1 2018-05-25 Richard Biener <rguenther@suse.de>
4 * tree-dump.c (dequeue_and_dump): Remove access to removed
5 operand 2 of a SWITCH_EXPR.
7 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
9 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
10 Add {q} suffix to insn mnemonic.
12 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
15 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
16 when memory input operand is handled.
18 2018-05-23 Bin Cheng <bin.cheng@arm.com>
20 Backport from mainline
21 2018-05-17 Bin Cheng <bin.cheng@arm.com>
22 Richard Biener <rguenther@suse.de>
24 PR tree-optimization/85793
25 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
28 2018-05-23 Richard Biener <rguenther@suse.de>
31 * tree-data-ref.c (create_runtime_alias_checks): Defer
32 and ignore overflow warnings.
34 2018-05-23 Martin Liska <mliska@suse.cz>
36 Backport from mainline
37 2018-05-23 Yury Gribov <tetra2005@gmail.com>
39 PR tree-optimization/85822
40 * tree-vrp.c (is_masked_range_test): Fix handling of negative
43 2018-05-23 Martin Liska <mliska@suse.cz>
45 Backport from mainline
46 2018-05-11 Martin Liska <mliska@suse.cz>
49 * doc/extend.texi: Document LLVM style format for no_sanitize
52 2018-05-22 Martin Sebor <msebor@redhat.com>
55 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
57 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
59 PR tree-optimization/85814
60 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
61 a null return from get_strinfo when unsharing the next
64 2018-05-21 Pat Haugen <pthaugen@us.ibm.com>
66 Backport from mainline
67 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
68 Segher Boessenkool <segher@kernel.crashing.org>
71 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest operand.
73 2018-05-18 Toon Moene <toon@moene.org>
75 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
76 directly after that of -floop-interchange. Indicate that both
77 options are enabled by default when specifying -O3.
79 2018-05-11 Martin Jambor <mjambor@suse.cz>
82 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
85 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
88 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
90 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
93 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
95 (cortex-m0): Use armv6s-m isa.
96 (cortex-m0plus): Likewise.
97 (cortex-m1): Likewise.
98 (cortex-m0.small-multiply): Likewise.
99 (cortex-m0plus.small-multiply): Likewise.
100 (cortex-m1.small-multiply): Likewise.
102 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
104 PR rtl-optimization/85638
105 * bb-reorder.c: Include common/common-target.h.
106 (create_forwarder_block): New function extracted from...
107 (fix_up_crossing_landing_pad): ...here. Rename into...
108 (dw2_fix_up_crossing_landing_pad): ...this.
109 (sjlj_fix_up_crossing_landing_pad): New function.
110 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
111 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
112 from both partitions and exit the loop after one iteration.
114 2018-05-08 Richard Biener <rguenther@suse.de>
116 Backport from mainline
117 2018-05-04 Richard Biener <rguenther@suse.de>
120 * fold-const.c (negate_expr_p): Restrict negation of operand
121 zero of a division to when we know that can happen without
123 (fold_negate_expr_1): Likewise.
125 2018-05-03 Richard Biener <rguenther@suse.de>
127 PR tree-optimization/85615
128 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
129 to loops not nested in BBs loop father to avoid creating multi-entry
132 2018-05-02 Richard Biener <rguenther@suse.de>
135 * gimplify.c (gimplify_save_expr): When in SSA form allow
136 SAVE_EXPRs to compute to SSA vars.
138 2018-05-02 Richard Biener <rguenther@suse.de>
140 PR tree-optimization/85597
141 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
142 do not use split vect_get_vec_defs call but call vect_get_slp_defs
145 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
148 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
149 (check_arch): Likewise.
150 (check_fpu): Return the result rather than printing it.
151 (end arch): Fix operator precedence.
153 (END): Print the result from check_fpu.
155 2018-05-08 Richard Biener <rguenther@suse.de>
157 Backport from mainline
158 2018-05-04 Richard Biener <rguenther@suse.de>
160 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
161 * gimple-ssa-store-merging.c
162 (imm_store_chain_info::output_merged_store): Remove redundant create,
163 release split_store vector contents on failure.
164 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
165 scalar stmt vector on cache hit.
167 2018-05-04 Richard Biener <rguenther@suse.de>
170 * tree-complex.c (tree_lower_complex): Handle removed blocks.
172 2018-05-07 Chung-Ju Wu <jasonwucj@gmail.com>
174 Backport from mainline.
175 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
177 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
178 GET_MODE_MASK before any checking.
179 (nds32_can_use_bset_p): Likewise.
180 (nds32_can_use_btgl_p): Likewise.
182 2018-05-07 Chung-Ju Wu <jasonwucj@gmail.com>
184 Backport from mainline.
185 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
187 * config/nds32/nds32-doubleword.md: New define_split pattern for
188 illegal register number.
190 2018-05-07 Chung-Ju Wu <jasonwucj@gmail.com>
192 Backport from mainline.
193 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
195 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
197 2018-05-07 Chung-Ju Wu <jasonwucj@gmail.com>
199 Backport from mainline.
200 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
202 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
204 2018-05-06 Jakub Jelinek <jakub@redhat.com>
207 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
208 the type is addressable. Don't force op into register if it has
211 2018-05-04 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
214 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
217 * config/aarch64/constraints.md (Usg): Limit to 31.
220 2018-05-02 Jakub Jelinek <jakub@redhat.com>
222 * BASE-VER: Set to 8.1.1.
224 2018-05-02 Release Manager
226 * GCC 8.1.0 released.
228 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
230 PR tree-optimization/85586
231 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
232 exit early for statements in the same group if the accesses are
235 2018-05-01 Jakub Jelinek <jakub@redhat.com>
238 * doc/install.texi2html: Replace _002d with - and _002a with * in
239 generated html files using sed.
241 2018-04-27 Martin Jambor <mjambor@suse.cz>
244 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
245 the jump function allows for passing through aggregate values.
247 2018-04-27 Jeff Law <law@redhat.com>
249 * config.gcc: Mark tile* targets as deprecated/obsolete.
251 2018-04-27 Jakub Jelinek <jakub@redhat.com>
253 PR tree-optimization/85529
254 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
255 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
256 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
257 zero extension or masking of the MSB bit.
258 (optimize_range_tests): Add FIRST_BB argument, pass it through
259 to optimize_range_tests_var_bound.
260 (maybe_optimize_range_tests, reassociate_bb): Adjust
261 optimize_range_tests callers.
263 2018-04-26 Richard Biener <rguenther@suse.de>
264 Jakub Jelinek <jakub@redhat.com>
266 * cgraph.h (symbol_table): Just declare debug method here.
267 * symtab.c (symbol_table::debug): Define.
269 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
271 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
273 2018-04-25 H.J. Lu <hongjiu.lu@intel.com>
275 Backport from mainline
276 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
279 * config/i386/i386.c (ix86_expand_builtin): Change memory
280 operand to XI, extend p0 to Pmode.
281 * config/i386/i386.md: Change unspec volatile and operand
282 1 mode to XI, change operand 0 mode to P.
284 2018-04-25 Jakub Jelinek <jakub@redhat.com>
286 * DEV-PHASE: Set to prerelease.
288 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
290 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
291 in __abskf2 and __powikf2.
293 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
296 * config/aarch64/constraints.md (Usg, Usj): New constraints.
297 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
298 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
299 Use the above on operand 2. Reindent.
300 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
302 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
305 * common/config/i386/i386-common.c (ix86_handle_option): Don't
307 * config/i386/i386.opt (mcet): Removed.
308 * doc/install.texi: Remove -mcet documentation.
309 * doc/invoke.texi: Likewise.
311 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
314 * doc/install.texi: Remove -mcet from bootstrap-cet.
316 2018-04-24 Jakub Jelinek <jakub@redhat.com>
319 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
320 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
324 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
325 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
326 containing a CONST_VECTOR.
328 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
330 * doc/install.texi: Update newlib dependency for nvptx.
332 2018-04-24 Jakub Jelinek <jakub@redhat.com>
335 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
336 instead of INTVAL when shifting x left.
338 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
340 PR tree-optimization/85478
341 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
342 vect_grouped_store_supported for single element vectors.
344 2018-04-24 Richard Biener <rguenther@suse.de>
347 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
348 load cost increase to the case of non-constant step.
350 2018-04-24 Jakub Jelinek <jakub@redhat.com>
353 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
354 destination if any_malformed_asm.
356 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
359 * expr.c (store_field): In the bitfield case, if the value comes from
360 a function call and is returned in registers by means of a PARALLEL,
361 do not change the mode of the temporary unless BLKmode and VOIDmode.
363 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
365 PR rtl-optimization/85423
366 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
367 dependencies to debug insns when the previous insn is non-debug.
369 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
371 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
372 enums into a single definition.
373 (fls): Fix predicates and printing.
376 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
378 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
379 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
380 and short u6 immediate.
381 (check_if_valid_sleep_operand): Remove.
382 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
384 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
386 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
387 flag_always_save_lp condition.
388 * config/nds32/nds32.opt (malways-save-lp): New option.
390 2018-04-22 Shiva Chen <shiva0217@gmail.com>
392 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
393 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
394 * config/nds32/nds32.h
395 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
396 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
398 2018-04-22 Shiva Chen <shiva0217@gmail.com>
400 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
401 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
403 2018-04-22 Shiva Chen <shiva0217@gmail.com>
404 Chung-Ju Wu <jasonwucj@gmail.com>
406 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
408 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
409 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
411 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
413 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
415 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
417 * config/nds32/nds32-protos.h (nds32_data_alignment,
418 nds32_local_alignment): Declare.
419 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
420 nds32_local_alignment): New functions.
421 (TARGET_CONSTANT_ALIGNMENT): Define.
422 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
424 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
426 * config/nds32/nds32.c
427 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
428 (TARGET_MODES_TIEABLE_P): Likewise.
430 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
432 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
435 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
436 Chung-Ju Wu <jasonwucj@gmail.com>
438 * config/nds32/constants.md (unspec_volatile_element): Add enum values
439 for unaligned access.
440 * config/nds32/nds32-intrinsic.c: Implementation of expanding
442 * config/nds32/nds32-intrinsic.md: Likewise.
443 * config/nds32/nds32_intrinsic.h: Likewise.
444 * config/nds32/nds32.h (nds32_builtins): Likewise.
445 * config/nds32/nds32.opt (munaligned-access): New option.
446 * config/nds32/nds32.c (nds32_asm_file_start): Display
447 flag_unaligned_access status.
449 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
451 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
452 -mno-relax is present.
453 * config/riscv/linux.h (LINK_SPEC): Ditto.
455 2018-04-20 Martin Sebor <msebor@redhat.com>
458 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
460 (gimple_fold_builtin_stxcpy_chk): Same.
