
803 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
804 (processor_type): Add PROCESSOR_SKYLAKE.
806 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
807 Jason Merrill <jason@redhat.com>
810 * convert.c (convert_to_integer_1): Use direct recursion for
811 enumeral types and types with a precision less than the number
812 of bits in their mode.
814 2018-04-16 Julia Koval <julia.koval@intel.com>
817 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
818 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
820 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
823 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
824 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
825 and -mno-direct-move.
827 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
830 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
831 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
832 Ensure negative shifts result in {0}.
834 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
836 PR rtl-optimization/79916
837 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
838 regs (if any) to define how to gnerate SD moves when LRA is in
841 2018-04-13 Jakub Jelinek <jakub@redhat.com>
843 PR rtl-optimization/85393
844 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
845 * except.c (expand_dw2_landing_pad_for_region): Make static.
846 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
847 a label and unconditional jump to old_bb, rather than
848 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
851 PR rtl-optimization/85376
852 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
853 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
854 instead of a specific value.
856 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
857 Bin Cheng <bin.cheng@arm.com>
859 PR tree-optimization/82965
860 PR tree-optimization/83991
861 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
862 by_profile_only parameter.
863 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
864 information if the loop was predicted to iterate too many times.
865 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
867 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
870 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
873 2018-04-13 Martin Liska <mliska@suse.cz>
874 Jakub Jelinek <jakub@redhat.com>
877 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
878 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
879 * builtins.c (expand_builtin_memory_copy_args): Use
880 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
881 handle dest_addr == pc_rtx.
883 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
886 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
887 asked to not generate direct moves.
888 (fix_trunc<mode>si2_stfiwx): Similar.
889 (fix_trunc<mode>si2_internal): Similar.
891 2018-04-12 Jakub Jelinek <jakub@redhat.com>
894 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
895 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
896 lookup if dest in some wider mode is known to be const0_rtx and
897 if so, record permanent equivalence for it to be ZERO_EXTEND of
898 the narrower mode destination.
900 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
902 * lto-streamer-out.c (output_function): Revert 259346.
903 * omp-expand.c (expand_oacc_for): Likewise.
905 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
907 PR rtl-optimization/85354
908 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
909 * sel-sched.c (sel_global_init): ... here.
911 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
914 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
915 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
916 mode for PE-COFF targets.
917 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
918 (i386_pe_asm_lto_end): Likewise.
919 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
920 (TARGET_ASM_LTO_END): Likewise.
921 * config/i386/winnt.c (saved_debug_info_level): New static variable.
922 (i386_pe_asm_lto_start): New function.
923 (i386_pe_asm_lto_end): Likewise.
925 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
926 Richard Biener <rguenther@suse.de>
929 * lto-streamer-out.c (output_function): Fix CFG loop state before
931 * omp-expand.c (expand_oacc_for): Handle calls to internal
932 functions like regular functions.
934 2018-04-12 Richard Biener <rguenther@suse.de>
937 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
938 for the early LTO debug to properly generate references to it
939 during DIE emission. Do not re-use that for the skeleton for
941 (dwarf2out_early_finish): Likewise.
943 2018-04-12 Jakub Jelinek <jakub@redhat.com>
947 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
948 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
949 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
950 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
951 and output is a reg, avoid creating invalid lowpart subreg, but
952 instead split into a 512-bit move. Don't split if not AVX512VL,
953 input is xmm16+ reg and output is a mem.
954 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
955 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
956 xmm16+ reg and output is a mem.
958 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
960 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
961 also for flag_dwarf2_cfi_asm.
963 2018-04-12 Jakub Jelinek <jakub@redhat.com>
965 PR rtl-optimization/85342
966 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
967 a bool scalar var inside of the loop instead. Don't try to update
968 recog_data.operand after failed apply_change_group.
970 2018-04-12 Tom de Vries <tom@codesourcery.com>
973 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
974 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
975 array with flexible array member as array without given dimension.
976 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
977 argument for undefined param to true.
979 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
982 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
983 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
984 from PowerPC section.
985 * config/rs6000/sysv4.opt (mcall-): Improve help text.
986 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
987 help text that is too long.
988 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
989 help text that is too long.
990 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
991 help text that is too long.
993 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
995 * config/alpha/alpha.md (stack_probe_internal): Rename
996 from "probe_stack". Update all callers.
998 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1000 PR rtl-optimization/84566
1001 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
1002 sched_macro_fuse_insns.
1004 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1007 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
1008 (compute_block_dependences): ... from here.
1010 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1012 PR tree-optimization/85331
1013 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
1014 from int to HOST_WIDE_INT.
1016 2018-04-11 Martin Jambor <mjambor@suse.cz>
1019 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
1020 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
1021 not the same as the source val.
1022 (cgraph_edge_brings_value_p): New parameter.
1023 (gather_edges_for_value): Pass destination value to
1024 cgraph_edge_brings_value_p.
1025 (perhaps_add_new_callers): Likewise.
1026 (get_info_about_necessary_edges): Likewise and exclude values brought
1027 only by self-recursive edges.
1028 (create_specialized_node): Redirect only clones of self-calling edges.
1029 (+self_recursive_pass_through_p): New function.
1030 (find_more_scalar_values_for_callers_subset): Use it.
1031 (find_aggregate_values_for_callers_subset): Likewise.
1032 (known_aggs_to_agg_replacement_list): Removed.
1033 (decide_whether_version_node): Re-calculate known constants for all
1034 remaining context clones.
1036 2018-04-11 Richard Biener <rguenther@suse.de>
1039 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
1040 from early DWARF output.
1041 (dwarf2out_early_finish): Output line info unconditionally into
1042 early DWARF and add reference to it.
1044 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1047 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
1048 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
1049 other than V2DFmode using iptr mode attribute.
1050 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
1052 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1054 PR rtl-optimization/84659
1055 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
1057 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1060 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
1062 (output_loc_list): Pass address of a dummy size variable even in the
1063 locview handling loop.
1064 (index_location_lists): Add comment on why skip_loc_list_entry can't
1067 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
1070 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
1073 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
1076 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1077 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
1078 and -mstring-compare-inline-limit.
1080 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1083 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
1084 for stack clash protection in a register whenever we need it to be in
1087 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1089 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
1090 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
1092 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1095 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
1097 (mlong-double-): Ditto.
1098 * config/rs6000/sysv4.opt (msdata=): Ditto.
1099 (mtls-size=): Ditto.
1101 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
1103 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1104 erroneous entries for
1105 "vector int vec_ldl (int, long int *)", and
1106 "vector unsigned int vec_ldl (int, unsigned long int *)".
1107 Add comments and entries for
1108 "vector bool char vec_ldl (int, bool char *)",
1109 "vector bool short vec_ldl (int, bool short *)",
1110 "vector bool int vec_ldl (int, bool int *)",
1111 "vector bool long long vec_ldl (int, bool long long *)",
1112 "vector pixel vec_ldl (int, pixel *)",
1113 "vector long long vec_ldl (int, long long *)",
1114 "vector unsigned long long vec_ldl (int, unsigned long long *)".
1115 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
1116 type tree bool_long_long_type_node and correct definition of
1117 bool_V2DI_type_node to make reference to this new type tree.
1118 (rs6000_mangle_type): Replace erroneous reference to
1119 bool_long_type_node with bool_long_long_type_node.
1120 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
1121 comments to emphasize sign distinctions for char and int types and
1122 replace RS6000_BTI_bool_long constant with
1123 RS6000_BTI_bool_long_long constant. Also add comment to restrict
1124 use of RS6000_BTI_pixel.
1125 (bool_long_type_node): Remove this macro definition.
1126 (bool_long_long_type_node): New macro definition
1128 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1130 PR rtl-optimization/85300
1131 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
1132 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
1133 simplify_unary_operation fails.
1135 2018-04-10 Martin Liska <mliska@suse.cz>
1137 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
1138 cgraph_edge and ipa_ref.
1140 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1144 * config/i386/sse.md
1145 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
1146 computation of the VEC_MERGE selector from mask.
1147 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
1148 Fix decoding of the VEC_MERGE selector into mask.
1150 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1152 PR tree-optimization/85286
1153 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
1155 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1157 * final.c (final_1): Set insn_last_address as well as
1158 insn_current_address.
1160 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1163 * explow.c (emit_stack_probe): Call validize_mem on memory location
1164 before passing it to gen_probe_stack. Create address operand and
1165 legitimize it for the probe_stack_address case.
1167 2018-04-09 Jan Hubicka <jh@suse.cz>
1170 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
1171 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
1172 * tree.c (free_lang_data_in_type): Fix handling of binfos;
1174 (free_lang_data): Rebuild type inheritance graph.
1176 2018-04-09 Martin Sebor <msebor@redhat.com>
1178 * invoke.texi (-finline-small-functions): Mention other optimization
1180 (-findirect-inlining, -fpartial-inlining): Same.
1181 (-finline-functions-called-once): Same.
1182 (-freorder-blocks-and-partition): Same.
1184 2018-04-09 Jan Hubicka <jh@suse.cz>
1187 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
1188 jumps; choose last target that matches the criteria (i.e.
1189 no partition changes for non-crossing jumps).
1190 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
1191 support for redirecting crossing jumps to non-crossing.
1193 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
1195 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
1196 also for naked functions.
1198 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
1200 * config/arc/arc.md (add_shift): New pattern.
1201 (add_shift2): Likewise.
1202 (sub_shift): Likewise.
1203 (sub_shift_cmp0_noout): Likewise.
1204 (compare_si_ashiftsi): Likewise.
1205 (xbfu_cmp0_noout): New combine pattern.
1206 (xbfu_cmp0"): Likewise.
1207 (movsi_set_cc_insn): Place the predicable variant first.
1208 (commutative_binary_cmp0_noout): Remove clobber.
1209 (commutative_binary_cmp0): New pattern.
1210 (noncommutative_binary_cmp0): Likewise.
1211 (noncommutative_binary_cmp0_noout): Likewise.
1212 (noncommutative_binary_comparison_result_used): Removed.
1213 (rsub_cmp0): New pattern.
1214 (rsub_cmp0_noout): Likewise.
1215 (extzvsi): Changed, keep only meaningful variants.
1216 (SQH, SEZ): New iterators.
1217 (SQH_postfix): New mode attribute.
1218 (SEZ_prefix): New code attribute.
1219 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
1220 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
1221 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
1223 (noncommutative_operator): Check the availability of barrel
1226 2018-04-09 Richard Biener <rguenther@suse.de>
1228 PR tree-optimization/85284
1229 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
1230 Only use the niter constraining form of simple_iv when the exit
1233 2018-04-09 Tom de Vries <tom@codesourcery.com>
1236 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1237 (define_expand "*memory_barrier"): New define_expand.
1238 (define_insn "memory_barrier"): New insn.
1240 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1242 PR rtl-optimization/80463
1243 PR rtl-optimization/83972
1244 PR rtl-optimization/83480
1246 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
1247 correct producer for the insn.
1248 (tidy_control_flow): Fixup seqnos in case of debug insns.
1250 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1252 PR rtl-optimization/83913
1254 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
1255 different sched-times when merging exprs.
1257 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1259 PR rtl-optimization/83962
1261 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
1262 tidy_fallthru_edge and tidy_control_flow.
1264 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1266 PR rtl-optimization/83530
1268 * sel-sched.c (force_next_insn): New global variable.
1269 (remove_insn_for_debug): When force_next_insn is true, also leave only
1270 next insn in the ready list.
1271 (sel_sched_region): When the region wasn't scheduled, make another pass
1272 over it with force_next_insn set to 1.
1274 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1276 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
1278 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1279 for interrupt control.
1280 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1281 functions for interrupt control.
1282 * config/nds32/nds32-intrinsic.md: Likewise.
1283 * config/nds32/nds32_intrinsic.h: Likewise.
1284 * config/nds32/nds32.h (nds32_builtins): Likewise.
1286 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1288 * config/nds32/nds32.c (nds32_init_machine_status,
1289 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
1290 strict_aligned_p field.
1291 (nds32_expand_to_rtl_hook): New function.
1292 (TARGET_EXPAND_TO_RTL_HOOK): Define.
1293 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
1295 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1296 Chung-Ju Wu <jasonwucj@gmail.com>
1298 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
1299 * config/nds32/nds32-n7.md: New file.
1300 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
1301 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
1303 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
1304 * config/nds32/nds32.md (pipeline_model): Add n7.
1305 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
1306 * config/nds32/pipelines.md: Include n7 settings.
1308 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1309 Chung-Ju Wu <jasonwucj@gmail.com>
1311 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
1312 * config/nds32/nds32-e8.md: New file.
1313 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
1314 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
1316 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
1317 * config/nds32/nds32.md (pipeline_model): Add e8.
1318 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
1319 * config/nds32/pipelines.md: Include e8 settings.
1321 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1322 Chung-Ju Wu <jasonwucj@gmail.com>
1324 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
1325 * config/nds32/nds32-n8.md: New file.
1326 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
1327 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
1329 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
1330 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
1331 * config/nds32/nds32.md (pipeline_model): Add n8.
1332 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
1333 * config/nds32/pipelines.md: Include n8 settings.
1335 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1336 Chung-Ju Wu <jasonwucj@gmail.com>
1338 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
1339 * config/nds32/nds32-n9-2r1w.md: New file.
1340 * config/nds32/nds32-n9-3r2w.md: New file.
1341 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
1342 nds32_register_ports): New or modify for cpu n9.
1343 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
1345 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
1346 * config/nds32/nds32-utils.c: New file.
1347 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
1348 TARGET_MUL_SLOW): Define.
1349 * config/nds32/nds32.md (pipeline_model): New attribute.
1350 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
1351 New options that support cpu n9.
1352 * config/nds32/pipelines.md: Include n9 settings.
1353 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
1355 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1357 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
1358 information if necessary.
1359 (output_cond_branch_compare_zero): Likewise.
1360 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
1361 (nds32_target_alignment): Refine for alignment.
1362 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
1363 (FUNCTION_BOUNDARY): Modify.
1364 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
1366 * config/nds32/nds32.opt (malways-align, malign-functions): New.
1368 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1370 * config/nds32/constants.md (unspec_volatile_element): Add values for
1371 TLB operation and data prefetch.
1372 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1373 functions for TLB operation and data prefetch.
1374 * config/nds32/nds32-intrinsic.md: Likewise.
1375 * config/nds32/nds32_intrinsic.h: Likewise.
1376 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
1377 (nds32_print_operand): Likewise.
1378 * config/nds32/nds32.h (nds32_builtins): Likewise.
1380 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
1381 Andrew Pinski <pinsika@gcc.gnu.org>
1384 * match.pd: Use constant_boolean_node of correct type instead of
1385 boolean_true_node or boolean_false_node for simplifying
1386 pointer comparisons to zero.
1388 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1390 PR tree-optimization/80021
1391 * tree.c (verify_type_variant): Make error call in verify_variant_match
1392 translatable and remove final full stop.
1394 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1396 * config/nds32/constants.md (unspec_volatile_element): Add
1397 UNSPEC_VOLATILE_EH_RETURN.
1398 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
1399 nds32_output_stack_pop): Support dwarf exception handling process.
1400 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
1401 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
1402 exception handling process.
1403 (nds32_compute_stack_frame): Likewise.
1404 (nds32_return_addr_rtx): Likewise.
1405 (nds32_initial_elimination_offset): Likewise.
1406 (nds32_expand_prologue): Likewise.
1407 (nds32_expand_epilogue): Likewise.
1408 (nds32_dynamic_chain_address): New function.
1409 * config/nds32/nds32.h (machine_function): Add fields for dwarf
1411 (DYNAMIC_CHAIN_ADDRESS): Define.
1412 (EH_RETURN_DATA_REGNO): Define.
1413 (EH_RETURN_STACKADJ_RTX): Define.
1414 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
1415 patterns for dwarf exception handling.
1417 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1419 * config/nds32/nds32.h: Clean up obsolete macros.
1421 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1423 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1424 Add enum values for particular instructions.
1425 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1426 particular intrinsic functions.
1427 * config/nds32/nds32-intrinsic.md: Likewise.
1428 * config/nds32/nds32_intrinsic.h: Likewise.
1429 * config/nds32/nds32.h (nds32_builtins): Likewise.
1430 * config/nds32/nds32.md (type): Add pbsad and pbsada.
1431 (btst, ave): New patterns for particular instructions.
1433 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1435 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1436 Add enum values for atomic load/store and memory sync.
1437 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
1439 * config/nds32/nds32-intrinsic.md: Likewise.
1440 * config/nds32/nds32_intrinsic.h: Likewise.
1441 * config/nds32/nds32.h (nds32_builtins): Likewise.
1443 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1445 PR tree-optimization/85257
1446 * fold-const.c (native_encode_vector): If not all elts could fit
1447 and off is -1, return 0 rather than offset.
1448 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
1449 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
1450 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
1451 adjust buffer in native_interpret_expr call.
1453 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1455 * config/nds32/constants.md (unspec_volatile_element): Add cache
1456 control enum values.
1457 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
1458 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
1459 * config/nds32/nds32.c (nds32_cctl_names): New.
1460 (nds32_print_operand): Handle cache control register names.
1461 * config/nds32/nds32.h (nds32_builtins): New enum values.
1462 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
1464 * config/nds32/nds32.md (type): Add mmu.
1465 * config/nds32/pipelines.md (simple_insn): Add mmu.
1467 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1469 * config/nds32/nds32.md (type): Remove call.
1470 * config/nds32/pipelines.md (simple_insn): Likewise.
1472 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1474 * config/nds32/constants.md (unspec_volatile_element): Add
1475 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
1476 UNSPEC_VOLATILE_FMFCFG.
1477 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
1478 description for fmfcfg and fmfcsr.
1479 (bdesc_1arg): Add fmtcsr.
1480 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
1481 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
1482 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
1483 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
1484 unspec_fmfcfg): New patterns.
1485 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
1486 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
1487 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
1488 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
1489 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
1490 __nds32__fmfcfg): Define.
1492 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1494 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
1495 intrinsic register names.
1496 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
1497 intrinsic register enum values and macros.
1499 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1501 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
1502 for load/store addressing form.
1503 (nds32_print_operand_address): Likewise.
1505 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
1508 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
1509 based on LABEL_REF. Remove useless assertion.
1510 (pic_address_needs_scratch): Fix formatting.
1511 (sparc_legitimize_pic_address): Minor tweaks.
1512 (sparc_delegitimize_address): Adjust assertion accordingly.
1513 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
1514 into symbolic_operand.
1515 (movsi_high_pic_label_ref): Likewise.
1516 (movsi_lo_sum_pic_label_ref): Likewise.
1517 (movdi_pic_label_ref): Likewise.
1518 (movdi_high_pic_label_ref): Likewise.
1519 (movdi_lo_sum_pic_label_ref): Likewise.
1521 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
1523 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
1524 custom LIB_SPEC setup.
1526 2018-04-06 Ruslan Bukin <br@bsdpad.com>
1527 Kito Cheng <kito.cheng@gmail.com>
1529 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
1530 * config/riscv/freebsd.h: New.
1532 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1534 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
1535 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
1538 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1539 Kito Cheng <kito.cheng@gmail.com>
1541 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
1542 nds32_output_call, nds32_symbol_binds_local_p): New functions.
1543 * config/nds32/nds32-protos.h (nds32_output_call,
1544 nds32_output_return): Declare.
1545 * config/nds32/nds32.md: Refine all the call and return patterns.
1547 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1550 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
1551 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
1553 PR rtl-optimization/84872
1554 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
1555 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
1558 2018-04-06 Tamar Christina <tamar.christina@arm.com>
1560 * expr.c (copy_blkmode_to_reg): Revert 254862.
1561 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
1563 2018-04-06 Richard Biener <rguenther@suse.de>
1566 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
1567 after seeing a component reference with an adjacent field. Treat
1568 refs to arrays at struct end of external decls similar to
1569 refs to unconstrained commons.
1571 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1574 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
1575 look through SAVE_EXPRs with non-side-effects argument. Adjust
1577 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
1580 2018-04-06 Richard Biener <rguenther@suse.de>
1583 * alias.c (find_base_term): New wrapper around find_base_term
1584 unwinding CSELIB_VAL_PTR changes.
1585 (find_base_term): Do not restore CSELIB_VAL_PTR during the
1588 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1590 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
1592 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
1593 constant definitions.
1594 ("nop"): lr 0,0 -> nopr r0
1595 ("nop_lr0", "nop_lr1"): New insn definitions.
1597 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1599 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
1600 NDS32_V3PUSH_AVAILABLE_P macro.
1602 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
1603 Chung-Ju Wu <jasonwucj@gmail.com>
1605 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
1606 (nds32*-*-*): Add float and fpu_config into supported_defaults.
1607 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
1608 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
1609 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
1610 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
1611 * config/nds32/constraints.md: New constraints and checking for hard
1612 float configuration.
1613 * config/nds32/iterators.md: New mode iterator and attribute for hard
1614 float configuration.
1615 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
1617 * config/nds32/nds32-fpu.md: New file.
1618 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
1619 deal with hard float code generation.
1620 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
1622 (abi_type, float_reg_number): New enum type.
1623 * config/nds32/nds32-predicates.c: New predicates for hard float.
1624 * config/nds32/nds32-protos.h: Declare functions for hard float.
1625 * config/nds32/nds32.c: Implementation for hard float configuration.
1626 * config/nds32/nds32.h: Definitions for hard float configuration.
1627 * config/nds32/nds32.md: Include hard float machine description and
1628 modify patterns for hard float configuration.
1629 * config/nds32/nds32.opt: New options for hard float configuration.
1630 * config/nds32/predicates.md: New predicates for hard float
1633 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1635 * common/config/nds32/nds32-common.c
1636 (nds32_option_optimization_table): Enable -mreleax-hint by default.
1638 2018-04-05 Jakub Jelinek <jakub@redhat.com>
1641 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
1642 CONSTRUCTOR_ELT (ctor, ...)->value.
1644 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
1647 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
1649 2018-04-05 Tom de Vries <tom@codesourcery.com>
1652 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
1655 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1656 Kito Cheng <kito.cheng@gmail.com>
1658 * config/nds32/constraints.md (U33): Fine-tune checking condition.
1659 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
1660 * config/nds32/nds32.h (nds32_16bit_address_type): Add
1661 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
1663 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1664 Kito Cheng <kito.cheng@gmail.com>
1666 * config/nds32/constraints.md (Ufe): New memory constraint.
1667 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
1668 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
1669 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
1671 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
1672 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
1674 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1676 * config/nds32/nds32.md: Use optimize_size in the condition for
1677 alu-shift instructions.
1679 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1681 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
1683 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1685 * config/nds32/nds32.md (negsi2): Refine pattern.
1687 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1688 Chung-Ju Wu <jasonwucj@gmail.com>
1690 * config/nds32/iterators.md (shift_rotate): New code iterator.
1691 (shift): New code attribute.
1692 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
1693 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
1694 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
1695 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
1696 bit-wise operations.
1697 (andsi3, *andsi3): Ditto.
1698 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
1699 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
1700 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
1701 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
1702 nds32_ior_operand, nds32_xor_operand): New predicates.
1704 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1706 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
1707 (addsi3, subsi3): ... this.
1709 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1711 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
1713 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1715 * config/nds32/nds32.md: Adjust indention.
1717 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1719 * config/nds32/nds32.md (feature): New attribute.
1721 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1723 * config/nds32/nds32.md (subtype): New attribute.
1725 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
1728 * config/arm/arm-builtins.c (arm_expand_builtin): Change
1729 expansion to perform a bitwise AND of the argument followed by a
1730 boolean negation of the result.
1732 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
1734 PR rtl-optimization/84878
1735 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
1736 the basic block. Assert the use reference is not artificial and that
1737 it has an associated insn.
1739 2018-04-04 Michael Matz <matz@suse.de>
1741 * builtins.c (compute_objsize): Pass correct operand
1742 to array_at_struct_end_p.
1744 2018-04-04 Richard Biener <rguenther@suse.de>
1747 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
1748 from contexts for DINFO_LEVEL_TERSE and below.
1750 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1752 * config/nds32/nds32-doubleword.md (move_<mode>): Require
1753 resiter_operand condition.
1754 * config/nds32/nds32.md (*move<mode>): Ditto.
1756 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1757 Monk Chiang <sh.chiang04@gmail.com>
1759 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
1761 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1763 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
1765 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1766 Kito Cheng <kito.cheng@gmail.com>
1768 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
1769 nds32_cond_code_str, output_cond_branch,
1770 output_cond_branch_compare_zero, nds32_expand_cbranch,
1771 nds32_expand_cstore, nds32_expand_movcc,
1772 nds32_output_cbranchsi4_equality_zero,
1773 nds32_output_cbranchsi4_equality_reg,
1774 nds32_output_cbranchsi4_equality_reg_or_const_int,
1775 nds32_output_cbranchsi4_greater_less_zero: New functions.
1776 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
1777 nds32_expand_cstore, nds32_expand_movcc,
1778 nds32_output_cbranchsi4_equality_zero,
1779 nds32_output_cbranchsi4_equality_reg,
1780 nds32_output_cbranchsi4_equality_reg_or_const_int,
1781 nds32_output_cbranchsi4_greater_less_zero): Declare.
1782 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
1783 nds32_rimm11s_operand): New predicates.
1784 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
1785 * config/nds32/nds32.md: Rewrite all the branch and conditional move
1788 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1790 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
1791 * config/nds32/nds32.md: Ditto.
1792 * config/nds32/pipelines.md: Ditto.
1794 2018-04-04 Richard Biener <rguenther@suse.de>
1796 PR tree-optimization/85168
1797 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
1798 propagating abnormals.
1800 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1802 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
1804 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1805 Kito Cheng <kito.cheng@gmail.com>
1807 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
1808 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
1809 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
1810 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
1811 * config/nds32/nds32.md (sibcall_internal): New.
1812 (sibcall_register): Remove.
1813 (sibcall_immediate): Remove.
1814 (sibcall_value_internal): New.
1815 (sibcall_value_register): Remove.
1816 (sibcall_value_immediate): Remove.
1817 * config/nds32/predicates.md (nds32_general_register_operand): New.
1818 (nds32_call_address_operand): New.
1820 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1822 PR rtl-optimization/85167
1823 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
1824 bb_defs if *split_p, instead preinitialize it to NULL.
1826 PR tree-optimization/85156
1827 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
1828 evaluating the argument multiple times.
1830 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
1832 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
1834 (_mm_cvtpd_ps): Likewise.
1835 (_mm_cvttpd_epi32): Likewise.
1836 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
1837 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
1838 vector, pixel, and bool following altivec.h include.
1840 2018-04-03 Martin Sebor <msebor@redhat.com>
1842 * doc/extend.texi (Common Function Attributes): Clarify.
1843 (const attribute): Likewise.
1844 (pure attribute): Likewise.
1846 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1849 * config/i386/i386.c (ix86_expand_vector_set): Use
1850 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
1852 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
1854 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
1855 instructions when changing rounding bits to preserve precision bits
1856 in the x87 control word.
1858 2018-04-03 Martin Liska <mliska@suse.cz>
1860 PR tree-optimization/82491
1861 * rtl.h (strip_offset_and_add): Replace += suboffset with
1862 poly_uint64 () + suboffset.
1864 2018-03-29 Martin Liska <mliska@suse.cz>
1865 Martin Jambor <mjambor@suse.cz>
1868 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
1869 param_type is not an integral or pointer type.
1871 2018-04-03 Richard Biener <rguenther@suse.de>
1873 * sese.h (recompute_all_dominators): Remove.
1875 2018-04-02 Martin Sebor <msebor@redhat.com>
1877 * doc/invoke.texi (-Wrestrict): Fix typos.
1879 2018-04-02 Jim Wilson <jimw@sifive.com>
1881 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
1882 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
1883 (<optab>di3, <optab>si3_extend): Likewise.
1884 (<optab>si3_mask, <optab>si3_mask_1): New.
1885 (<optab>di3_mask, <optab>di3_mask_1): New.
1886 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
1887 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
1888 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
1890 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
1892 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
1895 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
1897 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
1898 (nds32_canonicalize_comparison): New function.
1900 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1901 Kito Cheng <kito.cheng@gmail.com>
1902 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1904 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
1905 * config/nds32/constants.md (unspec_volatile_element): Add
1906 UNSPEC_VOLATILE_RELAX_GROUP.
1907 * config/nds32/nds32-relax-opt.c: New file.
1908 * config/nds32/nds32-predicates.c
1909 (nds32_symbol_load_store_p): New function.
1910 * config/nds32/nds32-protos.h
1911 (nds32_symbol_load_store_p): Declare function.
1912 (make_pass_nds32_relax_opt): Declare new rtl pass function.
1913 * config/nds32/nds32.c
1914 (nds32_register_pass): New function to register pass.
1915 (nds32_register_passes): New function to register passes.
1916 * config/nds32/nds32.md (relax_group): New pattern.
1917 * config/nds32/nds32.opt (mrelax-hint): New option.
1918 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
1920 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
1922 * config/nds32/t-nds32: Modify files dependency.
1924 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1926 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
1927 (PROFILE_HOOK): Define its implementation.
1929 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1931 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
1932 type and 32-bit size.
1934 2018-04-01 Jakub Jelinek <jakub@redhat.com>
1937 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
1938 (V_128_256): New mode iterator.
1939 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
1940 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
1941 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
1943 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
1944 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
1946 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
1949 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
1950 NaN inputs correctly.
1952 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
1955 * config/rs6000/vsx.md (??r): New mode attribute.
1956 (*vsx_mov<mode>_64bit): Use it.
1957 (*vsx_mov<mode>_32bit): Likewise.
1959 2018-03-30 Martin Sebor <msebor@redhat.com>
1961 PR tree-optimization/84818
1962 * builtins.c (check_access): Use warning_n.
1964 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1967 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
1969 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
1972 2018-03-30 Julia Koval <julia.koval@intel.com>
1975 * x86-tune.def (movx, partial_reg_dependency): Enable for
1978 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
1981 * lra-constraints.c (process_alt_operands): Move setting
1982 this_alternative_matches below.
1984 2018-03-29 Martin Liska <mliska@suse.cz>
1987 * doc/invoke.texi: Document how LTO works with debug info.
1988 Describe auto-load support of binutils. Mention 'x86-64'
1989 as valid option value of -march option.
1991 2018-03-29 Jakub Jelinek <jakub@redhat.com>
1993 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
1996 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
1997 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
1998 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
2001 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2004 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
2005 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
2006 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
2007 for RS6000_BTM_POWERPC64.
2008 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
2009 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
2010 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
2015 2018-03-28 Carl Love <cel@us.ibm.com>
2018 2017-09-27 Carl Love <cel@us.ibm.com>
2020 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
2021 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
2022 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
2025 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2027 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
2028 instead of __vector bool.
2029 (_mm_max_pu8): Likewise.
2030 (_mm_min_pi16): Likewise.
2032 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2035 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
2036 (DIVWEUO): Likewise.
2038 (DIVDEUO): Likewise.
2039 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
2040 DIVWEUO and DIVDEUO.
2041 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
2042 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
2043 (div_extend): Likewise.
2044 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
2046 (__builtin_divweuo): Likewise.
2047 (__builtin_divdeo): Likewise.
2048 (__builtin_divdeuo): Likewise.
2050 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2053 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
2054 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
2056 PR tree-optimization/82004
2057 * gimple-match-head.c (optimize_pow_to_exp): New function.
2058 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
2059 Don't fold to exp if optimize_pow_to_exp is false.
2061 2018-03-28 Martin Liska <mliska@suse.cz>
2064 * calls.c (initialize_argument_information): Fix trailing space.
2065 * common.opt: Fix typo and provide better explanation for
2066 -fsanitize-coverage option.
2067 * config/i386/i386.opt: Fix typo.
2069 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2070 Martin Liska <mliska@suse.cz>
2073 * gimplify.c (asan_poison_variable): Don't do the check for
2074 gimplify_omp_ctxp here.
2075 (gimplify_decl_expr): Do it here.
2076 (gimplify_target_expr): Likewise.
2078 2018-03-28 Martin Liska <mliska@suse.cz>
2081 * config/i386/i386.c (ix86_function_arg_advance): Do not call
2082 chkp_type_bounds_count if MPX is not enabled.
2084 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
2086 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
2088 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
2091 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
2092 function to create the function decl for complex long double
2093 multiply and divide for -mabi=ieeelongdouble.
2094 (init_float128_ieee): Call it.
2096 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
2099 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
2100 -fcf-protection=branch -mibt.
2101 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
2103 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2106 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
2108 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
2111 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
2112 extern array declarations.
2114 2018-03-27 Richard Biener <rguenther@suse.de>
2117 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
2118 explicit single_use checks.
2120 2018-03-27 Richard Biener <rguenther@suse.de>
2122 PR tree-optimization/85082
2123 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
2126 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2128 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
2129 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
2130 Turn on fasynchronous-unwind-tables and funwind-tables.
2132 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
2135 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
2136 (*bmi_blsr_<mode>_ccz): Ditto.
2138 2018-03-26 Tom de Vries <tom@codesourcery.com>
2140 PR tree-optimization/85063
2141 * omp-general.c (offloading_function_p): New function. Factor out
2143 * omp-offload.c (pass_omp_target_link::gate): ... here.
2144 * omp-general.h (offloading_function_p): Declare.
2145 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
2146 with attribute omp declare target for offloading functions.
2148 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
2150 PR tree-optimization/84005
2151 * tree-data-ref.h (get_base_for_alignment): Declare.
2152 * tree-data-ref.c (get_base_for_alignment_1): New function.
2153 (get_base_for_alignment): Likewise.
2154 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
2155 get_base_for_alignment to find a suitable base object, instead
2156 of always using drb->base_address.
2158 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2161 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
2162 known size by default.
2164 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
2167 * lra-constraints.c (process_alt_operands): Don't match BLKmode
2168 and non BLKmode operands.
2170 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2173 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
2174 Clean up attributes.
2176 2018-03-23 Richard Biener <rguenther@suse.de>
2179 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
2180 we are going to emit early debug for LTO.
2182 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2185 * function.c (match_asm_constraints_1): Don't optimize if input
2186 doesn't satisfy general_operand predicate for output's mode.
2189 * alias.c (write_dependence_p): Don't require for x_canonicalized
2190 non-VOIDmode if x has VOIDmode.
2193 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
2194 just don't try to optimize it rather than assert it never happens.
2196 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2198 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
2199 macro expansions for definition of ST_INTERNAL_<mode> and
2200 LD_INTERNAL_<mode> builtins.
2201 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
2203 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
2205 (altivec_expand_st_builtin): Likewise.
2206 (altivec_expand_builtin): Remove calls to deleted functions.
2207 (rs6000_address_for_altivec): Delete this function.
2208 * config/rs6000/vector.md: Remove expands for
2209 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
2211 2018-03-22 Sudakshina Das <sudi.das@arm.com>
2214 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
2215 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
2216 re-computing once computed.
2217 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
2218 (arm_init_machine_status): Initialize
2219 machine->static_chain_stack_bytes.
2221 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2224 * doc/extend.texi: Add four new prototypes for vec_ld.
2225 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
2226 definitions for more logical presentation.
2227 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
2228 entries for V1TI variants of __builtin_altivec_ld builtin.
2229 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
2230 handling of V1TI variant of LVX icode pattern.
2231 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
2232 (rs6000_gimple_fold_builtin): Likewise.
2233 (altivec_init_builtins): Add code to define
2234 __builtin_altivec_lvx_v1ti function.
2236 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2239 * function.c (match_asm_constraints_1): Don't do the optimization
2240 if input isn't a REG, SUBREG, MEM or constant.
2242 2018-03-22 Tom de Vries <tom@codesourcery.com>
2244 PR tree-optimization/84956
2245 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
2246 bb_has_abnormal_pred.
2248 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2251 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
2252 DECL_INITIAL (decl) to decl at the end.
2253 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
2256 2018-03-21 Joseph Myers <joseph@codesourcery.com>
2258 * doc/extend.texi (__builtin_tgmath): Document when complex
2259 integer types are treated as _Complex _Float64.
2261 2018-03-21 Tom de Vries <tom@codesourcery.com>
2263 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
2265 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2267 PR tree-optimization/84960
2268 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
2269 if it is ENTRY block, move them into single succ of ENTRY in that case.
2271 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
2273 PR tree-optimization/84811
2274 * poly-int.h (poly_span_traits): Remove the T3 parameter and
2275 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
2276 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
2277 (known_subrange_p): Update accordingly. Cast each value involved
2278 in the size comparison, rather than casting the result of the
2281 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2283 PR tree-optimization/84982
2284 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
2285 by flipping the least significant bit rather than all bits from
2286 bitpos to bitpos + bitsize - 1.
2288 2018-03-21 Nathan Sidwell <nathan@acm.org>
2290 * doc/extend.texi (Deprecated Features): Remove mention of
2291 long-deleted deprecations.
2293 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2296 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
2297 * configure: Regenerate.
2299 2018-03-21 Tom de Vries <tom@codesourcery.com>
2301 PR tree-optimization/83126
2302 * tree-parloops.c (num_phis): New function.
2303 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
2305 2018-03-21 Nathan Sidwell <nathan@acm.org>
2307 * doc/extend.texi (Deprecated Features): Update deprecated flags,
2308 mention anon-struct/union members and trailing attributes.
2310 2018-03-21 Bin Cheng <bin.cheng@arm.com>
2312 PR tree-optimization/84969
2313 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
2314 builtin memset partitions if they set different rhs values.
2316 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2318 PR rtl-optimization/84989
2319 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
2320 VEC_DUPLICATE with scalar result mode.
2322 2018-03-21 Martin Liska <mliska@suse.cz>
2325 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
2326 not intended return statement.
2328 2018-03-21 Martin Liska <mliska@suse.cz>
2331 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
2332 (chkp_find_bound_slots_1): Limit number of iterations.
2334 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
2337 * Minor grammar fixes for x86 options.
2339 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2342 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
2343 holding REG_CFA_RESTORE notes, instead turn them into a USE.
2345 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
2348 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
2349 (altivec_lvx_<mode>_1op): Likewise.
2350 (altivec_stvx_<mode>_2op): Likewise.
2351 (altivec_stvx_<mode>_1op): Likewise.
2352 (altivec_lvx_<VM2:mode>): New define_expand.
2353 (altivec_stvx_<VM2:mode>): Likewise.
2354 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
2355 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2356 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
2357 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2358 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
2359 (rs6000_gen_lvx): Likewise.
2360 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
2361 (altivec_expand_stv_builtin): Likewise.
2362 (altivec_expand_builtin): Likewise.
2363 * config/rs6000/vector.md: Likewise.
2365 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2368 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
2371 2018-03-20 Richard Biener <rguenther@suse.de>
2374 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
2375 sign-conversions as zero, fall back to standard scalar_stmt
2378 2018-03-20 Martin Liska <mliska@suse.cz>
2381 * predict.c (rebuild_frequencies): Handle case when we have
2382 PROFILE_ABSENT, but flag_guess_branch_prob is false.
2384 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2387 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
2388 flag_section_anchors.
2389 * varasm.c (use_blocks_for_decl_p): Remove hack for
2390 dw2_force_const_mem.
2393 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
2395 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
2396 be created, use lowpart_subreg of operands[0] rather than operands[0]
2398 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
2399 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
2400 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
2401 and n constraint instead of aarch64_shift_imm_di and Usd.
2402 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
2403 (*aarch64_<optab>_reg_minus<mode>3): ... this.
2405 2018-03-20 Sudakshina Das <sudi.das@arm.com>
2408 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
2409 to favor GPR over NEON registers.
2410 (<shift>di3_neon): Likewise.
2412 2018-03-20 Tom de Vries <tom@codesourcery.com>
2415 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
2416 (nvptx_process_pars): Emit bar.sync asap and alap.
2418 2018-03-20 Tom de Vries <tom@codesourcery.com>
2421 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
2422 seen_label if seen_label is already set.
2424 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2427 * config/i386/i386.c (fold_builtin_cpu): For features above 31
2428 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
2429 Use 1U instead of 1. Formatting fixes.
2432 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
2433 instead of TREE_TYPE (s1) for the return value.
2435 2018-03-19 Jakub Jelinek <jakub@redhat.com>
2437 PR tree-optimization/84946
2438 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
2439 bitsize + bitsize in poly_uint64 rather than poly_int64.
2442 * dwarf2asm.c: Include fold-const.c.
2443 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
2444 of decl rather than decl itself.
2446 PR rtl-optimization/84643
2447 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
2449 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
2452 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
2453 calling assemble_variable.
2455 2018-03-19 Sudakshina Das <sudi.das@arm.com>
2458 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
2459 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
2461 2018-03-19 Jim Wilson <jimw@sifive.com>
2464 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
2465 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
2466 (riscv_first_stack_step): Likewise.
2467 (riscv_option_override): Use STACK_BOUNDARY instead of
2469 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
2471 (BIGGEST_ALIGNMENT): Set to 128.
2472 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
2473 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
2476 2018-03-19 Richard Biener <rguenther@suse.de>
2478 PR tree-optimization/84933
2479 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
2480 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
2482 2018-03-19 Richard Biener <rguenther@suse.de>
2484 PR tree-optimization/84859
2485 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
2486 (cond_if_else_store_replacement): Perform sinking operation on
2487 single-store BBs regardless of MAX_STORES_TO_SINK setting.
2488 Generalize what a BB with a single eligible store is.
2490 2018-03-19 Richard Biener <rguenther@suse.de>
2492 PR tree-optimization/84929
2493 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
2494 chrec_is_positive against non-chrec arg.
2496 2018-03-19 Tamar Christina <tamar.christina@arm.com>
2499 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
2501 2018-03-18 Martin Liska <mliska@suse.cz>
2503 PR rtl-optimization/84635
2504 * regrename.c (build_def_use): Use matches_mode only when
2507 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
2509 PR tree-optimization/84913
2510 * tree-vect-loop.c (vectorizable_reduction): Don't try to
2511 vectorize chains of COND_EXPRs.
2513 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2515 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
2517 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2519 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
2521 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2523 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
2525 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2526 Kito Cheng <kito.cheng@gmail.com>
2528 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
2529 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
2530 (nds32_adjust_reg_alloc_order): New function.
2531 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
2533 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
2535 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
2536 nds32_print_operand, nds32_print_operand_address): Use
2537 HOST_WIDE_INT_PRINT_DEC instead.
2539 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2541 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
2543 2018-03-17 Jakub Jelinek <jakub@redhat.com>
2546 * config/i386/i386.c (initial_ix86_tune_features,
2547 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
2549 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
2550 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
2551 rather than 1u << ix86_tune. Formatting fix.
2552 (ix86_option_override_internal): Change ix86_arch_mask from
2553 unsigned int to unsigned HOST_WIDE_INT, initialize to
2554 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
2555 (ix86_function_specific_restore): Likewise.
2557 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2560 * postreload.c (reload_combine_recognize_pattern): Perform
2561 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
2562 truncate_int_for_mode the result for the destination's mode.
2565 * hsa-gen.c (mem_type_for_type): Fix comment typo.
2566 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
2568 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2571 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
2574 * lra-assigns.c (lra_split_hard_reg_for): Don't use
2575 regno_allocno_class_array and sorted_pseudos.
2576 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
2577 insns where regno is used.
2579 2018-03-16 Martin Liska <mliska@suse.cz>
2582 * multiple_target.c (create_dispatcher_calls): Redirect
2583 reference in the symbol table.
2585 2018-03-16 Martin Liska <mliska@suse.cz>
2588 * multiple_target.c (create_dispatcher_calls): Redirect also
2591 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2595 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
2596 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
2597 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
2599 2018-03-16 Julia Koval <julia.koval@intel.com>
2601 * doc/invoke.texi (Skylake Server): Add CLWB.
2602 Cannonlake): Remove CLWB.
2604 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2606 PR tree-optimization/84841
2607 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
2609 (FLOAT_ONE_CONST_TYPE): Define.
2610 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
2611 (sort_by_operand_rank): Put entries with higher constant_type last
2612 rather than first to match comments.
2614 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
2616 * config/nios2/nios2.md (movsi_internal): Fix thinko in
2619 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2622 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
2623 check and use address of target always.
2625 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
2628 * config/i386/i386.c (indirect_thunk_needed): Update comments.
2629 (indirect_thunk_bnd_needed): Likewise.
2630 (indirect_thunks_used): Likewise.
2631 (indirect_thunks_bnd_used): Likewise.
2632 (indirect_return_needed): New.
2633 (indirect_return_bnd_needed): Likewise.
2634 (output_indirect_thunk_function): Add a bool argument for
2636 (output_indirect_thunk_function): Don't generate alias for
2637 function return thunk.
2638 (ix86_code_end): Call output_indirect_thunk_function to generate
2639 function return thunks.
2640 (ix86_output_function_return): Set indirect_return_bnd_needed
2641 and indirect_return_needed instead of indirect_thunk_bnd_needed
2642 and indirect_thunk_needed.
2644 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
2646 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
2647 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
2648 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
2650 2018-03-15 David Malcolm <dmalcolm@redhat.com>
2651 Paul Hua <paul.hua.gm@gmail.com>
2654 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
2656 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
2658 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
2659 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
2662 2018-03-15 Tamar Christina <tamar.christina@arm.com>
2665 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
2666 instead of GET_MODE_SIZE when comparing Units.
2668 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2671 * varasm.c (hash_section): Return an unchangeble hash value
2672 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
2673 Return !aarch64_can_use_per_function_literal_pools_p ().
2675 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2678 * optabs.c (emit_conditional_move): Pass address of cmode's copy
2679 rather than address of cmode as last argument to prepare_cmp_insn.
2681 2018-03-15 Julia Koval <julia.koval@intel.com>
2683 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
2684 F_AVX512VNNI, F_AVX512BITALG): New.
2686 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
2689 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
2690 insn for floating-point loads and stores.
2692 2018-03-14 Carl Love <cel@us.ibm.com>
2694 * config/rs6000/rs6000-c.c: Add macro definitions for
2695 ALTIVEC_BUILTIN_VEC_PERMXOR.
2696 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
2697 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
2698 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
2699 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
2701 * config/doc/extend.texi: Add prototypes for vec_permxor.
2703 2018-03-14 David Malcolm <dmalcolm@redhat.com>
2706 * diagnostic-show-locus.c (class layout_point): Convert m_line
2707 from int to linenum_type.
2708 (line_span::comparator): Use linenum "compare" function when
2709 comparing line numbers.
2710 (test_line_span): New function.
2711 (layout_range::contains_point): Convert param "row" from int to
2713 (layout_range::intersects_line_p): Likewise.
2714 (layout::will_show_line_p): Likewise.
2715 (layout::print_source_line): Likewise.
2716 (layout::should_print_annotation_line_p): Likewise.
2717 (layout::print_annotation_line): Likewise.
2718 (layout::print_leading_fixits): Likewise.
2719 (layout::annotation_line_showed_range_p): Likewise.
2720 (struct line_corrections): Likewise for field m_row.
2721 (line_corrections::line_corrections): Likewise for param "row".
2722 (layout::print_trailing_fixits): Likewise.
2723 (layout::get_state_at_point): Likewise.
2724 (layout::get_x_bound_for_row): Likewise.
2725 (layout::print_line): Likewise.
2726 (diagnostic_show_locus): Likewise for locals "last_line" and
2728 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
2729 * input.c (selftest::test_linenum_comparisons): New function.
2730 (selftest::input_c_tests): Call it.
2731 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
2732 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
2733 * selftest.h (ASSERT_GT): New macro.
2734 (ASSERT_GT_AT): New macro.
2735 (ASSERT_LT): New macro.
2736 (ASSERT_LT_AT): New macro.
2738 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
2740 PR rtl-optimization/84780
2741 * combine.c (distribute_links): Don't make a link based on pc_rtx.
2743 2018-03-14 Martin Liska <mliska@suse.cz>
2745 * tree.c (record_node_allocation_statistics): Use
2746 get_stats_node_kind.
2747 (get_stats_node_kind): New function extracted from
2748 record_node_allocation_statistics.
2749 (free_node): Use get_stats_node_kind.
2751 2018-03-14 Richard Biener <rguenther@suse.de>
2753 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
2754 that the value-set of ANTIC_IN doesn't grow.
2757 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2759 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2760 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2762 2018-03-14 Julia Koval <julia.koval@intel.com>
2764 * config.gcc (icelake-client, icelake-server): New.
2766 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
2767 (initial_ix86_arch_features): Ditto.
2768 (PTA_SKYLAKE): Add SGX.
2769 (PTA_ICELAKE): Remove.
2770 (PTA_ICELAKE_CLIENT): New.
2771 (PTA_ICELAKE_SERVER): New.
2772 (ix86_option_override_internal): Split up icelake on icelake client and
2774 (get_builtin_code_for_version): Ditto.
2775 (fold_builtin_cpu): Ditto.
2776 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
2777 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
2778 * config/i386/i386.h (processor_type): Ditto.
2779 * doc/invoke.texi: Ditto.
2781 2018-03-14 Jakub Jelinek <jakub@redhat.com>
2784 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
2785 INTEGER_CST offset, add it together with bitpos / 8 and
2786 sign extend based on POINTER_SIZE.
2790 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
2793 * config/i386/constraints.md (Yc): New register constraint.
2794 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
2795 Use Yc constraint for alternative 2 of operand 0. Remove
2796 preferred_for_speed attribute.
2798 2018-03-14 Richard Biener <rguenther@suse.de>
2800 PR tree-optimization/84830
2801 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
2802 with the old one to avoid oscillations.
2804 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
2807 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
2809 (assign_by_spills): Return a flag of reload assignment failure.
2810 Do not process the reload assignment failures. Do not spill other
2811 reload pseudos if they has the same reg class. Update n if
2813 (lra_assign): Add a return arg. Set up from the result of
2814 assign_by_spills call.
2815 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2816 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2817 usage_insns if it is not NULL.
2818 (spill_hard_reg_in_range): New function.
2819 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2820 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2821 function prototypes.
2822 (lra_assign): Change prototype.
2823 * lra.c (lra): Add code to deal with fails by splitting hard reg
2826 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
2828 * config/riscv/riscv.opt (mrelax): New option.
2829 * config/riscv/riscv.c (riscv_file_start): Emit ".option
2830 "norelax" when riscv_mrelax is disabled.
2831 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
2833 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2836 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
2837 reassociation for int modes.
2839 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2841 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
2842 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
2844 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
2845 * config/aarch64/aarch64-sve.md
2846 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
2847 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
2848 (*extend<mode><Vwide>2): Rename to...
2849 (aarch64_sve_extend<mode><Vwide>2): ...this.
2850 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
2851 renaming the old pattern to...
2852 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
2854 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
2855 define_expand, renaming the old pattern to...
2856 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
2857 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
2858 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
2859 account when deciding which SVE instruction the optab should use.
2860 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
2862 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2864 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
2865 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
2866 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
2867 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
2868 (tlsdesc_small_<mode>): Turn a define_expand and use
2869 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
2870 (tlsdesc_small_advsimd_<mode>): ...this.
2871 (tlsdesc_small_sve_<mode>): New pattern.
2873 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2875 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
2876 (UNSPEC_UMUL_HIGHPART): New constants.
2877 (MUL_HIGHPART): New int iteraor.
2878 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
2879 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
2881 (*<su>mul<mode>3_highpart): New define_insn.
2883 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2886 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2889 2018-03-13 Martin Liska <mliska@suse.cz>
2892 * (sem_item_optimizer::sem_item_optimizer): Initialize new
2894 (sem_item_optimizer::~sem_item_optimizer): Release it.
2895 (sem_item_optimizer::merge_classes): Register variable aliases.
2896 (sem_item_optimizer::fixup_pt_set): New function.
2897 (sem_item_optimizer::fixup_points_to_sets): Likewise.
2898 * ipa-icf.h: Declare new variables and functions.
2900 2018-03-13 Jakub Jelinek <jakub@redhat.com>
2903 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
2904 integer_pow2p@2 and test integer_pow2p in condition.
2905 (A < 0 ? C : 0): Similarly for @1.
2908 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
2909 characters starting at p contain '\0' character, don't look beyond
2913 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
2914 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
2917 * reg-stack.c (change_stack): Change update_end var from int to
2918 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
2919 also call set_block_for_insn on the newly added insns and rescan.
2922 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
2923 on the last operand.
2926 * tree.c (stabilize_reference_1): Return save_expr (e) for
2927 STATEMENT_LIST even if it doesn't have side-effects.
2929 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
2931 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
2933 2018-03-12 Renlin Li <renlin.li@arm.com>
2935 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
2936 aarch64_output_scalar_simd_mov_immediate.
2938 2018-03-12 Martin Sebor <msebor@redhat.com>
2940 PR tree-optimization/83456
2941 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
2942 for perfectly overlapping calls to memcpy.
2943 (gimple_fold_builtin_memory_chk): Same.
2944 (gimple_fold_builtin_strcpy): Handle no-warning.
2945 (gimple_fold_builtin_stxcpy_chk): Same.
2946 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
2948 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2950 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
2951 parameter. Use it for SFmode.
2952 (rs6000_function_arg_advance_1): Adjust.
2953 (rs6000_function_arg): Adjust.
2954 (rs6000_gimplify_va_arg): Pass false for that new parameter.
2956 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2958 PR rtl-optimization/84169
2959 PR rtl-optimization/84780
2960 * combine.c (can_combine_p): Check for a 2-insn combination whether
2961 the destination register is used between the two insns, too.
2963 2018-03-12 Richard Biener <rguenther@suse.de>
2965 PR tree-optimization/84803
2966 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
2967 for refs DR analysis didn't process.
2969 2018-03-12 Richard Biener <rguenther@suse.de>
2971 PR tree-optimization/84777
2972 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
2973 force-vectorize loops ignore whether we are optimizing for size.
2975 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
2977 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
2978 (TARGET_MD_ASM_ADJUST): Define.
2980 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
2981 Kito Cheng <kito.cheng@gmail.com>
2982 Chung-Ju Wu <jasonwucj@gmail.com>
2984 * config/nds32/nds32.c (nds32_compute_stack_frame,
2985 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
2986 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
2987 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
2988 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
2989 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
2990 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
2991 * config/nds32/nds32.md (prologue, epilogue): Use macro
2992 NDS32_V3PUSH_AVAILABLE_P to do checking.
2994 2018-03-11 Jakub Jelinek <jakub@redhat.com>
2997 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
2998 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
2999 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
3000 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
3001 addition of most attributes on !orig_type_die or the attribute not
3002 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
3004 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3005 Chung-Ju Wu <jasonwucj@gmail.com>
3007 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
3009 * config/nds32/nds32.opt (mvh): New option.
3011 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3012 Chung-Ju Wu <jasonwucj@gmail.com>
3014 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
3016 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
3017 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
3020 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3021 Chung-Ju Wu <jasonwucj@gmail.com>
3023 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
3025 * config/nds32/nds32-multiple.md (strlensi): New pattern.
3026 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
3028 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
3029 Kito Cheng <kito.cheng@gmail.com>
3030 Chung-Ju Wu <jasonwucj@gmail.com>
3032 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
3033 UNSPEC_FFMISM and UNSPEC_FLMISM.
3034 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
3035 for ffb, ffmism and flmism.
3036 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
3037 (unspec_ffmism): Ditto.
3038 (unspec_flmism): Ditto.
3039 (nds32_expand_builtin_impl): Check if string extension is available.
3040 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
3041 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
3043 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
3046 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3049 * lra-assigns.c (assign_by_spills): Return a flag of reload
3050 assignment failure. Do not process the reload assignment
3051 failures. Do not spill other reload pseudos if they has the same
3053 (lra_assign): Add a return arg. Set up from the result of
3054 assign_by_spills call.
3055 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3056 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3057 usage_insns if it is not NULL.
3058 (spill_hard_reg_in_range): New function.
3059 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3060 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3061 function prototypes.
3062 (lra_assign): Change prototype.
3063 * lra.c (lra): Add code to deal with fails by splitting hard reg
3066 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
3069 * config/i386/i386.opt: Replace Enforcment with Enforcement.
3071 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
3074 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
3075 (dw_val_node): Add val_symbolic_view.
3076 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
3077 (symview_upper_bound): New.
3078 (new_line_info_table): Initialize symviews_since_reset.
3079 (dwarf2out_source_line): Count symviews_since_reset and set
3080 symview_upper_bound.
3081 (dw_val_equal_p): Handle symview.
3082 (add_AT_symview): New.
3083 (print_dw_val): Handle symview.
3084 (attr_checksum, attr_checksum_ordered): Likewise.
3085 (same_dw_val_p, size_of_die): Likewise.
3086 (value_format, output_die): Likewise.
3087 (add_high_low_attributes): Use add_AT_symview for entry_view.
3088 (dwarf2out_finish): Reset symview_upper_bound, clear
3091 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
3094 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
3095 Add strict argument and use it.
3096 (rs6000_split_multireg_move): Update for new strict argument.
3097 (mem_operand_gpr): Disallow all non-offsettable addresses.
3098 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
3100 2018-03-09 Jakub Jelinek <jakub@redhat.com>
3103 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
3104 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
3105 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
3108 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
3109 decl, use remap_type if we want to use the type.
3111 2018-03-09 Martin Sebor <msebor@redhat.com>
3113 PR tree-optimization/84526
3114 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3116 (builtin_access::generic_overlap): Be prepared to handle non-array
3119 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
3121 PR rtl-optimization/84682
3122 * lra-constraints.c (process_address_1): Check is_address flag
3123 for address constraints.
3124 (process_alt_operands): Likewise.
3125 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
3126 preprocess_constraints.
3127 * recog.h (preprocess_constraints): Add oploc parameter.
3129 * recog.c (preprocess_constraints): Test address_operand for
3130 CT_ADDRESS constraints.
3132 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3135 * lra-assigns.c (assign_by_spills): Return a flag of reload
3136 assignment failure. Do not process the reload assignment
3137 failures. Do not spill other reload pseudos if they has the same
3139 (lra_assign): Add a return arg. Set up from the result of
3140 assign_by_spills call.
3141 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3142 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3143 usage_insns if it is not NULL.
3144 (spill_hard_reg_in_range): New function.
3145 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3146 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3147 function prototypes.
3148 (lra_assign): Change prototype.
3149 * lra.c (lra): Add code to deal with fails by splitting hard reg
3152 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3155 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
3156 Accept complain bool parameter. Only emit errors if it is true.
3157 (arm_parse_cpu_option_name): Likewise.
3158 (arm_target_thumb_only): Adjust callers of the above.
3159 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
3160 prototype to take a default true bool parameter.
3161 (arm_parse_arch_option_name): Likewise.
3163 2018-03-09 David Malcolm <dmalcolm@redhat.com>
3164 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
3168 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
3169 * configure: Regenerate.
3170 * configure.ac ("linker --version-script option"): New.
3171 ("linker soname option"): New.
3173 2018-03-09 Richard Biener <rguenther@suse.de>
3175 PR tree-optimization/84775
3176 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
3177 immediate uses of predicate stmts and mark them modified.
3180 PR tree-optimization/84178
3181 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3183 (version_loop_for_if_conversion): Delay update_ssa call.
3184 (tree_if_conversion): Delay update_ssa until after predicate
3187 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
3190 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
3191 when the function accesses prior frames.
3193 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3196 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
3197 gen_llsym, otherwise call maybe_gen_llsym.
3200 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
3201 has ',' character inside of it.
3203 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3206 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
3207 as clobbering CC_REGNUM.
3209 2018-03-08 Richard Biener <rguenther@suse.de>
3212 * tree-scalar-evolution.c: Include tree-into-ssa.h.
3213 (follow_copies_to_constant): Do not follow SSA names registered
3216 2018-03-08 Richard Biener <rguenther@suse.de>
3218 PR tree-optimization/84178
3219 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3221 (version_loop_for_if_conversion): Delay update_ssa call.
3222 (tree_if_conversion): Delay update_ssa until after predicate
3225 2018-03-08 David Malcolm <dmalcolm@redhat.com>
3227 PR tree-optimization/84178
3228 * tree-if-conv.c (release_bb_predicate): Remove the
3229 the assertion that the stmts have NULL use_ops.
3230 Discard the statements, asserting that they haven't
3231 yet been added to a BB.
3233 2018-03-08 Richard Biener <rguenther@suse.de>
3235 PR tree-optimization/84746
3236 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
3237 (phi_translate): Pass in destination ANTIC_OUT set.
3238 (phi_translate_1): Likewise. For a simplified result lookup
3239 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
3240 (phi_translate_set): Adjust.
3241 (do_pre_regular_insertion): Likewise.
3242 (do_pre_partial_partial_insertion): Likewise.
3244 2018-03-08 Martin Liska <mliska@suse.cz>
3246 PR gcov-profile/84735
3247 * doc/gcov.texi: Document usage of profile files.
3248 * gcov-io.h: Document changes in the format.
3250 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
3254 * dwarf2out.c (struct dw_line_info_table): Update comments for
3256 (FORCE_RESET_NEXT_VIEW): New.
3257 (FORCE_RESETTING_VIEW_P): New.
3258 (RESETTING_VIEW_P): Check for -1 too.
3259 (ZERO_VIEW_P): Likewise.
3260 (new_line_info_table): Force-reset next view.
3261 (dwarf2out_begin_function): Likewise.
3262 (dwarf2out_source_line): Simplify zero_view_p initialization.
3263 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
3264 view directly. Omit view when omitting .loc at line 0.
3266 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3268 PR tree-optimization/84740
3269 * tree-switch-conversion.c (process_switch): Call build_constructors
3270 only if info.phi_count is non-zero.
3272 PR tree-optimization/84739
3273 * tree-tailcall.c (find_tail_calls): Check call arguments against
3274 DECL_ARGUMENTS (current_function_decl) rather than
3275 DECL_ARGUMENTS (func) when checking for tail recursion.
3277 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3279 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
3280 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
3281 Volker Reichelt's entry and add entries for people that perform
3282 GCC fuzzy testing and report numerous bugs.
3284 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
3287 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
3288 readonly data in sdata, if that is disabled.
3289 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
3290 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3291 -mreadonly-in-sdata option.
3293 2018-03-07 Martin Sebor <msebor@redhat.com>
3295 PR tree-optimization/84468
3296 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
3297 basic block when looking for nul assignment.
3299 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
3302 * except.h (output_function_exception_table): Adjust prototype.
3303 * except.c (output_function_exception_table): Remove FNNAME parameter
3304 and add SECTION parameter. Ouput one part of the table at a time.
3305 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
3306 the first part of the exception table and emit unwind directives.
3307 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
3308 (i386_pe_seh_cold_init): Likewise.
3309 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
3310 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
3311 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
3312 (ix86_output_call_insn): Emit a nop in one more case for SEH.
3313 * config/i386/winnt.c: Include except.h.
3314 (struct seh_frame_state): Add reg_offset, after_prologue and
3315 in_cold_section fields.
3316 (i386_pe_seh_end_prologue): Set seh->after_prologue.
3317 (i386_pe_seh_cold_init): New function.
3318 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
3319 to seh->in_cold_section.
3320 (seh_emit_push): Record the offset of the push.
3321 (seh_emit_save): Record the offet of the save.
3322 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
3323 Test seh->after_prologue to disregard the epilogue.
3324 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
3325 (i386_pe_end_cold_function): New function.
3327 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3330 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
3331 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
3334 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
3336 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
3337 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
3340 * multiple_target.c: Include tree-inline.h and intl.h.
3341 (expand_target_clones): Diagnose and fail if node->definition and
3342 !tree_versionable_function_p (node->decl).
3344 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
3346 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
3348 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
3349 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
3350 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
3352 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3355 * combine.c (try_combine): Use reg_or_subregno instead of handling
3356 just paradoxical SUBREGs and REGs.
3358 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
3360 * config/arc/arc.c (arc_finalize_pic): Remove function.
3361 (arc_must_save_register): We use single base PIC register, remove
3362 checks to save/restore the PIC register.
3363 (arc_expand_prologue): Likewise.
3364 * config/arc/arc-protos.h (arc_set_default_type_attributes):
3366 (arc_verify_short): Likewise.
3367 (arc_attr_type): Likewise.
3368 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
3369 (walk_stores): Likewise.
3370 (arc_address_cost): Make it static.
3371 (arc_verify_short): Likewise.
3372 (branch_dest): Likewise.
3373 (arc_attr_type): Likewise.
3374 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
3375 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
3376 (arc_final_prescan_insn): Remove inserting the nops due to
3377 hardware hazards. It is done in reorg step.
3378 (insn_length_variant_t): Remove.
3379 (insn_length_parameters_t): Likewise.
3380 (arc_insn_length_parameters): Likewise.
3381 (arc_get_insn_variants): Likewise.
3382 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
3384 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3387 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
3390 PR tree-optimization/84687
3391 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
3393 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
3395 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3397 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
3398 Rename to ppc_speculation_barrier.
3399 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
3400 __builtin_ppc_speculation_barrier.
3402 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3405 * combine.c (combine_simplify_rtx): Don't try to simplify if
3406 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
3409 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
3411 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
3412 to 32 bytes when compiling for POWER9.
3414 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3417 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
3418 regparm >= 3 with no arg reg available also for calls with
3419 flag_force_indirect_call. Pass decl to ix86_function_regparm.
3422 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
3424 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
3426 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
3429 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
3431 2018-03-05 Richard Biener <rguenther@suse.de>
3433 PR tree-optimization/84486
3434 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
3435 When inserting a __builtin_assume_aligned call set the LHS
3436 SSA name alignment info accordingly.
3438 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
3440 PR tree-optimization/84114
3441 * config/aarch64/aarch64.c (aarch64_reassociation_width)
3442 Avoid reassociation of FLOAT_MODE addition.
3444 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
3446 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
3447 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
3448 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
3449 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
3450 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
3451 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
3452 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
3454 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
3455 __builtin_ia32_wbinvd): New builtins.
3456 (SPECIAL_ARGS2): New.
3457 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
3458 (SPECIAL_ARGS2): New.
3459 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
3460 (ix86_valid_target_attribute_inner_p): Ditto.
3461 (ix86_init_mmx_sse_builtins): Add special_args2.
3462 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
3463 TARGET_WBNOINVD_P): New.
3464 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
3465 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
3466 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
3467 * config/i386/immintrin.h (_wbinvd): New intrinsic.
3468 * config/i386/pconfigintrin.h: New file.
3469 * config/i386/wbnoinvdintrin.h: Ditto.
3470 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
3471 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
3473 2018-03-05 Richard Biener <rguenther@suse.de>
3475 PR tree-optimization/84670
3476 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3478 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3479 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3480 (compute_antic_aux): Only assert the number of values in ANTIC_IN
3481 doesn't grow if all successors (recursively) were visited at least
3484 2018-03-05 Richard Biener <rguenther@suse.de>
3486 PR tree-optimization/84650
3487 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
3488 if executed in the loop pipeline.
3490 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
3492 * doc/configfiles.texi (Configuration Files): Move info about
3493 conditionalizing $target-protos.h to...
3494 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
3495 differs from $target-protos.h.
3497 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
3498 Chung-Ju Wu <jasonwucj@gmail.com>
3500 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
3501 * config/nds32/nds32-multiple.md (setmemsi): Define.
3502 * config/nds32/nds32-memory-manipulation.c
3503 (nds32_gen_dup_4_byte_to_word_value): New.
3504 (emit_setmem_word_loop): New.
3505 (emit_setmem_byte_loop): New.
3506 (nds32_expand_setmem_loop): New.
3507 (nds32_expand_setmem_loop_v3m): New.
3508 (nds32_expand_setmem_unroll): New.
3509 (nds32_expand_setmem): New.
3511 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3512 Chung-Ju Wu <jasonwucj@gmail.com>
3514 * config/nds32/nds32-memory-manipulation.c
3515 (nds32_emit_load_store): New.
3516 (nds32_emit_post_inc_load_store): New.
3517 (nds32_emit_mem_move): New.
3518 (nds32_emit_mem_move_block): New.
3519 (nds32_expand_movmemsi_loop_unknown_size): New.
3520 (nds32_expand_movmemsi_loop_known_size): New.
3521 (nds32_expand_movmemsi_loop): New.
3522 (nds32_expand_movmemsi_unroll): New.
3523 (nds32_expand_movmemqi): Rename ...
3524 (nds32_expand_movmemsi): ... to this.
3525 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
3526 (movmemsi): ... to this.
3527 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
3528 (nds32_expand_movmemsi): ... to this.
3530 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3531 Monk Chiang <sh.chiang04@gmail.com>
3532 Chung-Ju Wu <jasonwucj@gmail.com>
3534 * config/nds32/nds32-protos.h
3535 (nds32_expand_load_multiple): New arguments.
3536 (nds32_expand_store_multiple): Ditto.
3537 (nds32_valid_multiple_load_store): Rename ...
3538 (nds32_valid_multiple_load_store_p): ... to this.
3539 * config/nds32/nds32-memory-manipulation.c
3540 (nds32_expand_load_multiple): Refine implementation.
3541 (nds32_expand_store_multiple): Ditto.
3542 * config/nds32/nds32-multiple.md
3543 (load_multiple): Update nds32_expand_load_multiple interface.
3544 (store_multiple): Update nds32_expand_store_multiple interface.
3545 * config/nds32/nds32-predicates.c
3546 (nds32_valid_multiple_load_store): Rename ...
3547 (nds32_valid_multiple_load_store_p): ... to this and refine
3549 * config/nds32/predicates.md
3550 (nds32_load_multiple_and_update_address_operation): New predicate.
3551 (nds32_store_multiple_and_update_address_operation): New predicate.
3553 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3554 Chung-Ju Wu <jasonwucj@gmail.com>
3556 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
3557 (combo): New attribute.
3558 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
3560 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
3562 * config/nds32/nds32.opt: Change -mcmodel= default value.
3564 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
3565 Monk Chiang <sh.chiang04@gmail.com>
3566 Chung-Ju Wu <jasonwucj@gmail.com>
3568 * config/nds32/constants.md (unspec_element): New enum.
3569 * config/nds32/constraints.md (Umw): New constraint.
3570 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
3571 * config/nds32/nds32-intrinsic.md: Likewise.
3572 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
3573 (nds32_valid_smw_lwm_base_p): New.
3574 (nds32_output_smw_single_word): New.
3575 (nds32_output_lmw_single_word): New.
3576 (nds32_expand_unaligned_load): New.
3577 (nds32_expand_unaligned_store): New.
3578 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
3579 (nds32_output_smw_single_word): Declare.
3580 (nds32_output_lmw_single_word): Declare.
3581 (nds32_expand_unaligned_load): Declare.
3582 (nds32_expand_unaligned_store): Declare.
3583 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
3584 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
3585 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
3586 NDS32_BUILTIN_UASTORE_DW.
3587 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
3590 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
3591 Kito Cheng <kito.cheng@gmail.com>
3592 Chung-Ju Wu <jasonwucj@gmail.com>
3594 * config/nds32/nds32-intrinsic.c
3595 (nds32_expand_builtin_null_ftype_reg): Delete.
3596 (nds32_expand_builtin_reg_ftype_imm): Ditto.
3597 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
3598 (nds32_read_argument): New.
3599 (nds32_legitimize_target): Ditto.
3600 (nds32_legitimize_argument): Ditto.
3601 (nds32_check_constant_argument): Ditto.
3602 (nds32_expand_unop_builtin): Ditto.
3603 (nds32_expand_unopimm_builtin): Ditto.
3604 (nds32_expand_binop_builtin): Ditto.
3605 (nds32_builtin_decl_impl): Ditto.
3606 (builtin_description): Ditto.
3607 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
3608 (nds32_init_builtins_impl): Ditto.
3609 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
3610 (nds32_builtin_decl): New.
3611 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
3612 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
3614 2018-03-02 Jeff Law <law@redhat.com>
3616 * reorg.c (stop_search_p): Handle DEBUG_INSN.
3617 (redundant_insn, fill_simple_delay_slots): Likewise.
3618 (fill_slots_from_thread): Likewise.
3619 * resource.c (mark_referenced_resources): Likewise.
3620 (mark_set_resources, find_dead_or_set_registers): Likewise.
3622 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3624 * substring-locations.h (format_warning_va): Formatting fix for
3626 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
3628 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
3629 * substring-locations.c: Include intl.h.
3630 (format_warning_va): Turned into small wrapper around
3631 format_warning_n_va, renamed to ...
3632 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
3633 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
3635 (format_warning_at_substring_n): New function.
3636 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
3637 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
3638 format_warning_at_substring with just a shorter name instead of
3639 const function pointer.
3640 (fmtwarn_n): New function.
3641 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
3642 appropriate, get rid of all the fmtstr temporaries, move conditionals
3643 with G_() wrapped string literals directly into fmtwarn arguments,
3644 cast dir.len to (int), formatting fixes.
3646 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
3648 * doc/invoke.texi: Remove "Cilk Plus" references.
3650 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3651 Richard Biener <rguenther@suse.de>
3654 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
3655 for error or warning attributes if CALL_FROM_THUNK_P is set.
3658 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3661 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
3662 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
3665 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
3666 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
3668 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
3669 instead of -1U in last predictors element's probability member.
3671 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3674 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3675 arguments if they are comparable.
3677 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3679 PR tree-optimization/84634
3680 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
3681 masks and masked_loop_p with a single loop_masks, making sure it's
3682 null for bb vectorization.
3684 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3686 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3687 (vect_analyze_data_ref_access): Use loop->safe_len rather than
3688 loop->force_vectorize to check whether there is no alias.
3690 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3693 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
3695 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
3697 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
3698 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
3699 instead of a loop around prev_real_insn.
3700 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
3704 * config/i386/i386.c (ix86_print_operand): Use conditional
3705 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
3708 2018-03-02 Richard Biener <rguenther@suse.de>
3710 PR tree-optimization/84427
3711 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
3712 (bitmap_set_subtract_values): Rewrite to handle multiple
3715 (prune_clobbered_mems): Likewise.
3716 (phi_translate): Take edge instead of pred/phiblock.
3717 (phi_translate_1): Likewise.
3718 (phi_translate_set): Likewise. Insert all translated
3719 exprs for a value into the set, keeping possibly multiple
3720 expressions per value.
3721 (compute_antic_aux): Adjust for phi_translate changes.
3722 When intersecting union the expressions and prune those
3723 not in the final value set, keeping possibly multiple
3724 expressions per value. Do not use value-insertion
3725 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
3726 all expressions. Add verification that the value-sets
3727 only shrink during iteration.
3728 (compute_partial_antic_aux): Adjust for the phi_translate changes.
3729 (do_pre_regular_insertion): Likewise.
3730 (do_pre_partial_partial_insertion): Likewise.
3732 2018-03-02 Richard Biener <rguenther@suse.de>
3735 * config/darwin.c (saved_debug_info_level): New static global.
3736 (darwin_asm_lto_start): Disable debug info generation for LTO out.
3737 (darwin_asm_lto_end): Restore debug info generation settings.
3739 2018-03-01 Martin Liska <mliska@suse.cz>
3742 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
3745 2018-03-01 Richard Biener <rguenther@suse.de>
3748 * dwarf2out.c (gen_variable_die): Properly handle late VLA
3749 type annotation with LTO when debug was disabled at compile-time.
3751 2018-03-01 Matthew Fortune <mfortune@gmail.com>
3753 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
3755 (mips_final_postscan_insn): Likewise.
3757 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
3759 PR rtl-optimization/84528
3760 * alias.c (init_alias_target): Add commentary.
3761 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
3762 a unique base value if the frame pointer is not eliminated
3763 to the stack pointer.
3765 2018-03-01 Tom de Vries <tom@codesourcery.com>
3767 PR rtl-optimization/83327
3768 * lra-int.h (hard_regs_spilled_into): Declare.
3769 * lra.c (hard_regs_spilled_into): Define.
3770 (init_reg_info): Init hard_regs_spilled_into.
3771 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
3772 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
3773 (process_bb_lives): Handle hard_regs_spilled_into.
3774 (lra_create_live_ranges_1): Before doing liveness propagation, clear
3775 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
3777 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
3779 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
3780 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
3781 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
3782 * config/rs6000/aix72.h: New file.
3784 2018-02-28 Jakub Jelinek <jakub@redhat.com>
3786 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
3787 instead of warning_at with conditional singular and plural messages
3791 * stor-layout.c (update_alignment_for_field): For
3792 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
3793 && !DECL_PACKED (field), do the alignment update, just use
3794 only desired_align instead of MAX (type_align, desired_align)
3796 (place_field): Don't do known_align < desired_align handling
3797 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
3798 is non-NULL, instead do it after rli->prev_field handling and
3799 only if not within a bitfield word. For DECL_PACKED (field)
3800 use type_align of BITS_PER_UNIT.
3802 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
3804 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
3805 superfluous parentheses and trailing spaces.
3807 2018-02-28 Richard Biener <rguenther@suse.de>
3809 PR tree-optimization/84584
3810 * graphite-scop-detection.c (scop_detection::add_scop): Discard
3811 SCoPs with fake exit edge.
3813 2018-02-28 Martin Liska <mliska@suse.cz>
3816 * timevar.c (timer::print): Fix format to properly print 100%
3819 2018-02-28 Richard Biener <rguenther@suse.de>
3822 * genmatch.c (capture_info::walk_match): Do not mark
3823 captured expressions without operands as expr_p given
3824 they act more like predicates and should be subject to
3825 "lost tail" side-effect preserving.
3827 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
3829 PR rtl-optimization/81611
3830 * auto-inc-dec.c (attempt_change): Move dead note from
3831 mem_insn if it's the next use of regno
3832 (find_address): Take address use of reg holding
3833 non-incremented value. Add parm to limit search to the named
3835 (merge_in_block): Attempt to use a mem insn that is the next
3836 use of the original regno.
3838 2018-02-27 Martin Sebor <msebor@redhat.com>
3841 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
3842 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
3844 2018-02-27 Martin Sebor <msebor@redhat.com>
3846 PR translation/84207
3847 * diagnostic-core.h (warning_n, error_n, inform_n): Change
3848 n argument to unsigned HOST_WIDE_INT.