461 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
463 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
466 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
467 __powikf2 when long double is IEEE 128-bit.
469 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
471 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
472 step to make sure stack always aligned.
474 2018-04-20 Carl Love <cel@us.ibm.com>
477 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
480 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
481 Tom de Vries <tom@codesourcery.com>
484 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
485 Emit insns for calls too.
486 (nvptx_find_par): Always look for worker-level predecessor insn.
487 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
489 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
490 (nvptx_process_pars): Propagate frames for calls.
492 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
495 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
497 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
498 (ix86_handle_option): Don't handle OPT_mibt.
499 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
501 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
503 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
504 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
505 (ix86_target_macros): Define __CET__ with flag_cf_protection
507 * config/i386/i386.c (isa2_opts): Remove -mibt.
508 * config/i386/i386.h (TARGET_IBT): Removed.
509 (TARGET_IBT_P): Likewise.
510 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
511 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
512 * config/i386/i386.opt (mcet): Update help message.
515 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
516 -mcet as an alias for -mshstk.
518 2018-04-20 Richard Biener <rguenther@suse.de>
521 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
522 complexity by forcing a single use of the multiply operand.
524 2018-04-20 Martin Jambor <mjambor@suse.cz>
527 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
528 recursion dependency to only apply to non-clones.
530 2018-04-20 Martin Jambor <mjambor@suse.cz>
533 * ipa-cp.c (create_specialized_node): Check that clones of
534 self-recursive edges exist during IPA-CP.
536 2018-04-19 Toon Moene <toon@moene.org>
538 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
541 2018-04-19 Jakub Jelinek <jakub@redhat.com>
543 PR tree-optimization/85467
544 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
545 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
546 VECTOR_CST element to type.
548 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
551 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
552 * config/i386/i386.md (builtin_setjmp_setup): Removed.
553 (builtin_longjmp): Likewise.
554 (save_stack_nonlocal): New pattern.
555 (restore_stack_nonlocal): Likewise.
557 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
560 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
561 Replace ASM_OUTPUT_LABEL with fprintf.
563 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
566 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
567 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
568 * config/i386/i386-c.c (ix86_target_macros_internal): Also
569 define __IBT__ and __SHSTK__ for -fcf-protection.
570 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
572 (ix86_trampoline_init): Likewise.
573 (x86_output_mi_thunk): Likewise.
574 (ix86_notrack_prefixed_insn_p): Likewise.
575 (ix86_option_override_internal): Don't disallow -fcf-protection.
576 * config/i386/i386.md (rdssp<mode>): Also enable for
578 (incssp<mode>): Likewise.
579 (nop_endbr): Likewise.
580 * config/i386/i386.opt (mcet): Change help message to built-in
584 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
585 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
586 enable CET built-in functions.
588 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
590 * common/config/i386/i386-common.c
591 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
592 OPTION_MASK_ISA_MOVDIRI_UNSET,
593 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
594 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
595 * config.gcc (movdirintrin.h): New header.
596 * config/i386/cpuid.h (bit_MOVDIRI,
597 bit_MOVDIR64B): New bits.
598 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
600 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
601 (VOID, PVOID, PCVOID)): New function types.
602 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
603 __builtin_ia32_directstoreu_u64,
604 __builtin_ia32_movdir64b): New builtins.
605 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
606 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
608 (ix86_valid_target_attribute_inner_p): Ditto.
609 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
610 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
611 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
612 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
613 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
614 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
615 (movdiri<mode>, movdir64b_<mode>): New.
616 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
617 * config/i386/immintrin.h: Include movdirintrin.h.
618 * config/i386/movdirintrin.h: New file.
619 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
621 2018-04-19 Richard Biener <rguenther@suse.de>
624 * cfg.c (clear_bb_flags): When loop state says we have
625 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
627 2018-04-19 Richard Biener <rguenther@suse.de>
629 PR tree-optimization/84737
630 * tree-vect-data-refs.c (vect_copy_ref_info): New function
631 copying restrict info.
632 (vect_setup_realignment): Use it.
633 * tree-vectorizer.h (vect_copy_ref_info): Declare.
634 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
635 the first DR to all generated stores.
636 (vectorizable_load): Likewise for loads.
638 2018-04-19 Jakub Jelinek <jakub@redhat.com>
640 PR tree-optimization/85446
641 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
642 the integral and pointer types to have the same precision.
644 * doc/install.texi: Document --disable-cet being the default and
647 2018-04-18 Martin Liska <mliska@suse.cz>
649 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
652 2018-04-18 Martin Liska <mliska@suse.cz>
655 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
658 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
659 arguments if they are comparable.
661 2018-04-18 Martin Liska <mliska@suse.cz>
664 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
667 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
670 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
673 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
674 ENDBR after calling __morestack.
676 2018-04-18 David Malcolm <dmalcolm@redhat.com>
679 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
680 by using gcc_base_ver to generate a gcc_driver_version, and use
681 it when generating GCC_DRIVER_NAME.
682 * configure: Regenerate.
684 2018-04-18 Jakub Jelinek <jakub@redhat.com>
687 * config.gcc: Obsolete powerpc*-*-*spe*.
689 2018-04-17 Jakub Jelinek <jakub@redhat.com>
692 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
693 (stabstr_D): Change type of unum from unsigned int to
694 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
697 2018-04-17 Jim Wilson <jimw@sifive.com>
700 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
701 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
702 Set arg_pointer_offset after using pretend_args_size.
704 2018-04-17 Jakub Jelinek <jakub@redhat.com>
706 PR rtl-optimization/85431
707 * dse.c (record_store): Ignore zero width stores.
710 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
711 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
712 __builtin_stack_restore rather than after it.
713 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
714 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
715 argument instead of virtual_dynamic_stack_rtx.
717 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
719 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
721 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
722 Add note to error message to explain internal mapping of overloaded
723 built-in function name to non-overloaded built-in function name.
724 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
727 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
730 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
731 where the inputs overlap with the output.
733 2018-04-17 Jakub Jelinek <jakub@redhat.com>
735 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
736 (=v, v) alternative and explicit "memory" attribute.
737 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
738 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
740 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
741 "sselog1" type instead of "sselog".
742 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
743 "sselog". Remove explicit "memory" attribute.
744 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
745 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
747 (vec_extract_hi_v32hi): Merge all alternatives into one, use
748 "sselog1" type instead of "sselog". Remove explicit "memory"
750 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
751 use "sselog1" type instead of "sselog". Remove explicit "memory"
753 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
754 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
756 (vec_extract_hi_v64qi): Merge all alternatives into one, use
757 "sselog1" type instead of "sselog". Remove explicit "memory"
759 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
760 use "sselog1" type instead of "sselog". Remove explicit "memory"
764 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
767 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
770 2018-04-17 Martin Jambor <mjambor@suse.cz>
773 * ipa-cp.c (create_specialized_node): Call
774 expand_all_artificial_thunks if necessary.
776 2018-04-17 Martin Liska <mliska@suse.cz>
779 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
780 in message, remote space in between '_G' and '('.
782 2018-04-17 Jakub Jelinek <jakub@redhat.com>
785 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
786 avx512f_vmcmp<mode>3<round_saeonly_name>,
787 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
788 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
789 avx512f_rndscale<mode><round_saeonly_name>,
790 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
791 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
792 Use %<iptr>2 instead of %2 for -masm=intel.
793 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
794 avx512f_vcvttss2usi<round_saeonly_name>,
795 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
797 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
798 avx512f_vcvttsd2usi<round_saeonly_name>,
799 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
800 Use %q1 instead of %1 for -masm=intel.
801 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
802 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
803 of %3 for -masm=intel.
804 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
806 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
808 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
810 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
811 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
813 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
815 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
816 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
817 %g1 and one with %0 and %1.
818 (avx512er_vmrcp28<mode><round_saeonly_name>,
819 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
821 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
822 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
823 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
824 of %0 and %{%4%} for -masm=intel.
825 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
826 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
827 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
828 order of %0 and %{%5%}%{z%} for -masm=intel.
830 2018-04-17 Jan Hubicka <jh@suse.cz>
833 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
835 2018-04-17 Martin Liska <mliska@suse.cz>
838 * multiple_target.c (create_dispatcher_calls): Set apostrophes
839 for target_clone error message. Make default implementation
840 clone to be a local declaration.
841 (separate_attrs): Add new argument and check for an empty
843 (expand_target_clones): Handle it.
844 (ipa_target_clone): Make redirection just for target_clones
847 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
848 Tom de Vries <tom@codesourcery.com>
851 * omp-expand.c (expand_oacc_for): Add dummy false branch for
852 tiled basic blocks without omp continue statements.
854 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
857 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
858 vec_extract expression as having side effects to make sure it gets
861 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
864 * config/i386/i386.c (get_builtin_code_for_version): Check
867 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
870 * gcc/config.gcc: Support "skylake".
871 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
873 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
874 (processor_target_table): Add "skylake".
875 (ix86_option_override_internal): Add "skylake".
876 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
877 PROCESSOR_CANNONLAKE.
878 (get_builtin_code_for_version): Fix priority for
879 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
880 PROCESSOR_SKYLAKE-AVX512.
881 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
882 (processor_type): Add PROCESSOR_SKYLAKE.
884 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
885 Jason Merrill <jason@redhat.com>
888 * convert.c (convert_to_integer_1): Use direct recursion for
889 enumeral types and types with a precision less than the number
890 of bits in their mode.
892 2018-04-16 Julia Koval <julia.koval@intel.com>
895 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
896 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
898 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
901 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
902 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
903 and -mno-direct-move.
905 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
908 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
909 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
910 Ensure negative shifts result in {0}.
912 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
914 PR rtl-optimization/79916
915 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
916 regs (if any) to define how to gnerate SD moves when LRA is in
919 2018-04-13 Jakub Jelinek <jakub@redhat.com>
921 PR rtl-optimization/85393
922 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
923 * except.c (expand_dw2_landing_pad_for_region): Make static.
924 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
925 a label and unconditional jump to old_bb, rather than
926 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
929 PR rtl-optimization/85376
930 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
931 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
932 instead of a specific value.
934 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
935 Bin Cheng <bin.cheng@arm.com>
937 PR tree-optimization/82965
938 PR tree-optimization/83991
939 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
940 by_profile_only parameter.
941 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
942 information if the loop was predicted to iterate too many times.
943 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
945 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
948 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
951 2018-04-13 Martin Liska <mliska@suse.cz>
952 Jakub Jelinek <jakub@redhat.com>
955 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
956 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
957 * builtins.c (expand_builtin_memory_copy_args): Use
958 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
959 handle dest_addr == pc_rtx.
961 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
964 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
965 asked to not generate direct moves.
966 (fix_trunc<mode>si2_stfiwx): Similar.
967 (fix_trunc<mode>si2_internal): Similar.