3849 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
3850 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
3851 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
3852 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
3854 2018-02-27 Richard Biener <rguenther@suse.de>
3856 PR tree-optimization/84512
3857 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
3858 Do not use the estimate returned from record_stmt_cost for
3859 the scalar iteration cost but sum properly using add_stmt_cost.
3861 2018-02-27 Richard Biener <rguenther@suse.de>
3863 PR tree-optimization/84466
3864 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3865 Adjust last change to less strictly validate use operands.
3867 2018-02-27 Martin Liska <mliska@suse.cz>
3869 PR gcov-profile/84548
3870 * gcov.c (process_file): Allow partial overlap and consider it
3871 also as group functions.
3872 (output_lines): Properly calculate range of lines for a group.
3874 2018-02-27 Martin Liska <mliska@suse.cz>
3876 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
3877 'ggc' suffixes. Change first column width.
3878 (timer::print): Fix formatting of the column.
3880 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
3882 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
3883 preserve inline entry blocks for the sake of debug inline
3884 entry point markers alone.
3885 (remove_unused_locals): Suggest in comments a better place to
3886 force the preservation of inline entry blocks that are
3887 otherwise unused, but do not preserve them.
3889 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3891 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
3893 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3896 * config/i386/constraints.md (Bs): Replace
3897 ix86_indirect_branch_register with
3898 TARGET_INDIRECT_BRANCH_REGISTER.
3900 * config/i386/i386.md (indirect_jump): Likewise.
3901 (tablejump): Likewise.
3902 (*sibcall_memory): Likewise.
3903 (*sibcall_value_memory): Likewise.
3904 Peepholes of indirect call and jump via memory: Likewise.
3905 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
3906 (*sibcall_value_GOT_32): Likewise.
3907 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3908 (GOT_memory_operand): Likewise.
3909 (call_insn_operand): Likewise.
3910 (sibcall_insn_operand): Likewise.
3911 (GOT32_symbol_operand): Likewise.
3912 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
3914 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3916 PR rtl-optimization/83496
3917 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
3918 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
3919 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
3920 redundant insn, if any.
3921 (relax_delay_slots): Likewise.
3922 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
3924 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
3926 PR tree-optimization/83965
3927 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
3928 that grouped statements are part of a reduction chain. Return
3929 true if the statement is not marked as a reduction itself but
3931 (vect_recog_dot_prod_pattern): Don't check whether the statement
3932 is part of a group here.
3933 (vect_recog_sad_pattern): Likewise.
3934 (vect_recog_widen_sum_pattern): Likewise.
3936 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3939 * final.c (rest_of_clean_state): Also look for calls inside sequences.
3941 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3944 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
3946 (ix86_output_indirect_function_return): New prototype.
3947 (ix86_split_simple_return_pop_internal): Likewise.
3948 * config/i386/i386.c (indirect_return_via_cx): New.
3949 (indirect_return_via_cx_bnd): Likewise.
3950 (indirect_thunk_name): Handle return va CX_REG.
3951 (output_indirect_thunk_function): Create alias for
3952 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
3953 (ix86_output_indirect_jmp): Remove the bool argument.
3954 (ix86_output_indirect_function_return): New function.
3955 (ix86_split_simple_return_pop_internal): Likewise.
3956 * config/i386/i386.md (*indirect_jump): Don't pass false
3957 to ix86_output_indirect_jmp.
3958 (*tablejump_1): Likewise.
3959 (simple_return_pop_internal): Change it to define_insn_and_split.
3960 Call ix86_split_simple_return_pop_internal to split it for
3962 (simple_return_indirect_internal): Call
3963 ix86_output_indirect_function_return instead of
3964 ix86_output_indirect_jmp.
3966 2018-02-26 Jakub Jelinek <jakub@redhat.com>
3969 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
3970 memset and value initialization afterwards.
3972 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
3974 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
3976 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3979 * common/config/aarch64/aarch64-common.c
3980 (aarch_option_optimization_table[]): Switch
3981 off fomit-frame-pointer
3983 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3984 Chung-Ju Wu <jasonwucj@gmail.com>
3986 * config/nds32/nds32-multiple.md (load_multiple): Disallow
3988 (store_multiple): Ditto.
3990 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3992 * config.gcc: Add --with-cpu support for nds32 target.
3993 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
3994 * config/nds32/nds32.opt: Add -mcpu= option.
3996 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
3998 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
3999 isel=yes): Warn for these deprecated options.
4001 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
4003 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
4004 ISA_2_5_MASKS_EMBEDDED.
4006 2018-02-23 Jakub Jelinek <jakub@redhat.com>
4008 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
4009 p->max as pointers rather than using iterative_hash_expr.
4011 2018-02-23 Carl Love <cel@us.ibm.com>
4013 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
4014 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
4016 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
4017 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
4018 P8V_BUILTIN_VEC_VUNSIGNED2.
4020 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
4023 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
4024 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
4026 * lra-constraints.c (curr_insn_transform): Set up
4027 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
4029 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
4030 * lra-eliminations.c (spill_pseudos): Ditto.
4031 (process_insn_for_elimination): Ditto.
4032 * lra-lives.c (reg_early_clobber_p): Use the new macros.
4033 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
4034 LRA_NON_CLOBBERED_ALT.
4036 2018-02-22 Martin Sebor <msebor@redhat.com>
4038 PR tree-optimization/84480
4039 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
4040 to maybe_diag_stxncpy_trunc. Call it.
4041 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
4042 from gimple_fold_builtin_strcpy. Print inlining stack.
4043 (handle_builtin_stxncpy): Print inlining stack.
4044 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
4046 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
4049 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
4050 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
4051 and -fcheck-pointer-bounds are used together.
4052 (indirect_thunk_prefix): New enum.
4053 (indirect_thunk_need_prefix): New function.
4054 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
4055 "_nt" instead of "_bnd" for NOTRACK prefix.
4056 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
4057 (output_indirect_thunk_function): Likewise.
4059 (ix86_code_end): Update output_indirect_thunk_function calls.
4060 (ix86_output_indirect_branch_via_reg): Replace
4061 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
4062 (ix86_output_indirect_branch_via_push): Likewise.
4063 (ix86_output_function_return): Likewise.
4064 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
4065 incompatible with -fcf-protection=branch and
4066 -fcheck-pointer-bounds.
4068 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4071 * config/aarch64/aarch64.c (aarch64_print_address_internal):
4072 Change gcc_assert call to output_operand_lossage.
4074 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4076 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
4078 2018-02-22 DJ Delorie <dj@redhat.com>
4079 Sebastian Perta <sebastian.perta@renesas.com>
4080 Oleg Endo <olegendo@gcc.gnu.org>
4082 * config/rx/rx.c (rx_rtx_costs): New function.
4083 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
4085 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
4087 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
4089 2018-02-22 Martin Liska <mliska@suse.cz>
4092 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
4093 Add "native" as a possible value.
4095 2018-02-22 Martin Liska <mliska@suse.cz>
4098 * config/i386/i386.c (ix86_option_override_internal):
4099 Add "native" as a possible value for -march and -mtune.
4101 2018-02-22 Jakub Jelinek <jakub@redhat.com>
4104 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
4105 to all type variants.
4107 PR tree-optimization/84503
4108 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
4109 width as info->bitpos + info->bitsize - start.
4110 (merged_store_group::merge_overlapping): Simplify width computation.
4111 (check_no_overlap): New function.
4112 (imm_store_chain_info::try_coalesce_bswap): Compute expected
4113 start + width and last_order of the group, fail if check_no_overlap
4115 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
4116 to group if check_no_overlap fails.
4118 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4120 * config/rs6000/altivec.md: Delete contraint arguments to
4121 define_expand, define_split, and define_peephole2, and in
4122 define_insn_and_split if always unused.
4123 * config/rs6000/darwin.md: Ditto.
4124 * config/rs6000/dfp.md: Ditto.
4125 * config/rs6000/rs6000.md: Ditto.
4126 * config/rs6000/sync.md: Ditto.
4127 * config/rs6000/vector.md: Ditto.
4128 * config/rs6000/vsx.md: Ditto.
4130 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4132 * config/rs6000/altivec.md: Write output control strings as braced
4133 blocks instead of double-quoted strings.
4134 * config/rs6000/darwin.md: Ditto.
4135 * config/rs6000/rs6000.md: Ditto.
4136 * config/rs6000/vector.md: Ditto.
4137 * config/rs6000/vsx.md: Ditto.
4139 2018-02-21 Jason Merrill <jason@redhat.com>
4141 PR c++/84314 - ICE with templates and fastcall attribute.
4142 * attribs.c (build_type_attribute_qual_variant): Remove assert.
4144 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4146 * ipa-cp.c (determine_versionability): Fix comment typos.
4148 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4151 * ipa-cp.c (determine_versionability): Do not version functions caling
4154 2018-02-21 Martin Liska <mliska@suse.cz>
4157 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
4158 Add "native" as a possible value.
4159 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
4160 the macro when native cpu detection is available.
4162 2018-02-21 Martin Liska <mliska@suse.cz>
4165 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
4166 Add "native" as a possible value.
4167 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
4168 when native cpu detection is available.
4170 2018-02-21 Jakub Jelinek <jakub@redhat.com>
4171 Martin Sebor <msebor@redhat.com>
4173 PR tree-optimization/84478
4174 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
4176 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
4177 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
4178 support which is conservatively correct, for 2 only stay conservative
4179 for maxlen. Formatting and comment capitalization fixes. Add STRICT
4180 argument to the 2 argument get_range_strlen, adjust 6 arg
4181 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
4183 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
4184 (gimple_fold_builtin_strlen): Pass true as last argument to
4187 2018-02-20 Martin Sebor <msebor@redhat.com>
4190 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
4191 (builtin_memref::set_base_and_offset): Same. Handle inner references.
4192 (builtin_memref::builtin_memref): Factor out parts into
4193 set_base_and_offset and call it.
4195 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4198 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
4199 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
4200 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
4201 search at the associated MODE_INT.
4203 2018-02-20 Jeff Law <law@redhat.com>
4206 PR tree-optimization/81592
4208 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
4209 for range data rather than using global data.
4210 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
4211 range data rather than using global data.
4212 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
4213 pass it to children as needed.
4214 (struct directive::fmtresult): Similarly.
4215 (struct directive::set_width): Similarly.
4216 (struct directive::set_precision): Similarly.
4217 (format_integer, format_directive, parse_directive): Similarly.
4218 (format_none): Accept unnamed vr_values parameter.
4219 (format_percent, format_floating, format_character): Similarly.
4220 (format_string, format_plain): Similarly.
4221 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
4222 the EVRP range analyzer for range data rather than using global data.
4223 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
4224 gimple-ssa-evrp-analyze.h
4225 (class sprintf_dom_walker): Add after_dom_children member function.
4226 Add evrp_range_analyzer member.
4227 (sprintf_dom_walker::before_dom_children): Call into the EVRP
4228 range analyzer as needed.
4229 (sprintf_dom_walker::after_dom_children): New member function.
4230 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
4232 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
4233 (evrp_range_analyzer::pop_to_marker): Likewise.
4235 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4237 PR tree-optimization/84419
4238 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
4239 with the required type if its current type is compatible but
4242 2018-02-20 Jakub Jelinek <jakub@redhat.com>
4245 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
4246 after vectorization.
4248 2018-02-20 Martin Liska <mliska@suse.cz>
4251 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
4252 possible values if we don't have a hint.
4254 2018-02-20 Martin Liska <mliska@suse.cz>
4258 * final.c (shorten_branches): Build align_tab array with one
4260 * opts.c (finish_options): Add alignment option limit check.
4261 (MAX_CODE_ALIGN): Likewise.
4262 (MAX_CODE_ALIGN_VALUE): Likewise.
4263 * doc/invoke.texi: Document maximum allowed option value for
4264 all -falign-* options.
4266 2018-02-19 Jakub Jelinek <jakub@redhat.com>
4269 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
4270 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
4271 * var-tracking.c (emit_note_insn_var_location): Remove all references
4272 to NOTE_INSN_CALL_ARG_LOCATION.
4273 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
4274 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
4275 Use copy_rtx_if_shared.
4276 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
4277 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
4278 (dwarf2out_var_location): Remove handling of
4279 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
4281 * final.c (final_scan_insn): Remove all references to
4282 NOTE_INSN_CALL_ARG_LOCATION.
4283 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
4284 before dumping final insns.
4285 * except.c (emit_note_eh_region_end): Remove all references to
4286 NOTE_INSN_CALL_ARG_LOCATION.
4287 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
4288 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
4289 * config/arc/arc.c (hwloop_optimize): Likewise.
4290 * config/arm/arm.c (create_fix_barrier): Likewise.
4291 * config/s390/s390.c (s390_chunkify_start): Likewise.
4292 * config/sh/sh.c (find_barrier): Likewise.
4293 * config/i386/i386.c (rest_of_insert_endbranch,
4294 ix86_seh_fixup_eh_fallthru): Likewise.
4295 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
4296 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
4297 * config/frv/frv.c (frv_function_prologue): Likewise.
4298 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
4300 (note_outside_basic_block_p): Remove all references to
4301 NOTE_INSN_CALL_ARG_LOCATION.
4302 * gengtype.c (adjust_field_rtx_def): Likewise.
4303 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
4305 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
4306 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
4309 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
4312 PR tree-optimization/84452
4313 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
4314 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
4317 2018-02-19 Martin Liska <mliska@suse.cz>
4320 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
4322 2018-02-19 Martin Liska <mliska@suse.cz>
4323 Richard Sandiford <richard.sandiford@linaro.org>
4325 PR tree-optimization/82491
4326 * gimple-fold.c (get_base_constructor): Make earlier bail out
4329 2018-02-19 Carl Love <cel@us.ibm.com>
4331 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
4332 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
4334 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
4335 P8V_BUILTIN_VEC_NEG.
4337 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
4339 * config/rl78/rl78.md (movdf): New define expand.
4341 2018-02-19 Martin Liska <mliska@suse.cz>
4344 * doc/invoke.texi: Fix typo.
4345 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
4347 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
4349 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
4350 handle rs6000_single_float and rs6000_double_float specially for
4353 2018-02-16 Jeff Law <law@redhat.com>
4355 * config/rx/rx.c (add_pop_cfi_notes): New function.;
4358 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4361 * ipa-inline.c (inline_small_functions): Fix a typo.
4363 2018-02-16 Nathan Sidwell <nathan@acm.org>
4365 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
4367 2018-02-16 Carl Love <cel@us.ibm.com>
4369 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
4370 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
4371 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
4372 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
4373 expansion to P8V_BUILTIN_VEC_FLOAT2.
4375 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
4377 PR rtl-optimization/70023
4378 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
4379 src_regno into account.
4381 2018-02-16 Carl Love <cel@us.ibm.com>
4383 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
4384 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
4385 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
4386 * config/rs6000/rs6000.c: Remove case statements for
4387 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
4388 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
4389 and P9V_BUILTIN_VEC_VINSERT4B.
4390 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
4391 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
4392 * config/rs6000/vsx.md:
4393 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
4396 2018-02-16 Carl Love <cel@us.ibm.com>
4398 * config/rs6000/altivec.h: Add builtin names vec_extract4b
4400 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
4402 * config/rs6000/rs6000-c.c: Add the definitions for
4403 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
4404 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
4405 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
4406 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
4407 definition for insert4b and define insn *insert3b_internal.
4408 * doc/extend.texi: Add documentation for vec_extract4b.
4410 2018-02-16 Nathan Sidwell <nathan@acm.org>
4412 * doc/extend.texi (Backwards Compatibility): Mention friend
4413 injection. Note for-scope is deprecated.
4414 * doc/invoke.texi (-ffriend-injection): Deprecate.
4416 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
4418 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
4419 that moved to I2, also allow destinations that are a paradoxical
4420 subreg (instead of a normal reg).
4422 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
4425 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
4428 2018-02-16 Richard Biener <rguenther@suse.de>
4430 PR tree-optimization/84037
4431 PR tree-optimization/84016
4433 * config/i386/i386.c (ix86_builtin_vectorization_cost):
4434 Adjust vec_construct for the fact we need additional higher latency
4435 128bit inserts for AVX256 and AVX512 vector builds.
4436 (ix86_add_stmt_cost): Scale vector construction cost for
4439 2018-02-16 Richard Biener <rguenther@suse.de>
4441 PR tree-optimization/84417
4442 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
4443 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
4444 (non_rewritable_lvalue_p): Likewise, use poly-ints.
4446 2018-02-16 Martin Liska <mliska@suse.cz>
4449 * internal-fn.def (ASAN_CHECK): Set proper flags.
4450 (ASAN_MARK): Likewise.
4452 2018-02-16 Julia Koval <julia.koval@intel.com>
4454 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
4455 from PTA_CANNONLAKE.
4457 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4460 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
4461 Use ++iter rather than iter++ for std::list iterators.
4462 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
4463 defer deleting them until all nodes in the forest are processed. Do
4464 free even leaf nodes. Change to_process into auto_vec.
4467 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
4468 * vec.h (vec_default_construct): Use memset instead of placement new
4469 if BROKEN_VALUE_INITIALIZATION is defined.
4470 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
4471 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
4474 PR rtl-optimization/83723
4475 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
4476 * lra.c (lra_substitute_pseudo): Likewise. If true, use
4477 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
4479 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
4481 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
4483 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
4485 PR rtl-optimization/81443
4486 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
4487 from inner REGs to paradoxical SUBREGs.
4489 2018-02-16 Richard Biener <rguenther@suse.de>
4491 PR tree-optimization/84399
4492 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4493 For operands we can analyze at their definition make sure we can
4494 analyze them at each use as well.
4496 2018-02-16 Richard Biener <rguenther@suse.de>
4498 PR tree-optimization/84190
4499 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
4500 volatile accesses if the decl isn't volatile.
4502 2018-02-15 Jason Merrill <jason@redhat.com>
4504 PR c++/84314 - ICE with templates and fastcall attribute.
4505 * attribs.c (build_type_attribute_qual_variant): Don't clobber
4506 TYPE_CANONICAL on an existing type.
4508 2018-02-15 Jakub Jelinek <jakub@redhat.com>
4510 PR tree-optimization/84383
4511 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
4512 dstoff nor call operand_equal_p if dstbase is NULL.
4514 PR tree-optimization/84334
4515 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
4516 also a CONSTANT_CLASS_P, punt.
4518 2018-02-14 Jim Wilson <jimw@sifive.com>
4520 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
4521 first SMALL_OPERAND check. New local min_second_step. Move assert
4522 to where locals are set. Add TARGET_RVC support.
4523 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
4525 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
4527 * doc/invoke.texi: Correct -Wformat-overflow code sample.
4529 2018-02-14 Martin Sebor <msebor@redhat.com>
4531 PR tree-optimization/83698
4532 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
4533 arrays constrain the offset range to their bounds.
4534 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
4535 (builtin_access::overlap): Avoid setting the size of overlap if it's
4537 (maybe_diag_overlap): Also consider arrays when deciding what values
4538 of offsets to include in diagnostics.
4540 2018-02-14 Martin Sebor <msebor@redhat.com>
4543 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
4544 that correspond to the kind of a declaration.
4546 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
4549 * config/pa/pa.md: Load address of PIC label using the linkage table
4550 if the label is nonlocal.
4552 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
4554 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
4555 warning message if user requests -maltivec=be.
4556 * doc/invoke.texi: Document deprecation of -maltivec=be.
4558 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
4561 * config/rs6000/rs6000-c.c: Update definitions for
4562 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
4563 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
4565 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4568 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
4569 add _get_ssp intrinsics. Remove argument from
4570 __builtin_ia32_rdssp[d|q].
4571 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
4572 * config/i386/i386-builtin.def: Remove argument from
4573 __builtin_ia32_rdssp[d|q].
4574 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
4575 ix86_expand_special_args_builtin for _rdssp[d|q].
4576 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
4577 Clear register before usage.
4578 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
4579 Add documentation for new _get_ssp and _inc_ssp intrinsics.
4581 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
4583 PR tree-optimization/84357
4584 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
4585 operand 1 of an ARRAY_REF too.
4587 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
4590 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
4591 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4593 (set_of_reg): New struct.
4594 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
4595 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
4596 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4598 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
4599 Split into bitclr, bitset, bitinvert patterns if appropriate.
4600 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
4601 use rx_fuse_in_memory_bitop.
4602 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
4603 to named insn, correct maximum insn length.
4605 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
4608 * machmode.def: Define a complex mode for PARTIAL_INT.
4609 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
4611 * doc/rtl.texi: Document CSPImode.
4612 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
4614 (msp430_hard_regno_nregs_with_padding): Likewise.
4616 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
4619 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
4621 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
4623 PR rtl-optimization/84169
4624 * combine.c (try_combine): New variable split_i2i3. Set it to true if
4625 we generated a parallel as new i3 and we split that to new i2 and i3
4626 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
4627 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
4628 those to i2, not i1. Partially rewrite this scan code.
4630 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4633 * stor-layout.c (place_field): For variable length fields, adjust
4634 offset_align afterwards not just based on the field's alignment,
4635 but also on the size.
4638 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
4639 of exps and logs in the use_exp2 case.
4641 2018-02-13 Jeff Law <law@redhat.com>
4643 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
4646 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
4649 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
4653 * common.opt (gas-loc-support, gas-locview-support): New.
4654 (ginline-points, ginternal-reset-location-views): New.
4655 * doc/invoke.texi: Document them. Use @itemx where intended.
4656 (gvariable-location-views): Adjust.
4657 * target.def (reset_location_view): New.
4658 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
4659 (TARGET_RESET_LOCATION_VIEW): New.
4660 * doc/tm.texi: Rebuilt.
4661 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
4662 (dwarf2out_default_as_locview_support): New.
4663 (output_asm_line_debug_info): Use option variables.
4664 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
4665 (output_loc_list): Likewise.
4666 (add_high_low_attributes): Check option variables.
4667 Don't output entry view attribute in strict mode.
4668 (gen_inlined_subroutine_die): Check option variables.
4669 (dwarf2out_inline_entry): Likewise.
4670 (init_sections_and_labels): Likewise.
4671 (dwarf2out_early_finish): Likewise.
4672 (maybe_reset_location_view): New, from...
4673 (dwarf2out_var_location): ... here. Call it.
4674 * debug.h (dwarf2out_default_as_loc_support): Declare.
4675 (dwarf2out_default_as_locview_support): Declare.
4676 * hooks.c (hook_int_rtx_insn_0): New.
4677 * hooks.h (hook_int_rtx_insn_0): Declare.
4678 * toplev.c (process_options): Take -gas-loc-support and
4679 -gas-locview-support from dwarf2out. Enable
4680 -gvariable-location-views by default only with locview
4681 assembler support. Enable -ginternal-reset-location-views by
4682 default only if the target defines the corresponding hook.
4683 Enable -ginline-points by default if location views are
4684 enabled; force it disabled if statement frontiers are
4686 * tree-inline.c (expand_call_inline): Check option variables.
4687 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
4689 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
4691 PR tree-optimization/84321
4692 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
4693 handling. Also check whether the anti-range contains any values
4694 that satisfy the mask; switch to a VR_RANGE if not.
4696 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
4699 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
4701 2018-02-13 Martin Jambor <mjambor@suse.cz>
4704 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
4705 of call statements, also set location of a load to a temporary.
4707 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
4709 * config/rl78/rl78.c (add_vector_labels): New function.
4710 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
4711 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
4712 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
4713 which checks that no arguments are passed.
4714 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
4715 * doc/extend.texi: Documentation for the new attribute.
4717 2018-02-13 Andreas Schwab <schwab@suse.de>
4719 * config/riscv/linux.h (CPP_SPEC): Define.
4721 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4724 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
4725 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
4726 OPTION_MASK_ISA_AES as first argument to def_builtin_const
4727 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
4728 instead of OPTION_MASK_ISA_PCLMUL as first argument to
4729 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
4730 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
4731 temporarily for AES and PCLMUL builtins.
4733 PR tree-optimization/84339
4734 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
4735 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
4739 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
4740 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
4741 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
4743 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
4745 * omp-simd-clone.h: New file.
4746 * omp-simd-clone.c: Include omp-simd-clone.h.
4747 (expand_simd_clones): No longer static.
4748 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
4749 cgraph.h and omp-simd-clone.h.
4750 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
4751 (vect_recog_widen_shift_pattern): Formatting fix.
4752 (vect_pattern_recog_1): Don't check optab for calls.
4755 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
4756 operands[2] into a REG before using gen_lowpart on it.
4758 2018-02-12 Jeff Law <law@redhat.com>
4761 * config/sh/sh.c (find_barrier): Consider a sibling call
4764 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
4765 successfully back substituting a reg.
4767 2018-02-12 Richard Biener <rguenther@suse.de>
4769 PR tree-optimization/84037
4770 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
4771 parameter, move visited init to caller.
4772 (vect_slp_analyze_operations): Separate cost from validity
4773 check, initialize visited once for all instances.
4774 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
4776 * tree-vect-stmts.c (vect_model_simple_cost): Make early
4778 (vect_model_promotion_demotion_cost): Likewise.
4779 (vectorizable_bswap): Guard cost modeling with !slp_node
4780 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
4782 (vectorizable_call): Likewise.
4783 (vectorizable_conversion): Likewise.
4784 (vectorizable_assignment): Likewise.
4785 (vectorizable_shift): Likewise.
4786 (vectorizable_operation): Likewise.
4787 (vectorizable_store): Likewise.
4788 (vectorizable_load): Likewise.
4789 (vectorizable_condition): Likewise.
4790 (vectorizable_comparison): Likewise.
4792 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
4795 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
4796 (ASAN_MARK): Fix fnspec to account for return value, change pointer
4797 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
4799 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4802 * params.def (inline-min-speedup): Increase from 8 to 15.
4803 (max-inline-insns-auto): Decrease from 40 to 30.
4804 * ipa-split.c (consider_split): Add some buffer for function to
4805 be considered inlining candidate.
4806 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
4809 2018-02-12 Richard Biener <rguenther@suse.de>
4811 PR tree-optimization/84037
4812 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
4813 matched stmts if we cannot swap the non-matched ones.
4815 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4817 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
4818 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
4819 _mm_maskz_scalef_round_ss): New intrinsics.
4820 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
4821 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
4822 __builtin_ia32_scalefss_round): Remove.
4823 (__builtin_ia32_scalefsd_mask_round,
4824 __builtin_ia32_scalefss_mask_round): New intrinsics.
4825 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
4826 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
4827 ((match_operand:VF_128 2 "<round_nimm_predicate>"
4828 "<round_constraint>")): Changed to ...
4829 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
4830 "<round_scalar_constraint>")): ... this.
4831 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
4832 %0, %1, %2<round_op3>}"): Changed to ...
4833 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
4834 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
4835 %2<round_scalar_mask_op3>}"): ... this.
4836 * config/i386/subst.md (round_scalar_nimm_predicate): New.
4838 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4840 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
4841 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
4842 (_mm_maskz_sqrt_round_ss): New intrinsics.
4843 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
4844 (__builtin_ia32_sqrtsd_mask_round)
4845 (__builtin_ia32_sqrtss_mask_round): New builtins.
4846 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
4847 (__builtin_ia32_sqrtss_round): Remove.
4848 (__builtin_ia32_sqrtsd_mask_round)
4849 (__builtin_ia32_sqrtss_mask_round): New builtins.
4850 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
4851 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
4852 ((match_operand:VF_128 1 "vector_operand"
4853 "xBm,<round_constraint>")): Changed to ...
4854 ((match_operand:VF_128 1 "vector_operand"
4855 "xBm,<round_scalar_constraint>")): ... this.
4856 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
4857 %0, %2, %<iptr>1<round_op3>}): Changed to ...
4858 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
4859 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
4860 %<iptr>1<round_scalar_mask_op3>}): ... this.
4861 ((set_attr "prefix" "<round_prefix>")): Changed to ...
4862 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
4864 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
4867 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
4868 Cast vec_cmpeq result to correct type.
4869 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
4870 Cast vec_cmpgt result to correct type.
4872 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
4874 * final.c (final_scan_insn_1): Renamed from...
4875 (final_scan_insn): ... this. New wrapper, to recover
4876 seen from the outermost call in recursive ones.
4877 * config/sparc/sparc.c (output_return): Drop seen from call.
4878 (output_sibcall): Likewise.
4879 * config/visium/visium.c (output_branch): Likewise.
4881 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
4883 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
4886 2018-02-10 Alan Modra <amodra@gmail.com>
4889 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
4890 Specify LR as an input.
4892 2018-02-10 Jakub Jelinek <jakub@redhat.com>
4895 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
4896 remove_member_access_dummy_vars): New functions.
4897 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
4898 lower_omp_1, execute_lower_omp): Use them.
4900 PR rtl-optimization/84308
4901 * shrink-wrap.c (spread_components): Release todo vector.
4903 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
4905 PR rtl-optimization/57193
4906 * ira-color.c (struct allocno_color_data): Add member
4907 conflict_allocno_hard_prefs.
4908 (update_conflict_allocno_hard_prefs): New.
4909 (bucket_allocno_compare_func): Add a preference based on
4910 conflict_allocno_hard_prefs.
4911 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
4912 (color_allocnos): Remove a dead code. Initiate
4913 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
4915 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4918 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
4919 constraint from =wa to wa. Avoid a subreg on the output operand,
4920 instead use a pseudo and subreg it in a move.
4921 (p9_xxbrd_<mode>): Changed to ...
4922 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
4923 (p9_xxbrd_v2df): New expander.
4924 (p9_xxbrw_<mode>): Changed to ...
4925 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
4926 (p9_xxbrw_v4sf): New expander.
4928 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4930 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
4932 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
4935 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
4936 multiply in 32-bit mode.
4937 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
4938 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
4941 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4943 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
4944 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
4945 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
4946 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
4948 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
4951 * dwarf2out.c (is_trivial_indirect_ref): New function.
4952 (dwarf2out_late_global_decl): Do not generate a location
4953 attribute for variables that have a non-trivial DECL_VALUE_EXPR
4954 and that are not defined in the current unit.
4956 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4958 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
4959 instead of a libcall for UNORDERED.
4961 2018-02-09 Tamar Christina <tamar.christina@arm.com>
4964 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
4965 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
4967 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4970 * config/s390/s390.c (s390_set_current_function): Invoke
4971 s390_indirect_branch_settings also if fndecl didn't change.
4973 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4975 * config/rs6000/rs6000.md (blockage): Set length to zero.
4977 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4979 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
4981 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4984 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
4985 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
4989 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
4990 PARALLEL incoming that failed vt_get_decl_and_offset check.
4993 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
4994 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
4996 (get_variable_section): For decls in named .bss* sections pass true as
4997 second argument to bss_initializer_p.
4999 2018-02-09 Marek Polacek <polacek@redhat.com>
5000 Jakub Jelinek <jakub@redhat.com>
5003 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
5004 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
5005 Sync some changes from cxx_fold_indirect_ref.
5007 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5009 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
5011 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
5012 (BLOCK_INLINE_ENTRY_LABEL): New.
5013 (dwarf2out_var_location): Disregard inline entry markers.
5014 (inline_entry_data): New struct.
5015 (inline_entry_data_hasher): New hashtable type.
5016 (inline_entry_data_hasher::hash): New.
5017 (inline_entry_data_hasher::equal): New.
5018 (inline_entry_data_table): New variable.
5019 (add_high_low_attributes): Add DW_AT_entry_pc and
5020 DW_AT_GNU_entry_view attributes if a pending entry is found
5021 in inline_entry_data_table. Add old entry_pc attribute only
5022 if debug nonbinding markers are disabled.
5023 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
5024 markers are enabled.
5025 (block_within_block_p, dwarf2out_inline_entry): New.
5026 (dwarf2out_finish): Check that no entries remained in
5027 inline_entry_data_table.
5028 * final.c (reemit_insn_block_notes): Handle inline entry notes.
5029 (final_scan_insn, notice_source_line): Likewise.
5030 (rest_of_clean_state): Skip inline entry markers.
5031 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
5033 * gimple.c (gimple_build_debug_inline_entry): New.
5034 * gimple.h (enum gimple_debug_subcode): Add
5035 GIMPLE_DEBUG_INLINE_ENTRY.
5036 (gimple_build_debug_inline_entry): Declare.
5037 (gimple_debug_inline_entry_p): New.
5038 (gimple_debug_nonbind_marker_p): Adjust.
5039 * insn-notes.def (INLINE_ENTRY): New.
5040 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
5041 inline entry marker notes.
5042 (print_insn): Likewise.
5043 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
5044 (INSN_DEBUG_MARKER_KIND): Likewise.
5045 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
5046 * tree-inline.c (expand_call_inline): Build and insert
5047 debug_inline_entry stmt.
5048 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
5049 inline entry blocks early, if nonbind markers are enabled.
5050 (dump_scope_block): Dump fragment info.
5051 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
5052 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
5053 (gimple_build_debug_inline_entry): New.
5054 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
5055 Enable/disable inline entry points too.
5056 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
5057 (DEBUG_INSN): Describe inline entry markers.
5059 * common.opt (gvariable-location-views): New.
5060 (gvariable-location-views=incompat5): New.
5061 * config.in: Rebuilt.
5062 * configure: Rebuilt.
5063 * configure.ac: Test assembler for view support.
5064 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
5065 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
5066 * dwarf2out.c (var_loc_view): New typedef.
5067 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
5068 (dwarf2out_locviews_in_attribute): New.
5069 (dwarf2out_locviews_in_loclist): New.
5070 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
5071 (enum dw_line_info_opcode): Add LI_adv_address.
5072 (struct dw_line_info_table): Add view.
5073 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
5074 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
5075 (zero_view_p): New variable.
5076 (ZERO_VIEW_P): New macro.
5077 (output_asm_line_debug_info): New.
5078 (struct var_loc_node): Add view.
5079 (add_AT_view_list, AT_loc_list): New.
5080 (add_var_loc_to_decl): Add view param. Test it against last.
5081 (new_loc_list): Add view params. Record them.
5082 (AT_loc_list_ptr): Handle loc and view lists.
5083 (view_list_to_loc_list_val_node): New.
5084 (print_dw_val): Handle dw_val_class_view_list.
5085 (size_of_die): Likewise.
5086 (value_format): Likewise.
5087 (loc_list_has_views): New.
5088 (gen_llsym): Set vl_symbol too.
5089 (maybe_gen_llsym, skip_loc_list_entry): New.
5090 (dwarf2out_maybe_output_loclist_view_pair): New.
5091 (output_loc_list): Output view list or entries too.
5092 (output_view_list_offset): New.
5093 (output_die): Handle dw_val_class_view_list.
5094 (output_dwarf_version): New.
5095 (output_compilation_unit_header): Use it.
5096 (output_skeleton_debug_sections): Likewise.
5097 (output_rnglists, output_line_info): Likewise.
5098 (output_pubnames, output_aranges): Update version comments.
5099 (output_one_line_info_table): Output view numbers in asm comments.
5100 (dw_loc_list): Determine current endview, pass it to new_loc_list.
5101 Call maybe_gen_llsym.
5102 (loc_list_from_tree_1): Adjust.
5103 (add_AT_location_description): Create view list attribute if
5104 needed, check it's absent otherwise.
5105 (convert_cfa_to_fb_loc_list): Adjust.
5106 (maybe_emit_file): Call output_asm_line_debug_info for test.
5107 (dwarf2out_var_location): Reset views as needed. Precompute
5108 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
5109 attribute. Set view.
5110 (new_line_info_table): Reset next view.
5111 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
5112 (dwarf2out_source_line): Likewise. Output view resets and labels to
5113 the assembler, or select appropriate line info opcodes.
5114 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
5115 (optimize_string_length): Catch it. Adjust.
5116 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
5117 dw_val_class_view_list, and remove it if no longer needed.
5118 (hash_loc_list): Hash view numbers.
5119 (loc_list_hasher::equal): Compare them.
5120 (optimize_location_lists): Check whether a view list symbol is
5121 needed, and whether the locview attribute is present, and
5122 whether they match. Remove the locview attribute if no longer
5124 (index_location_lists): Call skip_loc_list_entry for test.
5125 (dwarf2out_finish): Call output_asm_line_debug_info for test.
5126 Use output_dwarf_version.
5127 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
5128 (struct dw_val_node): Add val_view_list.
5129 * final.c (SEEN_NEXT_VIEW): New.
5130 (set_next_view_needed): New.
5131 (clear_next_view_needed): New.
5132 (maybe_output_next_view): New.
5133 (final_start_function): Rename to...
5134 (final_start_function_1): ... this. Take pointer to FIRST,
5135 add SEEN parameter. Emit param bindings in the initial view.
5136 (final_start_function): Reintroduce SEEN-less interface.
5137 (final): Rename to...
5138 (final_1): ... this. Take SEEN parameter. Output final pending
5139 next view at the end.
5140 (final): Reintroduce seen-less interface.
5141 (final_scan_insn): Output pending next view before switching
5142 sections or ending a block. Mark the next view as needed when
5143 outputting variable locations. Notify debug backend of section
5144 changes, and of location view changes.
5145 (rest_of_handle_final): Adjust.
5146 * toplev.c (process_options): Autodetect value for debug variable
5147 location views option. Warn on incompat5 without -gdwarf-5.
5148 * doc/invoke.texi (gvariable-location-views): New.
5149 (gvariable-location-views=incompat5): New.
5150 (gno-variable-location-views): New.
5152 2018-02-08 David Malcolm <dmalcolm@redhat.com>
5154 PR tree-optimization/84136
5155 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
5156 that the result of find_edge is non-NULL.
5158 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5161 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
5162 storing integer register in SImode. Fix cost of 256 and 512
5163 byte aligned SSE register store.
5165 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5167 * config/i386/i386.c (ix86_multiplication_cost): Fix
5168 multiplication cost for TARGET_AVX512DQ.
5170 2018-02-08 Marek Polacek <polacek@redhat.com>
5172 PR tree-optimization/84238
5173 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
5176 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5178 PR tree-optimization/84265
5179 * tree-vect-stmts.c (vectorizable_store): Don't treat
5180 VMAT_CONTIGUOUS accesses as grouped.
5181 (vectorizable_load): Likewise.
5183 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5185 PR tree-optimization/81635
5186 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
5187 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
5188 (test_round_for_mask): New functions.
5189 (wide_int_cc_tests): Call test_round_for_mask.
5190 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
5191 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
5192 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
5193 range returned by get_range_info.
5195 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5198 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
5199 * symtab.c: Include builtins.h
5200 (symtab_node::output_to_lto_symbol_table_p): Move here
5201 from lto-streamer-out.c:output_symbol_p.
5202 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
5203 (output_symbol_p): Move all logic to symtab.c
5204 (produce_symtab): Update.
5206 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5208 * config/s390/s390-opts.h (enum indirect_branch): Define.
5209 * config/s390/s390-protos.h (s390_return_addr_from_memory)
5210 (s390_indirect_branch_via_thunk)
5211 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
5212 (enum s390_indirect_branch_type): Define.
5213 * config/s390/s390.c (struct s390_frame_layout, struct
5214 machine_function): Remove.
5215 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
5216 (indirect_branch_table_label_no, indirect_branch_table_name):
5218 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
5219 (enum s390_indirect_branch_option): Define.
5220 (s390_return_addr_from_memory): New function.
5221 (s390_handle_string_attribute): New function.
5222 (s390_attribute_table): Add new attribute handler.
5223 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
5224 (s390_indirect_branch_via_thunk): New function.
5225 (s390_indirect_branch_via_inline_thunk): New function.
5226 (s390_function_ok_for_sibcall): When jumping via thunk disallow
5227 sibling call optimization for non z10 compiles.
5228 (s390_emit_call): Force indirect branch target to be a single
5229 register. Add r1 clobber for non-z10 compiles.
5230 (s390_emit_epilogue): Emit return jump via return_use expander.
5231 (s390_reorg): Handle JUMP_INSNs as execute targets.
5232 (s390_option_override_internal): Perform validity checks for the
5233 new command line options.
5234 (s390_indirect_branch_attrvalue): New function.
5235 (s390_indirect_branch_settings): New function.
5236 (s390_set_current_function): Invoke s390_indirect_branch_settings.
5237 (s390_output_indirect_thunk_function): New function.
5238 (s390_code_end): Implement target hook.
5239 (s390_case_values_threshold): Implement target hook.
5240 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
5242 * config/s390/s390.h (struct s390_frame_layout)
5243 (struct machine_function): Move here from s390.c.
5244 (TARGET_INDIRECT_BRANCH_NOBP_RET)
5245 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
5246 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
5247 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
5248 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
5249 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
5250 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
5251 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
5252 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
5253 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
5254 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
5255 (mnemonic attribute): Add values which aren't recognized
5257 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
5258 pattern for branch conversion. Fix mnemonic attribute.
5259 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
5260 indirect branch via thunk if requested.
5261 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
5262 ("*indirect_jump"): Disable for branch conversion using out of
5264 ("indirect_jump_via_thunk<mode>_z10")
5265 ("indirect_jump_via_thunk<mode>")
5266 ("indirect_jump_via_inlinethunk<mode>_z10")
5267 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
5268 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
5269 ("casesi_jump_via_inlinethunk<mode>_z10")
5270 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
5271 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
5272 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
5273 ("*indirect2_jump"): Disable for branch conversion.
5274 ("casesi_jump"): Turn into expander and expand patterns for branch
5276 ("return_use"): New expander.
5277 ("*return"): Emit return via thunk and rename it to ...
5278 ("*return<mode>"): ... this one.
5279 * config/s390/s390.opt: Add new options and and enum for the
5282 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5284 * lra-constraints.c (match_reload): Unconditionally use
5285 gen_lowpart_SUBREG, rather than selecting between that
5286 and equivalent gen_rtx_SUBREG code.
5288 2018-02-08 Richard Biener <rguenther@suse.de>
5290 PR tree-optimization/84233
5291 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
5292 changed flag instead of boguously re-using phi_inserted.
5294 2018-02-08 Martin Jambor <mjambor@suse.cz>
5296 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
5297 static local variables.
5299 2018-02-08 Richard Biener <rguenther@suse.de>
5301 PR tree-optimization/84278
5302 * tree-vect-stmts.c (vectorizable_store): When looking for
5303 smaller vector types to perform grouped strided loads/stores
5304 make sure the mode is supported by the target.
5305 (vectorizable_load): Likewise.
5307 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5309 * config/aarch64/aarch64.c (aarch64_components_for_bb):
5310 Increase LDP/STP opportunities by adding adjacent callee-saves.
5312 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5314 PR rtl-optimization/84068
5315 PR rtl-optimization/83459
5316 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
5318 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
5320 PR tree-optimization/84224
5321 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
5322 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
5325 2018-02-07 Iain Sandoe <iain@codesourcery.com>
5328 * config/rs6000/altivec.md (*restore_world): Remove LR use.
5329 * config/rs6000/predicates.md (restore_world_operation): Adjust op
5330 count, remove one USE.
5332 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
5334 * doc/install.texi (Configuration): Document the
5335 --with-long-double-format={ibm,ieee} PowerPC configuration
5339 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
5340 Convert from define_expand to be define_insn_and_split. Rework
5341 float/double/_Float128 conversions to QI/HI/SImode to work with
5342 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
5343 conversions to QI/HImode types did a store and then a load to
5344 truncate the value. For conversions to VSX registers, don't split
5345 the insn, instead emit the code directly. Use the code iterator
5346 any_fix to combine signed and unsigned conversions.
5347 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
5348 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
5349 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5350 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
5351 (fix_<mode>di2_hw): Likewise.
5352 (fixuns_<mode>di2_hw): Likewise.
5353 (fix_<mode>si2_hw): Likewise.
5354 (fixuns_<mode>si2_hw): Likewise.
5355 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
5356 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5357 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
5358 fix<uns>_trunc<SFDF:mode>si2_p8.
5359 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
5361 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
5362 (fix<uns>_<mode>_mem): Likewise.
5363 (fctiw<u>z_<mode>_mem): Likewise.
5364 (fix<uns>_<mode>_mem): Likewise.
5365 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
5366 the register allocator from doing a direct move to the GPRs to do
5367 a store, and instead use the ISA 3.0 store byte/half-word from
5368 vector register instruction. For IEEE 128-bit floating point,
5369 also optimize stores of 32-bit ints.
5370 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
5372 2018-02-07 Alan Hayward <alan.hayward@arm.com>
5374 * genextract.c (push_pathstr_operand): New function to support
5376 (walk_rtx): Call push_pathstr_operand.
5377 (print_path): Support [a-zA-Z].
5379 2018-02-07 Richard Biener <rguenther@suse.de>
5381 PR tree-optimization/84037
5382 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
5383 (cse_and_gimplify_to_preheader): Declare.
5384 (vect_get_place_in_interleaving_chain): Likewise.
5385 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5387 (_loop_vec_info::~_loop_vec_info): Delete it.
5388 (cse_and_gimplify_to_preheader): New function.
5389 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
5390 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
5391 (vectorizable_load): Likewise. For grouped stores always base
5392 the IV on the first element.
5393 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
5394 condition before gimplifying.
5396 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5398 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
5399 *DIV_EXPR and *MOD_EXPR.
5401 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
5404 * config/i386/i386.c (ix86_option_override_internal): Mask out
5405 the CF_SET bit when checking -fcf-protection.
5407 2018-02-07 Tom de Vries <tom@codesourcery.com>
5410 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
5413 2018-02-07 Richard Biener <rguenther@suse.de>
5415 PR tree-optimization/84204
5416 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
5419 PR tree-optimization/84205
5420 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
5421 special-case isl_ast_op_zdiv_r.
5423 PR tree-optimization/84223
5424 * graphite-scop-detection.c (gather_bbs::before_dom_children):
5425 Only add conditions from within the region.
5426 (gather_bbs::after_dom_children): Adjust.
5428 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
5431 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
5432 * config/avr/avr.md: Only post-reload split REG-REG moves if
5433 either register is GENERAL_REG_P.
5435 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5437 PR tree-optimization/84235
5438 * tree-ssa-scopedtables.c
5439 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
5440 if the subtraction is performed in floating point type where NaNs are
5441 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
5442 build 1. Formatting fix.
5444 2018-02-06 Jakub Jelinek <jakub@redhat.com>
5447 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
5448 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
5449 and skip it regardless of bb boundaries. Use CALL_P macro,
5450 don't test INSN_P (insn) together with CALL_P or JUMP_P check
5451 unnecessarily, formatting fix.
5453 2018-02-06 Michael Collison <michael.collison@arm.com>
5455 * config/arm/thumb2.md:
5456 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
5457 (*thumb_mov_notscc): Ditto.
5459 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
5462 * config/rs6000/rs6000.md (su code attribute): Use "u" for
5463 unsigned_fix, not "s".
5465 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5467 * configure.ac (gcc_fn_eh_frame_ro): New function.
5468 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
5469 correct .eh_frame permissions.
5470 * configure: Regenerate.
5472 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
5474 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
5477 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5479 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5480 Display warning message for -mno-speculate-indirect-jumps.
5482 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
5484 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
5486 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
5488 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
5490 PR tree-optimization/84225
5491 * tree-eh.c (find_trapping_overflow): Only call
5492 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
5494 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5497 * config/i386/i386.c: Reimplement the check of possible options
5498 -mibt/-mshstk conbination. Change error messages.
5499 * doc/invoke.texi: Fix a typo: remove extra '='.
5501 2018-02-06 Marek Polacek <polacek@redhat.com>
5503 PR tree-optimization/84228
5504 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
5506 2018-02-06 Tamar Christina <tamar.christina@arm.com>
5509 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
5510 emitted arch directives.
5511 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
5512 __ARM_FEATURE_COPROC before changing architectures.
5514 2018-02-06 Richard Biener <rguenther@suse.de>
5516 * config/i386/i386.c (print_reg): Fix typo.
5517 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
5519 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
5521 * configure: Regenerate.
5523 2018-02-05 Martin Sebor <msebor@redhat.com>
5525 PR tree-optimization/83369
5526 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
5529 2018-02-05 Martin Liska <mliska@suse.cz>
5531 * doc/invoke.texi: Cherry-pick upstream r323995.
5533 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
5535 * ira.c (ira_init_register_move_cost): Adjust comment.
5537 2018-02-05 Martin Liska <mliska@suse.cz>
5539 PR gcov-profile/84137
5540 * doc/gcov.texi: Fix typo in documentation.
5542 2018-02-05 Martin Liska <mliska@suse.cz>
5544 PR gcov-profile/83879
5545 * doc/gcov.texi: Document necessity of --dynamic-list-data when
5546 using dlopen functionality.
5548 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
5550 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
5551 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
5552 _mm_maskz_range_ss, _mm_mask_range_round_ss,
5553 _mm_maskz_range_round_ss): New intrinsics.
5554 (__builtin_ia32_rangesd128_round)
5555 (__builtin_ia32_rangess128_round): Remove.
5556 (__builtin_ia32_rangesd128_mask_round,
5557 __builtin_ia32_rangess128_mask_round): New builtins.
5558 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
5559 __builtin_ia32_rangess128_round): Remove.
5560 (__builtin_ia32_rangesd128_mask_round,
5561 __builtin_ia32_rangess128_mask_round): New builtins.
5562 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
5563 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
5564 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
5565 "<round_saeonly_constraint>")): Changed to ...
5566 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
5567 "<round_saeonly_scalar_constraint>")): ... this.
5568 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
5569 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
5570 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
5571 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
5572 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
5574 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
5576 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
5578 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
5579 Remove all values except native, 8540 and 8548.
5581 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
5583 * config/i386/i386.c (ix86_output_function_return): Pass
5584 INVALID_REGNUM, instead of -1, as invalid register number to
5585 indirect_thunk_name and output_indirect_thunk.
5587 2018-02-02 Julia Koval <julia.koval@intel.com>
5589 * config.gcc: Add -march=icelake.
5590 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
5591 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
5592 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
5593 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
5594 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
5595 (processor_target_table): Add icelake.
5596 (ix86_option_override_internal): Handle new PTAs.
5597 (get_builtin_code_for_version): Handle icelake.
5598 (M_INTEL_COREI7_ICELAKE): New.
5599 (fold_builtin_cpu): Handle icelake.
5600 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
5601 * doc/invoke.texi: Add -march=icelake.
5603 2018-02-02 Julia Koval <julia.koval@intel.com>
5605 * config/i386/i386.c (ix86_option_override_internal): Change flags type
5606 to wide_int_bitmask.
5607 * wide-int-bitmask.h: New.
5609 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5612 * config/i386/i386.md: Replace Pmode with word_mode in
5613 builtin_setjmp_setup and builtin_longjmp to support x32.
5615 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
5619 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
5621 (rs6000_supported_cpu_names): New static variable.
5622 (linux_cpu_translation_table): Likewise.
5623 (elf_platform) <cpu>: Define new static variable and use it.
5624 Translate kernel AT_PLATFORM name to canonical name if needed.
5625 Error if platform name is unknown.
5627 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
5630 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
5632 2018-02-01 Jeff Law <law@redhat.com>
5635 * config/i386/i386.c (release_scratch_register_on_entry): Add new
5636 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
5637 the scratch if RELEASE_VIA_POP is false.
5638 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
5639 If we have to save a temporary register, decrement SIZE appropriately.
5640 Pass new arguments to release_scratch_register_on_entry.
5641 (ix86_adjust_stack_and_probe): Likewise.
5642 (ix86_emit_probe_stack_range): Pass new arguments to
5643 release_scratch_register_on_entry.
5645 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
5647 PR rtl-optimization/84157
5648 * combine.c (change_zero_ext): Use REG_P predicate in
5649 front of HARD_REGISTER_P predicate.
5651 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
5653 * config/avr/avr.c (avr_option_override): Move disabling of
5654 -fdelete-null-pointer-checks to...
5655 * common/config/avr/avr-common.c (avr_option_optimization_table):
5658 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5660 PR tree-optimization/81635
5661 * tree-data-ref.c (split_constant_offset_1): For types that
5662 wrap on overflow, try to use range info to prove that wrapping
5665 2018-02-01 Renlin Li <renlin.li@arm.com>
5668 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
5670 (aarch64_register_move_cost): Likewise.
5671 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
5673 (REG_CLASS_NAMES): Likewise.
5674 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
5675 TAILCALL_ADDR_REGS. Remove IP registers.
5676 * config/aarch64/aarch64.md (Ucs): Update register constraint.
5678 2018-02-01 Richard Biener <rguenther@suse.de>
5680 * domwalk.h (dom_walker::dom_walker): Add additional constructor
5681 for specifying RPO order and allow NULL for that.
5682 * domwalk.c (dom_walker::dom_walker): Likewise.
5683 (dom_walker::walk): Handle NULL RPO order.
5684 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
5686 (rewrite_update_dom_walker): Likewise.
5687 (mark_def_dom_walker): Likewise.
5689 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5691 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
5692 (aarch64_maybe_expand_sve_subreg_move): Declare.
5693 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
5694 * config/aarch64/predicates.md (aarch64_any_register_operand): New
5696 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
5697 that are semantically a reverse operation.
5698 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
5699 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
5700 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
5702 (aarch64_can_change_mode_class): For big-endian, forbid changes
5703 between two SVE modes if they have different element sizes.
5705 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5707 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
5708 the TImode handling for big-endian targets.
5710 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5712 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
5713 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
5715 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
5716 Remove BSWAP handing for big-endian targets and use the form of
5717 LD1RQ appropariate for the mode.
5719 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5721 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
5722 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
5725 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5728 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
5729 check for operands that need to go through aarch64_sve_reload_be.
5731 2018-02-01 Jakub Jelinek <jakub@redhat.com>
5733 PR tree-optimization/81661
5734 PR tree-optimization/84117
5735 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
5736 * tree-eh.c: Include gimplify.h.
5737 (find_trapping_overflow, replace_trapping_overflow,
5738 rewrite_to_non_trapping_overflow): New functions.
5739 * tree-vect-loop.c: Include tree-eh.h.
5740 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
5741 * tree-data-ref.c: Include tree-eh.h.
5742 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
5744 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
5746 PR rtl-optimization/84123
5747 * combine.c (change_zero_ext): Check if hard register satisfies
5748 can_change_dest_mode before calling gen_lowpart_SUBREG.
5750 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
5753 * ira.c (ira_init_register_move_cost): Remove assert.
5755 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5757 PR rtl-optimization/84071
5758 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
5759 * doc/tm.texi: Regenerate.
5761 2018-01-31 Richard Biener <rguenther@suse.de>
5763 PR tree-optimization/84132
5764 * tree-data-ref.c (analyze_miv_subscript): Properly
5765 check whether evolution_function_is_affine_multivariate_p
5766 before calling gcd_of_steps_may_divide_p.
5768 2018-01-31 Julia Koval <julia.koval@intel.com>
5771 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
5772 * config/i386/i386.md (rdpid_rex64) New.
5773 (rdpid): Make 32bit only.
5775 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
5778 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
5779 an IDENTIFIER_NODE for FUNCTION_TYPE's.
5781 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5784 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5786 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5788 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5790 PR rtl-optimization/84071
5791 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
5792 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
5794 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5796 * config/arc/arc.c (arc_handle_aux_attribute): New function.
5797 (arc_attribute_table): Add 'aux' attribute.
5798 (arc_in_small_data_p): Consider aux like variables.
5799 (arc_is_aux_reg_p): New function.
5800 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
5801 (arc_get_aux_arg): New function.
5802 (prepare_move_operands): Handle aux-register access.
5803 (arc_handle_aux_attribute): New function.
5804 * doc/extend.texi (ARC Variable attributes): Add subsection.
5806 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5808 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
5809 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
5810 (arc_attribute_table): Add 'uncached' attribute.
5811 (arc_print_operand): Print '.di' flag for uncached memory
5813 (arc_in_small_data_p): Do not consider for small data the uncached
5815 (arc_is_uncached_mem_p): New function.
5816 * config/arc/predicates.md (compact_store_memory_operand): Check
5817 for uncached memory accesses.
5818 (nonvol_nonimm_operand): Likewise.
5819 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
5821 2018-01-31 Jakub Jelinek <jakub@redhat.com>
5824 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
5825 falign-loops=): Add Optimization flag.
5827 2018-01-30 Jeff Law <law@redhat.com>
5830 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
5831 INT_REGISTERS_SAVED. Check it prior to calling
5832 get_scratch_register_on_entry.
5833 (ix86_adjust_stack_and_probe): Similarly.
5834 (ix86_emit_probe_stack_range): Similarly.
5835 (ix86_expand_prologue): Corresponding changes.
5837 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5840 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
5841 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
5843 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
5846 * lra-constraints.c (curr_insn_transform): Process AND in the
5849 2018-01-30 Jakub Jelinek <jakub@redhat.com>
5851 PR rtl-optimization/83986
5852 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
5853 dependence against last_pending_memory_flush in addition to
5856 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
5858 PR tree-optimization/81611
5859 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
5862 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5865 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
5868 2018-01-30 Richard Biener <rguenther@suse.de>
5869 Jakub Jelinek <jakub@redhat.com>
5871 PR tree-optimization/84111
5872 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
5873 inner loops added during recursion, as they don't have up-to-date
5876 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5879 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
5880 (can_inline_edge_by_limits_p): ... here.
5881 (can_early_inline_edge_p, check_callers,
5882 update_caller_keys, update_callee_keys, recursive_inlining,
5883 add_new_edges_to_heap, speculation_useful_p,
5884 inline_small_functions,
5885 inline_small_functions, flatten_function,
5886 inline_to_all_callers_1): Update.
5888 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5890 * profile-count.c (profile_count::combine_with_ipa_count): Handle
5893 2018-01-30 Richard Biener <rguenther@suse.de>
5895 PR tree-optimization/83008
5896 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
5897 invariant and constant vector uses in stmts when they need
5900 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5903 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
5904 * configure: Regenerate.
5906 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5908 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
5910 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
5911 Use gen_rtx_REG rather than gen_lowpart.
5913 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5915 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
5916 rather than 0 when creating partial subregs.
5918 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5920 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
5923 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
5926 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
5927 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
5928 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
5929 flags. This restores the settings used before the 2017-07-24.
5930 Turning off pre increment/decrement/modify allows IVOPTS to
5931 optimize DF/SF loops where the index is an int.
5933 2018-01-29 Richard Biener <rguenther@suse.de>
5934 Kelvin Nilsen <kelvin@gcc.gnu.org>
5937 * tree-vect-stmts.c (vectorizable_call): Don't call
5938 targetm.vectorize_builtin_md_vectorized_function if callee is
5941 2018-01-22 Carl Love <cel@us.ibm.com>
5943 * doc/extend.tex: Fix typo in second arg in
5944 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
5946 2018-01-29 Richard Biener <rguenther@suse.de>
5948 PR tree-optimization/84086
5949 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
5950 (flush_ssaname_freelist): When SSA names were released reset
5951 the SCEV hash table.
5953 2018-01-29 Richard Biener <rguenther@suse.de>
5955 PR tree-optimization/84057
5956 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
5957 removed paths when removing edges.
5959 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
5961 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
5962 -mfunction-return=@var{choice}.
5964 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
5967 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
5969 (layout::print_source_line): Likewise.
5970 (test_get_line_width_without_trailing_whitespace): Add test cases.
5972 2018-01-27 Jakub Jelinek <jakub@redhat.com>
5975 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
5978 2018-01-26 Jim Wilson <jimw@sifive.com>
5980 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
5982 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
5985 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5987 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
5988 and CMP + SUB-immediate -> SUBS.
5990 2018-01-26 Martin Sebor <msebor@redhat.com>
5992 PR tree-optimization/83896
5993 * tree-ssa-strlen.c (get_string_len): Rename...
5994 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
5995 Avoid assuming length is constant.
5996 (handle_char_store): Use HOST_WIDE_INT for string length.
5998 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
6001 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
6002 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
6004 2018-01-26 Richard Biener <rguenther@suse.de>
6006 PR rtl-optimization/84003
6007 * dse.c (record_store): Only record redundant stores when
6008 the earlier store aliases at least all accesses the later one does.
6010 2018-01-26 Jakub Jelinek <jakub@redhat.com>
6012 PR rtl-optimization/83985
6013 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
6014 REG_CFA_RESTORE insns.
6015 (delete_unmarked_insns): Don't ignore separate shrink wrapping
6016 REG_CFA_RESTORE insns here.
6019 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
6020 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
6022 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6024 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
6025 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
6026 (arc_init): Likewise.
6027 (arc_override_options): Likewise.
6028 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
6030 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
6032 * config/arc/arc.h (TARGET_DBNZ): Define.
6033 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
6034 properly set the tune attribute.
6035 (dbnz): Use TARGET_DBNZ guard.
6036 * config/arc/arc.opt (mtune): Add core3 option.
6038 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6040 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
6041 recognize new pic like addresses.
6042 (arc_delegitimize_address): Clean up.
6044 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6046 * config/arc/arc-arches.def: Option mrf16 valid for all
6048 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
6049 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
6050 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
6051 * config/arc/arc-tables.opt: Regenerate.
6052 * config/arc/arc.c (arc_conditional_register_usage): Handle
6053 reduced register file case.
6054 (arc_file_start): Set must have build attributes.
6055 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
6057 * config/arc/arc.opt (mrf16): Add new option.
6058 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
6059 * config/arc/genmultilib.awk: Handle new mrf16 option.
6060 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
6061 * config/arc/t-multilib: Regenerate.
6062 * doc/invoke.texi (ARC Options): Document mrf16 option.
6064 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6066 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
6067 * config/arc/arc.c (arc_handle_secure_attribute): New function.
6068 (arc_attribute_table): Add 'secure_call' attribute.
6069 (arc_print_operand): Print secure call operand.
6070 (arc_function_ok_for_sibcall): Don't optimize tail calls when
6072 (arc_is_secure_call_p): New function. * config/arc/arc.md
6073 (call_i): Add support for sjli instruction.
6074 (call_value_i): Likewise.
6075 * config/arc/constraints.md (Csc): New constraint.
6077 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6078 John Eric Martin <John.Martin@emmicro-us.com>
6080 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
6081 * config/arc/arc.c (_arc_jli_section): New struct.
6082 (arc_jli_section): New type.
6083 (rc_jli_sections): New static variable.
6084 (arc_handle_jli_attribute): New function.
6085 (arc_attribute_table): Add jli_always and jli_fixed attribute.
6086 (arc_file_end): New function.
6087 (TARGET_ASM_FILE_END): Define.
6088 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
6089 (arc_add_jli_section): New function.
6090 (jli_call_scan): Likewise.
6091 (arc_reorg): Call jli_call_scan.
6092 (arc_output_addsi): Remove 'S' from printing asm operand.
6093 (arc_is_jli_call_p): New function.
6094 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
6096 (movhi_insn): Likewise.
6097 (movsi_insn): Likewise.
6098 (movsi_set_cc_insn): Likewise.
6099 (loadqi_update): Likewise.
6100 (load_zeroextendqisi_update): Likewise.
6101 (load_signextendqisi_update): Likewise.
6102 (loadhi_update): Likewise.
6103 (load_zeroextendhisi_update): Likewise.
6104 (load_signextendhisi_update): Likewise.
6105 (loadsi_update): Likewise.
6106 (loadsf_update): Likewise.
6107 (movsicc_insn): Likewise.
6108 (bset_insn): Likewise.
6109 (bxor_insn): Likewise.
6110 (bclr_insn): Likewise.
6111 (bmsk_insn): Likewise.
6112 (bicsi3_insn): Likewise.
6113 (cmpsi_cc_c_insn): Likewise.
6114 (movsi_ne): Likewise.
6115 (movsi_cond_exec): Likewise.
6116 (clrsbsi2): Likewise.
6127 (call_i): Remove 'S' asm letter, add jli instruction.
6128 (call_value_i): Likewise.
6129 * config/arc/arc.op (mjli-always): New option.
6130 * config/arc/constraints.md (Cji): New constraint.
6131 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
6133 (subsf3_fpx): Likewise.
6134 (mulsf3_fpx): Likewise.
6135 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
6137 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
6139 * doc/invoke.texi (ARC): Document mjli-always option.
6141 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
6143 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
6144 avoid addition with 0 and use incw and decw where possible.
6146 2018-01-26 Richard Biener <rguenther@suse.de>
6148 PR tree-optimization/81082
6149 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
6150 association if it requires casting to unsigned.
6151 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
6152 from fold_plusminus_mult_expr to catch important cases late when
6153 range info is available.
6155 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6157 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
6158 * configure.ac (hidden_linkonce): New test.
6159 * configure: Regenerate.
6160 * config.in: Regenerate.
6162 2018-01-26 Julia Koval <julia.koval@intel.com>
6164 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
6165 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
6166 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
6167 _mm_mask_bitshuffle_epi64_mask): Fix type.
6168 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
6169 USI_FTYPE_V4DI_V4DI_USI): Remove.
6170 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
6171 __builtin_ia32_vpshufbitqmb256_mask,
6172 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
6173 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
6174 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
6176 2018-01-26 Alan Modra <amodra@gmail.com>
6179 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
6180 UNSPEC_VBPERMQ. Sort other unspecs.
6182 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
6184 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
6186 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
6189 * predict.c (drop_profile): Do not push/pop cfun; update also
6191 (handle_missing_profiles): Fix logic looking for zero profiles.
6193 2018-01-25 Jakub Jelinek <jakub@redhat.com>
6196 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
6197 on functions with #pragma omp declare simd or functions with simd
6199 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
6200 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
6201 Remove trailing \n from warning_at calls.
6203 2018-01-25 Tom de Vries <tom@codesourcery.com>
6206 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6207 for neutered workers.
6209 2018-01-24 Joseph Myers <joseph@codesourcery.com>
6212 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
6213 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
6215 2018-01-24 Jeff Law <law@redhat.com>
6218 * i386.c (get_probe_interval): Move to earlier point.
6219 (ix86_compute_frame_layout): If -fstack-clash-protection and
6220 the frame is larger than the probe interval, then use pushes
6221 to save registers rather than reg->mem moves.
6222 (ix86_expand_prologue): Remove conditional for int_registers_saved
6225 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
6228 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
6229 min/max for never referenced object.
6231 2018-01-24 Jakub Jelinek <jakub@redhat.com>
6234 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
6236 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
6237 attributes from DECL_ATTRIBUTES (decl) without affecting
6238 DECL_ATTRIBUTES (current_function_decl).
6239 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
6240 functions with non-NULL DECL_ABSTRACT_ORIGIN.
6242 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
6244 PR tree-optimization/83979
6245 * fold-const.c (fold_comparison): Use constant_boolean_node
6246 instead of boolean_{true,false}_node.
6248 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
6250 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
6253 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6255 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6256 Simplify the clause that sets the length attribute.
6257 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6258 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
6259 clause that sets the length attribute.
6260 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6262 2018-01-24 Tom de Vries <tom@codesourcery.com>
6265 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
6266 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
6267 Add strict parameter.
6268 (prevent_branch_around_nothing): Insert dummy insn between branch to
6269 label and label with no ptx insn inbetween.
6270 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
6272 2018-01-24 Tom de Vries <tom@codesourcery.com>
6275 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6276 for neutered threads in warp.
6277 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
6279 2018-01-24 Richard Biener <rguenther@suse.de>
6281 PR tree-optimization/83176
6282 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
6285 2018-01-24 Richard Biener <rguenther@suse.de>
6287 PR tree-optimization/82819
6288 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
6289 code generating pluses that are no-ops in the target precision.
6291 2018-01-24 Richard Biener <rguenther@suse.de>
6294 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
6296 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6298 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
6299 to merge probabilities.
6300 * predict.c (probably_never_executed): Also mark as cold functions
6301 with global 0 profile and guessed local profile.
6302 * profile-count.c (profile_probability::combine_with_count): New
6304 * profile-count.h (profile_probability::operator*,
6305 profile_probability::operator*=, profile_probability::operator/,
6306 profile_probability::operator/=): Reduce precision to adjusted
6307 and set value to guessed on contradictory divisions.
6308 (profile_probability::combine_with_freq): Remove.
6309 (profile_probability::combine_wiht_count): Declare.
6310 (profile_count::force_nonzero):: Set to adjusted.
6311 (profile_count::probability_in):: Set quality to adjusted.
6312 * tree-ssa-tail-merge.c (replace_block_by): Use
6315 2018-01-23 Andrew Waterman <andrew@sifive.com>
6316 Jim Wilson <jimw@sifive.com>
6318 * config/riscv/riscv.c (riscv_stack_boundary): New.
6319 (riscv_option_override): Set riscv_stack_boundary. Handle
6320 riscv_preferred_stack_boundary_arg.
6321 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
6322 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
6323 (STACK_BOUNDARY): Set to riscv_stack_boundary.
6324 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
6325 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
6326 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
6328 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
6331 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
6332 of struct ix86_frame.
6333 (ix86_expand_epilogue): Likewise. Add a local variable for
6334 the reg_save_offset field in struct ix86_frame.
6336 2018-01-23 Bin Cheng <bin.cheng@arm.com>
6338 PR tree-optimization/82604
6339 * tree-loop-distribution.c (enum partition_kind): New enum item
6340 PKIND_PARTIAL_MEMSET.
6341 (partition_builtin_p): Support above new enum item.
6342 (generate_code_for_partition): Ditto.
6343 (compute_access_range): Differentiate cases that equality can be
6344 proven at all loops, the innermost loops or no loops.
6345 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
6346 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
6347 (finalize_partitions, distribute_loop): Don't fuse partition of
6348 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
6349 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
6352 2018-01-23 Martin Liska <mliska@suse.cz>
6354 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
6355 order to ignore the predictor.
6356 (PRED_POLYMORPHIC_CALL): Likewise.
6357 (PRED_RECURSIVE_CALL): Likewise.
6359 2018-01-23 Martin Liska <mliska@suse.cz>
6361 * tree-profile.c (tree_profiling): Print function header to
6362 aware reader which function we are working on.
6363 * value-prof.c (gimple_find_values_to_profile): Do not print
6364 not interesting value histograms.
6366 2018-01-23 Martin Liska <mliska@suse.cz>
6368 * profile-count.h (enum profile_quality): Add
6369 profile_uninitialized as the first value. Do not number values
6370 as they are zero based.
6371 (profile_count::verify): Update sanity check.
6372 (profile_probability::verify): Likewise.
6374 2018-01-23 Nathan Sidwell <nathan@acm.org>
6376 * doc/invoke.texi (ffor-scope): Deprecate.
6378 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6380 PR tree-optimization/83510
6381 * domwalk.c (set_all_edges_as_executable): New function.
6382 (dom_walker::dom_walker): Convert bool param
6383 "skip_unreachable_blocks" to enum reachability. Move setup of
6384 edge flags to set_all_edges_as_executable and only do it when
6385 reachability is REACHABLE_BLOCKS.
6386 * domwalk.h (enum dom_walker::reachability): New enum.
6387 (dom_walker::dom_walker): Convert bool param
6388 "skip_unreachable_blocks" to enum reachability.
6389 (set_all_edges_as_executable): New decl.
6390 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
6391 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
6393 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
6394 but converting true to REACHABLE_BLOCKS.
6395 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
6397 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
6398 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
6399 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
6401 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
6402 if check_all_array_refs will be called.
6404 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6406 * tree.c (selftest::test_location_wrappers): Add more test
6409 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6411 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
6412 (selftest::test_bit_in_range): Likewise.
6414 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6417 * doc/sourcebuild.texi (vect_float): Say that the selector
6418 only describes the situation when -funsafe-math-optimizations is on.
6419 (vect_float_strict): Document.
6421 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6423 PR tree-optimization/83965
6424 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
6425 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
6426 instead of checking only for a reduction.
6427 (vect_recog_widen_sum_pattern): Likewise.
6429 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6431 * predict.c (probably_never_executed): Only use precise profile info.
6432 (compute_function_frequency): Skip after inlining hack since we now
6433 have quality checking.
6435 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6437 * profile-count.h (profile_probability::very_unlikely,
6438 profile_probability::unlikely, profile_probability::even): Set
6439 precision to guessed.
6441 2018-01-23 Richard Biener <rguenther@suse.de>
6443 PR tree-optimization/83963
6444 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
6445 Properly terminate dominator walk when crossing the exit edge not
6446 when visiting its source block.
6448 2018-01-23 Jakub Jelinek <jakub@redhat.com>
6451 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
6452 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
6454 2018-01-22 Jakub Jelinek <jakub@redhat.com>
6456 PR tree-optimization/83957
6457 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
6458 semicolon after for body surrounded by braces.
6460 PR tree-optimization/83081
6461 * profile-count.h (profile_probability::split): New method.
6462 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
6463 Use profile_probability::split.
6464 (do_compare_rtx_and_jump): Fix adjustment of probabilities
6465 when splitting a single conditional jump into 2.
6467 2018-01-22 David Malcolm <dmalcolm@redhat.com>
6469 PR tree-optimization/69452
6470 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
6473 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6475 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
6476 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
6477 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
6479 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6481 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
6482 * config/rl78/rl78.md (movdi): New define_expand.
6483 * config/rl78/rl78.c (rl78_split_movdi): New function.
6485 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
6488 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
6490 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
6491 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
6492 128-bit to produce an UNSPEC move to get the double word with the
6493 signbit and then a shift directly to do signbit.
6494 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
6495 implementation with a new version that just does either a direct
6496 move or a regular move. Move memory interface to separate insns.
6497 Move insns so they are next to the expander.
6498 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
6499 with signbit move. Split big and little endian case.
6500 (signbit<mode>2_dm_mem_le): Likewise.
6501 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
6502 (signbit<mode>2_dm2): Likewise.
6504 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6506 * config/rl78/rl78.md (anddi3): New define_expand.
6508 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6510 * config/rl78/rl78.md (umindi3): New define_expand.
6512 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6514 * config/rl78/rl78.md (smindi3): New define_expand.
6516 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6518 * config/rl78/rl78.md (smaxdi3): New define_expand.
6520 2018-01-22 Carl Love <cel@us.ibm.com>
6522 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
6523 LVX_V1TI): Add macro expansion.
6524 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
6525 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
6526 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
6527 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
6528 Change check to determine if the instruction is a byte reversing
6529 entry. Fix typo in comment.
6530 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
6531 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
6532 Add def_builtin calls for new builtins.
6533 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
6534 Add define_insn expansion.
6536 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6538 * config/rl78/rl78.md (umaxdi3): New define_expand.
6540 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6542 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
6543 for non-QImode registers.
6545 2018-01-22 Richard Biener <rguenther@suse.de>
6547 PR tree-optimization/83963
6548 * graphite-scop-detection.c (scop_detection::get_sese): Delay
6549 including the loop exit block.
6550 (scop_detection::merge_sese): Likewise.
6551 (scop_detection::add_scop): Do it here instead.
6553 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6555 * doc/sourcebuild.texi (arm_softfloat): Document.
6557 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
6560 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
6561 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
6562 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
6564 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6565 David Edelsohn <dje.gcc@gmail.com>
6568 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6569 Change "crset eq" to "crset 2".
6570 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6571 (*call_indirect_aix<mode>_nospec): Likewise.
6572 (*call_value_indirect_aix<mode>_nospec): Likewise.
6573 (*call_indirect_elfv2<mode>_nospec): Likewise.
6574 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
6575 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
6576 change assembly output from . to $.
6577 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6578 (indirect_jump<mode>_nospec): Change assembly output from . to $.
6579 (*tablejump<mode>_internal1_nospec): Likewise.
6581 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
6584 * config/sh/sh_optimize_sett_clrt.cc:
6585 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
6587 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
6589 PR tree-optimization/83940
6590 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
6591 offset_dt to vect_constant_def rather than vect_unknown_def_type.
6592 (vect_check_load_store_mask): Add a mask_dt_out parameter and
6593 use it to pass back the definition type.
6594 (vect_check_store_rhs): Likewise rhs_dt_out.
6595 (vect_build_gather_load_calls): Add a mask_dt argument and use
6596 it instead of a call to vect_is_simple_use.
6597 (vectorizable_store): Update calls to vect_check_load_store_mask
6598 and vect_check_store_rhs. Use the dt returned by the latter instead
6599 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
6600 instead of calls to vect_is_simple_use. Pass the scalar rather
6601 than the vector operand to vect_is_simple_use when handling
6602 second and subsequent copies of an rhs value.
6603 (vectorizable_load): Update calls to vect_check_load_store_mask
6604 and vect_build_gather_load_calls. Use the cached mask_dt and
6605 gs_info.offset_dt instead of calls to vect_is_simple_use.
6607 2018-01-20 Jakub Jelinek <jakub@redhat.com>
6610 * tree-emutls.c: Include gimplify.h.
6611 (lower_emutls_2): New function.
6612 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
6613 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
6614 it before further processing.
6617 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
6618 UINTVAL (trueop1) instead of INTVAL (op1).
6620 2018-01-19 Jakub Jelinek <jakub@redhat.com>
6624 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
6625 INCOMING_FRAME_SP_OFFSET if not defined.
6626 (scan_trace): Add ENTRY argument. If true and
6627 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
6628 emit a note to adjust the CFA offset.
6629 (create_cfi_notes): Adjust scan_trace callers.
6630 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
6631 INCOMING_FRAME_SP_OFFSET in the CIE.
6632 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
6633 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
6635 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
6636 * doc/tm.texi: Regenerated.
6638 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6640 PR rtl-optimization/83147
6641 * lra-constraints.c (remove_inheritance_pseudos): Use
6642 lra_substitute_pseudo_within_insn.
6644 2018-01-19 Tom de Vries <tom@codesourcery.com>
6645 Cesar Philippidis <cesar@codesourcery.com>
6648 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
6650 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
6653 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
6654 spaces for function labels.
6656 2018-01-19 Martin Liska <mliska@suse.cz>
6658 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
6659 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
6660 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
6661 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
6662 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
6663 (PRED_CONST_RETURN): Change from 69 to 65.
6664 (PRED_NULL_RETURN): Change from 91 to 71.
6665 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
6666 (PRED_LOOP_GUARD): Change from 66 to 73.
6668 2018-01-19 Martin Liska <mliska@suse.cz>
6670 * predict.c (predict_insn_def): Add new assert.
6671 (struct branch_predictor): Change type to signed integer.