969 2018-04-12 Jakub Jelinek <jakub@redhat.com>
972 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
973 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
974 lookup if dest in some wider mode is known to be const0_rtx and
975 if so, record permanent equivalence for it to be ZERO_EXTEND of
976 the narrower mode destination.
978 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
980 * lto-streamer-out.c (output_function): Revert 259346.
981 * omp-expand.c (expand_oacc_for): Likewise.
983 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
985 PR rtl-optimization/85354
986 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
987 * sel-sched.c (sel_global_init): ... here.
989 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
992 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
993 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
994 mode for PE-COFF targets.
995 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
996 (i386_pe_asm_lto_end): Likewise.
997 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
998 (TARGET_ASM_LTO_END): Likewise.
999 * config/i386/winnt.c (saved_debug_info_level): New static variable.
1000 (i386_pe_asm_lto_start): New function.
1001 (i386_pe_asm_lto_end): Likewise.
1003 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1004 Richard Biener <rguenther@suse.de>
1007 * lto-streamer-out.c (output_function): Fix CFG loop state before
1009 * omp-expand.c (expand_oacc_for): Handle calls to internal
1010 functions like regular functions.
1012 2018-04-12 Richard Biener <rguenther@suse.de>
1015 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
1016 for the early LTO debug to properly generate references to it
1017 during DIE emission. Do not re-use that for the skeleton for
1019 (dwarf2out_early_finish): Likewise.
1021 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1024 * config/i386/sse.md
1025 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
1026 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
1027 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
1028 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
1029 and output is a reg, avoid creating invalid lowpart subreg, but
1030 instead split into a 512-bit move. Don't split if not AVX512VL,
1031 input is xmm16+ reg and output is a mem.
1032 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
1033 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
1034 xmm16+ reg and output is a mem.
1036 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1038 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
1039 also for flag_dwarf2_cfi_asm.
1041 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1043 PR rtl-optimization/85342
1044 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
1045 a bool scalar var inside of the loop instead. Don't try to update
1046 recog_data.operand after failed apply_change_group.
1048 2018-04-12 Tom de Vries <tom@codesourcery.com>
1051 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
1052 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
1053 array with flexible array member as array without given dimension.
1054 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
1055 argument for undefined param to true.
1057 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
1060 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1061 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
1062 from PowerPC section.
1063 * config/rs6000/sysv4.opt (mcall-): Improve help text.
1064 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
1065 help text that is too long.
1066 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
1067 help text that is too long.
1068 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
1069 help text that is too long.
1071 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
1073 * config/alpha/alpha.md (stack_probe_internal): Rename
1074 from "probe_stack". Update all callers.
1076 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1078 PR rtl-optimization/84566
1079 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
1080 sched_macro_fuse_insns.
1082 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1085 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
1086 (compute_block_dependences): ... from here.
1088 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1090 PR tree-optimization/85331
1091 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
1092 from int to HOST_WIDE_INT.
1094 2018-04-11 Martin Jambor <mjambor@suse.cz>
1097 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
1098 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
1099 not the same as the source val.
1100 (cgraph_edge_brings_value_p): New parameter.
1101 (gather_edges_for_value): Pass destination value to
1102 cgraph_edge_brings_value_p.
1103 (perhaps_add_new_callers): Likewise.
1104 (get_info_about_necessary_edges): Likewise and exclude values brought
1105 only by self-recursive edges.
1106 (create_specialized_node): Redirect only clones of self-calling edges.
1107 (+self_recursive_pass_through_p): New function.
1108 (find_more_scalar_values_for_callers_subset): Use it.
1109 (find_aggregate_values_for_callers_subset): Likewise.
1110 (known_aggs_to_agg_replacement_list): Removed.
1111 (decide_whether_version_node): Re-calculate known constants for all
1112 remaining context clones.
1114 2018-04-11 Richard Biener <rguenther@suse.de>
1117 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
1118 from early DWARF output.
1119 (dwarf2out_early_finish): Output line info unconditionally into
1120 early DWARF and add reference to it.
1122 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1125 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
1126 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
1127 other than V2DFmode using iptr mode attribute.
1128 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
1130 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1132 PR rtl-optimization/84659
1133 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
1135 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1138 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
1140 (output_loc_list): Pass address of a dummy size variable even in the
1141 locview handling loop.
1142 (index_location_lists): Add comment on why skip_loc_list_entry can't
1145 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
1148 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
1151 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
1154 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1155 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
1156 and -mstring-compare-inline-limit.
1158 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1161 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
1162 for stack clash protection in a register whenever we need it to be in
1165 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1167 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
1168 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
1170 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1173 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
1175 (mlong-double-): Ditto.
1176 * config/rs6000/sysv4.opt (msdata=): Ditto.
1177 (mtls-size=): Ditto.
1179 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
1181 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1182 erroneous entries for
1183 "vector int vec_ldl (int, long int *)", and
1184 "vector unsigned int vec_ldl (int, unsigned long int *)".
1185 Add comments and entries for
1186 "vector bool char vec_ldl (int, bool char *)",
1187 "vector bool short vec_ldl (int, bool short *)",
1188 "vector bool int vec_ldl (int, bool int *)",
1189 "vector bool long long vec_ldl (int, bool long long *)",
1190 "vector pixel vec_ldl (int, pixel *)",
1191 "vector long long vec_ldl (int, long long *)",
1192 "vector unsigned long long vec_ldl (int, unsigned long long *)".
1193 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
1194 type tree bool_long_long_type_node and correct definition of
1195 bool_V2DI_type_node to make reference to this new type tree.
1196 (rs6000_mangle_type): Replace erroneous reference to
1197 bool_long_type_node with bool_long_long_type_node.
1198 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
1199 comments to emphasize sign distinctions for char and int types and
1200 replace RS6000_BTI_bool_long constant with
1201 RS6000_BTI_bool_long_long constant. Also add comment to restrict
1202 use of RS6000_BTI_pixel.
1203 (bool_long_type_node): Remove this macro definition.
1204 (bool_long_long_type_node): New macro definition
1206 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1208 PR rtl-optimization/85300
1209 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
1210 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
1211 simplify_unary_operation fails.
1213 2018-04-10 Martin Liska <mliska@suse.cz>
1215 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
1216 cgraph_edge and ipa_ref.
1218 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1222 * config/i386/sse.md
1223 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
1224 computation of the VEC_MERGE selector from mask.
1225 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
1226 Fix decoding of the VEC_MERGE selector into mask.
1228 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1230 PR tree-optimization/85286
1231 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
1233 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1235 * final.c (final_1): Set insn_last_address as well as
1236 insn_current_address.
1238 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1241 * explow.c (emit_stack_probe): Call validize_mem on memory location
1242 before passing it to gen_probe_stack. Create address operand and
1243 legitimize it for the probe_stack_address case.
1245 2018-04-09 Jan Hubicka <jh@suse.cz>
1248 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
1249 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
1250 * tree.c (free_lang_data_in_type): Fix handling of binfos;
1252 (free_lang_data): Rebuild type inheritance graph.
1254 2018-04-09 Martin Sebor <msebor@redhat.com>
1256 * invoke.texi (-finline-small-functions): Mention other optimization
1258 (-findirect-inlining, -fpartial-inlining): Same.
1259 (-finline-functions-called-once): Same.
1260 (-freorder-blocks-and-partition): Same.
1262 2018-04-09 Jan Hubicka <jh@suse.cz>
1265 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
1266 jumps; choose last target that matches the criteria (i.e.
1267 no partition changes for non-crossing jumps).
1268 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
1269 support for redirecting crossing jumps to non-crossing.
1271 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
1273 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
1274 also for naked functions.
1276 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
1278 * config/arc/arc.md (add_shift): New pattern.
1279 (add_shift2): Likewise.
1280 (sub_shift): Likewise.
1281 (sub_shift_cmp0_noout): Likewise.
1282 (compare_si_ashiftsi): Likewise.
1283 (xbfu_cmp0_noout): New combine pattern.
1284 (xbfu_cmp0"): Likewise.
1285 (movsi_set_cc_insn): Place the predicable variant first.
1286 (commutative_binary_cmp0_noout): Remove clobber.
1287 (commutative_binary_cmp0): New pattern.
1288 (noncommutative_binary_cmp0): Likewise.
1289 (noncommutative_binary_cmp0_noout): Likewise.
1290 (noncommutative_binary_comparison_result_used): Removed.
1291 (rsub_cmp0): New pattern.
1292 (rsub_cmp0_noout): Likewise.
1293 (extzvsi): Changed, keep only meaningful variants.
1294 (SQH, SEZ): New iterators.
1295 (SQH_postfix): New mode attribute.
1296 (SEZ_prefix): New code attribute.
1297 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
1298 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
1299 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
1301 (noncommutative_operator): Check the availability of barrel
1304 2018-04-09 Richard Biener <rguenther@suse.de>
1306 PR tree-optimization/85284
1307 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
1308 Only use the niter constraining form of simple_iv when the exit
1311 2018-04-09 Tom de Vries <tom@codesourcery.com>
1314 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1315 (define_expand "*memory_barrier"): New define_expand.
1316 (define_insn "memory_barrier"): New insn.
1318 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1320 PR rtl-optimization/80463
1321 PR rtl-optimization/83972
1322 PR rtl-optimization/83480
1324 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
1325 correct producer for the insn.
1326 (tidy_control_flow): Fixup seqnos in case of debug insns.
1328 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1330 PR rtl-optimization/83913
1332 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
1333 different sched-times when merging exprs.
1335 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1337 PR rtl-optimization/83962
1339 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
1340 tidy_fallthru_edge and tidy_control_flow.
1342 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1344 PR rtl-optimization/83530
1346 * sel-sched.c (force_next_insn): New global variable.
1347 (remove_insn_for_debug): When force_next_insn is true, also leave only
1348 next insn in the ready list.
1349 (sel_sched_region): When the region wasn't scheduled, make another pass
1350 over it with force_next_insn set to 1.
1352 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1354 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
1356 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1357 for interrupt control.
1358 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1359 functions for interrupt control.
1360 * config/nds32/nds32-intrinsic.md: Likewise.
1361 * config/nds32/nds32_intrinsic.h: Likewise.
1362 * config/nds32/nds32.h (nds32_builtins): Likewise.
1364 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1366 * config/nds32/nds32.c (nds32_init_machine_status,
1367 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
1368 strict_aligned_p field.
1369 (nds32_expand_to_rtl_hook): New function.
1370 (TARGET_EXPAND_TO_RTL_HOOK): Define.
1371 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
1373 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1374 Chung-Ju Wu <jasonwucj@gmail.com>
1376 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
1377 * config/nds32/nds32-n7.md: New file.
1378 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
1379 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
1381 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
1382 * config/nds32/nds32.md (pipeline_model): Add n7.