6672 (test_prediction_value_range): Amend test to cover
6674 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
6675 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
6676 (PRED_LOOP_ITERATIONS_MAX): Likewise.
6677 (PRED_LOOP_IV_COMPARE): Likewise.
6678 * predict.h (PROB_UNINITIALIZED): Define new constant.
6680 2018-01-19 Martin Liska <mliska@suse.cz>
6682 * predict.c (dump_prediction): Add new format for
6683 analyze_brprob.py script which is enabled with -details
6685 * profile-count.h (precise_p): New function.
6687 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6689 PR tree-optimization/83922
6690 * tree-vect-loop.c (vect_verify_full_masking): Return false if
6691 there are no statements that need masking.
6692 (vect_active_double_reduction_p): New function.
6693 (vect_analyze_loop_operations): Use it when handling phis that
6694 are not in the loop header.
6696 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6698 PR tree-optimization/83914
6699 * tree-vect-loop.c (vectorizable_induction): Don't convert
6700 init_expr or apply the peeling adjustment for inductions
6701 that are nested within the vectorized loop.
6703 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6705 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
6708 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6712 * function.h (gimplify_parameters): Add gimple_seq * argument.
6713 * function.c: Include gimple.h and options.h.
6714 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
6715 for the added local temporaries if needed.
6716 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
6717 if there are any parameter cleanups, wrap whole body into a
6718 try/finally with the cleanups.
6720 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
6723 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
6724 Use GET_MODE_CLASS for scalar floating point.
6726 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6730 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
6731 Fix call of call_cgraph_insertion_hooks.
6733 2018-01-18 Martin Sebor <msebor@redhat.com>
6735 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
6737 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6740 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
6743 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
6746 * common.opt: (-ffile-prefix-map): New option.
6747 * opts.c (common_handle_option): Defer it.
6748 * opts-global.c (handle_common_deferred_options): Handle it.
6749 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
6750 * file-prefix-map.h: New file.
6751 (remap_debug_filename, add_debug_prefix_map): ...here.
6752 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
6753 * final.c (debug_prefix_map, add_debug_prefix_map
6754 remap_debug_filename): Move to...
6755 * file-prefix-map.c: New file.
6756 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
6757 generalize, get rid of alloca(), use strrchr() instead of strchr().
6758 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
6759 Implement in terms of add_prefix_map().
6760 (remap_macro_filename, remap_debug_filename): Implement in term of
6762 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
6763 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
6764 * dbxout.c: Include file-prefix-map.h.
6765 * varasm.c: Likewise.
6766 * vmsdbgout.c: Likewise.
6767 * xcoffout.c: Likewise.
6768 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
6769 * doc/cppopts.texi (-fmacro-prefix-map): Document.
6770 * doc/invoke.texi (-ffile-prefix-map): Document.
6771 (-fdebug-prefix-map): Update description.
6773 2018-01-18 Martin Liska <mliska@suse.cz>
6775 * config/i386/i386.c (indirect_thunk_name): Document that also
6777 (output_indirect_thunk): Document why both instructions
6778 (pause and lfence) are generated.
6780 2018-01-18 Richard Biener <rguenther@suse.de>
6782 PR tree-optimization/83887
6783 * graphite-scop-detection.c
6784 (scop_detection::get_nearest_dom_with_single_entry): Remove.
6785 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
6786 (scop_detection::merge_sese): Re-implement with a flood-fill
6787 algorithm that properly finds a SESE region if it exists.
6789 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6792 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
6793 pointer_diff optimizations use view_convert instead of convert.
6795 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6797 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6798 Generate different code for -mno-speculate-indirect-jumps.
6799 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6800 (*call_indirect_aix<mode>): Disable for
6801 -mno-speculate-indirect-jumps.
6802 (*call_indirect_aix<mode>_nospec): New define_insn.
6803 (*call_value_indirect_aix<mode>): Disable for
6804 -mno-speculate-indirect-jumps.
6805 (*call_value_indirect_aix<mode>_nospec): New define_insn.
6806 (*sibcall_nonlocal_sysv<mode>): Generate different code for
6807 -mno-speculate-indirect-jumps.
6808 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6810 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6812 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
6813 long double type, set the flags for noting the default long double
6814 type, even if we don't pass or return a long double type.
6816 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
6819 * ipa-inline.c (flatten_function): Do not overwrite final inlining
6822 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
6824 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
6825 support for merge[hl].
6826 (fold_mergehl_helper): New helper function.
6827 (tree-vector-builder.h): New #include for tree_vector_builder usage.
6828 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
6829 (altivec_vmrglw_direct): Add xxmrglw insn.
6831 2018-01-17 Andrew Waterman <andrew@sifive.com>
6833 * config/riscv/riscv.c (riscv_conditional_register_usage): If
6834 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
6836 2018-01-17 David Malcolm <dmalcolm@redhat.com>
6839 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
6840 call the lto_location_cache before reading the
6841 DECL_SOURCE_LOCATION of the types.
6843 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
6844 Richard Sandiford <richard.sandiford@linaro.org>
6846 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
6847 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
6848 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
6849 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
6850 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
6852 * config/aarch64/constraints.md (aarch64_movti_operand):
6854 * config/aarch64/predicates.md (Uti): Add new constraint.
6856 2018-01-17 Carl Love <cel@us.ibm.com>
6858 * config/rs6000/vsx.md (define_expand xl_len_r,
6859 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
6860 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
6862 (define_expand, define_insn): Move the shift left from the
6863 define_insn to the define_expand for lxvl and stxvl instructions.
6864 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
6865 and XL_LEN_R definitions to PURE.
6867 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
6869 * config/i386/i386.c (indirect_thunk_name): Declare regno
6870 as unsigned int. Compare regno with INVALID_REGNUM.
6871 (output_indirect_thunk): Ditto.
6872 (output_indirect_thunk_function): Ditto.
6873 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
6874 in the call to output_indirect_thunk_function.
6876 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
6879 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
6880 rather than the size of inner_type to determine the stack slot size
6881 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
6883 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
6886 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
6889 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
6891 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
6892 endian Linux systems to optionally enable multilibs for selecting
6893 the long double type if the user configured an explicit type.
6894 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
6895 have no long double multilibs if not defined.
6896 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
6897 warn if the user used -mabi={ieee,ibm}longdouble and we built
6898 multilibs for long double.
6899 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
6900 appropriate multilib option.
6901 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
6903 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
6904 for building long double multilibs.
6905 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
6907 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
6909 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
6912 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
6914 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
6917 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
6920 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
6923 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6925 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
6926 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
6928 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
6930 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
6931 different rtl trees depending on TARGET_64BIT.
6932 (rs6000_gen_lvx): Likewise.
6934 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6936 * config/visium/visium.md (nop): Tweak comment.
6937 (hazard_nop): Likewise.
6939 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6941 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
6942 -mspeculate-indirect-jumps.
6943 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
6944 for -mno-speculate-indirect-jumps.
6945 (*call_indirect_elfv2<mode>_nospec): New define_insn.
6946 (*call_value_indirect_elfv2<mode>): Disable for
6947 -mno-speculate-indirect-jumps.
6948 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
6949 (indirect_jump): Emit different RTL for
6950 -mno-speculate-indirect-jumps.
6951 (*indirect_jump<mode>): Disable for
6952 -mno-speculate-indirect-jumps.
6953 (*indirect_jump<mode>_nospec): New define_insn.
6954 (tablejump): Emit different RTL for
6955 -mno-speculate-indirect-jumps.
6956 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
6957 (tablejumpsi_nospec): New define_expand.
6958 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
6959 (tablejumpdi_nospec): New define_expand.
6960 (*tablejump<mode>_internal1): Disable for
6961 -mno-speculate-indirect-jumps.
6962 (*tablejump<mode>_internal1_nospec): New define_insn.
6963 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
6966 2018-01-16 Artyom Skrobov tyomitch@gmail.com
6968 * caller-save.c (insert_save): Drop unnecessary parameter. All
6971 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6972 Richard Biener <rguenth@suse.de>
6975 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
6976 return early, inline manually is_gimple_sizepos. Make sure if we
6977 call gimplify_expr we don't end up with a gimple constant.
6978 * tree.c (variably_modified_type_p): Don't return true for
6979 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
6980 * gimplify.h (is_gimple_sizepos): Remove.
6982 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6984 PR tree-optimization/83857
6985 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
6986 vectorizable_live_operation for pure SLP statements.
6987 (vectorizable_live_operation): Handle PHIs.
6989 2018-01-16 Richard Biener <rguenther@suse.de>
6991 PR tree-optimization/83867
6992 * tree-vect-stmts.c (vect_transform_stmt): Precompute
6993 nested_in_vect_loop_p since the scalar stmt may get invalidated.
6995 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6998 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
6999 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
7000 If off is not INTEGER_CST, issue a may not be aligned warning
7001 rather than isn't aligned. Use isn%'t rather than isn't.
7002 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
7004 <case MULT_EXPR>: Improve the case when bottom and one of the
7005 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
7006 operand, in that case check if the other operand is multiple of
7007 bottom divided by the INTEGER_CST operand.
7009 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7012 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
7013 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
7014 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
7015 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
7016 * config/pa/pa.c (pa_function_arg_advance): Likewise.
7017 (pa_function_arg, pa_arg_partial_bytes): Likewise.
7018 (pa_function_arg_size): New function.
7020 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7022 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
7023 in a separate statement.
7025 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7027 PR tree-optimization/83847
7028 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
7029 group gathers and scatters.
7031 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7033 PR rtl-optimization/86620
7034 * params.def (max-sched-ready-insns): Bump minimum value to 1.
7036 PR rtl-optimization/83213
7037 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
7038 to last if both are JUMP_INSNs.
7040 PR tree-optimization/83843
7041 * gimple-ssa-store-merging.c
7042 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
7043 store_immediate_info for bswap/nop orig_stores.
7045 2018-01-15 Andrew Waterman <andrew@sifive.com>
7047 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
7049 <UDIV>: Increase cost if !TARGET_DIV.
7051 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
7053 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
7054 (define_attr "cr_logical_3op"): New.
7055 (cceq_ior_compare): Adjust.
7056 (cceq_ior_compare_complement): Adjust.
7057 (*cceq_rev_compare): Adjust.
7058 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
7059 (is_cracked_insn): Adjust.
7060 (insn_must_be_first_in_group): Adjust.
7061 * config/rs6000/40x.md: Adjust.
7062 * config/rs6000/440.md: Adjust.
7063 * config/rs6000/476.md: Adjust.
7064 * config/rs6000/601.md: Adjust.
7065 * config/rs6000/603.md: Adjust.
7066 * config/rs6000/6xx.md: Adjust.
7067 * config/rs6000/7450.md: Adjust.
7068 * config/rs6000/7xx.md: Adjust.
7069 * config/rs6000/8540.md: Adjust.
7070 * config/rs6000/cell.md: Adjust.
7071 * config/rs6000/e300c2c3.md: Adjust.
7072 * config/rs6000/e500mc.md: Adjust.
7073 * config/rs6000/e500mc64.md: Adjust.
7074 * config/rs6000/e5500.md: Adjust.
7075 * config/rs6000/e6500.md: Adjust.
7076 * config/rs6000/mpc.md: Adjust.
7077 * config/rs6000/power4.md: Adjust.
7078 * config/rs6000/power5.md: Adjust.
7079 * config/rs6000/power6.md: Adjust.
7080 * config/rs6000/power7.md: Adjust.
7081 * config/rs6000/power8.md: Adjust.
7082 * config/rs6000/power9.md: Adjust.
7083 * config/rs6000/rs64.md: Adjust.
7084 * config/rs6000/titan.md: Adjust.
7086 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7088 * config/i386/predicates.md (indirect_branch_operand): Rewrite
7089 ix86_indirect_branch_register logic.
7091 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7093 * config/i386/constraints.md (Bs): Update
7094 ix86_indirect_branch_register check. Don't check
7095 ix86_indirect_branch_register with GOT_memory_operand.
7097 * config/i386/predicates.md (GOT_memory_operand): Don't check
7098 ix86_indirect_branch_register here.
7099 (GOT32_symbol_operand): Likewise.
7101 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7103 * config/i386/predicates.md (constant_call_address_operand):
7104 Rewrite ix86_indirect_branch_register logic.
7105 (sibcall_insn_operand): Likewise.
7107 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7109 * config/i386/constraints.md (Bs): Replace
7110 ix86_indirect_branch_thunk_register with
7111 ix86_indirect_branch_register.
7113 * config/i386/i386.md (indirect_jump): Likewise.
7114 (tablejump): Likewise.
7115 (*sibcall_memory): Likewise.
7116 (*sibcall_value_memory): Likewise.
7117 Peepholes of indirect call and jump via memory: Likewise.
7118 * config/i386/i386.opt: Likewise.
7119 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7120 (GOT_memory_operand): Likewise.
7121 (call_insn_operand): Likewise.
7122 (sibcall_insn_operand): Likewise.
7123 (GOT32_symbol_operand): Likewise.
7125 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7128 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
7129 type rather than type addr's type points to.
7130 (expand_omp_atomic_mutex): Likewise.
7131 (expand_omp_atomic): Likewise.
7133 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7136 * config/i386/i386.c (output_indirect_thunk_function): Use
7137 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
7138 for __x86_return_thunk.
7140 2018-01-15 Richard Biener <rguenther@suse.de>
7143 * expmed.c (extract_bit_field_1): Fix typo.
7145 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7148 * config/arm/iterators.md (VF): New mode iterator.
7149 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
7150 Remove integer-related logic from pattern.
7151 (neon_vabd<mode>_3): Likewise.
7153 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7156 * common.opt (fstrict-overflow): No longer an alias.
7157 (fwrapv-pointer): New option.
7158 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
7159 also for pointer types based on flag_wrapv_pointer.
7160 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
7161 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
7162 opts->x_flag_wrapv got set.
7163 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
7164 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
7165 POINTER_TYPE_OVERFLOW_UNDEFINED.
7166 * match.pd: Likewise in address comparison pattern.
7167 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
7169 2018-01-15 Richard Biener <rguenther@suse.de>
7172 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
7173 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
7174 Reset type names to their identifier if their TYPE_DECL doesn't
7175 have linkage (and thus is used for ODR and devirt).
7176 (save_debug_info_for_decl): Remove.
7177 (save_debug_info_for_type): Likewise.
7178 (add_tree_to_fld_list): Adjust.
7179 * tree-pretty-print.c (dump_generic_node): Make dumping of
7180 type names more robust.
7182 2018-01-15 Richard Biener <rguenther@suse.de>
7184 * BASE-VER: Bump to 8.0.1.
7186 2018-01-14 Martin Sebor <msebor@redhat.com>
7189 * builtins.c (check_access): Avoid warning when the no-warning bit
7192 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
7194 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
7195 * ira-color (allocno_hard_regs_compare): Likewise.
7197 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
7200 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
7201 Use .pushsection/.popsection.
7203 2018-01-14 Martin Sebor <msebor@redhat.com>
7206 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
7208 2018-01-14 Jakub Jelinek <jakub@redhat.com>
7210 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
7211 entry from extra_headers.
7212 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
7213 extra_headers, make the list bitwise identical to the i?86-*-* one.
7215 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7217 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
7218 -mcmodel=large with -mindirect-branch=thunk,
7219 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
7220 -mfunction-return=thunk-extern.
7221 * doc/invoke.texi: Document -mcmodel=large is incompatible with
7222 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
7223 -mfunction-return=thunk and -mfunction-return=thunk-extern.
7225 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7227 * config/i386/i386.c (print_reg): Print the name of the full
7228 integer register without '%'.
7229 (ix86_print_operand): Handle 'V'.
7230 * doc/extend.texi: Document 'V' modifier.
7232 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7234 * config/i386/constraints.md (Bs): Disallow memory operand for
7235 -mindirect-branch-register.
7237 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7238 (GOT_memory_operand): Likewise.
7239 (call_insn_operand): Likewise.
7240 (sibcall_insn_operand): Likewise.
7241 (GOT32_symbol_operand): Likewise.
7242 * config/i386/i386.md (indirect_jump): Call convert_memory_address
7243 for -mindirect-branch-register.
7244 (tablejump): Likewise.
7245 (*sibcall_memory): Likewise.
7246 (*sibcall_value_memory): Likewise.
7247 Disallow peepholes of indirect call and jump via memory for
7248 -mindirect-branch-register.
7249 (*call_pop): Replace m with Bw.
7250 (*call_value_pop): Likewise.
7251 (*sibcall_pop_memory): Replace m with Bs.
7252 * config/i386/i386.opt (mindirect-branch-register): New option.
7253 * doc/invoke.texi: Document -mindirect-branch-register option.
7255 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7257 * config/i386/i386-protos.h (ix86_output_function_return): New.
7258 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
7259 set function_return_type.
7260 (indirect_thunk_name): Add ret_p to indicate thunk for function
7262 (output_indirect_thunk_function): Pass false to
7263 indirect_thunk_name.
7264 (ix86_output_indirect_branch_via_reg): Likewise.
7265 (ix86_output_indirect_branch_via_push): Likewise.
7266 (output_indirect_thunk_function): Create alias for function
7267 return thunk if regno < 0.
7268 (ix86_output_function_return): New function.
7269 (ix86_handle_fndecl_attribute): Handle function_return.
7270 (ix86_attribute_table): Add function_return.
7271 * config/i386/i386.h (machine_function): Add
7272 function_return_type.
7273 * config/i386/i386.md (simple_return_internal): Use
7274 ix86_output_function_return.
7275 (simple_return_internal_long): Likewise.
7276 * config/i386/i386.opt (mfunction-return=): New option.
7277 (indirect_branch): Mention -mfunction-return=.
7278 * doc/extend.texi: Document function_return function attribute.
7279 * doc/invoke.texi: Document -mfunction-return= option.
7281 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7283 * config/i386/i386-opts.h (indirect_branch): New.
7284 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
7285 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
7286 with local indirect jump when converting indirect call and jump.
7287 (ix86_set_indirect_branch_type): New.
7288 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
7289 (indirectlabelno): New.
7290 (indirect_thunk_needed): Likewise.
7291 (indirect_thunk_bnd_needed): Likewise.
7292 (indirect_thunks_used): Likewise.
7293 (indirect_thunks_bnd_used): Likewise.
7294 (INDIRECT_LABEL): Likewise.
7295 (indirect_thunk_name): Likewise.
7296 (output_indirect_thunk): Likewise.
7297 (output_indirect_thunk_function): Likewise.
7298 (ix86_output_indirect_branch_via_reg): Likewise.
7299 (ix86_output_indirect_branch_via_push): Likewise.
7300 (ix86_output_indirect_branch): Likewise.
7301 (ix86_output_indirect_jmp): Likewise.
7302 (ix86_code_end): Call output_indirect_thunk_function if needed.
7303 (ix86_output_call_insn): Call ix86_output_indirect_branch if
7305 (ix86_handle_fndecl_attribute): Handle indirect_branch.
7306 (ix86_attribute_table): Add indirect_branch.
7307 * config/i386/i386.h (machine_function): Add indirect_branch_type
7308 and has_local_indirect_jump.
7309 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
7311 (tablejump): Likewise.
7312 (*indirect_jump): Use ix86_output_indirect_jmp.
7313 (*tablejump_1): Likewise.
7314 (simple_return_indirect_internal): Likewise.
7315 * config/i386/i386.opt (mindirect-branch=): New option.
7316 (indirect_branch): New.
7319 (thunk-inline): Likewise.
7320 (thunk-extern): Likewise.
7321 * doc/extend.texi: Document indirect_branch function attribute.
7322 * doc/invoke.texi: Document -mindirect-branch= option.
7324 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
7327 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
7329 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
7331 * ipa-inline.c (want_inline_small_function_p): Return false if
7332 inlining has already failed with CIF_FINAL_ERROR.
7333 (update_caller_keys): Call want_inline_small_function_p before
7335 (update_callee_keys): Likewise.
7337 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7339 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
7341 (rs6000_quadword_masked_address_p): Likewise.
7342 (quad_aligned_load_p): Likewise.
7343 (quad_aligned_store_p): Likewise.
7344 (const_load_sequence_p): Add comment to describe the outer-most loop.
7345 (mimic_memory_attributes_and_flags): New function.
7346 (rs6000_gen_stvx): Likewise.
7347 (replace_swapped_aligned_store): Likewise.
7348 (rs6000_gen_lvx): Likewise.
7349 (replace_swapped_aligned_load): Likewise.
7350 (replace_swapped_load_constant): Capitalize argument name in
7351 comment describing this function.
7352 (rs6000_analyze_swaps): Add a third pass to search for vector loads
7353 and stores that access quad-word aligned addresses and replace
7354 with stvx or lvx instructions when appropriate.
7355 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
7356 New function prototype.
7357 (rs6000_quadword_masked_address_p): Likewise.
7358 (rs6000_gen_lvx): Likewise.
7359 (rs6000_gen_stvx): Likewise.
7360 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
7361 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
7362 when memory address is aligned.
7363 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
7364 this split to select lvx instruction when memory address is aligned.
7365 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
7366 instruction when memory address is aligned.
7367 (*vsx_le_perm_load_v16qi): Likewise.
7368 (four unnamed splitters): Modify to select the stvx instruction
7369 when memory is aligned.
7371 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
7373 * predict.c (determine_unlikely_bbs): Handle correctly BBs
7374 which appears in the queue multiple times.
7376 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7377 Alan Hayward <alan.hayward@arm.com>
7378 David Sherwood <david.sherwood@arm.com>
7380 * tree-vectorizer.h (vec_lower_bound): New structure.
7381 (_loop_vec_info): Add check_nonzero and lower_bounds.
7382 (LOOP_VINFO_CHECK_NONZERO): New macro.
7383 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
7384 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
7385 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
7386 fields. Make seg_len the distance travelled, not including the
7388 (dr_direction_indicator): Declare.
7389 (dr_zero_step_indicator): Likewise.
7390 (dr_known_forward_stride_p): Likewise.
7391 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
7393 (runtime_alias_check_p): Allow runtime alias checks with
7395 (operator ==): Compare access_size and align.
7396 (prune_runtime_alias_test_list): Rework for new distinction between
7397 the access_size and seg_len.
7398 (create_intersect_range_checks_index): Likewise. Cope with polynomial
7400 (get_segment_min_max): New function.
7401 (create_intersect_range_checks): Use it.
7402 (dr_step_indicator): New function.
7403 (dr_direction_indicator): Likewise.
7404 (dr_zero_step_indicator): Likewise.
7405 (dr_known_forward_stride_p): Likewise.
7406 * tree-loop-distribution.c (data_ref_segment_size): Return
7407 DR_STEP * (niters - 1).
7408 (compute_alias_check_pairs): Update call to the dr_with_seg_len
7410 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
7411 (vect_preserves_scalar_order_p): New function, split out from...
7412 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
7413 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
7414 (vect_vfa_access_size): New function.
7415 (vect_vfa_align): Likewise.
7416 (vect_compile_time_alias): Take access_size_a and access_b arguments.
7417 (dump_lower_bound): New function.
7418 (vect_check_lower_bound): Likewise.
7419 (vect_small_gap_p): Likewise.
7420 (vectorizable_with_step_bound_p): Likewise.
7421 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
7422 depencies if the vectorization factor is 1. Convert the checks
7423 for nonzero steps into checks on the bounds of DR_STEP. Try using
7424 a bunds check for variable steps if the minimum required step is
7425 relatively small. Update calls to the dr_with_seg_len
7426 constructor and to vect_compile_time_alias.
7427 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
7429 (vect_loop_versioning): Call it.
7430 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
7432 (vect_estimate_min_profitable_iters): Account for any bounds checks.
7434 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7435 Alan Hayward <alan.hayward@arm.com>
7436 David Sherwood <david.sherwood@arm.com>
7438 * doc/sourcebuild.texi (vect_scatter_store): Document.
7439 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
7441 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
7443 * genopinit.c (main): Add supports_vec_scatter_store and
7444 supports_vec_scatter_store_cached to target_optabs.
7445 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
7446 IFN_MASK_SCATTER_STORE.
7447 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
7449 * internal-fn.h (internal_store_fn_p): Declare.
7450 (internal_fn_stored_value_index): Likewise.
7451 * internal-fn.c (scatter_store_direct): New macro.
7452 (expand_scatter_store_optab_fn): New function.
7453 (direct_scatter_store_optab_supported_p): New macro.
7454 (internal_store_fn_p): New function.
7455 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
7456 IFN_MASK_SCATTER_STORE.
7457 (internal_fn_mask_index): Likewise.
7458 (internal_fn_stored_value_index): New function.
7459 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
7461 * optabs-query.h (supports_vec_scatter_store_p): Declare.
7462 * optabs-query.c (supports_vec_scatter_store_p): New function.
7463 * tree-vectorizer.h (vect_get_store_rhs): Declare.
7464 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
7465 true for scatter stores.
7466 (vect_gather_scatter_fn_p): Handle scatter stores too.
7467 (vect_check_gather_scatter): Consider using scatter stores if
7468 supports_vec_scatter_store_p.
7469 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
7471 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7472 internal_fn_stored_value_index.
7473 (check_load_store_masking): Handle scatter stores too.
7474 (vect_get_store_rhs): Make public.
7475 (vectorizable_call): Use internal_store_fn_p.
7476 (vectorizable_store): Handle scatter store internal functions.
7477 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
7478 when deciding whether the end of the group has been reached.
7479 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
7480 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
7481 (mask_scatter_store<mode>): New insns.
7483 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7484 Alan Hayward <alan.hayward@arm.com>
7485 David Sherwood <david.sherwood@arm.com>
7487 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
7488 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
7489 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
7491 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
7492 Use vect_truncate_gather_scatter_offset if we can't treat the
7493 operation as a normal gather load or scatter store.
7494 (get_group_load_store_type): Take the gather_scatter_info
7495 as argument. Try using a gather load or scatter store for
7496 single-element groups.
7497 (get_load_store_type): Update calls to get_group_load_store_type
7498 and vect_use_strided_gather_scatters_p.
7500 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7501 Alan Hayward <alan.hayward@arm.com>
7502 David Sherwood <david.sherwood@arm.com>
7504 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
7505 optional tree argument.
7506 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
7508 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
7509 but continue to use the current value as a fallback.
7510 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
7511 to compare the updates.
7512 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
7513 (get_load_store_type): Use it when handling a strided access.
7514 (vect_get_strided_load_store_ops): New function.
7515 (vect_get_data_ptr_increment): Likewise.
7516 (vectorizable_load): Handle strided gather loads. Always pass
7517 a step to vect_create_data_ref_ptr and bump_vector_ptr.
7519 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7520 Alan Hayward <alan.hayward@arm.com>
7521 David Sherwood <david.sherwood@arm.com>
7523 * doc/md.texi (gather_load@var{m}): Document.
7524 (mask_gather_load@var{m}): Likewise.
7525 * genopinit.c (main): Add supports_vec_gather_load and
7526 supports_vec_gather_load_cached to target_optabs.
7527 * optabs-tree.c (init_tree_optimization_optabs): Use
7528 ggc_cleared_alloc to allocate target_optabs.
7529 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
7530 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
7532 * internal-fn.h (internal_load_fn_p): Declare.
7533 (internal_gather_scatter_fn_p): Likewise.
7534 (internal_fn_mask_index): Likewise.
7535 (internal_gather_scatter_fn_supported_p): Likewise.
7536 * internal-fn.c (gather_load_direct): New macro.
7537 (expand_gather_load_optab_fn): New function.
7538 (direct_gather_load_optab_supported_p): New macro.
7539 (direct_internal_fn_optab): New function.
7540 (internal_load_fn_p): Likewise.
7541 (internal_gather_scatter_fn_p): Likewise.
7542 (internal_fn_mask_index): Likewise.
7543 (internal_gather_scatter_fn_supported_p): Likewise.
7544 * optabs-query.c (supports_at_least_one_mode_p): New function.
7545 (supports_vec_gather_load_p): Likewise.
7546 * optabs-query.h (supports_vec_gather_load_p): Declare.
7547 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
7548 and memory_type field.
7549 (NUM_PATTERNS): Bump to 15.
7550 * tree-vect-data-refs.c: Include internal-fn.h.
7551 (vect_gather_scatter_fn_p): New function.
7552 (vect_describe_gather_scatter_call): Likewise.
7553 (vect_check_gather_scatter): Try using internal functions for
7554 gather loads. Recognize existing calls to a gather load function.
7555 (vect_analyze_data_refs): Consider using gather loads if
7556 supports_vec_gather_load_p.
7557 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
7558 (vect_get_gather_scatter_offset_type): Likewise.
7559 (vect_convert_mask_for_vectype): Likewise.
7560 (vect_add_conversion_to_patterm): Likewise.
7561 (vect_try_gather_scatter_pattern): Likewise.
7562 (vect_recog_gather_scatter_pattern): New pattern recognizer.
7563 (vect_vect_recog_func_ptrs): Add it.
7564 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7565 internal_fn_mask_index and internal_gather_scatter_fn_p.
7566 (check_load_store_masking): Take the gather_scatter_info as an
7567 argument and handle gather loads.
7568 (vect_get_gather_scatter_ops): New function.
7569 (vectorizable_call): Check internal_load_fn_p.
7570 (vectorizable_load): Likewise. Handle gather load internal
7572 (vectorizable_store): Update call to check_load_store_masking.
7573 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
7574 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
7575 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
7576 (aarch64_gather_scale_operand_d): New predicates.
7577 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
7578 (mask_gather_load<mode>): New insns.
7580 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7581 Alan Hayward <alan.hayward@arm.com>
7582 David Sherwood <david.sherwood@arm.com>
7584 * optabs.def (fold_left_plus_optab): New optab.
7585 * doc/md.texi (fold_left_plus_@var{m}): Document.
7586 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
7587 * internal-fn.c (fold_left_direct): Define.
7588 (expand_fold_left_optab_fn): Likewise.
7589 (direct_fold_left_optab_supported_p): Likewise.
7590 * fold-const-call.c (fold_const_fold_left): New function.
7591 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
7592 * tree-parloops.c (valid_reduction_p): New function.
7593 (gather_scalar_reductions): Use it.
7594 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
7595 (vect_finish_replace_stmt): Declare.
7596 * tree-vect-loop.c (fold_left_reduction_fn): New function.
7597 (needs_fold_left_reduction_p): New function, split out from...
7598 (vect_is_simple_reduction): ...here. Accept reductions that
7599 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
7600 (vect_force_simple_reduction): Also store the reduction type in
7601 the assignment's STMT_VINFO_REDUC_TYPE.
7602 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
7603 (merge_with_identity): New function.
7604 (vect_expand_fold_left): Likewise.
7605 (vectorize_fold_left_reduction): Likewise.
7606 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
7607 scalar phi in place for it. Check for target support and reject
7608 cases that would reassociate the operation. Defer the transform
7609 phase to vectorize_fold_left_reduction.
7610 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
7611 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
7612 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
7614 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7616 * tree-if-conv.c (predicate_mem_writes): Remove redundant
7617 call to ifc_temp_var.
7619 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7620 Alan Hayward <alan.hayward@arm.com>
7621 David Sherwood <david.sherwood@arm.com>
7623 * target.def (legitimize_address_displacement): Take the original
7624 offset as a poly_int.
7625 * targhooks.h (default_legitimize_address_displacement): Update
7627 * targhooks.c (default_legitimize_address_displacement): Likewise.
7628 * doc/tm.texi: Regenerate.
7629 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
7630 as an argument, moving assert of ad->disp == ad->disp_term to...
7631 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
7632 Try calling targetm.legitimize_address_displacement before expanding
7633 the address rather than afterwards, and adjust for the new interface.
7634 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
7635 Match the new hook interface. Handle SVE addresses.
7636 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
7639 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7641 * Makefile.in (OBJS): Add early-remat.o.
7642 * target.def (select_early_remat_modes): New hook.
7643 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
7644 * doc/tm.texi: Regenerate.
7645 * targhooks.h (default_select_early_remat_modes): Declare.
7646 * targhooks.c (default_select_early_remat_modes): New function.
7647 * timevar.def (TV_EARLY_REMAT): New timevar.
7648 * passes.def (pass_early_remat): New pass.
7649 * tree-pass.h (make_pass_early_remat): Declare.
7650 * early-remat.c: New file.
7651 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
7653 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
7655 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7656 Alan Hayward <alan.hayward@arm.com>
7657 David Sherwood <david.sherwood@arm.com>
7659 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
7660 vfm1 with a bound_epilog parameter.
7661 (vect_do_peeling): Update calls accordingly, and move the prologue
7662 call earlier in the function. Treat the base bound_epilog as 0 for
7663 fully-masked loops and retain vf - 1 for other loops. Add 1 to
7664 this base when peeling for gaps.
7665 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
7666 with fully-masked loops.
7667 (vect_estimate_min_profitable_iters): Handle the single peeled
7668 iteration in that case.
7670 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7671 Alan Hayward <alan.hayward@arm.com>
7672 David Sherwood <david.sherwood@arm.com>
7674 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
7675 single-element interleaving even if the size is not a power of 2.
7676 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
7677 accesses for single-element interleaving if the group size is
7680 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7681 Alan Hayward <alan.hayward@arm.com>
7682 David Sherwood <david.sherwood@arm.com>
7684 * doc/md.texi (fold_extract_last_@var{m}): Document.
7685 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
7686 * optabs.def (fold_extract_last_optab): New optab.
7687 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
7688 * internal-fn.c (fold_extract_direct): New macro.
7689 (expand_fold_extract_optab_fn): Likewise.
7690 (direct_fold_extract_optab_supported_p): Likewise.
7691 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
7692 * tree-vect-loop.c (vect_model_reduction_cost): Handle
7693 EXTRACT_LAST_REDUCTION.
7694 (get_initial_def_for_reduction): Do not create an initial vector
7695 for EXTRACT_LAST_REDUCTION reductions.
7696 (vectorizable_reduction): Leave the scalar phi in place for
7697 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
7698 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
7699 epilogue code for EXTRACT_LAST_REDUCTION and defer the
7700 transform phase to vectorizable_condition.
7701 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
7703 (vect_finish_stmt_generation): ...here.
7704 (vect_finish_replace_stmt): New function.
7705 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
7706 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
7708 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
7710 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7711 Alan Hayward <alan.hayward@arm.com>
7712 David Sherwood <david.sherwood@arm.com>
7714 * doc/md.texi (extract_last_@var{m}): Document.
7715 * optabs.def (extract_last_optab): New optab.
7716 * internal-fn.def (EXTRACT_LAST): New internal function.
7717 * internal-fn.c (cond_unary_direct): New macro.
7718 (expand_cond_unary_optab_fn): Likewise.
7719 (direct_cond_unary_optab_supported_p): Likewise.
7720 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
7721 loops using EXTRACT_LAST.
7722 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
7723 (extract_last_<mode>): ...this optab.
7724 (vec_extract<mode><Vel>): Update accordingly.
7726 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7727 Alan Hayward <alan.hayward@arm.com>
7728 David Sherwood <david.sherwood@arm.com>
7730 * target.def (empty_mask_is_expensive): New hook.
7731 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
7732 * doc/tm.texi: Regenerate.
7733 * targhooks.h (default_empty_mask_is_expensive): Declare.
7734 * targhooks.c (default_empty_mask_is_expensive): New function.
7735 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
7736 if the target says that empty masks are expensive.
7737 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
7739 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
7741 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7742 Alan Hayward <alan.hayward@arm.com>
7743 David Sherwood <david.sherwood@arm.com>
7745 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
7746 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
7747 (vect_use_loop_mask_for_alignment_p): New function.
7748 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
7749 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
7750 niters_skip argument. Make sure that the first niters_skip elements
7751 of the first iteration are inactive.
7752 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
7753 Update call to vect_set_loop_masks_directly.
7754 (get_misalign_in_elems): New function, split out from...
7755 (vect_gen_prolog_loop_niters): ...here.
7756 (vect_update_init_of_dr): Take a code argument that specifies whether
7757 the adjustment should be added or subtracted.
7758 (vect_update_init_of_drs): Likewise.
7759 (vect_prepare_for_masked_peels): New function.
7760 (vect_do_peeling): Skip prologue peeling if we're using a mask
7761 instead. Update call to vect_update_inits_of_drs.
7762 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7764 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
7765 alignment. Do not include the number of peeled iterations in
7766 the minimum threshold in that case.
7767 (vectorizable_induction): Adjust the start value down by
7768 LOOP_VINFO_MASK_SKIP_NITERS iterations.
7769 (vect_transform_loop): Call vect_prepare_for_masked_peels.
7770 Take the number of skipped iterations into account when calculating
7772 * tree-vect-stmts.c (vect_gen_while_not): New function.
7774 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7775 Alan Hayward <alan.hayward@arm.com>
7776 David Sherwood <david.sherwood@arm.com>
7778 * doc/sourcebuild.texi (vect_fully_masked): Document.
7779 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
7781 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
7783 (vect_analyze_loop_2): ...here. Don't check the vectorization
7784 factor against the number of loop iterations if the loop is
7787 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7788 Alan Hayward <alan.hayward@arm.com>
7789 David Sherwood <david.sherwood@arm.com>
7791 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
7792 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
7793 (dump_groups): Update accordingly.
7794 (iv_use::mem_type): New member variable.
7795 (address_p): New function.
7796 (record_use): Add a mem_type argument and initialize the new
7798 (record_group_use): Add a mem_type argument. Use address_p.
7799 Remove obsolete null checks of base_object. Update call to record_use.
7800 (find_interesting_uses_op): Update call to record_group_use.
7801 (find_interesting_uses_cond): Likewise.
7802 (find_interesting_uses_address): Likewise.
7803 (get_mem_type_for_internal_fn): New function.
7804 (find_address_like_use): Likewise.
7805 (find_interesting_uses_stmt): Try find_address_like_use before
7806 calling find_interesting_uses_op.
7807 (addr_offset_valid_p): Use the iv mem_type field as the type
7808 of the addressed memory.
7809 (add_autoinc_candidates): Likewise.
7810 (get_address_cost): Likewise.
7811 (split_small_address_groups_p): Use address_p.
7812 (split_address_groups): Likewise.
7813 (add_iv_candidate_for_use): Likewise.
7814 (autoinc_possible_for_pair): Likewise.
7815 (rewrite_groups): Likewise.
7816 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
7817 (determine_group_iv_cost): Update after split of USE_ADDRESS.
7818 (get_alias_ptr_type_for_ptr_address): New function.
7819 (rewrite_use_address): Rewrite address uses in calls that were
7820 identified by find_address_like_use.
7822 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7823 Alan Hayward <alan.hayward@arm.com>
7824 David Sherwood <david.sherwood@arm.com>
7826 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
7828 * gimple-expr.h (is_gimple_addressable: Likewise.
7829 * gimple-expr.c (is_gimple_address): Likewise.
7830 * internal-fn.c (expand_call_mem_ref): New function.
7831 (expand_mask_load_optab_fn): Use it.
7832 (expand_mask_store_optab_fn): Likewise.
7834 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7835 Alan Hayward <alan.hayward@arm.com>
7836 David Sherwood <david.sherwood@arm.com>
7838 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
7839 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
7840 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
7841 (cond_umax@var{mode}): Document.
7842 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
7843 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
7844 (cond_umin_optab, cond_umax_optab): New optabs.
7845 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
7846 (COND_IOR, COND_XOR): New internal functions.
7847 * internal-fn.h (get_conditional_internal_fn): Declare.
7848 * internal-fn.c (cond_binary_direct): New macro.
7849 (expand_cond_binary_optab_fn): Likewise.
7850 (direct_cond_binary_optab_supported_p): Likewise.
7851 (get_conditional_internal_fn): New function.
7852 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
7853 Cope with reduction statements that are vectorized as calls rather
7855 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
7856 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
7857 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
7858 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
7859 (UNSPEC_COND_EOR): New unspecs.
7860 (optab): Add mappings for them.
7861 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
7862 (sve_int_op, sve_fp_op): New int attributes.
7864 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7865 Alan Hayward <alan.hayward@arm.com>
7866 David Sherwood <david.sherwood@arm.com>
7868 * optabs.def (while_ult_optab): New optab.
7869 * doc/md.texi (while_ult@var{m}@var{n}): Document.
7870 * internal-fn.def (WHILE_ULT): New internal function.
7871 * internal-fn.h (direct_internal_fn_supported_p): New override
7872 that takes two types as argument.
7873 * internal-fn.c (while_direct): New macro.
7874 (expand_while_optab_fn): New function.
7875 (convert_optab_supported_p): Likewise.
7876 (direct_while_optab_supported_p): New macro.
7877 * wide-int.h (wi::udiv_ceil): New function.
7878 * tree-vectorizer.h (rgroup_masks): New structure.
7879 (vec_loop_masks): New typedef.
7880 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
7882 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
7883 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
7884 (vect_max_vf): New function.
7885 (slpeel_make_loop_iterate_ntimes): Delete.
7886 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
7887 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
7888 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
7889 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
7890 internal-fn.h, stor-layout.h and optabs-query.h.
7891 (vect_set_loop_mask): New function.
7892 (add_preheader_seq): Likewise.
7893 (add_header_seq): Likewise.
7894 (interleave_supported_p): Likewise.
7895 (vect_maybe_permute_loop_masks): Likewise.
7896 (vect_set_loop_masks_directly): Likewise.
7897 (vect_set_loop_condition_masked): Likewise.
7898 (vect_set_loop_condition_unmasked): New function, split out from
7899 slpeel_make_loop_iterate_ntimes.
7900 (slpeel_make_loop_iterate_ntimes): Rename to..
7901 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
7902 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
7903 (vect_do_peeling): Update call accordingly.
7904 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
7906 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7907 mask_compare_type, can_fully_mask_p and fully_masked_p.
7908 (release_vec_loop_masks): New function.
7909 (_loop_vec_info): Use it to free the loop masks.
7910 (can_produce_all_loop_masks_p): New function.
7911 (vect_get_max_nscalars_per_iter): Likewise.
7912 (vect_verify_full_masking): Likewise.
7913 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
7914 retries, and free the mask rgroups before retrying. Check loop-wide
7915 reasons for disallowing fully-masked loops. Make the final decision
7916 about whether use a fully-masked loop or not.
7917 (vect_estimate_min_profitable_iters): Do not assume that peeling
7918 for the number of iterations will be needed for fully-masked loops.
7919 (vectorizable_reduction): Disable fully-masked loops.
7920 (vectorizable_live_operation): Likewise.
7921 (vect_halve_mask_nunits): New function.
7922 (vect_double_mask_nunits): Likewise.
7923 (vect_record_loop_mask): Likewise.
7924 (vect_get_loop_mask): Likewise.
7925 (vect_transform_loop): Handle the case in which the final loop
7926 iteration might handle a partial vector. Call vect_set_loop_condition
7927 instead of slpeel_make_loop_iterate_ntimes.
7928 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
7929 (check_load_store_masking): New function.
7930 (prepare_load_store_mask): Likewise.
7931 (vectorizable_store): Handle fully-masked loops.
7932 (vectorizable_load): Likewise.
7933 (supportable_widening_operation): Use vect_halve_mask_nunits for
7935 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
7936 (vect_gen_while): New function.
7937 * config/aarch64/aarch64.md (umax<mode>3): New expander.
7938 (aarch64_uqdec<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 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
7945 (reduc_xor_scal_optab): New optabs.
7946 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
7947 (reduc_xor_scal_@var{m}): Document.
7948 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
7949 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
7951 * fold-const-call.c (fold_const_call): Handle them.
7952 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
7953 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
7954 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
7955 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
7956 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
7957 (UNSPEC_XORV): New unspecs.
7958 (optab): Add entries for them.
7959 (BITWISEV): New int iterator.
7960 (bit_reduc_op): New int attributes.
7962 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7963 Alan Hayward <alan.hayward@arm.com>
7964 David Sherwood <david.sherwood@arm.com>
7966 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
7967 * internal-fn.def (VEC_SHL_INSERT): New internal function.
7968 * optabs.def (vec_shl_insert_optab): New optab.
7969 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
7970 (duplicate_and_interleave): Likewise.
7971 * tree-vect-loop.c: Include internal-fn.h.
7972 (neutral_op_for_slp_reduction): New function, split out from
7973 get_initial_defs_for_reduction.
7974 (get_initial_def_for_reduction): Handle option 2 for variable-length
7975 vectors by loading the neutral value into a vector and then shifting
7976 the initial value into element 0.
7977 (get_initial_defs_for_reduction): Replace the code argument with
7978 the neutral value calculated by neutral_op_for_slp_reduction.
7979 Use gimple_build_vector for constant-length vectors.
7980 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
7981 but the first group_size elements have a neutral value.
7982 Use duplicate_and_interleave otherwise.
7983 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
7984 Update call to get_initial_defs_for_reduction. Handle SLP
7985 reductions for variable-length vectors by creating one vector
7986 result for each scalar result, with the elements associated
7987 with other scalar results stubbed out with the neutral value.
7988 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
7989 Require IFN_VEC_SHL_INSERT for double reductions on
7990 variable-length vectors, or SLP reductions that have
7991 a neutral value. Require can_duplicate_and_interleave_p
7992 support for variable-length unchained SLP reductions if there
7993 is no neutral value, such as for MIN/MAX reductions. Also require
7994 the number of vector elements to be a multiple of the number of
7995 SLP statements when doing variable-length unchained SLP reductions.
7996 Update call to vect_create_epilog_for_reduction.
7997 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
7998 and remove initial values.
7999 (duplicate_and_interleave): Make public.
8000 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
8001 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
8003 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8004 Alan Hayward <alan.hayward@arm.com>
8005 David Sherwood <david.sherwood@arm.com>
8007 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
8008 (can_duplicate_and_interleave_p): New function.
8009 (vect_get_and_check_slp_defs): Take the vector of statements
8010 rather than just the current one. Remove excess parentheses.
8011 Restriction rejectinon of vect_constant_def and vect_external_def
8012 for variable-length vectors to boolean types, or types for which
8013 can_duplicate_and_interleave_p is false.
8014 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
8015 (duplicate_and_interleave): New function.
8016 (vect_get_constant_vectors): Use gimple_build_vector for
8017 constant-length vectors and suitable variable-length constant
8018 vectors. Use duplicate_and_interleave for other variable-length
8019 vectors. Don't defer the update when inserting new statements.
8021 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8022 Alan Hayward <alan.hayward@arm.com>
8023 David Sherwood <david.sherwood@arm.com>
8025 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
8026 min_profitable_iters doesn't go negative.
8028 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8029 Alan Hayward <alan.hayward@arm.com>
8030 David Sherwood <david.sherwood@arm.com>
8032 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
8033 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
8034 * optabs.def (vec_mask_load_lanes_optab): New optab.
8035 (vec_mask_store_lanes_optab): Likewise.
8036 * internal-fn.def (MASK_LOAD_LANES): New internal function.
8037 (MASK_STORE_LANES): Likewise.
8038 * internal-fn.c (mask_load_lanes_direct): New macro.
8039 (mask_store_lanes_direct): Likewise.
8040 (expand_mask_load_optab_fn): Handle masked operations.
8041 (expand_mask_load_lanes_optab_fn): New macro.
8042 (expand_mask_store_optab_fn): Handle masked operations.
8043 (expand_mask_store_lanes_optab_fn): New macro.
8044 (direct_mask_load_lanes_optab_supported_p): Likewise.
8045 (direct_mask_store_lanes_optab_supported_p): Likewise.
8046 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
8048 (vect_load_lanes_supported): Likewise.
8049 * tree-vect-data-refs.c (strip_conversion): New function.
8050 (can_group_stmts_p): Likewise.
8051 (vect_analyze_data_ref_accesses): Use it instead of checking
8052 for a pair of assignments.
8053 (vect_store_lanes_supported): Take a masked_p parameter.
8054 (vect_load_lanes_supported): Likewise.
8055 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
8056 vect_store_lanes_supported and vect_load_lanes_supported.
8057 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8058 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
8059 parameter. Don't allow gaps for masked accesses.
8060 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
8061 and vect_load_lanes_supported.
8062 (get_load_store_type): Take a masked_p parameter and update
8063 call to get_group_load_store_type.
8064 (vectorizable_store): Update call to get_load_store_type.
8065 Handle IFN_MASK_STORE_LANES.
8066 (vectorizable_load): Update call to get_load_store_type.
8067 Handle IFN_MASK_LOAD_LANES.
8069 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8070 Alan Hayward <alan.hayward@arm.com>
8071 David Sherwood <david.sherwood@arm.com>
8073 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
8075 * config/aarch64/aarch64-protos.h
8076 (aarch64_sve_struct_memory_operand_p): Declare.
8077 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
8078 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
8079 (VPRED, vpred): Handle SVE structure modes.
8080 * config/aarch64/constraints.md (Utx): New constraint.
8081 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
8082 (aarch64_sve_struct_nonimmediate_operand): New predicates.
8083 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
8084 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
8085 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
8086 structure modes. Split into pieces after RA.
8087 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
8088 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
8090 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
8091 SVE structure modes.
8092 (aarch64_classify_address): Likewise.
8093 (sizetochar): Move earlier in file.
8094 (aarch64_print_operand): Handle SVE register lists.
8095 (aarch64_array_mode): New function.
8096 (aarch64_sve_struct_memory_operand_p): Likewise.
8097 (TARGET_ARRAY_MODE): Redefine.
8099 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8100 Alan Hayward <alan.hayward@arm.com>
8101 David Sherwood <david.sherwood@arm.com>
8103 * target.def (array_mode): New target hook.
8104 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
8105 * doc/tm.texi: Regenerate.
8106 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
8107 * hooks.c (hook_optmode_mode_uhwi_none): New function.
8108 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
8110 * stor-layout.c (mode_for_array): Likewise. Support polynomial
8113 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8114 Alan Hayward <alan.hayward@arm.com>
8115 David Sherwood <david.sherwood@arm.com>
8117 * fold-const.c (fold_binary_loc): Check the argument types
8118 rather than the result type when testing for a vector operation.
8120 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8122 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
8123 * doc/tm.texi: Regenerate.
8125 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8126 Alan Hayward <alan.hayward@arm.com>
8127 David Sherwood <david.sherwood@arm.com>
8129 * doc/invoke.texi (-msve-vector-bits=): Document new option.
8130 (sve): Document new AArch64 extension.
8131 * doc/md.texi (w): Extend the description of the AArch64
8132 constraint to include SVE vectors.
8133 (Upl, Upa): Document new AArch64 predicate constraints.
8134 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
8136 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
8137 (msve-vector-bits=): New option.
8138 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
8139 SVE when these are disabled.
8140 (sve): New extension.
8141 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
8142 modes. Adjust their number of units based on aarch64_sve_vg.
8143 (MAX_BITSIZE_MODE_ANY_MODE): Define.
8144 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
8145 aarch64_addr_query_type.
8146 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
8147 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
8148 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
8149 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
8150 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
8151 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
8152 (aarch64_simd_imm_zero_p): Delete.
8153 (aarch64_check_zero_based_sve_index_immediate): Declare.
8154 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8155 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8156 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8157 (aarch64_sve_float_mul_immediate_p): Likewise.
8158 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8160 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
8161 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
8162 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
8163 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
8164 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
8165 (aarch64_regmode_natural_size): Likewise.
8166 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
8167 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
8169 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
8170 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
8171 for VG and the SVE predicate registers.
8172 (V_ALIASES): Add a "z"-prefixed alias.
8173 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
8174 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
8175 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
8176 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
8177 (REG_CLASS_NAMES): Add entries for them.
8178 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
8179 and the predicate registers.
8180 (aarch64_sve_vg): Declare.
8181 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
8182 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
8183 (REGMODE_NATURAL_SIZE): Define.
8184 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
8186 * config/aarch64/aarch64.c: Include cfgrtl.h.
8187 (simd_immediate_info): Add a constructor for series vectors,
8188 and an associated step field.
8189 (aarch64_sve_vg): New variable.
8190 (aarch64_dbx_register_number): Handle VG and the predicate registers.
8191 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
8192 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
8193 (VEC_ANY_DATA, VEC_STRUCT): New constants.
8194 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
8195 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
8196 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
8197 (aarch64_get_mask_mode): New functions.
8198 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
8199 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
8200 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
8201 predicate modes and predicate registers. Explicitly restrict
8202 GPRs to modes of 16 bytes or smaller. Only allow FP registers
8203 to store a vector mode if it is recognized by
8204 aarch64_classify_vector_mode.
8205 (aarch64_regmode_natural_size): New function.
8206 (aarch64_hard_regno_caller_save_mode): Return the original mode
8208 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
8209 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
8210 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
8211 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
8213 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
8214 does not overlap dest if the function is frame-related. Handle
8216 (aarch64_split_add_offset): New function.
8217 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
8218 them aarch64_add_offset.
8219 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
8220 and update call to aarch64_sub_sp.
8221 (aarch64_add_cfa_expression): New function.
8222 (aarch64_expand_prologue): Pass extra temporary registers to the
8223 functions above. Handle the case in which we need to emit new
8224 DW_CFA_expressions for registers that were originally saved
8225 relative to the stack pointer, but now have to be expressed
8226 relative to the frame pointer.
8227 (aarch64_output_mi_thunk): Pass extra temporary registers to the
8229 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
8230 IP0 and IP1 values for SVE frames.
8231 (aarch64_expand_vec_series): New function.
8232 (aarch64_expand_sve_widened_duplicate): Likewise.
8233 (aarch64_expand_sve_const_vector): Likewise.
8234 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
8235 Handle SVE constants. Use emit_move_insn to move a force_const_mem
8236 into the register, rather than emitting a SET directly.
8237 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
8238 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
8239 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
8240 (offset_9bit_signed_scaled_p): New functions.
8241 (aarch64_replicate_bitmask_imm): New function.
8242 (aarch64_bitmask_imm): Use it.
8243 (aarch64_cannot_force_const_mem): Reject expressions involving
8244 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
8245 (aarch64_classify_index): Handle SVE indices, by requiring
8246 a plain register index with a scale that matches the element size.
8247 (aarch64_classify_address): Handle SVE addresses. Assert that
8248 the mode of the address is VOIDmode or an integer mode.
8249 Update call to aarch64_classify_symbol.
8250 (aarch64_classify_symbolic_expression): Update call to
8251 aarch64_classify_symbol.
8252 (aarch64_const_vec_all_in_range_p): New function.
8253 (aarch64_print_vector_float_operand): Likewise.
8254 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
8255 "vN" for FP registers with SVE modes. Handle (const ...) vectors
8256 and the FP immediates 1.0 and 0.5.
8257 (aarch64_print_address_internal): Handle SVE addresses.
8258 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
8259 (aarch64_regno_regclass): Handle predicate registers.
8260 (aarch64_secondary_reload): Handle big-endian reloads of SVE
8262 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
8263 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
8264 (aarch64_convert_sve_vector_bits): New function.
8265 (aarch64_override_options): Use it to handle -msve-vector-bits=.
8266 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8268 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
8269 Handle SVE vector and predicate modes. Accept VL-based constants
8270 that need only one temporary register, and VL offsets that require
8271 no temporary registers.
8272 (aarch64_conditional_register_usage): Mark the predicate registers
8273 as fixed if SVE isn't available.
8274 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
8275 Return true for SVE vector and predicate modes.
8276 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
8277 rather than an unsigned int. Handle SVE modes.
8278 (aarch64_preferred_simd_mode): Update call accordingly. Handle
8280 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
8282 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8283 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8284 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8285 (aarch64_sve_float_mul_immediate_p): New functions.
8286 (aarch64_sve_valid_immediate): New function.
8287 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
8288 Explicitly reject structure modes. Check for INDEX constants.
8289 Handle PTRUE and PFALSE constants.
8290 (aarch64_check_zero_based_sve_index_immediate): New function.
8291 (aarch64_simd_imm_zero_p): Delete.
8292 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
8293 vector modes. Accept constants in the range of CNT[BHWD].
8294 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
8295 ask for an Advanced SIMD mode.
8296 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
8297 (aarch64_simd_vector_alignment): Handle SVE predicates.
8298 (aarch64_vectorize_preferred_vector_alignment): New function.
8299 (aarch64_simd_vector_alignment_reachable): Use it instead of
8301 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
8302 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
8304 (MAX_VECT_LEN): Delete.
8305 (expand_vec_perm_d): Add a vec_flags field.
8306 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
8307 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
8308 (aarch64_evpc_ext): Don't apply a big-endian lane correction
8310 (aarch64_evpc_rev): Rename to...
8311 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
8312 (aarch64_evpc_rev_global): New function.
8313 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
8314 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
8316 (aarch64_evpc_sve_tbl): New function.
8317 (aarch64_expand_vec_perm_const_1): Update after rename of
8318 aarch64_evpc_rev. Handle SVE permutes too, trying
8319 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
8320 than aarch64_evpc_tbl.
8321 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
8322 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
8323 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
8324 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
8325 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
8326 (aarch64_expand_sve_vcond): New functions.
8327 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
8328 of aarch64_vector_mode_p.
8329 (aarch64_dwarf_poly_indeterminate_value): New function.
8330 (aarch64_compute_pressure_classes): Likewise.
8331 (aarch64_can_change_mode_class): Likewise.
8332 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
8333 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
8334 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
8335 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
8336 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
8337 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
8338 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
8339 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
8341 (Dn, Dl, Dr): Accept const as well as const_vector.
8342 (Dz): Likewise. Compare against CONST0_RTX.
8343 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
8344 of "vector" where appropriate.
8345 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
8346 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
8347 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
8348 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
8349 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
8350 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
8351 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
8352 (v_int_equiv): Extend to SVE modes.
8353 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
8355 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
8356 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
8357 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
8358 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
8359 (SVE_COND_FP_CMP): New int iterators.
8360 (perm_hilo): Handle the new unpack unspecs.
8361 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
8363 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
8364 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
8365 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
8366 (aarch64_equality_operator, aarch64_constant_vector_operand)
8367 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
8368 (aarch64_sve_nonimmediate_operand): Likewise.
8369 (aarch64_sve_general_operand): Likewise.
8370 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
8371 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
8372 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
8373 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
8374 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
8375 (aarch64_sve_float_arith_immediate): Likewise.
8376 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
8377 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
8378 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
8379 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
8380 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
8381 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
8382 (aarch64_sve_float_arith_operand): Likewise.
8383 (aarch64_sve_float_arith_with_sub_operand): Likewise.
8384 (aarch64_sve_float_mul_operand): Likewise.
8385 (aarch64_sve_vec_perm_operand): Likewise.
8386 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
8387 (aarch64_mov_operand): Accept const_poly_int and const_vector.
8388 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
8389 as well as const_vector.
8390 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
8391 in file. Use CONST0_RTX and CONSTM1_RTX.
8392 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
8393 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
8394 Use aarch64_simd_imm_zero.
8395 * config/aarch64/aarch64-sve.md: New file.
8396 * config/aarch64/aarch64.md: Include it.
8397 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
8398 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
8399 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
8400 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
8401 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
8402 (sve): New attribute.
8403 (enabled): Disable instructions with the sve attribute unless
8405 (movqi, movhi): Pass CONST_POLY_INT operaneds through
8406 aarch64_expand_mov_immediate.
8407 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
8408 CNT[BHSD] immediates.
8409 (movti): Split CONST_POLY_INT moves into two halves.
8410 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
8411 Split additions that need a temporary here if the destination
8412 is the stack pointer.
8413 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
8414 (*add<mode>3_poly_1): New instruction.
8415 (set_clobber_cc): New expander.
8417 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8419 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
8420 parameter and use it instead of GET_MODE_SIZE (innermode). Use
8421 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
8422 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
8423 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
8424 Change innermode from fixed_mode_size to machine_mode.
8425 (simplify_subreg): Update call accordingly. Handle a constant-sized
8426 subreg of a variable-length CONST_VECTOR.
8428 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8429 Alan Hayward <alan.hayward@arm.com>
8430 David Sherwood <david.sherwood@arm.com>
8432 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
8433 (add_offset_to_base): New function, split out from...
8434 (create_mem_ref): ...here. When handling a scale other than 1,
8435 check first whether the address is valid without the offset.
8436 Add it into the base if so, leaving the index and scale as-is.
8438 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8441 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
8442 fold_for_warn before checking if arg2 is INTEGER_CST.
8444 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
8446 * config/rs6000/predicates.md (load_multiple_operation): Delete.
8447 (store_multiple_operation): Delete.
8448 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
8449 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
8450 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
8451 guarded by TARGET_STRING.
8452 (rs6000_output_load_multiple): Delete.
8453 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
8454 OPTION_MASK_STRING / TARGET_STRING handling.
8455 (print_operand) <'N', 'O'>: Add comment that these are unused now.
8456 (const rs6000_opt_masks) <"string">: Change mask to 0.
8457 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
8458 (MASK_STRING): Delete.
8459 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
8461 (load_multiple): Delete.
8468 (store_multiple): Delete.
8475 (movmemsi_8reg): Delete.
8476 (corresponding unnamed define_insn): Delete.
8477 (movmemsi_6reg): Delete.
8478 (corresponding unnamed define_insn): Delete.
8479 (movmemsi_4reg): Delete.
8480 (corresponding unnamed define_insn): Delete.
8481 (movmemsi_2reg): Delete.
8482 (corresponding unnamed define_insn): Delete.
8483 (movmemsi_1reg): Delete.
8484 (corresponding unnamed define_insn): Delete.
8485 * config/rs6000/rs6000.opt (mno-string): New.
8486 (mstring): Replace by deprecation warning stub.
8487 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
8489 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8491 * regrename.c (regrename_do_replace): If replacing the same
8492 reg multiple times, try to reuse last created gen_raw_REG.
8495 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
8496 main to workaround a bug in GDB.
8498 2018-01-12 Tom de Vries <tom@codesourcery.com>
8501 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
8503 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
8505 PR rtl-optimization/80481
8506 * ira-color.c (get_cap_member): New function.
8507 (allocnos_conflict_by_live_ranges_p): Use it.
8508 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
8509 (setup_slot_coalesced_allocno_live_ranges): Ditto.
8511 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
8514 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
8515 (*saddl_se_1): Ditto.
8517 (*ssubl_se_1): Ditto.
8519 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8521 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
8522 rather than wi::to_widest for DR_INITs.
8523 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
8524 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
8525 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
8527 (vect_analyze_group_access_1): Note that here.
8529 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8531 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
8532 polynomial type sizes.
8534 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8536 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
8537 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
8538 (gimple_add_tmp_var): Likewise.
8540 2018-01-12 Martin Liska <mliska@suse.cz>
8542 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
8543 (gimple_alloc_sizes): Likewise.
8544 (dump_gimple_statistics): Use PRIu64 in printf format.
8545 * gimple.h: Change uint64_t to int.
8547 2018-01-12 Martin Liska <mliska@suse.cz>
8549 * tree-core.h: Use uint64_t instead of int.
8550 * tree.c (tree_node_counts): Likewise.
8551 (tree_node_sizes): Likewise.
8552 (dump_tree_statistics): Use PRIu64 in printf format.
8554 2018-01-12 Martin Liska <mliska@suse.cz>
8556 * Makefile.in: As qsort_chk is implemented in vec.c, add
8557 vec.o to linkage of gencfn-macros.
8558 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
8559 passing the info to record_node_allocation_statistics.
8560 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
8562 * ggc-common.c (struct ggc_usage): Add operator== and use
8563 it in operator< and compare function.
8564 * mem-stats.h (struct mem_usage): Likewise.
8565 * vec.c (struct vec_usage): Remove operator< and compare
8566 function. Can be simply inherited.
8568 2018-01-12 Martin Jambor <mjambor@suse.cz>
8571 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
8572 * tree-ssa-math-opts.c: Include domwalk.h.
8573 (convert_mult_to_fma_1): New function.
8574 (fma_transformation_info): New type.
8575 (fma_deferring_state): Likewise.
8576 (cancel_fma_deferring): New function.
8577 (result_of_phi): Likewise.
8578 (last_fma_candidate_feeds_initial_phi): Likewise.
8579 (convert_mult_to_fma): Added deferring logic, split actual
8580 transformation to convert_mult_to_fma_1.
8581 (math_opts_dom_walker): New type.
8582 (math_opts_dom_walker::after_dom_children): New method, body moved
8583 here from pass_optimize_widening_mul::execute, added deferring logic
8585 (pass_optimize_widening_mul::execute): Moved most of code to
8586 math_opts_dom_walker::after_dom_children.
8587 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
8588 * config/i386/i386.c (ix86_option_override_internal): Added
8589 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
8591 2018-01-12 Richard Biener <rguenther@suse.de>
8594 * dwarf2out.c (gen_variable_die): Do not reset old_die for
8595 inline instance vars.
8597 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
8600 * config/rx/rx.c (rx_is_restricted_memory_address):
8603 2018-01-12 Richard Biener <rguenther@suse.de>
8605 PR tree-optimization/80846
8606 * target.def (split_reduction): New target hook.
8607 * targhooks.c (default_split_reduction): New function.
8608 * targhooks.h (default_split_reduction): Declare.
8609 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
8610 target requests first reduce vectors by combining low and high
8612 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
8613 (get_vectype_for_scalar_type_and_size): Export.
8614 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
8615 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
8616 * doc/tm.texi: Regenerate.
8617 * config/i386/i386.c (ix86_split_reduction): Implement
8618 TARGET_VECTORIZE_SPLIT_REDUCTION.
8620 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8623 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
8624 in PIC mode except for TARGET_VXWORKS_RTP.
8625 * config/sparc/sparc.c: Include cfgrtl.h.
8626 (TARGET_INIT_PIC_REG): Define.
8627 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
8628 (sparc_pic_register_p): New predicate.
8629 (sparc_legitimate_address_p): Use it.
8630 (sparc_legitimize_pic_address): Likewise.
8631 (sparc_delegitimize_address): Likewise.
8632 (sparc_mode_dependent_address_p): Likewise.
8633 (gen_load_pcrel_sym): Remove 4th parameter.
8634 (load_got_register): Adjust call to above. Remove obsolete stuff.
8635 (sparc_expand_prologue): Do not call load_got_register here.
8636 (sparc_flat_expand_prologue): Likewise.
8637 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
8638 (sparc_use_pseudo_pic_reg): New function.
8639 (sparc_init_pic_reg): Likewise.
8640 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8641 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
8643 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
8645 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
8646 Add item for branch_cost.
8648 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8650 PR rtl-optimization/83565
8651 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
8652 not extend the result to a larger mode for rotate operations.
8653 (num_sign_bit_copies1): Likewise.
8655 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8658 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
8660 Use values-Xc.o for -pedantic.
8661 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
8663 2018-01-12 Martin Liska <mliska@suse.cz>
8666 * ipa-devirt.c (final_warning_record::grow_type_warnings):
8668 (possible_polymorphic_call_targets): Use it.
8669 (ipa_devirt): Likewise.
8671 2018-01-12 Martin Liska <mliska@suse.cz>
8673 * profile-count.h (enum profile_quality): Use 0 as invalid
8674 enum value of profile_quality.
8676 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
8678 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
8679 -mext-string options.
8681 2018-01-12 Richard Biener <rguenther@suse.de>
8683 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
8684 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
8685 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
8687 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
8689 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
8691 * configure.ac (--with-long-double-format): Add support for the
8692 configuration option to change the default long double format on
8694 * config.gcc (powerpc*-linux*-*): Likewise.
8695 * configure: Regenerate.
8696 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
8697 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
8698 used without modification.
8700 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8702 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
8703 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
8704 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
8705 MISC_BUILTIN_SPEC_BARRIER.
8706 (rs6000_init_builtins): Likewise.
8707 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
8709 (speculation_barrier): New define_insn.
8710 * doc/extend.texi: Document __builtin_speculation_barrier.
8712 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8715 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
8716 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
8717 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
8719 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
8720 integral modes instead of "ss" and "sd".
8721 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
8722 vectors with 32-bit and 64-bit elements.
8723 (vecdupssescalarmodesuffix): New mode attribute.
8724 (vec_dup<mode>): Use it.
8726 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
8729 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
8730 frame if argument is passed on stack.
8732 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8735 * ree.c (combine_reaching_defs): Optimize also
8736 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
8737 reg2=any_extend(exp); reg1=reg2;, formatting fix.
8739 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8742 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
8744 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8747 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
8748 after they are computed.
8750 2018-01-11 Bin Cheng <bin.cheng@arm.com>
8752 PR tree-optimization/83695
8753 * gimple-loop-linterchange.cc
8754 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
8755 reset cached scev information after interchange.
8756 (pass_linterchange::execute): Remove call to scev_reset_htab.
8758 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8760 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
8761 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
8762 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
8763 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
8764 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
8765 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
8766 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
8767 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
8768 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
8769 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
8770 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
8771 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
8772 (V_lane_reg): Likewise.
8773 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
8775 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
8776 (vfmal_lane_low<mode>_intrinsic,
8777 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
8778 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
8779 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
8780 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
8781 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
8782 vfmsl_lane_high<mode>_intrinsic): New define_insns.
8784 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8786 * config/arm/arm-cpus.in (fp16fml): New feature.
8787 (ALL_SIMD): Add fp16fml.
8788 (armv8.2-a): Add fp16fml as an option.
8789 (armv8.3-a): Likewise.
8790 (armv8.4-a): Add fp16fml as part of fp16.
8791 * config/arm/arm.h (TARGET_FP16FML): Define.
8792 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
8794 * config/arm/arm-modes.def (V2HF): Define.
8795 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
8796 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
8797 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
8798 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
8799 vfmsl_low, vfmsl_high): New set of builtins.
8800 * config/arm/iterators.md (PLUSMINUS): New code iterator.
8801 (vfml_op): New code attribute.
8802 (VFMLHALVES): New int iterator.
8803 (VFML, VFMLSEL): New mode attributes.
8804 (V_reg): Define mapping for V2HF.
8805 (V_hi, V_lo): New mode attributes.
8806 (VF_constraint): Likewise.
8807 (vfml_half, vfml_half_selector): New int attributes.
8808 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
8810 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
8811 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
8813 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
8814 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
8815 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
8816 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
8818 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
8819 Document new effective target and option set.
8821 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8823 * config/arm/arm-cpus.in (armv8_4): New feature.
8824 (ARMv8_4a): New fgroup.
8825 (armv8.4-a): New arch.
8826 * config/arm/arm-tables.opt: Regenerate.
8827 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
8828 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
8829 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
8830 Add matching rules for -march=armv8.4-a and extensions.
8831 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
8833 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
8836 * config/rx/rx.md (BW): New mode attribute.
8837 (sync_lock_test_and_setsi): Add mode suffix to insn output.
8839 2018-01-11 Richard Biener <rguenther@suse.de>
8841 PR tree-optimization/83435
8842 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
8843 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
8844 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
8846 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8847 Alan Hayward <alan.hayward@arm.com>
8848 David Sherwood <david.sherwood@arm.com>
8850 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
8852 (aarch64_classify_address): Initialize it. Track polynomial offsets.
8853 (aarch64_print_address_internal): Use it to check for a zero offset.
8855 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8856 Alan Hayward <alan.hayward@arm.com>
8857 David Sherwood <david.sherwood@arm.com>
8859 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
8860 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
8861 Return a poly_int64 rather than a HOST_WIDE_INT.
8862 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
8863 rather than a HOST_WIDE_INT.
8864 * config/aarch64/aarch64.h (aarch64_frame): Protect with
8865 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
8866 hard_fp_offset, frame_size, initial_adjust, callee_offset and
8867 final_offset from HOST_WIDE_INT to poly_int64.
8868 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
8869 to_constant when getting the number of units in an Advanced SIMD
8871 (aarch64_builtin_vectorized_function): Check for a constant number
8873 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
8875 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
8876 attribute instead of GET_MODE_NUNITS.
8877 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
8878 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
8879 GET_MODE_SIZE for fixed-size registers.
8880 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
8881 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
8882 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
8883 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
8884 (aarch64_print_operand, aarch64_print_address_internal)
8885 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
8886 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
8887 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
8888 Handle polynomial GET_MODE_SIZE.
8889 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
8890 wider than SImode without modification.
8891 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
8892 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
8893 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
8894 passing and returning SVE modes.
8895 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
8896 rather than GEN_INT.
8897 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
8898 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
8899 (aarch64_allocate_and_probe_stack_space): Likewise.
8900 (aarch64_layout_frame): Cope with polynomial offsets.
8901 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
8902 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
8904 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
8905 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
8906 poly_int64 rather than a HOST_WIDE_INT.
8907 (aarch64_get_separate_components, aarch64_process_components)
8908 (aarch64_expand_prologue, aarch64_expand_epilogue)
8909 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
8910 (aarch64_anchor_offset): New function, split out from...
8911 (aarch64_legitimize_address): ...here.
8912 (aarch64_builtin_vectorization_cost): Handle polynomial
8913 TYPE_VECTOR_SUBPARTS.
8914 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
8916 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
8917 number of elements from the PARALLEL rather than the mode.
8918 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
8919 rather than GET_MODE_BITSIZE.
8920 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
8921 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
8922 (aarch64_expand_vec_perm_const_1): Handle polynomial
8923 d->perm.length () and d->perm elements.
8924 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
8925 Apply to_constant to d->perm elements.
8926 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
8927 polynomial CONST_VECTOR_NUNITS.
8928 (aarch64_move_pointer): Take amount as a poly_int64 rather
8930 (aarch64_progress_pointer): Avoid temporary variable.
8931 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
8932 the mode attribute instead of GET_MODE.
8934 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8935 Alan Hayward <alan.hayward@arm.com>
8936 David Sherwood <david.sherwood@arm.com>
8938 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
8939 x exists before using it.
8940 (aarch64_add_constant_internal): Rename to...
8941 (aarch64_add_offset_1): ...this. Replace regnum with separate
8942 src and dest rtxes. Handle the case in which they're different,
8943 including when the offset is zero. Replace scratchreg with an rtx.
8944 Use 2 additions if there is no spare register into which we can
8945 move a 16-bit constant.
8946 (aarch64_add_constant): Delete.
8947 (aarch64_add_offset): Replace reg with separate src and dest
8948 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
8949 Use aarch64_add_offset_1.
8950 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
8951 an rtx rather than an int. Take the delta as a poly_int64
8952 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
8953 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
8954 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
8955 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
8956 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
8958 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
8959 aarch64_add_constant.
8961 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8963 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
8964 Use scalar_float_mode.
8966 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8968 * config/aarch64/aarch64-simd.md
8969 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
8970 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
8971 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
8972 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
8973 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
8974 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
8975 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
8976 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
8977 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
8978 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
8980 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8983 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
8984 targ_options->x_arm_arch_string is non NULL.
8986 2018-01-11 Tamar Christina <tamar.christina@arm.com>
8988 * config/aarch64/aarch64.h
8989 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
8991 2018-01-11 Sudakshina Das <sudi.das@arm.com>
8994 * expmed.c (emit_store_flag_force): Swap if const op0
8995 and change VOIDmode to mode of op0.
8997 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8999 PR rtl-optimization/83761
9000 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
9001 than bytes to mode_for_size.
9003 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9006 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
9007 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
9010 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9013 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
9014 when in layout mode.
9015 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
9016 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
9019 2018-01-10 Michael Collison <michael.collison@arm.com>
9021 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
9022 * config/aarch64/aarch64-option-extension.def: Add
9023 AARCH64_OPT_EXTENSION of 'fp16fml'.
9024 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9025 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
9026 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
9027 * config/aarch64/constraints.md (Ui7): New constraint.
9028 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
9029 (VFMLA_SEL_W): Ditto.
9032 (VFMLA16_LOW): New int iterator.
9033 (VFMLA16_HIGH): Ditto.
9034 (UNSPEC_FMLAL): New unspec.
9035 (UNSPEC_FMLSL): Ditto.
9036 (UNSPEC_FMLAL2): Ditto.
9037 (UNSPEC_FMLSL2): Ditto.
9038 (f16mac): New code attribute.
9039 * config/aarch64/aarch64-simd-builtins.def
9040 (aarch64_fmlal_lowv2sf): Ditto.
9041 (aarch64_fmlsl_lowv2sf): Ditto.
9042 (aarch64_fmlalq_lowv4sf): Ditto.
9043 (aarch64_fmlslq_lowv4sf): Ditto.
9044 (aarch64_fmlal_highv2sf): Ditto.
9045 (aarch64_fmlsl_highv2sf): Ditto.
9046 (aarch64_fmlalq_highv4sf): Ditto.
9047 (aarch64_fmlslq_highv4sf): Ditto.
9048 (aarch64_fmlal_lane_lowv2sf): Ditto.
9049 (aarch64_fmlsl_lane_lowv2sf): Ditto.
9050 (aarch64_fmlal_laneq_lowv2sf): Ditto.
9051 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
9052 (aarch64_fmlalq_lane_lowv4sf): Ditto.
9053 (aarch64_fmlsl_lane_lowv4sf): Ditto.
9054 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
9055 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
9056 (aarch64_fmlal_lane_highv2sf): Ditto.
9057 (aarch64_fmlsl_lane_highv2sf): Ditto.
9058 (aarch64_fmlal_laneq_highv2sf): Ditto.
9059 (aarch64_fmlsl_laneq_highv2sf): Ditto.
9060 (aarch64_fmlalq_lane_highv4sf): Ditto.
9061 (aarch64_fmlsl_lane_highv4sf): Ditto.
9062 (aarch64_fmlalq_laneq_highv4sf): Ditto.
9063 (aarch64_fmlsl_laneq_highv4sf): Ditto.
9064 * config/aarch64/aarch64-simd.md:
9065 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
9066 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9067 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
9068 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9069 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
9070 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
9071 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
9072 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
9073 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
9074 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
9075 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
9076 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
9077 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
9078 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
9079 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
9080 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
9081 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
9082 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
9083 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
9084 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
9085 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
9086 (vfmlsl_low_u32): Ditto.
9087 (vfmlalq_low_u32): Ditto.
9088 (vfmlslq_low_u32): Ditto.
9089 (vfmlal_high_u32): Ditto.
9090 (vfmlsl_high_u32): Ditto.
9091 (vfmlalq_high_u32): Ditto.
9092 (vfmlslq_high_u32): Ditto.