1383 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
1384 * config/nds32/pipelines.md: Include n7 settings.
1386 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1387 Chung-Ju Wu <jasonwucj@gmail.com>
1389 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
1390 * config/nds32/nds32-e8.md: New file.
1391 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
1392 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
1394 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
1395 * config/nds32/nds32.md (pipeline_model): Add e8.
1396 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
1397 * config/nds32/pipelines.md: Include e8 settings.
1399 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1400 Chung-Ju Wu <jasonwucj@gmail.com>
1402 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
1403 * config/nds32/nds32-n8.md: New file.
1404 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
1405 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
1407 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
1408 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
1409 * config/nds32/nds32.md (pipeline_model): Add n8.
1410 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
1411 * config/nds32/pipelines.md: Include n8 settings.
1413 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1414 Chung-Ju Wu <jasonwucj@gmail.com>
1416 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
1417 * config/nds32/nds32-n9-2r1w.md: New file.
1418 * config/nds32/nds32-n9-3r2w.md: New file.
1419 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
1420 nds32_register_ports): New or modify for cpu n9.
1421 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
1423 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
1424 * config/nds32/nds32-utils.c: New file.
1425 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
1426 TARGET_MUL_SLOW): Define.
1427 * config/nds32/nds32.md (pipeline_model): New attribute.
1428 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
1429 New options that support cpu n9.
1430 * config/nds32/pipelines.md: Include n9 settings.
1431 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
1433 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1435 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
1436 information if necessary.
1437 (output_cond_branch_compare_zero): Likewise.
1438 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
1439 (nds32_target_alignment): Refine for alignment.
1440 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
1441 (FUNCTION_BOUNDARY): Modify.
1442 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
1444 * config/nds32/nds32.opt (malways-align, malign-functions): New.
1446 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1448 * config/nds32/constants.md (unspec_volatile_element): Add values for
1449 TLB operation and data prefetch.
1450 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1451 functions for TLB operation and data prefetch.
1452 * config/nds32/nds32-intrinsic.md: Likewise.
1453 * config/nds32/nds32_intrinsic.h: Likewise.
1454 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
1455 (nds32_print_operand): Likewise.
1456 * config/nds32/nds32.h (nds32_builtins): Likewise.
1458 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
1459 Andrew Pinski <pinsika@gcc.gnu.org>
1462 * match.pd: Use constant_boolean_node of correct type instead of
1463 boolean_true_node or boolean_false_node for simplifying
1464 pointer comparisons to zero.
1466 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1468 PR tree-optimization/80021
1469 * tree.c (verify_type_variant): Make error call in verify_variant_match
1470 translatable and remove final full stop.
1472 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1474 * config/nds32/constants.md (unspec_volatile_element): Add
1475 UNSPEC_VOLATILE_EH_RETURN.
1476 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
1477 nds32_output_stack_pop): Support dwarf exception handling process.
1478 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
1479 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
1480 exception handling process.
1481 (nds32_compute_stack_frame): Likewise.
1482 (nds32_return_addr_rtx): Likewise.
1483 (nds32_initial_elimination_offset): Likewise.
1484 (nds32_expand_prologue): Likewise.
1485 (nds32_expand_epilogue): Likewise.
1486 (nds32_dynamic_chain_address): New function.
1487 * config/nds32/nds32.h (machine_function): Add fields for dwarf
1489 (DYNAMIC_CHAIN_ADDRESS): Define.
1490 (EH_RETURN_DATA_REGNO): Define.
1491 (EH_RETURN_STACKADJ_RTX): Define.
1492 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
1493 patterns for dwarf exception handling.
1495 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1497 * config/nds32/nds32.h: Clean up obsolete macros.
1499 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1501 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1502 Add enum values for particular instructions.
1503 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1504 particular intrinsic functions.
1505 * config/nds32/nds32-intrinsic.md: Likewise.
1506 * config/nds32/nds32_intrinsic.h: Likewise.
1507 * config/nds32/nds32.h (nds32_builtins): Likewise.
1508 * config/nds32/nds32.md (type): Add pbsad and pbsada.
1509 (btst, ave): New patterns for particular instructions.
1511 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1513 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1514 Add enum values for atomic load/store and memory sync.
1515 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
1517 * config/nds32/nds32-intrinsic.md: Likewise.
1518 * config/nds32/nds32_intrinsic.h: Likewise.
1519 * config/nds32/nds32.h (nds32_builtins): Likewise.
1521 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1523 PR tree-optimization/85257
1524 * fold-const.c (native_encode_vector): If not all elts could fit
1525 and off is -1, return 0 rather than offset.
1526 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
1527 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
1528 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
1529 adjust buffer in native_interpret_expr call.
1531 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1533 * config/nds32/constants.md (unspec_volatile_element): Add cache
1534 control enum values.
1535 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
1536 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
1537 * config/nds32/nds32.c (nds32_cctl_names): New.
1538 (nds32_print_operand): Handle cache control register names.
1539 * config/nds32/nds32.h (nds32_builtins): New enum values.
1540 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
1542 * config/nds32/nds32.md (type): Add mmu.
1543 * config/nds32/pipelines.md (simple_insn): Add mmu.
1545 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1547 * config/nds32/nds32.md (type): Remove call.
1548 * config/nds32/pipelines.md (simple_insn): Likewise.
1550 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1552 * config/nds32/constants.md (unspec_volatile_element): Add
1553 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
1554 UNSPEC_VOLATILE_FMFCFG.
1555 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
1556 description for fmfcfg and fmfcsr.
1557 (bdesc_1arg): Add fmtcsr.
1558 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
1559 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
1560 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
1561 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
1562 unspec_fmfcfg): New patterns.
1563 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
1564 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
1565 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
1566 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
1567 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
1568 __nds32__fmfcfg): Define.
1570 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1572 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
1573 intrinsic register names.
1574 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
1575 intrinsic register enum values and macros.
1577 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1579 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
1580 for load/store addressing form.
1581 (nds32_print_operand_address): Likewise.
1583 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
1586 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
1587 based on LABEL_REF. Remove useless assertion.
1588 (pic_address_needs_scratch): Fix formatting.
1589 (sparc_legitimize_pic_address): Minor tweaks.
1590 (sparc_delegitimize_address): Adjust assertion accordingly.
1591 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
1592 into symbolic_operand.
1593 (movsi_high_pic_label_ref): Likewise.
1594 (movsi_lo_sum_pic_label_ref): Likewise.
1595 (movdi_pic_label_ref): Likewise.
1596 (movdi_high_pic_label_ref): Likewise.
1597 (movdi_lo_sum_pic_label_ref): Likewise.
1599 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
1601 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
1602 custom LIB_SPEC setup.
1604 2018-04-06 Ruslan Bukin <br@bsdpad.com>
1605 Kito Cheng <kito.cheng@gmail.com>
1607 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
1608 * config/riscv/freebsd.h: New.
1610 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1612 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
1613 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
1616 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1617 Kito Cheng <kito.cheng@gmail.com>
1619 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
1620 nds32_output_call, nds32_symbol_binds_local_p): New functions.
1621 * config/nds32/nds32-protos.h (nds32_output_call,
1622 nds32_output_return): Declare.
1623 * config/nds32/nds32.md: Refine all the call and return patterns.
1625 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1628 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
1629 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
1631 PR rtl-optimization/84872
1632 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
1633 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
1636 2018-04-06 Tamar Christina <tamar.christina@arm.com>
1638 * expr.c (copy_blkmode_to_reg): Revert 254862.
1639 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
1641 2018-04-06 Richard Biener <rguenther@suse.de>
1644 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
1645 after seeing a component reference with an adjacent field. Treat
1646 refs to arrays at struct end of external decls similar to
1647 refs to unconstrained commons.
1649 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1652 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
1653 look through SAVE_EXPRs with non-side-effects argument. Adjust
1655 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
1658 2018-04-06 Richard Biener <rguenther@suse.de>
1661 * alias.c (find_base_term): New wrapper around find_base_term
1662 unwinding CSELIB_VAL_PTR changes.
1663 (find_base_term): Do not restore CSELIB_VAL_PTR during the
1666 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1668 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
1670 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
1671 constant definitions.
1672 ("nop"): lr 0,0 -> nopr r0
1673 ("nop_lr0", "nop_lr1"): New insn definitions.
1675 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1677 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
1678 NDS32_V3PUSH_AVAILABLE_P macro.
1680 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
1681 Chung-Ju Wu <jasonwucj@gmail.com>
1683 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
1684 (nds32*-*-*): Add float and fpu_config into supported_defaults.
1685 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
1686 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
1687 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
1688 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
1689 * config/nds32/constraints.md: New constraints and checking for hard
1690 float configuration.
1691 * config/nds32/iterators.md: New mode iterator and attribute for hard
1692 float configuration.
1693 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
1695 * config/nds32/nds32-fpu.md: New file.
1696 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
1697 deal with hard float code generation.
1698 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
1700 (abi_type, float_reg_number): New enum type.
1701 * config/nds32/nds32-predicates.c: New predicates for hard float.
1702 * config/nds32/nds32-protos.h: Declare functions for hard float.
1703 * config/nds32/nds32.c: Implementation for hard float configuration.
1704 * config/nds32/nds32.h: Definitions for hard float configuration.
1705 * config/nds32/nds32.md: Include hard float machine description and
1706 modify patterns for hard float configuration.
1707 * config/nds32/nds32.opt: New options for hard float configuration.
1708 * config/nds32/predicates.md: New predicates for hard float
1711 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1713 * common/config/nds32/nds32-common.c
1714 (nds32_option_optimization_table): Enable -mreleax-hint by default.
1716 2018-04-05 Jakub Jelinek <jakub@redhat.com>
1719 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
1720 CONSTRUCTOR_ELT (ctor, ...)->value.
1722 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
1725 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
1727 2018-04-05 Tom de Vries <tom@codesourcery.com>
1730 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
1733 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1734 Kito Cheng <kito.cheng@gmail.com>
1736 * config/nds32/constraints.md (U33): Fine-tune checking condition.
1737 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
1738 * config/nds32/nds32.h (nds32_16bit_address_type): Add
1739 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
1741 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1742 Kito Cheng <kito.cheng@gmail.com>
1744 * config/nds32/constraints.md (Ufe): New memory constraint.
1745 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
1746 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
1747 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
1749 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
1750 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
1752 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1754 * config/nds32/nds32.md: Use optimize_size in the condition for
1755 alu-shift instructions.
1757 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1759 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
1761 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1763 * config/nds32/nds32.md (negsi2): Refine pattern.
1765 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1766 Chung-Ju Wu <jasonwucj@gmail.com>
1768 * config/nds32/iterators.md (shift_rotate): New code iterator.
1769 (shift): New code attribute.
1770 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
1771 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
1772 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
1773 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
1774 bit-wise operations.