9093 (vfmlal_lane_low_u32): Ditto.
9094 (vfmlsl_lane_low_u32): Ditto.
9095 (vfmlal_laneq_low_u32): Ditto.
9096 (vfmlsl_laneq_low_u32): Ditto.
9097 (vfmlalq_lane_low_u32): Ditto.
9098 (vfmlslq_lane_low_u32): Ditto.
9099 (vfmlalq_laneq_low_u32): Ditto.
9100 (vfmlslq_laneq_low_u32): Ditto.
9101 (vfmlal_lane_high_u32): Ditto.
9102 (vfmlsl_lane_high_u32): Ditto.
9103 (vfmlal_laneq_high_u32): Ditto.
9104 (vfmlsl_laneq_high_u32): Ditto.
9105 (vfmlalq_lane_high_u32): Ditto.
9106 (vfmlslq_lane_high_u32): Ditto.
9107 (vfmlalq_laneq_high_u32): Ditto.
9108 (vfmlslq_laneq_high_u32): Ditto.
9109 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
9110 (AARCH64_FL_FOR_ARCH8_4): New.
9111 (AARCH64_ISA_F16FML): New ISA flag.
9112 (TARGET_F16FML): New feature flag for fp16fml.
9113 (doc/invoke.texi): Document new fp16fml option.
9115 2018-01-10 Michael Collison <michael.collison@arm.com>
9117 * config/aarch64/aarch64-builtins.c:
9118 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
9119 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9120 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
9121 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
9122 (AARCH64_ISA_SHA3): New ISA flag.
9123 (TARGET_SHA3): New feature flag for sha3.
9124 * config/aarch64/iterators.md (sha512_op): New int attribute.
9125 (CRYPTO_SHA512): New int iterator.
9126 (UNSPEC_SHA512H): New unspec.
9127 (UNSPEC_SHA512H2): Ditto.
9128 (UNSPEC_SHA512SU0): Ditto.
9129 (UNSPEC_SHA512SU1): Ditto.
9130 * config/aarch64/aarch64-simd-builtins.def
9131 (aarch64_crypto_sha512hqv2di): New builtin.
9132 (aarch64_crypto_sha512h2qv2di): Ditto.
9133 (aarch64_crypto_sha512su0qv2di): Ditto.
9134 (aarch64_crypto_sha512su1qv2di): Ditto.
9135 (aarch64_eor3qv8hi): Ditto.
9136 (aarch64_rax1qv2di): Ditto.
9137 (aarch64_xarqv2di): Ditto.
9138 (aarch64_bcaxqv8hi): Ditto.
9139 * config/aarch64/aarch64-simd.md:
9140 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
9141 (aarch64_crypto_sha512su0qv2di): Ditto.
9142 (aarch64_crypto_sha512su1qv2di): Ditto.
9143 (aarch64_eor3qv8hi): Ditto.
9144 (aarch64_rax1qv2di): Ditto.
9145 (aarch64_xarqv2di): Ditto.
9146 (aarch64_bcaxqv8hi): Ditto.
9147 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
9148 (vsha512h2q_u64): Ditto.
9149 (vsha512su0q_u64): Ditto.
9150 (vsha512su1q_u64): Ditto.
9151 (veor3q_u16): Ditto.
9152 (vrax1q_u64): Ditto.
9154 (vbcaxq_u16): Ditto.
9155 * config/arm/types.md (crypto_sha512): New type attribute.
9156 (crypto_sha3): Ditto.
9157 (doc/invoke.texi): Document new sha3 option.
9159 2018-01-10 Michael Collison <michael.collison@arm.com>
9161 * config/aarch64/aarch64-builtins.c:
9162 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
9163 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9164 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
9165 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
9166 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
9167 (AARCH64_ISA_SM4): New ISA flag.
9168 (TARGET_SM4): New feature flag for sm4.
9169 * config/aarch64/aarch64-simd-builtins.def
9170 (aarch64_sm3ss1qv4si): Ditto.
9171 (aarch64_sm3tt1aq4si): Ditto.
9172 (aarch64_sm3tt1bq4si): Ditto.
9173 (aarch64_sm3tt2aq4si): Ditto.
9174 (aarch64_sm3tt2bq4si): Ditto.
9175 (aarch64_sm3partw1qv4si): Ditto.
9176 (aarch64_sm3partw2qv4si): Ditto.
9177 (aarch64_sm4eqv4si): Ditto.
9178 (aarch64_sm4ekeyqv4si): Ditto.
9179 * config/aarch64/aarch64-simd.md:
9180 (aarch64_sm3ss1qv4si): Ditto.
9181 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
9182 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
9183 (aarch64_sm4eqv4si): Ditto.
9184 (aarch64_sm4ekeyqv4si): Ditto.
9185 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
9186 (sm3part_op): Ditto.
9187 (CRYPTO_SM3TT): Ditto.
9188 (CRYPTO_SM3PART): Ditto.
9189 (UNSPEC_SM3SS1): New unspec.
9190 (UNSPEC_SM3TT1A): Ditto.
9191 (UNSPEC_SM3TT1B): Ditto.
9192 (UNSPEC_SM3TT2A): Ditto.
9193 (UNSPEC_SM3TT2B): Ditto.
9194 (UNSPEC_SM3PARTW1): Ditto.
9195 (UNSPEC_SM3PARTW2): Ditto.
9196 (UNSPEC_SM4E): Ditto.
9197 (UNSPEC_SM4EKEY): Ditto.
9198 * config/aarch64/constraints.md (Ui2): New constraint.
9199 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
9200 * config/arm/types.md (crypto_sm3): New type attribute.
9201 (crypto_sm4): Ditto.
9202 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
9203 (vsm3tt1aq_u32): Ditto.
9204 (vsm3tt1bq_u32): Ditto.
9205 (vsm3tt2aq_u32): Ditto.
9206 (vsm3tt2bq_u32): Ditto.
9207 (vsm3partw1q_u32): Ditto.
9208 (vsm3partw2q_u32): Ditto.
9209 (vsm4eq_u32): Ditto.
9210 (vsm4ekeyq_u32): Ditto.
9211 (doc/invoke.texi): Document new sm4 option.
9213 2018-01-10 Michael Collison <michael.collison@arm.com>
9215 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
9216 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
9217 (AARCH64_FL_FOR_ARCH8_4): New.
9218 (AARCH64_FL_V8_4): New flag.
9219 (doc/invoke.texi): Document new armv8.4-a option.
9221 2018-01-10 Michael Collison <michael.collison@arm.com>
9223 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9224 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
9225 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
9226 * config/aarch64/aarch64-option-extension.def: Add
9227 AARCH64_OPT_EXTENSION of 'sha2'.
9228 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
9229 (crypto): Disable sha2 and aes if crypto disabled.
9230 (crypto): Enable aes and sha2 if enabled.
9231 (simd): Disable sha2 and aes if simd disabled.
9232 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
9234 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
9235 (TARGET_SHA2): New feature flag for sha2.
9236 (TARGET_AES): New feature flag for aes.
9237 * config/aarch64/aarch64-simd.md:
9238 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
9239 conditional on TARGET_AES.
9240 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
9241 (aarch64_crypto_sha1hsi): Make pattern conditional
9243 (aarch64_crypto_sha1hv4si): Ditto.
9244 (aarch64_be_crypto_sha1hv4si): Ditto.
9245 (aarch64_crypto_sha1su1v4si): Ditto.
9246 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
9247 (aarch64_crypto_sha1su0v4si): Ditto.
9248 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
9249 (aarch64_crypto_sha256su0v4si): Ditto.
9250 (aarch64_crypto_sha256su1v4si): Ditto.
9251 (doc/invoke.texi): Document new aes and sha2 options.
9253 2018-01-10 Martin Sebor <msebor@redhat.com>
9255 PR tree-optimization/83781
9256 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
9259 2018-01-11 Martin Sebor <msebor@gmail.com>
9260 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9262 PR tree-optimization/83501
9263 PR tree-optimization/81703
9265 * tree-ssa-strlen.c (get_string_cst): Rename...
9266 (get_string_len): ...to this. Handle global constants.
9267 (handle_char_store): Adjust.
9269 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
9270 Jim Wilson <jimw@sifive.com>
9272 * config/riscv/riscv-protos.h (riscv_output_return): New.
9273 * config/riscv/riscv.c (struct machine_function): New naked_p field.
9274 (riscv_attribute_table, riscv_output_return),
9275 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
9276 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
9277 (riscv_compute_frame_info): Only compute frame->mask if not a naked
9279 (riscv_expand_prologue): Add early return for naked function.
9280 (riscv_expand_epilogue): Likewise.
9281 (riscv_function_ok_for_sibcall): Return false for naked function.
9282 (riscv_set_current_function): New.
9283 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
9284 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
9285 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
9286 * doc/extend.texi (RISC-V Function Attributes): New.
9288 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
9290 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
9291 check for 128-bit long double before checking TCmode.
9292 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
9293 128-bit long doubles before checking TFmode or TCmode.
9294 (FLOAT128_IBM_P): Likewise.
9296 2018-01-10 Martin Sebor <msebor@redhat.com>
9298 PR tree-optimization/83671
9299 * builtins.c (c_strlen): Unconditionally return zero for the empty
9301 Use -Warray-bounds for warnings.
9302 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
9303 for non-constant array indices with COMPONENT_REF, arrays of
9304 arrays, and pointers to arrays.
9305 (gimple_fold_builtin_strlen): Determine and set length range for
9306 non-constant character arrays.
9308 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
9311 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
9314 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
9316 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
9318 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9321 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
9322 VECTOR_MEM_ALTIVEC_OR_VSX_P.
9323 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
9324 indexed_or_indirect_operand predicate.
9325 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
9326 (*vsx_le_perm_load_v8hi): Likewise.
9327 (*vsx_le_perm_load_v16qi): Likewise.
9328 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
9329 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
9330 (*vsx_le_perm_store_v8hi): Likewise.
9331 (*vsx_le_perm_store_v16qi): Likewise.
9332 (eight unnamed splitters): Likewise.
9334 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9336 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
9337 * config/rs6000/emmintrin.h: Likewise.
9338 * config/rs6000/mmintrin.h: Likewise.
9339 * config/rs6000/xmmintrin.h: Likewise.
9341 2018-01-10 David Malcolm <dmalcolm@redhat.com>
9344 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
9346 * tree.c (tree_nop_conversion): Return true for location wrapper
9348 (maybe_wrap_with_location): New function.
9349 (selftest::check_strip_nops): New function.
9350 (selftest::test_location_wrappers): New function.
9351 (selftest::tree_c_tests): Call it.
9352 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
9353 (maybe_wrap_with_location): New decl.
9354 (EXPR_LOCATION_WRAPPER_P): New macro.
9355 (location_wrapper_p): New inline function.
9356 (tree_strip_any_location_wrapper): New inline function.
9358 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
9361 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
9362 stack_realign_offset for the largest alignment of stack slot
9364 (ix86_find_max_used_stack_alignment): New function.
9365 (ix86_finalize_stack_frame_flags): Use it. Set
9366 max_used_stack_alignment if we don't realign stack.
9367 * config/i386/i386.h (machine_function): Add
9368 max_used_stack_alignment.
9370 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
9372 * config/arm/arm.opt (-mbranch-cost): New option.
9373 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
9376 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
9379 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
9380 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
9382 2018-01-10 Richard Biener <rguenther@suse.de>
9385 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
9386 early out so it also covers the case where we have a non-NULL
9389 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9391 PR tree-optimization/83753
9392 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
9393 for non-strided grouped accesses if the number of elements is 1.
9395 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9398 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
9399 * i386.h (TARGET_USE_GATHER): Define.
9400 * x86-tune.def (X86_TUNE_USE_GATHER): New.
9402 2018-01-10 Martin Liska <mliska@suse.cz>
9405 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
9406 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
9408 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
9409 CLEANUP_NO_PARTITIONING is not set.
9411 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9413 * doc/rtl.texi: Remove documentation of (const ...) wrappers
9414 for vectors, as a partial revert of r254296.
9415 * rtl.h (const_vec_p): Delete.
9416 (const_vec_duplicate_p): Don't test for vector CONSTs.
9417 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
9418 * expmed.c (make_tree): Likewise.
9421 * common.md (E, F): Use CONSTANT_P instead of checking for
9423 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9424 checking for CONST_VECTOR.
9426 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9429 * predict.c (force_edge_cold): Handle in more sane way edges
9432 2018-01-09 Carl Love <cel@us.ibm.com>
9434 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
9436 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
9437 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
9438 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
9439 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
9440 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
9441 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
9442 * config/rs6000/rs6000-protos.h: Add extern defition for
9443 rs6000_generate_float2_double_code.
9444 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
9446 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
9447 (float2_v2df): Add define_expand.
9449 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
9452 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
9453 op_mode in the force_to_mode call.
9455 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9457 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
9458 instead of checking each element individually.
9459 (aarch64_evpc_uzp): Likewise.
9460 (aarch64_evpc_zip): Likewise.
9461 (aarch64_evpc_ext): Likewise.
9462 (aarch64_evpc_rev): Likewise.
9463 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
9464 instead of checking each element individually. Return true without
9466 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
9467 whether all selected elements come from the same input, instead of
9468 checking each element individually. Remove calls to gen_rtx_REG,
9469 start_sequence and end_sequence and instead assert that no rtl is
9472 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9474 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
9475 order of HIGH and CONST checks.
9477 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9479 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
9480 if the destination isn't an SSA_NAME.
9482 2018-01-09 Richard Biener <rguenther@suse.de>
9484 PR tree-optimization/83668
9485 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
9487 (canonicalize_loop_form): ... here, renamed from ...
9488 (canonicalize_loop_closed_ssa_form): ... this and amended to
9489 swap successor edges for loop exit blocks to make us use
9490 the RPO order we need for initial schedule generation.
9492 2018-01-09 Joseph Myers <joseph@codesourcery.com>
9494 PR tree-optimization/64811
9495 * match.pd: When optimizing comparisons with Inf, avoid
9496 introducing or losing exceptions from comparisons with NaN.
9498 2018-01-09 Martin Liska <mliska@suse.cz>
9501 * asan.c (shadow_mem_size): Add gcc_assert.
9503 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
9505 Don't save registers in main().
9508 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
9509 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
9510 * config/avr/avr.c (avr_set_current_function): Don't error if
9511 naked, OS_task or OS_main are specified at the same time.
9512 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
9514 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
9516 * common/config/avr/avr-common.c (avr_option_optimization_table):
9517 Switch on -mmain-is-OS_task for optimizing compilations.
9519 2018-01-09 Richard Biener <rguenther@suse.de>
9521 PR tree-optimization/83572
9522 * graphite.c: Include cfganal.h.
9523 (graphite_transform_loops): Connect infinite loops to exit
9524 and remove fake edges at the end.
9526 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9528 * ipa-inline.c (edge_badness): Revert accidental checkin.
9530 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9533 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
9534 symbols; not inline clones.
9536 2018-01-09 Jakub Jelinek <jakub@redhat.com>
9539 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
9540 hard registers. Formatting fixes.
9542 PR preprocessor/83722
9543 * gcc.c (try_generate_repro): Pass
9544 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
9545 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
9548 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
9549 Kito Cheng <kito.cheng@gmail.com>
9551 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
9552 (riscv_leaf_function_p): Delete.
9553 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
9555 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9557 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
9559 (do_ifelse): New function.
9560 (do_isel): New function.
9561 (do_sub3): New function.
9562 (do_add3): New function.
9563 (do_load_mask_compare): New function.
9564 (do_overlap_load_compare): New function.
9565 (expand_compare_loop): New function.
9566 (expand_block_compare): Call expand_compare_loop() when appropriate.
9567 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
9569 (-mblock-compare-inline-loop-limit): New option.
9571 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9574 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
9575 Reverse order of second and third operands in first alternative.
9576 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
9577 of first and second elements in UNSPEC_VPERMR vector.
9578 (altivec_expand_vec_perm_le): Likewise.
9580 2018-01-08 Jeff Law <law@redhat.com>
9582 PR rtl-optimizatin/81308
9583 * tree-switch-conversion.c (cfg_altered): New file scoped static.
9584 (process_switch): If group_case_labels makes a change, then set
9586 (pass_convert_switch::execute): If a switch is converted, then
9587 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
9589 PR rtl-optimization/81308
9590 * recog.c (split_all_insns): Conditionally cleanup the CFG after
9593 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
9595 PR target/83663 - Revert r255946
9596 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
9597 generation for cases where splatting a value is not useful.
9598 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
9599 across a vec_duplicate and a paradoxical subreg forming a vector
9600 mode to a vec_concat.
9602 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9604 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
9605 -march=armv8.3-a variants.
9606 * config/arm/t-multilib: Likewise.
9607 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
9609 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9611 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
9613 (cceq_ior_compare_complement): Give it a name so I can use it, and
9614 change boolean_or_operator predicate to boolean_operator so it can
9615 be used to generate a crand.
9616 (eqne): New code iterator.
9617 (bd/bd_neg): New code_attrs.
9618 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
9619 a single define_insn.
9620 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
9621 decrement (bdnzt/bdnzf/bdzt/bdzf).
9622 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
9623 with the new names of the branch decrement patterns, and added the
9624 names of the branch decrement conditional patterns.
9626 2018-01-08 Richard Biener <rguenther@suse.de>
9628 PR tree-optimization/83563
9629 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
9632 2018-01-08 Richard Biener <rguenther@suse.de>
9635 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
9637 2018-01-08 Richard Biener <rguenther@suse.de>
9639 PR tree-optimization/83685
9640 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
9641 references to abnormals.
9643 2018-01-08 Richard Biener <rguenther@suse.de>
9646 * dwarf2out.c (output_indirect_strings): Handle empty
9647 skeleton_debug_str_hash.
9648 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
9650 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9652 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
9653 (emit_store_direct): Likewise.
9654 (arc_trampoline_adjust_address): Likewise.
9655 (arc_asm_trampoline_template): New function.
9656 (arc_initialize_trampoline): Use asm_trampoline_template.
9657 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
9658 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
9659 * config/arc/arc.md (flush_icache): Delete pattern.
9661 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9663 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
9664 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
9667 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9670 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
9671 by not USED_FOR_TARGET.
9672 (make_pass_resolve_sw_modes): Likewise.
9674 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9676 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
9679 2018-01-08 Richard Biener <rguenther@suse.de>
9682 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
9684 2018-01-08 Richard Biener <rguenther@suse.de>
9687 * match.pd ((t * 2) / 2) -> t): Add missing :c.
9689 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
9692 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
9693 basic blocks with a small number of successors.
9694 (convert_control_dep_chain_into_preds): Improve handling of
9696 (dump_predicates): Split apart into...
9697 (dump_pred_chain): ...here...
9698 (dump_pred_info): ...and here.
9699 (can_one_predicate_be_invalidated_p): Add debugging printfs.
9700 (can_chain_union_be_invalidated_p): Improve check for invalidation
9702 (uninit_uses_cannot_happen): Avoid unnecessary if
9703 convert_control_dep_chain_into_preds yielded nothing.
9705 2018-01-06 Martin Sebor <msebor@redhat.com>
9707 PR tree-optimization/83640
9708 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
9709 subtracting negative offset from size.
9710 (builtin_access::overlap): Adjust offset bounds of the access to fall
9711 within the size of the object if possible.
9713 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
9715 PR rtl-optimization/83699
9716 * expmed.c (extract_bit_field_1): Restrict the vector usage of
9717 extract_bit_field_as_subreg to cases in which the extracted
9718 value is also a vector.
9720 * lra-constraints.c (process_alt_operands): Test for the equivalence
9721 substitutions when detecting a possible reload cycle.
9723 2018-01-06 Jakub Jelinek <jakub@redhat.com>
9726 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
9727 by default if flag_selective_schedling{,2}. Formatting fixes.
9729 PR rtl-optimization/83682
9730 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
9731 if it has non-VECTOR_MODE element mode.
9732 (vec_duplicate_p): Likewise.
9735 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
9736 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
9738 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9741 * config/i386/i386-builtin.def
9742 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
9743 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
9744 Require also OPTION_MASK_ISA_AVX512F in addition to
9745 OPTION_MASK_ISA_GFNI.
9746 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
9747 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
9748 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
9749 to OPTION_MASK_ISA_GFNI.
9750 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
9751 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
9752 OPTION_MASK_ISA_AVX512BW.
9753 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
9754 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
9755 addition to OPTION_MASK_ISA_GFNI.
9756 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
9757 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
9758 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
9759 to OPTION_MASK_ISA_GFNI.
9760 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
9761 a requirement for all ISAs rather than any of them with a few
9763 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
9765 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
9766 bitmasks to be enabled with 3 exceptions, instead of requiring any
9767 enabled ISA with lots of exceptions.
9768 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
9769 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
9770 Change avx512bw in isa attribute to avx512f.
9771 * config/i386/sgxintrin.h: Add license boilerplate.
9772 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
9773 to __AVX512F__ and __AVX512VL to __AVX512VL__.
9774 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
9775 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
9777 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
9778 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
9779 temporarily sse2 rather than sse if not enabled already.
9782 * config/i386/sse.md (VI248_VLBW): Rename to ...
9783 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
9784 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
9785 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
9786 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
9787 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
9788 mode iterator instead of VI248_VLBW.
9790 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
9792 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
9793 (record_modified): Skip clobbers; add debug output.
9794 (param_change_prob): Use sreal frequencies.
9796 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9798 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
9799 punt for user-aligned variables.
9801 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9803 * tree-chrec.c (chrec_contains_symbols): Return true for
9806 2018-01-05 Sudakshina Das <sudi.das@arm.com>
9809 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
9810 of (x|y) == x for BICS pattern.
9812 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9814 PR tree-optimization/83605
9815 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
9816 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
9819 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
9821 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
9822 * config/epiphany/rtems.h: New file.
9824 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9825 Uros Bizjak <ubizjak@gmail.com>
9828 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
9829 QIreg_operand instead of register_operand predicate.
9830 * config/i386/i386.c (ix86_rop_should_change_byte_p,
9831 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
9832 comments instead of -fmitigate[-_]rop.
9834 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9837 * cgraphunit.c (symbol_table::compile): Switch to text_section
9838 before calling assembly_start debug hook.
9839 * run-rtl-passes.c (run_rtl_passes): Likewise.
9842 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9844 * tree-vrp.c (extract_range_from_binary_expr_1): Check
9845 range_int_cst_p rather than !symbolic_range_p before calling
9846 extract_range_from_multiplicative_op_1.
9848 2018-01-04 Jeff Law <law@redhat.com>
9850 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
9851 redundant test in assertion.
9853 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9855 * doc/rtl.texi: Document machine_mode wrapper classes.
9857 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9859 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
9862 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9864 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
9865 the VEC_PERM_EXPR fold to fail.
9867 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9870 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
9871 to switched_sections.
9873 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9876 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
9879 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
9882 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
9883 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
9885 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9888 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
9889 is BLKmode and bitpos not zero or mode change is needed.
9891 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9894 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
9897 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
9900 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
9901 instead of MULT rtx. Update all corresponding splitters.
9903 (*ssub<modesuffix>): Ditto.
9905 (*cmp_sadd_di): Update split patterns.
9906 (*cmp_sadd_si): Ditto.
9907 (*cmp_sadd_sidi): Ditto.
9908 (*cmp_ssub_di): Ditto.
9909 (*cmp_ssub_si): Ditto.
9910 (*cmp_ssub_sidi): Ditto.
9911 * config/alpha/predicates.md (const23_operand): New predicate.
9912 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
9913 Look for ASHIFT, not MULT inner operand.
9914 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
9916 2018-01-04 Martin Liska <mliska@suse.cz>
9918 PR gcov-profile/83669
9919 * gcov.c (output_intermediate_file): Add version to intermediate
9921 * doc/gcov.texi: Document new field 'version' in intermediate
9922 file format. Fix location of '-k' option of gcov command.
9924 2018-01-04 Martin Liska <mliska@suse.cz>
9927 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
9929 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9931 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
9933 2018-01-03 Martin Sebor <msebor@redhat.com>
9935 PR tree-optimization/83655
9936 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
9937 checking calls with invalid arguments.
9939 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9941 * tree-vect-stmts.c (vect_get_store_rhs): New function.
9942 (vectorizable_mask_load_store): Delete.
9943 (vectorizable_call): Return false for masked loads and stores.
9944 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
9945 instead of gimple_assign_rhs1.
9946 (vectorizable_load): Handle IFN_MASK_LOAD.
9947 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
9949 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9951 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
9953 (vectorizable_mask_load_store): ...here.
9954 (vectorizable_load): ...and here.
9956 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9958 * tree-vect-stmts.c (vect_build_all_ones_mask)
9959 (vect_build_zero_merge_argument): New functions, split out from...
9960 (vectorizable_load): ...here.
9962 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9964 * tree-vect-stmts.c (vect_check_store_rhs): New function,
9966 (vectorizable_mask_load_store): ...here.
9967 (vectorizable_store): ...and here.
9969 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9971 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
9973 (vectorizable_mask_load_store): ...here.
9975 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9977 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
9978 (vect_model_store_cost): Take a vec_load_store_type instead of a
9980 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
9981 (vect_model_store_cost): Take a vec_load_store_type instead of a
9983 (vectorizable_mask_load_store): Update accordingly.
9984 (vectorizable_store): Likewise.
9985 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
9987 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9989 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
9990 IFN_MASK_LOAD calls here rather than...
9991 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
9993 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9994 Alan Hayward <alan.hayward@arm.com>
9995 David Sherwood <david.sherwood@arm.com>
9997 * expmed.c (extract_bit_field_1): For vector extracts,
9998 fall back to extract_bit_field_as_subreg if vec_extract
10001 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10002 Alan Hayward <alan.hayward@arm.com>
10003 David Sherwood <david.sherwood@arm.com>
10005 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
10006 they are variable or constant sized.
10007 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
10008 slots for constant-sized data.
10010 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10011 Alan Hayward <alan.hayward@arm.com>
10012 David Sherwood <david.sherwood@arm.com>
10014 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
10015 handling COND_EXPRs with boolean comparisons, try to find a better
10016 basis for the mask type than the boolean itself.
10018 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10020 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
10021 is calculated and how it can be overridden.
10022 * genmodes.c (max_bitsize_mode_any_mode): New variable.
10023 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
10025 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
10028 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10029 Alan Hayward <alan.hayward@arm.com>
10030 David Sherwood <david.sherwood@arm.com>
10032 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
10033 Remove the mode argument.
10034 (aarch64_simd_valid_immediate): Remove the mode and inverse
10036 * config/aarch64/iterators.md (bitsize): New iterator.
10037 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
10038 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
10039 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
10040 aarch64_simd_valid_immediate.
10041 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
10042 (aarch64_reg_or_bic_imm): Likewise.
10043 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
10044 with an insn_type enum and msl with a modifier_type enum.
10045 Replace element_width with a scalar_mode. Change the shift
10046 to unsigned int. Add constructors for scalar_float_mode and
10047 scalar_int_mode elements.
10048 (aarch64_vect_float_const_representable_p): Delete.
10049 (aarch64_can_const_movi_rtx_p)
10050 (aarch64_simd_scalar_immediate_valid_for_move)
10051 (aarch64_simd_make_constant): Update call to
10052 aarch64_simd_valid_immediate.
10053 (aarch64_advsimd_valid_immediate_hs): New function.
10054 (aarch64_advsimd_valid_immediate): Likewise.
10055 (aarch64_simd_valid_immediate): Remove mode and inverse
10056 arguments. Rewrite to use the above. Use const_vec_duplicate_p
10057 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
10058 and aarch64_float_const_representable_p on the result.
10059 (aarch64_output_simd_mov_immediate): Remove mode argument.
10060 Update call to aarch64_simd_valid_immediate and use of
10061 simd_immediate_info.
10062 (aarch64_output_scalar_simd_mov_immediate): Update call
10065 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10066 Alan Hayward <alan.hayward@arm.com>
10067 David Sherwood <david.sherwood@arm.com>
10069 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
10070 (mode_nunits): Likewise CONST_MODE_NUNITS.
10071 * machmode.def (ADJUST_NUNITS): Document.
10072 * genmodes.c (mode_data::need_nunits_adj): New field.
10073 (blank_mode): Update accordingly.
10074 (adj_nunits): New variable.
10075 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
10077 (emit_mode_size_inline): Set need_bytesize_adj for all modes
10078 listed in adj_nunits.
10079 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
10080 listed in adj_nunits. Don't emit case statements for such modes.
10081 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
10082 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
10083 nothing if adj_nunits is nonnull.
10084 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
10085 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
10086 (emit_mode_fbit): Update use of print_maybe_const_decl.
10087 (emit_move_size): Likewise. Treat the array as non-const
10089 (emit_mode_adjustments): Handle adj_nunits.
10091 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10093 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
10094 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
10095 (VECTOR_MODES): Use it.
10096 (make_vector_modes): Take the prefix as an argument.
10098 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10099 Alan Hayward <alan.hayward@arm.com>
10100 David Sherwood <david.sherwood@arm.com>
10102 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
10103 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
10104 for MODE_VECTOR_BOOL.
10105 * machmode.def (VECTOR_BOOL_MODE): Document.
10106 * genmodes.c (VECTOR_BOOL_MODE): New macro.
10107 (make_vector_bool_mode): New function.
10108 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
10110 * lto-streamer-in.c (lto_input_mode_table): Likewise.
10111 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
10113 * stor-layout.c (int_mode_for_mode): Likewise.
10114 * tree.c (build_vector_type_for_mode): Likewise.
10115 * varasm.c (output_constant_pool_2): Likewise.
10116 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
10117 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
10118 for MODE_VECTOR_BOOL.
10119 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
10120 of mode class checks.
10121 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
10122 instead of a list of mode class checks.
10123 (expand_vector_scalar_condition): Likewise.
10124 (type_for_widest_vector_mode): Handle BImode as an inner mode.
10126 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10127 Alan Hayward <alan.hayward@arm.com>
10128 David Sherwood <david.sherwood@arm.com>
10130 * machmode.h (mode_size): Change from unsigned short to
10132 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
10133 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10134 or if measurement_type is not polynomial.
10135 (fixed_size_mode::includes_p): Check for constant-sized modes.
10136 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
10137 return a poly_uint16 rather than an unsigned short.
10138 (emit_mode_size): Change the type of mode_size from unsigned short
10139 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
10140 (emit_mode_adjustments): Cope with polynomial vector sizes.
10141 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10143 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10145 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
10146 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
10147 * caller-save.c (setup_save_areas): Likewise.
10148 (replace_reg_with_saved_mem): Likewise.
10149 * calls.c (emit_library_call_value_1): Likewise.
10150 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
10151 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
10152 (gen_lowpart_for_combine): Likewise.
10153 * convert.c (convert_to_integer_1): Likewise.
10154 * cse.c (equiv_constant, cse_insn): Likewise.
10155 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
10156 (cselib_subst_to_values): Likewise.
10157 * dce.c (word_dce_process_block): Likewise.
10158 * df-problems.c (df_word_lr_mark_ref): Likewise.
10159 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
10160 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
10161 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
10162 (rtl_for_decl_location): Likewise.
10163 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
10164 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
10165 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
10166 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
10167 (expand_expr_real_1): Likewise.
10168 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
10169 (pad_below): Likewise.
10170 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10171 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
10172 * ira.c (get_subreg_tracking_sizes): Likewise.
10173 * ira-build.c (ira_create_allocno_objects): Likewise.
10174 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
10175 (ira_sort_regnos_for_alter_reg): Likewise.
10176 * ira-costs.c (record_operand_costs): Likewise.
10177 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
10178 (resolve_simple_move): Likewise.
10179 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
10180 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
10181 (lra_constraints): Likewise.
10182 (CONST_POOL_OK_P): Reject variable-sized modes.
10183 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
10184 (add_pseudo_to_slot, lra_spill): Likewise.
10185 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10186 * optabs-query.c (get_best_extraction_insn): Likewise.
10187 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10188 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
10189 (expand_mult_highpart, valid_multiword_target_p): Likewise.
10190 * recog.c (offsettable_address_addr_space_p): Likewise.
10191 * regcprop.c (maybe_mode_change): Likewise.
10192 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
10193 * regrename.c (build_def_use): Likewise.
10194 * regstat.c (dump_reg_info): Likewise.
10195 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
10196 (find_reloads, find_reloads_subreg_address): Likewise.
10197 * reload1.c (eliminate_regs_1): Likewise.
10198 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
10199 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
10200 (simplify_binary_operation_1, simplify_subreg): Likewise.
10201 * targhooks.c (default_function_arg_padding): Likewise.
10202 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
10203 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
10204 (verify_gimple_assign_ternary): Likewise.
10205 * tree-inline.c (estimate_move_cost): Likewise.
10206 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10207 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
10208 (get_address_cost_ainc): Likewise.
10209 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
10210 (vect_supportable_dr_alignment): Likewise.
10211 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10212 (vectorizable_reduction): Likewise.
10213 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
10214 (vectorizable_operation, vectorizable_load): Likewise.
10215 * tree.c (build_same_sized_truth_vector_type): Likewise.
10216 * valtrack.c (cleanup_auto_inc_dec): Likewise.
10217 * var-tracking.c (emit_note_insn_var_location): Likewise.
10218 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
10219 (ADDR_VEC_ALIGN): Likewise.
10221 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10222 Alan Hayward <alan.hayward@arm.com>
10223 David Sherwood <david.sherwood@arm.com>
10225 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
10227 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10228 or if measurement_type is polynomial.
10229 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
10230 * combine.c (make_extraction): Likewise.
10231 * dse.c (find_shift_sequence): Likewise.
10232 * dwarf2out.c (mem_loc_descriptor): Likewise.
10233 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
10234 (extract_bit_field, extract_low_bits): Likewise.
10235 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
10236 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
10237 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
10238 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
10239 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10240 * reload.c (find_reloads): Likewise.
10241 * reload1.c (alter_reg): Likewise.
10242 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
10243 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
10244 * tree-if-conv.c (predicate_mem_writes): Likewise.
10245 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
10246 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
10247 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
10248 * valtrack.c (dead_debug_insert_temp): Likewise.
10249 * varasm.c (mergeable_constant_section): Likewise.
10250 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
10252 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10253 Alan Hayward <alan.hayward@arm.com>
10254 David Sherwood <david.sherwood@arm.com>
10256 * expr.c (expand_assignment): Cope with polynomial mode sizes
10257 when assigning to a CONCAT.
10259 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10260 Alan Hayward <alan.hayward@arm.com>
10261 David Sherwood <david.sherwood@arm.com>
10263 * machmode.h (mode_precision): Change from unsigned short to
10265 (mode_to_precision): Return a poly_uint16 rather than an unsigned
10267 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
10268 or if measurement_type is not polynomial.
10269 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
10270 in which the mode is already known to be a scalar_int_mode.
10271 * genmodes.c (emit_mode_precision): Change the type of mode_precision
10272 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
10274 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10275 for GET_MODE_PRECISION.
10276 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10277 for GET_MODE_PRECISION.
10278 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
10280 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
10281 (expand_field_assignment, make_extraction): Likewise.
10282 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
10283 (get_last_value): Likewise.
10284 * convert.c (convert_to_integer_1): Likewise.
10285 * cse.c (cse_insn): Likewise.
10286 * expr.c (expand_expr_real_1): Likewise.
10287 * lra-constraints.c (simplify_operand_subreg): Likewise.
10288 * optabs-query.c (can_atomic_load_p): Likewise.
10289 * optabs.c (expand_atomic_load): Likewise.
10290 (expand_atomic_store): Likewise.
10291 * ree.c (combine_reaching_defs): Likewise.
10292 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
10293 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
10294 * tree.h (type_has_mode_precision_p): Likewise.
10295 * ubsan.c (instrument_si_overflow): Likewise.
10297 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10298 Alan Hayward <alan.hayward@arm.com>
10299 David Sherwood <david.sherwood@arm.com>
10301 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
10302 polynomial numbers of units.
10303 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
10304 (valid_vector_subparts_p): New function.
10305 (build_vector_type): Remove temporary shim and take the number
10306 of units as a poly_uint64 rather than an int.
10307 (build_opaque_vector_type): Take the number of units as a
10308 poly_uint64 rather than an int.
10309 * tree.c (build_vector_from_ctor): Handle polynomial
10310 TYPE_VECTOR_SUBPARTS.
10311 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
10312 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
10313 (build_vector_from_val): If the number of units is variable,
10314 use build_vec_duplicate_cst for constant operands and
10315 VEC_DUPLICATE_EXPR otherwise.
10316 (make_vector_type): Remove temporary is_constant ().
10317 (build_vector_type, build_opaque_vector_type): Take the number of
10318 units as a poly_uint64 rather than an int.
10319 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
10321 * cfgexpand.c (expand_debug_expr): Likewise.
10322 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
10323 (store_constructor, expand_expr_real_1): Likewise.
10324 (const_scalar_mask_from_tree): Likewise.
10325 * fold-const-call.c (fold_const_reduction): Likewise.
10326 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
10327 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
10328 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
10329 (fold_relational_const): Likewise.
10330 (native_interpret_vector): Likewise. Change the size from an
10331 int to an unsigned int.
10332 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
10333 TYPE_VECTOR_SUBPARTS.
10334 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
10335 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
10336 duplicating a non-constant operand into a variable-length vector.
10337 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
10338 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
10339 * ipa-icf.c (sem_variable::equals): Likewise.
10340 * match.pd: Likewise.
10341 * omp-simd-clone.c (simd_clone_subparts): Likewise.
10342 * print-tree.c (print_node): Likewise.
10343 * stor-layout.c (layout_type): Likewise.
10344 * targhooks.c (default_builtin_vectorization_cost): Likewise.
10345 * tree-cfg.c (verify_gimple_comparison): Likewise.
10346 (verify_gimple_assign_binary): Likewise.
10347 (verify_gimple_assign_ternary): Likewise.
10348 (verify_gimple_assign_single): Likewise.
10349 * tree-pretty-print.c (dump_generic_node): Likewise.
10350 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10351 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
10352 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
10353 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
10354 (vect_shift_permute_load_chain): Likewise.
10355 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
10356 (expand_vector_condition, optimize_vector_constructor): Likewise.
10357 (lower_vec_perm, get_compute_type): Likewise.
10358 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10359 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
10360 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
10361 (vect_recog_mask_conversion_pattern): Likewise.
10362 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
10363 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
10364 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10365 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
10366 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
10367 (vectorizable_shift, vectorizable_operation, vectorizable_store)
10368 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
10369 (supportable_widening_operation): Likewise.
10370 (supportable_narrowing_operation): Likewise.
10371 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
10373 * varasm.c (output_constant): Likewise.
10375 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10376 Alan Hayward <alan.hayward@arm.com>
10377 David Sherwood <david.sherwood@arm.com>
10379 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
10380 so that both the length == 3 and length != 3 cases set up their
10381 own permute vectors. Add comments explaining why we know the
10382 number of elements is constant.
10383 (vect_permute_load_chain): Likewise.
10385 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10386 Alan Hayward <alan.hayward@arm.com>
10387 David Sherwood <david.sherwood@arm.com>
10389 * machmode.h (mode_nunits): Change from unsigned char to
10391 (ONLY_FIXED_SIZE_MODES): New macro.