1775 (andsi3, *andsi3): Ditto.
1776 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
1777 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
1778 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
1779 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
1780 nds32_ior_operand, nds32_xor_operand): New predicates.
1782 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1784 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
1785 (addsi3, subsi3): ... this.
1787 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1789 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
1791 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1793 * config/nds32/nds32.md: Adjust indention.
1795 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1797 * config/nds32/nds32.md (feature): New attribute.
1799 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1801 * config/nds32/nds32.md (subtype): New attribute.
1803 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
1806 * config/arm/arm-builtins.c (arm_expand_builtin): Change
1807 expansion to perform a bitwise AND of the argument followed by a
1808 boolean negation of the result.
1810 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
1812 PR rtl-optimization/84878
1813 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
1814 the basic block. Assert the use reference is not artificial and that
1815 it has an associated insn.
1817 2018-04-04 Michael Matz <matz@suse.de>
1819 * builtins.c (compute_objsize): Pass correct operand
1820 to array_at_struct_end_p.
1822 2018-04-04 Richard Biener <rguenther@suse.de>
1825 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
1826 from contexts for DINFO_LEVEL_TERSE and below.
1828 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1830 * config/nds32/nds32-doubleword.md (move_<mode>): Require
1831 resiter_operand condition.
1832 * config/nds32/nds32.md (*move<mode>): Ditto.
1834 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1835 Monk Chiang <sh.chiang04@gmail.com>
1837 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
1839 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1841 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
1843 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1844 Kito Cheng <kito.cheng@gmail.com>
1846 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
1847 nds32_cond_code_str, output_cond_branch,
1848 output_cond_branch_compare_zero, nds32_expand_cbranch,
1849 nds32_expand_cstore, nds32_expand_movcc,
1850 nds32_output_cbranchsi4_equality_zero,
1851 nds32_output_cbranchsi4_equality_reg,
1852 nds32_output_cbranchsi4_equality_reg_or_const_int,
1853 nds32_output_cbranchsi4_greater_less_zero: New functions.
1854 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
1855 nds32_expand_cstore, nds32_expand_movcc,
1856 nds32_output_cbranchsi4_equality_zero,
1857 nds32_output_cbranchsi4_equality_reg,
1858 nds32_output_cbranchsi4_equality_reg_or_const_int,
1859 nds32_output_cbranchsi4_greater_less_zero): Declare.
1860 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
1861 nds32_rimm11s_operand): New predicates.
1862 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
1863 * config/nds32/nds32.md: Rewrite all the branch and conditional move
1866 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1868 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
1869 * config/nds32/nds32.md: Ditto.
1870 * config/nds32/pipelines.md: Ditto.
1872 2018-04-04 Richard Biener <rguenther@suse.de>
1874 PR tree-optimization/85168
1875 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
1876 propagating abnormals.
1878 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1880 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
1882 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1883 Kito Cheng <kito.cheng@gmail.com>
1885 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
1886 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
1887 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
1888 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
1889 * config/nds32/nds32.md (sibcall_internal): New.
1890 (sibcall_register): Remove.
1891 (sibcall_immediate): Remove.
1892 (sibcall_value_internal): New.
1893 (sibcall_value_register): Remove.
1894 (sibcall_value_immediate): Remove.
1895 * config/nds32/predicates.md (nds32_general_register_operand): New.
1896 (nds32_call_address_operand): New.
1898 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1900 PR rtl-optimization/85167
1901 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
1902 bb_defs if *split_p, instead preinitialize it to NULL.
1904 PR tree-optimization/85156
1905 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
1906 evaluating the argument multiple times.
1908 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
1910 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
1912 (_mm_cvtpd_ps): Likewise.
1913 (_mm_cvttpd_epi32): Likewise.
1914 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
1915 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
1916 vector, pixel, and bool following altivec.h include.
1918 2018-04-03 Martin Sebor <msebor@redhat.com>
1920 * doc/extend.texi (Common Function Attributes): Clarify.
1921 (const attribute): Likewise.
1922 (pure attribute): Likewise.
1924 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1927 * config/i386/i386.c (ix86_expand_vector_set): Use
1928 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
1930 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
1932 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
1933 instructions when changing rounding bits to preserve precision bits
1934 in the x87 control word.
1936 2018-04-03 Martin Liska <mliska@suse.cz>
1938 PR tree-optimization/82491
1939 * rtl.h (strip_offset_and_add): Replace += suboffset with
1940 poly_uint64 () + suboffset.
1942 2018-03-29 Martin Liska <mliska@suse.cz>
1943 Martin Jambor <mjambor@suse.cz>
1946 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
1947 param_type is not an integral or pointer type.
1949 2018-04-03 Richard Biener <rguenther@suse.de>
1951 * sese.h (recompute_all_dominators): Remove.
1953 2018-04-02 Martin Sebor <msebor@redhat.com>
1955 * doc/invoke.texi (-Wrestrict): Fix typos.
1957 2018-04-02 Jim Wilson <jimw@sifive.com>
1959 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
1960 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
1961 (<optab>di3, <optab>si3_extend): Likewise.
1962 (<optab>si3_mask, <optab>si3_mask_1): New.
1963 (<optab>di3_mask, <optab>di3_mask_1): New.
1964 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
1965 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
1966 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
1968 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
1970 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
1973 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
1975 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
1976 (nds32_canonicalize_comparison): New function.
1978 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1979 Kito Cheng <kito.cheng@gmail.com>
1980 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1982 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
1983 * config/nds32/constants.md (unspec_volatile_element): Add
1984 UNSPEC_VOLATILE_RELAX_GROUP.
1985 * config/nds32/nds32-relax-opt.c: New file.
1986 * config/nds32/nds32-predicates.c
1987 (nds32_symbol_load_store_p): New function.
1988 * config/nds32/nds32-protos.h
1989 (nds32_symbol_load_store_p): Declare function.
1990 (make_pass_nds32_relax_opt): Declare new rtl pass function.
1991 * config/nds32/nds32.c
1992 (nds32_register_pass): New function to register pass.
1993 (nds32_register_passes): New function to register passes.
1994 * config/nds32/nds32.md (relax_group): New pattern.
1995 * config/nds32/nds32.opt (mrelax-hint): New option.
1996 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
1998 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
2000 * config/nds32/t-nds32: Modify files dependency.
2002 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2004 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
2005 (PROFILE_HOOK): Define its implementation.
2007 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2009 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
2010 type and 32-bit size.
2012 2018-04-01 Jakub Jelinek <jakub@redhat.com>
2015 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
2016 (V_128_256): New mode iterator.
2017 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
2018 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
2019 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
2021 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
2022 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
2024 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
2027 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
2028 NaN inputs correctly.
2030 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
2033 * config/rs6000/vsx.md (??r): New mode attribute.
2034 (*vsx_mov<mode>_64bit): Use it.
2035 (*vsx_mov<mode>_32bit): Likewise.
2037 2018-03-30 Martin Sebor <msebor@redhat.com>
2039 PR tree-optimization/84818
2040 * builtins.c (check_access): Use warning_n.
2042 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2045 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
2047 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
2050 2018-03-30 Julia Koval <julia.koval@intel.com>
2053 * x86-tune.def (movx, partial_reg_dependency): Enable for
2056 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
2059 * lra-constraints.c (process_alt_operands): Move setting
2060 this_alternative_matches below.
2062 2018-03-29 Martin Liska <mliska@suse.cz>
2065 * doc/invoke.texi: Document how LTO works with debug info.
2066 Describe auto-load support of binutils. Mention 'x86-64'
2067 as valid option value of -march option.
2069 2018-03-29 Jakub Jelinek <jakub@redhat.com>
2071 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
2074 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
2075 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
2076 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
2079 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2082 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
2083 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
2084 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
2085 for RS6000_BTM_POWERPC64.
2086 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
2087 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
2088 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
2093 2018-03-28 Carl Love <cel@us.ibm.com>
2096 2017-09-27 Carl Love <cel@us.ibm.com>
2098 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
2099 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
2100 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
2103 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2105 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
2106 instead of __vector bool.
2107 (_mm_max_pu8): Likewise.
2108 (_mm_min_pi16): Likewise.
2110 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2113 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
2114 (DIVWEUO): Likewise.
2116 (DIVDEUO): Likewise.
2117 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
2118 DIVWEUO and DIVDEUO.
2119 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
2120 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
2121 (div_extend): Likewise.
2122 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
2124 (__builtin_divweuo): Likewise.
2125 (__builtin_divdeo): Likewise.
2126 (__builtin_divdeuo): Likewise.
2128 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2131 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
2132 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
2134 PR tree-optimization/82004
2135 * gimple-match-head.c (optimize_pow_to_exp): New function.
2136 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
2137 Don't fold to exp if optimize_pow_to_exp is false.
2139 2018-03-28 Martin Liska <mliska@suse.cz>
2142 * calls.c (initialize_argument_information): Fix trailing space.
2143 * common.opt: Fix typo and provide better explanation for
2144 -fsanitize-coverage option.
2145 * config/i386/i386.opt: Fix typo.
2147 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2148 Martin Liska <mliska@suse.cz>
2151 * gimplify.c (asan_poison_variable): Don't do the check for
2152 gimplify_omp_ctxp here.
2153 (gimplify_decl_expr): Do it here.
2154 (gimplify_target_expr): Likewise.
2156 2018-03-28 Martin Liska <mliska@suse.cz>
2159 * config/i386/i386.c (ix86_function_arg_advance): Do not call
2160 chkp_type_bounds_count if MPX is not enabled.
2162 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
2164 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
2166 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
2169 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
2170 function to create the function decl for complex long double
2171 multiply and divide for -mabi=ieeelongdouble.
2172 (init_float128_ieee): Call it.
2174 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
2177 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
2178 -fcf-protection=branch -mibt.
2179 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
2181 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2184 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
2186 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
2189 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
2190 extern array declarations.
2192 2018-03-27 Richard Biener <rguenther@suse.de>
2195 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
2196 explicit single_use checks.
2198 2018-03-27 Richard Biener <rguenther@suse.de>
2200 PR tree-optimization/85082
2201 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
2204 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2206 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
2207 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
2208 Turn on fasynchronous-unwind-tables and funwind-tables.
2210 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
2213 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
2214 (*bmi_blsr_<mode>_ccz): Ditto.
2216 2018-03-26 Tom de Vries <tom@codesourcery.com>
2218 PR tree-optimization/85063
2219 * omp-general.c (offloading_function_p): New function. Factor out
2221 * omp-offload.c (pass_omp_target_link::gate): ... here.
2222 * omp-general.h (offloading_function_p): Declare.
2223 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
2224 with attribute omp declare target for offloading functions.
2226 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
2228 PR tree-optimization/84005
2229 * tree-data-ref.h (get_base_for_alignment): Declare.
2230 * tree-data-ref.c (get_base_for_alignment_1): New function.
2231 (get_base_for_alignment): Likewise.
2232 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
2233 get_base_for_alignment to find a suitable base object, instead
2234 of always using drb->base_address.
2236 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2239 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
2240 known size by default.
2242 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
2245 * lra-constraints.c (process_alt_operands): Don't match BLKmode
2246 and non BLKmode operands.
2248 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2251 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
2252 Clean up attributes.
2254 2018-03-23 Richard Biener <rguenther@suse.de>
2257 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
2258 we are going to emit early debug for LTO.
2260 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2263 * function.c (match_asm_constraints_1): Don't optimize if input
2264 doesn't satisfy general_operand predicate for output's mode.
2267 * alias.c (write_dependence_p): Don't require for x_canonicalized
2268 non-VOIDmode if x has VOIDmode.
2271 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
2272 just don't try to optimize it rather than assert it never happens.
2274 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2276 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
2277 macro expansions for definition of ST_INTERNAL_<mode> and
2278 LD_INTERNAL_<mode> builtins.
2279 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
2281 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
2283 (altivec_expand_st_builtin): Likewise.
2284 (altivec_expand_builtin): Remove calls to deleted functions.
2285 (rs6000_address_for_altivec): Delete this function.
2286 * config/rs6000/vector.md: Remove expands for
2287 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
2289 2018-03-22 Sudakshina Das <sudi.das@arm.com>
2292 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
2293 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
2294 re-computing once computed.
2295 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
2296 (arm_init_machine_status): Initialize
2297 machine->static_chain_stack_bytes.
2299 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2302 * doc/extend.texi: Add four new prototypes for vec_ld.
2303 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
2304 definitions for more logical presentation.
2305 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
2306 entries for V1TI variants of __builtin_altivec_ld builtin.
2307 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
2308 handling of V1TI variant of LVX icode pattern.
2309 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
2310 (rs6000_gimple_fold_builtin): Likewise.
2311 (altivec_init_builtins): Add code to define
2312 __builtin_altivec_lvx_v1ti function.
2314 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2317 * function.c (match_asm_constraints_1): Don't do the optimization
2318 if input isn't a REG, SUBREG, MEM or constant.
2320 2018-03-22 Tom de Vries <tom@codesourcery.com>
2322 PR tree-optimization/84956
2323 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
2324 bb_has_abnormal_pred.
2326 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2329 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
2330 DECL_INITIAL (decl) to decl at the end.
2331 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
2334 2018-03-21 Joseph Myers <joseph@codesourcery.com>
2336 * doc/extend.texi (__builtin_tgmath): Document when complex
2337 integer types are treated as _Complex _Float64.
2339 2018-03-21 Tom de Vries <tom@codesourcery.com>
2341 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
2343 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2345 PR tree-optimization/84960
2346 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
2347 if it is ENTRY block, move them into single succ of ENTRY in that case.
2349 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
2351 PR tree-optimization/84811
2352 * poly-int.h (poly_span_traits): Remove the T3 parameter and
2353 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
2354 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
2355 (known_subrange_p): Update accordingly. Cast each value involved
2356 in the size comparison, rather than casting the result of the
2359 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2361 PR tree-optimization/84982
2362 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
2363 by flipping the least significant bit rather than all bits from
2364 bitpos to bitpos + bitsize - 1.
2366 2018-03-21 Nathan Sidwell <nathan@acm.org>
2368 * doc/extend.texi (Deprecated Features): Remove mention of
2369 long-deleted deprecations.
2371 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2374 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
2375 * configure: Regenerate.
2377 2018-03-21 Tom de Vries <tom@codesourcery.com>
2379 PR tree-optimization/83126
2380 * tree-parloops.c (num_phis): New function.
2381 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
2383 2018-03-21 Nathan Sidwell <nathan@acm.org>
2385 * doc/extend.texi (Deprecated Features): Update deprecated flags,
2386 mention anon-struct/union members and trailing attributes.
2388 2018-03-21 Bin Cheng <bin.cheng@arm.com>
2390 PR tree-optimization/84969
2391 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
2392 builtin memset partitions if they set different rhs values.
2394 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2396 PR rtl-optimization/84989
2397 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
2398 VEC_DUPLICATE with scalar result mode.
2400 2018-03-21 Martin Liska <mliska@suse.cz>
2403 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
2404 not intended return statement.
2406 2018-03-21 Martin Liska <mliska@suse.cz>
2409 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
2410 (chkp_find_bound_slots_1): Limit number of iterations.
2412 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
2415 * Minor grammar fixes for x86 options.
2417 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2420 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
2421 holding REG_CFA_RESTORE notes, instead turn them into a USE.
2423 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
2426 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
2427 (altivec_lvx_<mode>_1op): Likewise.
2428 (altivec_stvx_<mode>_2op): Likewise.
2429 (altivec_stvx_<mode>_1op): Likewise.
2430 (altivec_lvx_<VM2:mode>): New define_expand.
2431 (altivec_stvx_<VM2:mode>): Likewise.
2432 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
2433 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2434 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
2435 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2436 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
2437 (rs6000_gen_lvx): Likewise.
2438 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
2439 (altivec_expand_stv_builtin): Likewise.
2440 (altivec_expand_builtin): Likewise.
2441 * config/rs6000/vector.md: Likewise.
2443 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2446 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
2449 2018-03-20 Richard Biener <rguenther@suse.de>
2452 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
2453 sign-conversions as zero, fall back to standard scalar_stmt
2456 2018-03-20 Martin Liska <mliska@suse.cz>
2459 * predict.c (rebuild_frequencies): Handle case when we have
2460 PROFILE_ABSENT, but flag_guess_branch_prob is false.
2462 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2465 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
2466 flag_section_anchors.
2467 * varasm.c (use_blocks_for_decl_p): Remove hack for
2468 dw2_force_const_mem.
2471 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
2473 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
2474 be created, use lowpart_subreg of operands[0] rather than operands[0]
2476 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
2477 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
2478 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
2479 and n constraint instead of aarch64_shift_imm_di and Usd.
2480 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
2481 (*aarch64_<optab>_reg_minus<mode>3): ... this.
2483 2018-03-20 Sudakshina Das <sudi.das@arm.com>
2486 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
2487 to favor GPR over NEON registers.
2488 (<shift>di3_neon): Likewise.
2490 2018-03-20 Tom de Vries <tom@codesourcery.com>
2493 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
2494 (nvptx_process_pars): Emit bar.sync asap and alap.
2496 2018-03-20 Tom de Vries <tom@codesourcery.com>
2499 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
2500 seen_label if seen_label is already set.
2502 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2505 * config/i386/i386.c (fold_builtin_cpu): For features above 31
2506 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
2507 Use 1U instead of 1. Formatting fixes.
2510 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
2511 instead of TREE_TYPE (s1) for the return value.
2513 2018-03-19 Jakub Jelinek <jakub@redhat.com>
2515 PR tree-optimization/84946
2516 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
2517 bitsize + bitsize in poly_uint64 rather than poly_int64.
2520 * dwarf2asm.c: Include fold-const.c.
2521 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
2522 of decl rather than decl itself.
2524 PR rtl-optimization/84643
2525 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
2527 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
2530 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
2531 calling assemble_variable.
2533 2018-03-19 Sudakshina Das <sudi.das@arm.com>
2536 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
2537 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
2539 2018-03-19 Jim Wilson <jimw@sifive.com>
2542 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
2543 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
2544 (riscv_first_stack_step): Likewise.
2545 (riscv_option_override): Use STACK_BOUNDARY instead of
2547 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
2549 (BIGGEST_ALIGNMENT): Set to 128.
2550 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
2551 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
2554 2018-03-19 Richard Biener <rguenther@suse.de>
2556 PR tree-optimization/84933
2557 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
2558 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
2560 2018-03-19 Richard Biener <rguenther@suse.de>
2562 PR tree-optimization/84859
2563 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
2564 (cond_if_else_store_replacement): Perform sinking operation on
2565 single-store BBs regardless of MAX_STORES_TO_SINK setting.
2566 Generalize what a BB with a single eligible store is.
2568 2018-03-19 Richard Biener <rguenther@suse.de>
2570 PR tree-optimization/84929
2571 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
2572 chrec_is_positive against non-chrec arg.
2574 2018-03-19 Tamar Christina <tamar.christina@arm.com>
2577 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
2579 2018-03-18 Martin Liska <mliska@suse.cz>
2581 PR rtl-optimization/84635
2582 * regrename.c (build_def_use): Use matches_mode only when
2585 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
2587 PR tree-optimization/84913
2588 * tree-vect-loop.c (vectorizable_reduction): Don't try to
2589 vectorize chains of COND_EXPRs.
2591 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2593 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
2595 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2597 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
2599 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2601 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
2603 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2604 Kito Cheng <kito.cheng@gmail.com>
2606 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
2607 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
2608 (nds32_adjust_reg_alloc_order): New function.
2609 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
2611 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
2613 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
2614 nds32_print_operand, nds32_print_operand_address): Use
2615 HOST_WIDE_INT_PRINT_DEC instead.
2617 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2619 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
2621 2018-03-17 Jakub Jelinek <jakub@redhat.com>
2624 * config/i386/i386.c (initial_ix86_tune_features,
2625 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
2627 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
2628 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
2629 rather than 1u << ix86_tune. Formatting fix.
2630 (ix86_option_override_internal): Change ix86_arch_mask from
2631 unsigned int to unsigned HOST_WIDE_INT, initialize to
2632 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
2633 (ix86_function_specific_restore): Likewise.
2635 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2638 * postreload.c (reload_combine_recognize_pattern): Perform
2639 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
2640 truncate_int_for_mode the result for the destination's mode.
2643 * hsa-gen.c (mem_type_for_type): Fix comment typo.
2644 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
2646 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2649 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
2652 * lra-assigns.c (lra_split_hard_reg_for): Don't use
2653 regno_allocno_class_array and sorted_pseudos.
2654 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
2655 insns where regno is used.
2657 2018-03-16 Martin Liska <mliska@suse.cz>
2660 * multiple_target.c (create_dispatcher_calls): Redirect
2661 reference in the symbol table.
2663 2018-03-16 Martin Liska <mliska@suse.cz>
2666 * multiple_target.c (create_dispatcher_calls): Redirect also
2669 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2673 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
2674 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
2675 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
2677 2018-03-16 Julia Koval <julia.koval@intel.com>
2679 * doc/invoke.texi (Skylake Server): Add CLWB.
2680 Cannonlake): Remove CLWB.
2682 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2684 PR tree-optimization/84841
2685 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
2687 (FLOAT_ONE_CONST_TYPE): Define.
2688 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
2689 (sort_by_operand_rank): Put entries with higher constant_type last
2690 rather than first to match comments.