10392 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
10393 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
10394 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
10396 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
10397 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
10398 or if measurement_type is not polynomial.
10399 * genmodes.c (ZERO_COEFFS): New macro.
10400 (emit_mode_nunits_inline): Make mode_nunits_inline return a
10402 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
10403 Use ZERO_COEFFS when emitting initializers.
10404 * data-streamer.h (bp_pack_poly_value): New function.
10405 (bp_unpack_poly_value): Likewise.
10406 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10407 for GET_MODE_NUNITS.
10408 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10409 for GET_MODE_NUNITS.
10410 * tree.c (make_vector_type): Remove temporary shim and make
10411 the real function take the number of units as a poly_uint64
10412 rather than an int.
10413 (build_vector_type_for_mode): Handle polynomial nunits.
10414 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
10415 * emit-rtl.c (const_vec_series_p_1): Likewise.
10416 (gen_rtx_CONST_VECTOR): Likewise.
10417 * fold-const.c (test_vec_duplicate_folding): Likewise.
10418 * genrecog.c (validate_pattern): Likewise.
10419 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
10420 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10421 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
10422 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
10423 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
10424 * rtlanal.c (subreg_get_info): Likewise.
10425 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10426 (vect_grouped_load_supported): Likewise.
10427 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
10428 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
10429 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
10430 (simplify_const_unary_operation, simplify_binary_operation_1)
10431 (simplify_const_binary_operation, simplify_ternary_operation)
10432 (test_vector_ops_duplicate, test_vector_ops): Likewise.
10433 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
10434 instead of CONST_VECTOR_NUNITS.
10435 * varasm.c (output_constant_pool_2): Likewise.
10436 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
10437 explicit-encoded elements in the XVEC for variable-length vectors.
10439 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10441 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
10443 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10444 Alan Hayward <alan.hayward@arm.com>
10445 David Sherwood <david.sherwood@arm.com>
10447 * coretypes.h (fixed_size_mode): Declare.
10448 (fixed_size_mode_pod): New typedef.
10449 * builtins.h (target_builtins::x_apply_args_mode)
10450 (target_builtins::x_apply_result_mode): Change type to
10451 fixed_size_mode_pod.
10452 * builtins.c (apply_args_size, apply_result_size, result_vector)
10453 (expand_builtin_apply_args_1, expand_builtin_apply)
10454 (expand_builtin_return): Update accordingly.
10456 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10458 * cse.c (hash_rtx_cb): Hash only the encoded elements.
10459 * cselib.c (cselib_hash_rtx): Likewise.
10460 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
10461 CONST_VECTOR encoding.
10463 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10464 Jeff Law <law@redhat.com>
10467 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
10468 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
10469 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
10470 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
10473 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
10474 explicitly probe *sp in a noreturn function if there were any callee
10475 register saves or frame pointer is needed.
10477 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10480 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
10481 BLKmode for ternary, binary or unary expressions.
10484 * var-tracking.c (delete_vta_debug_insn): New inline function.
10485 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
10486 insns from get_insns () to NULL instead of each bb separately.
10487 Use delete_vta_debug_insn. No longer static.
10488 (vt_debug_insns_local, variable_tracking_main_1): Adjust
10489 delete_vta_debug_insns callers.
10490 * rtl.h (delete_vta_debug_insns): Declare.
10491 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
10492 instead of variable_tracking_main.
10494 2018-01-03 Martin Sebor <msebor@redhat.com>
10496 PR tree-optimization/83603
10497 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
10498 arguments past the endof the argument list in functions declared
10499 without a prototype.
10500 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
10501 Avoid checking when arguments are null.
10503 2018-01-03 Martin Sebor <msebor@redhat.com>
10506 * doc/extend.texi (attribute const): Fix a typo.
10507 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
10508 issuing -Wsuggest-attribute for void functions.
10510 2018-01-03 Martin Sebor <msebor@redhat.com>
10512 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
10513 offset_int::from instead of wide_int::to_shwi.
10514 (maybe_diag_overlap): Remove assertion.
10515 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
10516 * gimple-ssa-sprintf.c (format_directive): Same.
10517 (parse_directive): Same.
10518 (sprintf_dom_walker::compute_format_length): Same.
10519 (try_substitute_return_value): Same.
10521 2018-01-03 Jeff Law <law@redhat.com>
10523 PR middle-end/83654
10524 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
10525 non-constant residual for zero at runtime and avoid probing in
10526 that case. Reorganize code for trailing problem to mirror handling
10529 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10531 PR tree-optimization/83501
10532 * tree-ssa-strlen.c (get_string_cst): New.
10533 (handle_char_store): Call get_string_cst.
10535 2018-01-03 Martin Liska <mliska@suse.cz>
10537 PR tree-optimization/83593
10538 * tree-ssa-strlen.c: Include tree-cfg.h.
10539 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
10540 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
10541 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
10543 (strlen_dom_walker::before_dom_children): Call
10544 gimple_purge_dead_eh_edges. Dump tranformation with details
10546 (strlen_dom_walker::before_dom_children): Update call by adding
10547 new argument cleanup_eh.
10548 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
10550 2018-01-03 Martin Liska <mliska@suse.cz>
10553 * cif-code.def (VARIADIC_THUNK): New enum value.
10554 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
10557 2018-01-03 Jan Beulich <jbeulich@suse.com>
10559 * sse.md (mov<mode>_internal): Tighten condition for when to use
10560 vmovdqu<ssescalarsize> for TI and OI modes.
10562 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10564 Update copyright years.
10566 2018-01-03 Martin Liska <mliska@suse.cz>
10569 * ipa-visibility.c (function_and_variable_visibility): Skip
10570 functions with noipa attribure.
10572 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10574 * gcc.c (process_command): Update copyright notice dates.
10575 * gcov-dump.c (print_version): Ditto.
10576 * gcov.c (print_version): Ditto.
10577 * gcov-tool.c (print_version): Ditto.
10578 * gengtype.c (create_file): Ditto.
10579 * doc/cpp.texi: Bump @copying's copyright year.
10580 * doc/cppinternals.texi: Ditto.
10581 * doc/gcc.texi: Ditto.
10582 * doc/gccint.texi: Ditto.
10583 * doc/gcov.texi: Ditto.
10584 * doc/install.texi: Ditto.
10585 * doc/invoke.texi: Ditto.
10587 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10589 * vector-builder.h (vector_builder::m_full_nelts): Change from
10590 unsigned int to poly_uint64.
10591 (vector_builder::full_nelts): Update prototype accordingly.
10592 (vector_builder::new_vector): Likewise.
10593 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
10594 (vector_builder::operator ==): Likewise.
10595 (vector_builder::finalize): Likewise.
10596 * int-vector-builder.h (int_vector_builder::int_vector_builder):
10597 Take the number of elements as a poly_uint64 rather than an
10599 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
10600 from unsigned int to poly_uint64.
10601 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
10602 (vec_perm_indices::new_vector): Likewise.
10603 (vec_perm_indices::length): Likewise.
10604 (vec_perm_indices::nelts_per_input): Likewise.
10605 (vec_perm_indices::input_nelts): Likewise.
10606 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
10607 number of elements per input as a poly_uint64 rather than an
10608 unsigned int. Use the original encoding for variable-length
10609 vectors, rather than clamping each individual element.
10610 For the second and subsequent elements in each pattern,
10611 clamp the step and base before clamping their sum.
10612 (vec_perm_indices::series_p): Handle polynomial element counts.
10613 (vec_perm_indices::all_in_range_p): Likewise.
10614 (vec_perm_indices_to_tree): Likewise.
10615 (vec_perm_indices_to_rtx): Likewise.
10616 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
10617 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
10618 (tree_vector_builder::new_binary_operation): Handle polynomial
10619 element counts. Return false if we need to know the number
10620 of elements at compile time.
10621 * fold-const.c (fold_vec_perm): Punt if the number of elements
10622 isn't known at compile time.
10624 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10626 * vec-perm-indices.h (vec_perm_builder): Change element type
10627 from HOST_WIDE_INT to poly_int64.
10628 (vec_perm_indices::element_type): Update accordingly.
10629 (vec_perm_indices::clamp): Handle polynomial element_types.
10630 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10631 (vec_perm_indices::all_in_range_p): Likewise.
10632 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
10634 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
10635 polynomial vec_perm_indices element types.
10636 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
10637 * fold-const.c (fold_vec_perm): Likewise.
10638 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
10639 * tree-vect-generic.c (lower_vec_perm): Likewise.
10640 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10641 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
10642 element type to HOST_WIDE_INT.
10644 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10645 Alan Hayward <alan.hayward@arm.com>
10646 David Sherwood <david.sherwood@arm.com>
10648 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
10649 rather than an int. Use plus_constant.
10650 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
10651 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
10653 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10654 Alan Hayward <alan.hayward@arm.com>
10655 David Sherwood <david.sherwood@arm.com>
10657 * calls.c (emit_call_1, expand_call): Change struct_value_size from
10658 a HOST_WIDE_INT to a poly_int64.
10660 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10661 Alan Hayward <alan.hayward@arm.com>
10662 David Sherwood <david.sherwood@arm.com>
10664 * calls.c (load_register_parameters): Cope with polynomial
10665 mode sizes. Require a constant size for BLKmode parameters
10666 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
10667 forces a parameter to be padded at the lsb end in order to
10668 fill a complete number of words, require the parameter size
10669 to be ordered wrt UNITS_PER_WORD.
10671 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10672 Alan Hayward <alan.hayward@arm.com>
10673 David Sherwood <david.sherwood@arm.com>
10675 * reload1.c (spill_stack_slot_width): Change element type
10676 from unsigned int to poly_uint64_pod.
10677 (alter_reg): Treat mode sizes as polynomial.
10679 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10680 Alan Hayward <alan.hayward@arm.com>
10681 David Sherwood <david.sherwood@arm.com>
10683 * reload.c (complex_word_subreg_p): New function.
10684 (reload_inner_reg_of_subreg, push_reload): Use it.
10686 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10687 Alan Hayward <alan.hayward@arm.com>
10688 David Sherwood <david.sherwood@arm.com>
10690 * lra-constraints.c (process_alt_operands): Reject matched
10691 operands whose sizes aren't ordered.
10692 (match_reload): Refer to this check here.
10694 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10695 Alan Hayward <alan.hayward@arm.com>
10696 David Sherwood <david.sherwood@arm.com>
10698 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
10699 that the mode size is in the set {1, 2, 4, 8, 16}.
10701 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10702 Alan Hayward <alan.hayward@arm.com>
10703 David Sherwood <david.sherwood@arm.com>
10705 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
10706 Use plus_constant instead of gen_rtx_PLUS.
10708 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10709 Alan Hayward <alan.hayward@arm.com>
10710 David Sherwood <david.sherwood@arm.com>
10712 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
10713 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
10714 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
10715 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
10716 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
10717 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
10718 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
10719 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
10720 * config/i386/i386.c (ix86_push_rounding): ...this new function.
10721 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
10723 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
10724 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
10725 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
10726 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
10727 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
10728 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
10729 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
10730 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
10731 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
10732 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
10734 * expr.c (emit_move_resolve_push): Treat the input and result
10735 of PUSH_ROUNDING as a poly_int64.
10736 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
10737 (emit_push_insn): Likewise.
10738 * lra-eliminations.c (mark_not_eliminable): Likewise.
10739 * recog.c (push_operand): Likewise.
10740 * reload1.c (elimination_effects): Likewise.
10741 * rtlanal.c (nonzero_bits1): Likewise.
10742 * calls.c (store_one_arg): Likewise. Require the padding to be
10743 known at compile time.
10745 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10746 Alan Hayward <alan.hayward@arm.com>
10747 David Sherwood <david.sherwood@arm.com>
10749 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
10750 Use plus_constant instead of gen_rtx_PLUS.
10752 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10753 Alan Hayward <alan.hayward@arm.com>
10754 David Sherwood <david.sherwood@arm.com>
10756 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
10757 rather than an int.
10759 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10760 Alan Hayward <alan.hayward@arm.com>
10761 David Sherwood <david.sherwood@arm.com>
10763 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
10764 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
10765 via stack temporaries. Treat the mode size as polynomial too.
10767 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10768 Alan Hayward <alan.hayward@arm.com>
10769 David Sherwood <david.sherwood@arm.com>
10771 * expr.c (expand_expr_real_2): When handling conversions involving
10772 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
10773 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
10774 as a poly_uint64 too.
10776 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10777 Alan Hayward <alan.hayward@arm.com>
10778 David Sherwood <david.sherwood@arm.com>
10780 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
10782 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10783 Alan Hayward <alan.hayward@arm.com>
10784 David Sherwood <david.sherwood@arm.com>
10786 * combine.c (can_change_dest_mode): Handle polynomial
10787 REGMODE_NATURAL_SIZE.
10788 * expmed.c (store_bit_field_1): Likewise.
10789 * expr.c (store_constructor): Likewise.
10790 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
10791 and polynomial REGMODE_NATURAL_SIZE.
10792 (gen_lowpart_common): Likewise.
10793 * reginfo.c (record_subregs_of_mode): Likewise.
10794 * rtlanal.c (read_modify_subreg_p): Likewise.
10796 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10797 Alan Hayward <alan.hayward@arm.com>
10798 David Sherwood <david.sherwood@arm.com>
10800 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
10801 numbers of elements.
10803 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10804 Alan Hayward <alan.hayward@arm.com>
10805 David Sherwood <david.sherwood@arm.com>
10807 * match.pd: Cope with polynomial numbers of vector elements.
10809 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10810 Alan Hayward <alan.hayward@arm.com>
10811 David Sherwood <david.sherwood@arm.com>
10813 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
10814 in a POINTER_PLUS_EXPR.
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 * omp-simd-clone.c (simd_clone_subparts): New function.
10821 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
10822 (ipa_simd_modify_function_body): Likewise.
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-generic.c (nunits_for_known_piecewise_op): New function.
10829 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
10830 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
10831 (expand_vector_condition, vector_element): Likewise.
10832 (subparts_gt): New function.
10833 (get_compute_type): Use subparts_gt.
10834 (count_type_subparts): Delete.
10835 (expand_vector_operations_1): Use subparts_gt instead of
10836 count_type_subparts.
10838 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10839 Alan Hayward <alan.hayward@arm.com>
10840 David Sherwood <david.sherwood@arm.com>
10842 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
10843 (vect_compile_time_alias): ...this new function. Do the calculation
10844 on poly_ints rather than trees.
10845 (vect_prune_runtime_alias_test_list): Update call accordingly.
10847 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10848 Alan Hayward <alan.hayward@arm.com>
10849 David Sherwood <david.sherwood@arm.com>
10851 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
10853 (vect_schedule_slp_instance): Likewise.
10855 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10856 Alan Hayward <alan.hayward@arm.com>
10857 David Sherwood <david.sherwood@arm.com>
10859 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
10860 constant and extern definitions for variable-length vectors.
10861 (vect_get_constant_vectors): Note that the number of units
10862 is known to be constant.
10864 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10865 Alan Hayward <alan.hayward@arm.com>
10866 David Sherwood <david.sherwood@arm.com>
10868 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
10869 of units as polynomial. Choose between WIDE and NARROW based
10872 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10873 Alan Hayward <alan.hayward@arm.com>
10874 David Sherwood <david.sherwood@arm.com>
10876 * tree-vect-stmts.c (simd_clone_subparts): New function.
10877 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
10879 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10880 Alan Hayward <alan.hayward@arm.com>
10881 David Sherwood <david.sherwood@arm.com>
10883 * tree-vect-stmts.c (vectorizable_call): Treat the number of
10884 vectors as polynomial. Use build_index_vector for
10885 IFN_GOMP_SIMD_LANE.
10887 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10888 Alan Hayward <alan.hayward@arm.com>
10889 David Sherwood <david.sherwood@arm.com>
10891 * tree-vect-stmts.c (get_load_store_type): Treat the number of
10892 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
10893 for variable-length vectors.
10894 (vectorizable_mask_load_store): Treat the number of units as
10895 polynomial, asserting that it is constant if the condition has
10896 already been enforced.
10897 (vectorizable_store, vectorizable_load): Likewise.
10899 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10900 Alan Hayward <alan.hayward@arm.com>
10901 David Sherwood <david.sherwood@arm.com>
10903 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
10904 of units as polynomial. Punt if we can't tell at compile time
10905 which vector contains the final result.
10907 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10908 Alan Hayward <alan.hayward@arm.com>
10909 David Sherwood <david.sherwood@arm.com>
10911 * tree-vect-loop.c (vectorizable_induction): Treat the number
10912 of units as polynomial. Punt on SLP inductions. Use an integer
10913 VEC_SERIES_EXPR for variable-length integer reductions. Use a
10914 cast of such a series for variable-length floating-point
10917 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10918 Alan Hayward <alan.hayward@arm.com>
10919 David Sherwood <david.sherwood@arm.com>
10921 * tree.h (build_index_vector): Declare.
10922 * tree.c (build_index_vector): New function.
10923 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
10924 of units as polynomial, forcibly converting it to a constant if
10925 vectorizable_reduction has already enforced the condition.
10926 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
10927 to create a {1,2,3,...} vector.
10928 (vectorizable_reduction): Treat the number of units as polynomial.
10929 Choose vectype_in based on the largest scalar element size rather
10930 than the smallest number of units. Enforce the restrictions
10933 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10934 Alan Hayward <alan.hayward@arm.com>
10935 David Sherwood <david.sherwood@arm.com>
10937 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
10938 number of units as polynomial.
10940 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10941 Alan Hayward <alan.hayward@arm.com>
10942 David Sherwood <david.sherwood@arm.com>
10944 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
10945 * target.def (autovectorize_vector_sizes): Return the vector sizes
10946 by pointer, using vector_sizes rather than a bitmask.
10947 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
10948 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
10949 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
10951 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
10952 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
10953 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
10954 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
10955 * omp-general.c (omp_max_vf): Likewise.
10956 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10957 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
10958 * tree-vect-loop.c (vect_analyze_loop): Likewise.
10959 * tree-vect-slp.c (vect_slp_bb): Likewise.
10960 * doc/tm.texi: Regenerate.
10961 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
10963 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
10964 the vector size as a poly_uint64 rather than an unsigned int.
10965 (current_vector_size): Change from an unsigned int to a poly_uint64.
10966 (get_vectype_for_scalar_type): Update accordingly.
10967 * tree.h (build_truth_vector_type): Take the size and number of
10968 units as a poly_uint64 rather than an unsigned int.
10969 (build_vector_type): Add a temporary overload that takes
10970 the number of units as a poly_uint64 rather than an unsigned int.
10971 * tree.c (make_vector_type): Likewise.
10972 (build_truth_vector_type): Take the number of units as a poly_uint64
10973 rather than an unsigned int.
10975 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10976 Alan Hayward <alan.hayward@arm.com>
10977 David Sherwood <david.sherwood@arm.com>
10979 * target.def (get_mask_mode): Take the number of units and length
10980 as poly_uint64s rather than unsigned ints.
10981 * targhooks.h (default_get_mask_mode): Update accordingly.
10982 * targhooks.c (default_get_mask_mode): Likewise.
10983 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
10984 * doc/tm.texi: Regenerate.
10986 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10987 Alan Hayward <alan.hayward@arm.com>
10988 David Sherwood <david.sherwood@arm.com>
10990 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
10991 * omp-general.c (omp_max_vf): Likewise.
10992 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
10993 (expand_omp_simd): Handle polynomial safelen.
10994 * omp-low.c (omplow_simd_context): Add a default constructor.
10995 (omplow_simd_context::max_vf): Change from int to poly_uint64.
10996 (lower_rec_simd_input_clauses): Update accordingly.
10997 (lower_rec_input_clauses): Likewise.
10999 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11000 Alan Hayward <alan.hayward@arm.com>
11001 David Sherwood <david.sherwood@arm.com>
11003 * tree-vectorizer.h (vect_nunits_for_cost): New function.
11004 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
11005 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
11006 (vect_analyze_slp_cost): Likewise.
11007 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
11008 (vect_model_load_cost): Likewise.
11010 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11011 Alan Hayward <alan.hayward@arm.com>
11012 David Sherwood <david.sherwood@arm.com>
11014 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
11015 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
11016 from an unsigned int * to a poly_uint64_pod *.
11017 (calculate_unrolling_factor): New function.
11018 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
11020 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11021 Alan Hayward <alan.hayward@arm.com>
11022 David Sherwood <david.sherwood@arm.com>
11024 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
11025 from an unsigned int to a poly_uint64.
11026 (_loop_vec_info::slp_unrolling_factor): Likewise.
11027 (_loop_vec_info::vectorization_factor): Change from an int
11029 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
11030 (vect_get_num_vectors): New function.
11031 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
11032 (vect_get_num_copies): Use vect_get_num_vectors.
11033 (vect_analyze_data_ref_dependences): Change max_vf from an int *
11034 to an unsigned int *.
11035 (vect_analyze_data_refs): Change min_vf from an int * to a
11037 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11038 than an unsigned HOST_WIDE_INT.
11039 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
11040 (vect_analyze_data_ref_dependence): Change max_vf from an int *
11041 to an unsigned int *.
11042 (vect_analyze_data_ref_dependences): Likewise.
11043 (vect_compute_data_ref_alignment): Handle polynomial vf.
11044 (vect_enhance_data_refs_alignment): Likewise.
11045 (vect_prune_runtime_alias_test_list): Likewise.
11046 (vect_shift_permute_load_chain): Likewise.
11047 (vect_supportable_dr_alignment): Likewise.
11048 (dependence_distance_ge_vf): Take the vectorization factor as a
11049 poly_uint64 rather than an unsigned HOST_WIDE_INT.
11050 (vect_analyze_data_refs): Change min_vf from an int * to a
11052 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
11053 vfm1 as a poly_uint64 rather than an int. Make the same change
11054 for the returned bound_scalar.
11055 (vect_gen_vector_loop_niters): Handle polynomial vf.
11056 (vect_do_peeling): Likewise. Update call to
11057 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
11058 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
11060 * tree-vect-loop.c (vect_determine_vectorization_factor)
11061 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
11062 (vect_get_known_peeling_cost): Likewise.
11063 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
11064 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
11065 (vect_transform_loop): Likewise. Use the lowest possible VF when
11066 updating the upper bounds of the loop.
11067 (vect_min_worthwhile_factor): Make static. Return an unsigned int
11068 rather than an int.
11069 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
11070 polynomial unroll factors.
11071 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
11072 (vect_make_slp_decision): Likewise.
11073 (vect_supported_load_permutation_p): Likewise, and polynomial
11075 (vect_analyze_slp_cost): Handle polynomial vf.
11076 (vect_slp_analyze_node_operations): Likewise.
11077 (vect_slp_analyze_bb_1): Likewise.
11078 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11079 than an unsigned HOST_WIDE_INT.
11080 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
11081 (vectorizable_load): Handle polynomial vf.
11082 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
11084 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
11086 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11087 Alan Hayward <alan.hayward@arm.com>
11088 David Sherwood <david.sherwood@arm.com>
11090 * match.pd: Handle bit operations involving three constants
11091 and try to fold one pair.
11093 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11095 * tree-vect-loop-manip.c: Include gimple-fold.h.
11096 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
11097 niters_maybe_zero parameters. Handle other cases besides a step of 1.
11098 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
11099 Add a path that uses a step of VF instead of 1, but disable it
11101 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
11102 and niters_no_overflow parameters. Update calls to
11103 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
11104 Create a new SSA name if the latter choses to use a ste other
11105 than zero, and return it via niters_vector_mult_vf_var.
11106 * tree-vect-loop.c (vect_transform_loop): Update calls to
11107 vect_do_peeling, vect_gen_vector_loop_niters and
11108 slpeel_make_loop_iterate_ntimes.
11109 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
11110 (vect_gen_vector_loop_niters): Update declarations after above changes.
11112 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
11114 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
11115 128-bit round to integer instructions.
11116 (ceil<mode>2): Likewise.
11117 (btrunc<mode>2): Likewise.
11118 (round<mode>2): Likewise.
11120 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11122 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
11123 unaligned VSX load/store on P8/P9.
11124 (expand_block_clear): Allow the use of unaligned VSX
11125 load/store on P8/P9.
11127 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11129 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
11131 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
11132 swap associated with both a load and a store.
11134 2018-01-02 Andrew Waterman <andrew@sifive.com>
11136 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
11137 * config/riscv/riscv.md (clear_cache): Use it.
11139 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
11141 * web.c: Remove out-of-date comment.
11143 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11145 * expr.c (fixup_args_size_notes): Check that any existing
11146 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
11147 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
11148 (emit_single_push_insn): ...here.
11150 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11152 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
11153 (const_vector_encoded_nelts): New function.
11154 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
11155 (const_vector_int_elt, const_vector_elt): Declare.
11156 * emit-rtl.c (const_vector_int_elt_1): New function.
11157 (const_vector_elt): Likewise.
11158 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
11159 of CONST_VECTOR_ELT.
11161 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11163 * expr.c: Include rtx-vector-builder.h.
11164 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
11165 directly on the tree encoding.
11166 (const_vector_from_tree): Likewise.
11167 * optabs.c: Include rtx-vector-builder.h.
11168 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
11169 sequence of "u" values.
11170 * vec-perm-indices.c: Include rtx-vector-builder.h.
11171 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
11172 directly on the vec_perm_indices encoding.
11174 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11176 * doc/rtl.texi (const_vector): Describe new encoding scheme.
11177 * Makefile.in (OBJS): Add rtx-vector-builder.o.
11178 * rtx-vector-builder.h: New file.
11179 * rtx-vector-builder.c: Likewise.
11180 * rtl.h (rtx_def::u2): Add a const_vector field.
11181 (CONST_VECTOR_NPATTERNS): New macro.
11182 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
11183 (CONST_VECTOR_DUPLICATE_P): Likewise.
11184 (CONST_VECTOR_STEPPED_P): Likewise.
11185 (CONST_VECTOR_ENCODED_ELT): Likewise.
11186 (const_vec_duplicate_p): Check for a duplicated vector encoding.
11187 (unwrap_const_vec_duplicate): Likewise.
11188 (const_vec_series_p): Check for a non-duplicated vector encoding.
11189 Say that the function only returns true for integer vectors.
11190 * emit-rtl.c: Include rtx-vector-builder.h.
11191 (gen_const_vec_duplicate_1): Delete.
11192 (gen_const_vector): Call gen_const_vec_duplicate instead of
11193 gen_const_vec_duplicate_1.
11194 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
11195 (gen_const_vec_duplicate): Use rtx_vector_builder.
11196 (gen_const_vec_series): Likewise.
11197 (gen_rtx_CONST_VECTOR): Likewise.
11198 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
11199 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11200 Build a new vector rather than modifying a CONST_VECTOR in-place.
11201 (handle_special_swappables): Update call accordingly.
11202 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
11203 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11204 Build a new vector rather than modifying a CONST_VECTOR in-place.
11205 (handle_special_swappables): Update call accordingly.
11207 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11209 * simplify-rtx.c (simplify_const_binary_operation): Use
11210 CONST_VECTOR_ELT instead of XVECEXP.
11212 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11214 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
11215 the selector elements to be different from the data elements
11216 if the selector is a VECTOR_CST.
11217 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
11218 ssizetype for the selector.
11220 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11222 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
11223 before testing each element individually.
11224 * tree-vect-generic.c (lower_vec_perm): Likewise.
11226 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11228 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
11229 * selftest-run-tests.c (selftest::run_tests): Call it.
11230 * vector-builder.h (vector_builder::operator ==): New function.
11231 (vector_builder::operator !=): Likewise.
11232 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
11233 (vec_perm_indices::all_from_input_p): New function.
11234 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11235 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
11236 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
11237 instead of reading the VECTOR_CST directly. Detect whether both
11238 vector inputs are the same before constructing the vec_perm_indices,
11239 and update the number of inputs argument accordingly. Use the
11240 utility functions added above. Only construct sel2 if we need to.
11242 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11244 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
11245 the broadcast of the low byte.
11246 (expand_mult_highpart): Use an explicit encoding for the permutes.
11247 * optabs-query.c (can_mult_highpart_p): Likewise.
11248 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
11249 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11250 (vectorizable_bswap): Likewise.
11251 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
11252 explicit encoding for the power-of-2 permutes.
11253 (vect_permute_store_chain): Likewise.
11254 (vect_grouped_load_supported): Likewise.
11255 (vect_permute_load_chain): Likewise.
11257 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11259 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
11260 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
11261 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
11262 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11263 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
11264 (vect_gen_perm_mask_any): Likewise.
11266 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11268 * int-vector-builder.h: New file.
11269 * vec-perm-indices.h: Include int-vector-builder.h.
11270 (vec_perm_indices): Redefine as an int_vector_builder.
11271 (auto_vec_perm_indices): Delete.
11272 (vec_perm_builder): Redefine as a stand-alone class.
11273 (vec_perm_indices::vec_perm_indices): New function.
11274 (vec_perm_indices::clamp): Likewise.
11275 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
11276 (vec_perm_indices::new_vector): New function.
11277 (vec_perm_indices::new_expanded_vector): Update for new
11278 vec_perm_indices class.
11279 (vec_perm_indices::rotate_inputs): New function.
11280 (vec_perm_indices::all_in_range_p): Operate directly on the
11281 encoded form, without computing elided elements.
11282 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
11283 encoding. Update for new vec_perm_indices class.
11284 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
11285 the given vec_perm_builder.
11286 (expand_vec_perm_var): Update vec_perm_builder constructor.
11287 (expand_mult_highpart): Use vec_perm_builder instead of
11288 auto_vec_perm_indices.
11289 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
11290 vec_perm_indices instead of auto_vec_perm_indices. Use a single
11291 or double series encoding as appropriate.
11292 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
11293 vec_perm_indices instead of auto_vec_perm_indices.
11294 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11295 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11296 (vect_permute_store_chain): Likewise.
11297 (vect_grouped_load_supported): Likewise.
11298 (vect_permute_load_chain): Likewise.
11299 (vect_shift_permute_load_chain): Likewise.
11300 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11301 (vect_transform_slp_perm_load): Likewise.
11302 (vect_schedule_slp_instance): Likewise.
11303 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11304 (vectorizable_mask_load_store): Likewise.
11305 (vectorizable_bswap): Likewise.
11306 (vectorizable_store): Likewise.
11307 (vectorizable_load): Likewise.
11308 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
11309 vec_perm_indices instead of auto_vec_perm_indices. Use
11310 tree_to_vec_perm_builder to read the vector from a tree.
11311 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
11312 vec_perm_builder instead of a vec_perm_indices.
11313 (have_whole_vector_shift): Use vec_perm_builder and
11314 vec_perm_indices instead of auto_vec_perm_indices. Leave the
11315 truncation to calc_vec_perm_mask_for_shift.
11316 (vect_create_epilog_for_reduction): Likewise.
11317 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
11318 from auto_vec_perm_indices to vec_perm_indices.
11319 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11320 instead of changing individual elements.
11321 (aarch64_vectorize_vec_perm_const): Use new_vector to install
11322 the vector in d.perm.
11323 * config/arm/arm.c (expand_vec_perm_d::perm): Change
11324 from auto_vec_perm_indices to vec_perm_indices.
11325 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11326 instead of changing individual elements.
11327 (arm_vectorize_vec_perm_const): Use new_vector to install
11328 the vector in d.perm.
11329 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
11330 Update vec_perm_builder constructor.
11331 (rs6000_expand_interleave): Likewise.
11332 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
11333 (rs6000_expand_interleave): Likewise.
11335 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11337 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
11338 to qimode could truncate the indices.
11339 * optabs.c (expand_vec_perm_var): Likewise.
11341 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11343 * Makefile.in (OBJS): Add vec-perm-indices.o.
11344 * vec-perm-indices.h: New file.
11345 * vec-perm-indices.c: Likewise.
11346 * target.h (vec_perm_indices): Replace with a forward class
11348 (auto_vec_perm_indices): Move to vec-perm-indices.h.
11349 * optabs.h: Include vec-perm-indices.h.
11350 (expand_vec_perm): Delete.
11351 (selector_fits_mode_p, expand_vec_perm_var): Declare.
11352 (expand_vec_perm_const): Declare.
11353 * target.def (vec_perm_const_ok): Replace with...
11354 (vec_perm_const): ...this new hook.
11355 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
11356 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
11357 * doc/tm.texi: Regenerate.
11358 * optabs.def (vec_perm_const): Delete.
11359 * doc/md.texi (vec_perm_const): Likewise.
11360 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
11361 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
11362 expand_vec_perm for constant permutation vectors. Assert that
11363 the mode of variable permutation vectors is the integer equivalent
11364 of the mode that is being permuted.
11365 * optabs-query.h (selector_fits_mode_p): Declare.
11366 * optabs-query.c: Include vec-perm-indices.h.
11367 (selector_fits_mode_p): New function.
11368 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
11369 is defined, instead of checking whether the vec_perm_const_optab
11370 exists. Use targetm.vectorize.vec_perm_const instead of
11371 targetm.vectorize.vec_perm_const_ok. Check whether the indices
11372 fit in the vector mode before using a variable permute.
11373 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
11374 vec_perm_indices instead of an rtx.
11375 (expand_vec_perm): Replace with...
11376 (expand_vec_perm_const): ...this new function. Take the selector
11377 as a vec_perm_indices rather than an rtx. Also take the mode of
11378 the selector. Update call to shift_amt_for_vec_perm_mask.
11379 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
11380 Use vec_perm_indices::new_expanded_vector to expand the original
11381 selector into bytes. Check whether the indices fit in the vector
11382 mode before using a variable permute.
11383 (expand_vec_perm_var): Make global.
11384 (expand_mult_highpart): Use expand_vec_perm_const.
11385 * fold-const.c: Includes vec-perm-indices.h.
11386 * tree-ssa-forwprop.c: Likewise.
11387 * tree-vect-data-refs.c: Likewise.
11388 * tree-vect-generic.c: Likewise.
11389 * tree-vect-loop.c: Likewise.
11390 * tree-vect-slp.c: Likewise.
11391 * tree-vect-stmts.c: Likewise.
11392 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
11394 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
11395 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
11396 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
11397 (aarch64_vectorize_vec_perm_const): ...this new function.
11398 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11399 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11400 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
11401 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
11402 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11403 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11404 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
11406 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
11407 check for NEON modes.
11408 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
11409 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
11410 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
11411 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
11413 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
11414 the old VEC_PERM_CONST conditions.
11415 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
11416 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
11417 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
11418 (ia64_vectorize_vec_perm_const_ok): Merge into...
11419 (ia64_vectorize_vec_perm_const): ...this new function.
11420 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
11421 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
11422 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
11423 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
11424 * config/mips/mips.c (mips_expand_vec_perm_const)
11425 (mips_vectorize_vec_perm_const_ok): Merge into...
11426 (mips_vectorize_vec_perm_const): ...this new function.
11427 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
11428 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
11429 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
11430 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
11431 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
11432 (rs6000_expand_vec_perm_const): Delete.
11433 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
11435 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11436 (altivec_expand_vec_perm_const_le): Take each operand individually.
11437 Operate on constant selectors rather than rtxes.
11438 (altivec_expand_vec_perm_const): Likewise. Update call to
11439 altivec_expand_vec_perm_const_le.
11440 (rs6000_expand_vec_perm_const): Delete.
11441 (rs6000_vectorize_vec_perm_const_ok): Delete.
11442 (rs6000_vectorize_vec_perm_const): New function.
11443 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11444 an element count and rtx array.
11445 (rs6000_expand_extract_even): Update call accordingly.
11446 (rs6000_expand_interleave): Likewise.
11447 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
11448 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
11449 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
11450 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
11451 (rs6000_expand_vec_perm_const): Delete.
11452 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11453 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11454 (altivec_expand_vec_perm_const_le): Take each operand individually.
11455 Operate on constant selectors rather than rtxes.
11456 (altivec_expand_vec_perm_const): Likewise. Update call to
11457 altivec_expand_vec_perm_const_le.
11458 (rs6000_expand_vec_perm_const): Delete.
11459 (rs6000_vectorize_vec_perm_const_ok): Delete.
11460 (rs6000_vectorize_vec_perm_const): New function. Remove stray
11461 reference to the SPE evmerge intructions.
11462 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11463 an element count and rtx array.
11464 (rs6000_expand_extract_even): Update call accordingly.
11465 (rs6000_expand_interleave): Likewise.
11466 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
11467 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
11469 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11471 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11473 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
11474 vector mode and that that mode matches the mode of the data
11476 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
11477 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
11478 directly using expand_vec_perm_1 when forcing selectors into
11480 (expand_vec_perm_var): New function, split out from expand_vec_perm.
11482 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11484 * optabs-query.h (can_vec_perm_p): Delete.
11485 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
11486 * optabs-query.c (can_vec_perm_p): Split into...
11487 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
11488 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
11489 particular selector is valid.
11490 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11491 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11492 (vect_grouped_load_supported): Likewise.
11493 (vect_shift_permute_load_chain): Likewise.
11494 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11495 (vect_transform_slp_perm_load): Likewise.
11496 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11497 (vectorizable_bswap): Likewise.
11498 (vect_gen_perm_mask_checked): Likewise.
11499 * fold-const.c (fold_ternary_loc): Likewise. Don't take
11500 implementations of variable permutation vectors into account
11501 when deciding which selector to use.
11502 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
11503 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
11504 with a false third argument.
11505 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
11506 to test whether the constant selector is valid and can_vec_perm_var_p
11507 to test whether a variable selector is valid.
11509 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11511 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
11512 * optabs-query.c (can_vec_perm_p): Likewise.
11513 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
11514 instead of vec_perm_indices.
11515 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
11516 (vect_gen_perm_mask_checked): Likewise,
11517 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
11518 (vect_gen_perm_mask_checked): Likewise,
11520 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11522 * optabs-query.h (qimode_for_vec_perm): Declare.
11523 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
11524 (qimode_for_vec_perm): ...this new function.
11525 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
11527 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11529 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
11530 does not have a conditional at the top.
11532 2018-01-02 Richard Biener <rguenther@suse.de>
11534 * ipa-inline.c (big_speedup_p): Fix expression.
11536 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11539 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
11542 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11546 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
11547 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
11548 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
11549 cond_taken_branch_cost 3->4.
11551 2018-01-01 Jakub Jelinek <jakub@redhat.com>
11553 PR tree-optimization/83581
11554 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
11555 TODO_cleanup_cfg if any changes have been made.
11557 PR middle-end/83608
11558 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
11559 convert_modes if target mode has the right side, but different mode
11562 PR middle-end/83609
11563 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
11564 last argument when extracting from CONCAT. If either from_real or
11565 from_imag is NULL, use expansion through memory. If result is not
11566 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
11567 the parts directly to inner mode, if even that fails, use expansion
11570 PR middle-end/83623
11571 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
11572 check for bswap in mode rather than HImode and use that in expand_unop
11575 Copyright (C) 2018 Free Software Foundation, Inc.
11577 Copying and distribution of this file, with or without modification,
11578 are permitted in any medium without royalty provided the copyright
11579 notice and this notice are preserved.