2692 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
2694 * config/nios2/nios2.md (movsi_internal): Fix thinko in
2697 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2700 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
2701 check and use address of target always.
2703 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
2706 * config/i386/i386.c (indirect_thunk_needed): Update comments.
2707 (indirect_thunk_bnd_needed): Likewise.
2708 (indirect_thunks_used): Likewise.
2709 (indirect_thunks_bnd_used): Likewise.
2710 (indirect_return_needed): New.
2711 (indirect_return_bnd_needed): Likewise.
2712 (output_indirect_thunk_function): Add a bool argument for
2714 (output_indirect_thunk_function): Don't generate alias for
2715 function return thunk.
2716 (ix86_code_end): Call output_indirect_thunk_function to generate
2717 function return thunks.
2718 (ix86_output_function_return): Set indirect_return_bnd_needed
2719 and indirect_return_needed instead of indirect_thunk_bnd_needed
2720 and indirect_thunk_needed.
2722 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
2724 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
2725 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
2726 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
2728 2018-03-15 David Malcolm <dmalcolm@redhat.com>
2729 Paul Hua <paul.hua.gm@gmail.com>
2732 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
2734 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
2736 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
2737 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
2740 2018-03-15 Tamar Christina <tamar.christina@arm.com>
2743 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
2744 instead of GET_MODE_SIZE when comparing Units.
2746 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2749 * varasm.c (hash_section): Return an unchangeble hash value
2750 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
2751 Return !aarch64_can_use_per_function_literal_pools_p ().
2753 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2756 * optabs.c (emit_conditional_move): Pass address of cmode's copy
2757 rather than address of cmode as last argument to prepare_cmp_insn.
2759 2018-03-15 Julia Koval <julia.koval@intel.com>
2761 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
2762 F_AVX512VNNI, F_AVX512BITALG): New.
2764 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
2767 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
2768 insn for floating-point loads and stores.
2770 2018-03-14 Carl Love <cel@us.ibm.com>
2772 * config/rs6000/rs6000-c.c: Add macro definitions for
2773 ALTIVEC_BUILTIN_VEC_PERMXOR.
2774 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
2775 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
2776 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
2777 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
2779 * config/doc/extend.texi: Add prototypes for vec_permxor.
2781 2018-03-14 David Malcolm <dmalcolm@redhat.com>
2784 * diagnostic-show-locus.c (class layout_point): Convert m_line
2785 from int to linenum_type.
2786 (line_span::comparator): Use linenum "compare" function when
2787 comparing line numbers.
2788 (test_line_span): New function.
2789 (layout_range::contains_point): Convert param "row" from int to
2791 (layout_range::intersects_line_p): Likewise.
2792 (layout::will_show_line_p): Likewise.
2793 (layout::print_source_line): Likewise.
2794 (layout::should_print_annotation_line_p): Likewise.
2795 (layout::print_annotation_line): Likewise.
2796 (layout::print_leading_fixits): Likewise.
2797 (layout::annotation_line_showed_range_p): Likewise.
2798 (struct line_corrections): Likewise for field m_row.
2799 (line_corrections::line_corrections): Likewise for param "row".
2800 (layout::print_trailing_fixits): Likewise.
2801 (layout::get_state_at_point): Likewise.
2802 (layout::get_x_bound_for_row): Likewise.
2803 (layout::print_line): Likewise.
2804 (diagnostic_show_locus): Likewise for locals "last_line" and
2806 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
2807 * input.c (selftest::test_linenum_comparisons): New function.
2808 (selftest::input_c_tests): Call it.
2809 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
2810 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
2811 * selftest.h (ASSERT_GT): New macro.
2812 (ASSERT_GT_AT): New macro.
2813 (ASSERT_LT): New macro.
2814 (ASSERT_LT_AT): New macro.
2816 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
2818 PR rtl-optimization/84780
2819 * combine.c (distribute_links): Don't make a link based on pc_rtx.
2821 2018-03-14 Martin Liska <mliska@suse.cz>
2823 * tree.c (record_node_allocation_statistics): Use
2824 get_stats_node_kind.
2825 (get_stats_node_kind): New function extracted from
2826 record_node_allocation_statistics.
2827 (free_node): Use get_stats_node_kind.
2829 2018-03-14 Richard Biener <rguenther@suse.de>
2831 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
2832 that the value-set of ANTIC_IN doesn't grow.
2835 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2837 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2838 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2840 2018-03-14 Julia Koval <julia.koval@intel.com>
2842 * config.gcc (icelake-client, icelake-server): New.
2844 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
2845 (initial_ix86_arch_features): Ditto.
2846 (PTA_SKYLAKE): Add SGX.
2847 (PTA_ICELAKE): Remove.
2848 (PTA_ICELAKE_CLIENT): New.
2849 (PTA_ICELAKE_SERVER): New.
2850 (ix86_option_override_internal): Split up icelake on icelake client and
2852 (get_builtin_code_for_version): Ditto.
2853 (fold_builtin_cpu): Ditto.
2854 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
2855 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
2856 * config/i386/i386.h (processor_type): Ditto.
2857 * doc/invoke.texi: Ditto.
2859 2018-03-14 Jakub Jelinek <jakub@redhat.com>
2862 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
2863 INTEGER_CST offset, add it together with bitpos / 8 and
2864 sign extend based on POINTER_SIZE.
2868 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
2871 * config/i386/constraints.md (Yc): New register constraint.
2872 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
2873 Use Yc constraint for alternative 2 of operand 0. Remove
2874 preferred_for_speed attribute.
2876 2018-03-14 Richard Biener <rguenther@suse.de>
2878 PR tree-optimization/84830
2879 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
2880 with the old one to avoid oscillations.
2882 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
2885 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
2887 (assign_by_spills): Return a flag of reload assignment failure.
2888 Do not process the reload assignment failures. Do not spill other
2889 reload pseudos if they has the same reg class. Update n if
2891 (lra_assign): Add a return arg. Set up from the result of
2892 assign_by_spills call.
2893 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2894 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2895 usage_insns if it is not NULL.
2896 (spill_hard_reg_in_range): New function.
2897 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2898 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2899 function prototypes.
2900 (lra_assign): Change prototype.
2901 * lra.c (lra): Add code to deal with fails by splitting hard reg
2904 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
2906 * config/riscv/riscv.opt (mrelax): New option.
2907 * config/riscv/riscv.c (riscv_file_start): Emit ".option
2908 "norelax" when riscv_mrelax is disabled.
2909 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
2911 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2914 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
2915 reassociation for int modes.
2917 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2919 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
2920 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
2922 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
2923 * config/aarch64/aarch64-sve.md
2924 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
2925 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
2926 (*extend<mode><Vwide>2): Rename to...
2927 (aarch64_sve_extend<mode><Vwide>2): ...this.
2928 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
2929 renaming the old pattern to...
2930 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
2932 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
2933 define_expand, renaming the old pattern to...
2934 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
2935 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
2936 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
2937 account when deciding which SVE instruction the optab should use.
2938 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
2940 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2942 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
2943 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
2944 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
2945 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
2946 (tlsdesc_small_<mode>): Turn a define_expand and use
2947 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
2948 (tlsdesc_small_advsimd_<mode>): ...this.
2949 (tlsdesc_small_sve_<mode>): New pattern.
2951 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2953 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
2954 (UNSPEC_UMUL_HIGHPART): New constants.
2955 (MUL_HIGHPART): New int iteraor.
2956 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
2957 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
2959 (*<su>mul<mode>3_highpart): New define_insn.
2961 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2964 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2967 2018-03-13 Martin Liska <mliska@suse.cz>
2970 * (sem_item_optimizer::sem_item_optimizer): Initialize new
2972 (sem_item_optimizer::~sem_item_optimizer): Release it.
2973 (sem_item_optimizer::merge_classes): Register variable aliases.
2974 (sem_item_optimizer::fixup_pt_set): New function.
2975 (sem_item_optimizer::fixup_points_to_sets): Likewise.
2976 * ipa-icf.h: Declare new variables and functions.
2978 2018-03-13 Jakub Jelinek <jakub@redhat.com>
2981 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
2982 integer_pow2p@2 and test integer_pow2p in condition.
2983 (A < 0 ? C : 0): Similarly for @1.
2986 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
2987 characters starting at p contain '\0' character, don't look beyond
2991 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
2992 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
2995 * reg-stack.c (change_stack): Change update_end var from int to
2996 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
2997 also call set_block_for_insn on the newly added insns and rescan.
3000 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
3001 on the last operand.
3004 * tree.c (stabilize_reference_1): Return save_expr (e) for
3005 STATEMENT_LIST even if it doesn't have side-effects.
3007 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
3009 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
3011 2018-03-12 Renlin Li <renlin.li@arm.com>
3013 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
3014 aarch64_output_scalar_simd_mov_immediate.
3016 2018-03-12 Martin Sebor <msebor@redhat.com>
3018 PR tree-optimization/83456
3019 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
3020 for perfectly overlapping calls to memcpy.
3021 (gimple_fold_builtin_memory_chk): Same.
3022 (gimple_fold_builtin_strcpy): Handle no-warning.
3023 (gimple_fold_builtin_stxcpy_chk): Same.
3024 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
3026 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3028 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
3029 parameter. Use it for SFmode.
3030 (rs6000_function_arg_advance_1): Adjust.
3031 (rs6000_function_arg): Adjust.
3032 (rs6000_gimplify_va_arg): Pass false for that new parameter.
3034 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3036 PR rtl-optimization/84169
3037 PR rtl-optimization/84780
3038 * combine.c (can_combine_p): Check for a 2-insn combination whether
3039 the destination register is used between the two insns, too.
3041 2018-03-12 Richard Biener <rguenther@suse.de>
3043 PR tree-optimization/84803
3044 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
3045 for refs DR analysis didn't process.
3047 2018-03-12 Richard Biener <rguenther@suse.de>
3049 PR tree-optimization/84777
3050 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
3051 force-vectorize loops ignore whether we are optimizing for size.
3053 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
3055 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
3056 (TARGET_MD_ASM_ADJUST): Define.
3058 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
3059 Kito Cheng <kito.cheng@gmail.com>
3060 Chung-Ju Wu <jasonwucj@gmail.com>
3062 * config/nds32/nds32.c (nds32_compute_stack_frame,
3063 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
3064 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
3065 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
3066 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
3067 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
3068 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
3069 * config/nds32/nds32.md (prologue, epilogue): Use macro
3070 NDS32_V3PUSH_AVAILABLE_P to do checking.
3072 2018-03-11 Jakub Jelinek <jakub@redhat.com>
3075 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
3076 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
3077 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
3078 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
3079 addition of most attributes on !orig_type_die or the attribute not
3080 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
3082 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3083 Chung-Ju Wu <jasonwucj@gmail.com>
3085 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
3087 * config/nds32/nds32.opt (mvh): New option.
3089 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3090 Chung-Ju Wu <jasonwucj@gmail.com>
3092 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
3094 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
3095 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
3098 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3099 Chung-Ju Wu <jasonwucj@gmail.com>
3101 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
3103 * config/nds32/nds32-multiple.md (strlensi): New pattern.
3104 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
3106 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
3107 Kito Cheng <kito.cheng@gmail.com>
3108 Chung-Ju Wu <jasonwucj@gmail.com>
3110 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
3111 UNSPEC_FFMISM and UNSPEC_FLMISM.
3112 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
3113 for ffb, ffmism and flmism.
3114 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
3115 (unspec_ffmism): Ditto.
3116 (unspec_flmism): Ditto.
3117 (nds32_expand_builtin_impl): Check if string extension is available.
3118 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
3119 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
3121 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
3124 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3127 * lra-assigns.c (assign_by_spills): Return a flag of reload
3128 assignment failure. Do not process the reload assignment
3129 failures. Do not spill other reload pseudos if they has the same
3131 (lra_assign): Add a return arg. Set up from the result of
3132 assign_by_spills call.
3133 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3134 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3135 usage_insns if it is not NULL.
3136 (spill_hard_reg_in_range): New function.
3137 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3138 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3139 function prototypes.
3140 (lra_assign): Change prototype.
3141 * lra.c (lra): Add code to deal with fails by splitting hard reg
3144 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
3147 * config/i386/i386.opt: Replace Enforcment with Enforcement.
3149 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
3152 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
3153 (dw_val_node): Add val_symbolic_view.
3154 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
3155 (symview_upper_bound): New.
3156 (new_line_info_table): Initialize symviews_since_reset.
3157 (dwarf2out_source_line): Count symviews_since_reset and set
3158 symview_upper_bound.
3159 (dw_val_equal_p): Handle symview.
3160 (add_AT_symview): New.
3161 (print_dw_val): Handle symview.
3162 (attr_checksum, attr_checksum_ordered): Likewise.
3163 (same_dw_val_p, size_of_die): Likewise.
3164 (value_format, output_die): Likewise.
3165 (add_high_low_attributes): Use add_AT_symview for entry_view.
3166 (dwarf2out_finish): Reset symview_upper_bound, clear
3169 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
3172 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
3173 Add strict argument and use it.
3174 (rs6000_split_multireg_move): Update for new strict argument.
3175 (mem_operand_gpr): Disallow all non-offsettable addresses.
3176 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
3178 2018-03-09 Jakub Jelinek <jakub@redhat.com>
3181 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
3182 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
3183 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
3186 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
3187 decl, use remap_type if we want to use the type.
3189 2018-03-09 Martin Sebor <msebor@redhat.com>
3191 PR tree-optimization/84526
3192 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3194 (builtin_access::generic_overlap): Be prepared to handle non-array
3197 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
3199 PR rtl-optimization/84682
3200 * lra-constraints.c (process_address_1): Check is_address flag
3201 for address constraints.
3202 (process_alt_operands): Likewise.
3203 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
3204 preprocess_constraints.
3205 * recog.h (preprocess_constraints): Add oploc parameter.
3207 * recog.c (preprocess_constraints): Test address_operand for
3208 CT_ADDRESS constraints.
3210 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3213 * lra-assigns.c (assign_by_spills): Return a flag of reload
3214 assignment failure. Do not process the reload assignment
3215 failures. Do not spill other reload pseudos if they has the same
3217 (lra_assign): Add a return arg. Set up from the result of
3218 assign_by_spills call.
3219 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3220 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3221 usage_insns if it is not NULL.
3222 (spill_hard_reg_in_range): New function.
3223 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3224 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3225 function prototypes.
3226 (lra_assign): Change prototype.
3227 * lra.c (lra): Add code to deal with fails by splitting hard reg
3230 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3233 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
3234 Accept complain bool parameter. Only emit errors if it is true.
3235 (arm_parse_cpu_option_name): Likewise.
3236 (arm_target_thumb_only): Adjust callers of the above.
3237 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
3238 prototype to take a default true bool parameter.
3239 (arm_parse_arch_option_name): Likewise.
3241 2018-03-09 David Malcolm <dmalcolm@redhat.com>
3242 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
3246 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
3247 * configure: Regenerate.
3248 * configure.ac ("linker --version-script option"): New.
3249 ("linker soname option"): New.
3251 2018-03-09 Richard Biener <rguenther@suse.de>
3253 PR tree-optimization/84775
3254 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
3255 immediate uses of predicate stmts and mark them modified.
3258 PR tree-optimization/84178
3259 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3261 (version_loop_for_if_conversion): Delay update_ssa call.
3262 (tree_if_conversion): Delay update_ssa until after predicate
3265 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
3268 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
3269 when the function accesses prior frames.
3271 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3274 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
3275 gen_llsym, otherwise call maybe_gen_llsym.
3278 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
3279 has ',' character inside of it.
3281 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3284 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
3285 as clobbering CC_REGNUM.
3287 2018-03-08 Richard Biener <rguenther@suse.de>
3290 * tree-scalar-evolution.c: Include tree-into-ssa.h.
3291 (follow_copies_to_constant): Do not follow SSA names registered
3294 2018-03-08 Richard Biener <rguenther@suse.de>
3296 PR tree-optimization/84178
3297 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3299 (version_loop_for_if_conversion): Delay update_ssa call.
3300 (tree_if_conversion): Delay update_ssa until after predicate
3303 2018-03-08 David Malcolm <dmalcolm@redhat.com>
3305 PR tree-optimization/84178
3306 * tree-if-conv.c (release_bb_predicate): Remove the
3307 the assertion that the stmts have NULL use_ops.
3308 Discard the statements, asserting that they haven't
3309 yet been added to a BB.
3311 2018-03-08 Richard Biener <rguenther@suse.de>
3313 PR tree-optimization/84746
3314 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
3315 (phi_translate): Pass in destination ANTIC_OUT set.
3316 (phi_translate_1): Likewise. For a simplified result lookup
3317 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
3318 (phi_translate_set): Adjust.
3319 (do_pre_regular_insertion): Likewise.
3320 (do_pre_partial_partial_insertion): Likewise.
3322 2018-03-08 Martin Liska <mliska@suse.cz>
3324 PR gcov-profile/84735
3325 * doc/gcov.texi: Document usage of profile files.
3326 * gcov-io.h: Document changes in the format.
3328 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
3332 * dwarf2out.c (struct dw_line_info_table): Update comments for
3334 (FORCE_RESET_NEXT_VIEW): New.
3335 (FORCE_RESETTING_VIEW_P): New.
3336 (RESETTING_VIEW_P): Check for -1 too.
3337 (ZERO_VIEW_P): Likewise.
3338 (new_line_info_table): Force-reset next view.
3339 (dwarf2out_begin_function): Likewise.
3340 (dwarf2out_source_line): Simplify zero_view_p initialization.
3341 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
3342 view directly. Omit view when omitting .loc at line 0.
3344 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3346 PR tree-optimization/84740
3347 * tree-switch-conversion.c (process_switch): Call build_constructors
3348 only if info.phi_count is non-zero.
3350 PR tree-optimization/84739
3351 * tree-tailcall.c (find_tail_calls): Check call arguments against
3352 DECL_ARGUMENTS (current_function_decl) rather than
3353 DECL_ARGUMENTS (func) when checking for tail recursion.
3355 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3357 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
3358 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
3359 Volker Reichelt's entry and add entries for people that perform
3360 GCC fuzzy testing and report numerous bugs.
3362 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
3365 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
3366 readonly data in sdata, if that is disabled.
3367 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
3368 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3369 -mreadonly-in-sdata option.
3371 2018-03-07 Martin Sebor <msebor@redhat.com>
3373 PR tree-optimization/84468
3374 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
3375 basic block when looking for nul assignment.
3377 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
3380 * except.h (output_function_exception_table): Adjust prototype.
3381 * except.c (output_function_exception_table): Remove FNNAME parameter
3382 and add SECTION parameter. Ouput one part of the table at a time.
3383 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
3384 the first part of the exception table and emit unwind directives.
3385 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
3386 (i386_pe_seh_cold_init): Likewise.
3387 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
3388 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
3389 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
3390 (ix86_output_call_insn): Emit a nop in one more case for SEH.
3391 * config/i386/winnt.c: Include except.h.
3392 (struct seh_frame_state): Add reg_offset, after_prologue and
3393 in_cold_section fields.
3394 (i386_pe_seh_end_prologue): Set seh->after_prologue.
3395 (i386_pe_seh_cold_init): New function.
3396 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
3397 to seh->in_cold_section.
3398 (seh_emit_push): Record the offset of the push.
3399 (seh_emit_save): Record the offet of the save.
3400 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
3401 Test seh->after_prologue to disregard the epilogue.
3402 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
3403 (i386_pe_end_cold_function): New function.
3405 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3408 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
3409 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
3412 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
3414 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
3415 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
3418 * multiple_target.c: Include tree-inline.h and intl.h.
3419 (expand_target_clones): Diagnose and fail if node->definition and
3420 !tree_versionable_function_p (node->decl).
3422 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
3424 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
3426 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
3427 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
3428 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
3430 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3433 * combine.c (try_combine): Use reg_or_subregno instead of handling
3434 just paradoxical SUBREGs and REGs.
3436 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
3438 * config/arc/arc.c (arc_finalize_pic): Remove function.
3439 (arc_must_save_register): We use single base PIC register, remove
3440 checks to save/restore the PIC register.
3441 (arc_expand_prologue): Likewise.
3442 * config/arc/arc-protos.h (arc_set_default_type_attributes):
3444 (arc_verify_short): Likewise.
3445 (arc_attr_type): Likewise.
3446 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
3447 (walk_stores): Likewise.
3448 (arc_address_cost): Make it static.
3449 (arc_verify_short): Likewise.
3450 (branch_dest): Likewise.
3451 (arc_attr_type): Likewise.
3452 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
3453 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
3454 (arc_final_prescan_insn): Remove inserting the nops due to
3455 hardware hazards. It is done in reorg step.
3456 (insn_length_variant_t): Remove.
3457 (insn_length_parameters_t): Likewise.
3458 (arc_insn_length_parameters): Likewise.
3459 (arc_get_insn_variants): Likewise.
3460 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
3462 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3465 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
3468 PR tree-optimization/84687
3469 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
3471 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
3473 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3475 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
3476 Rename to ppc_speculation_barrier.
3477 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
3478 __builtin_ppc_speculation_barrier.
3480 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3483 * combine.c (combine_simplify_rtx): Don't try to simplify if
3484 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
3487 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
3489 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
3490 to 32 bytes when compiling for POWER9.
3492 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3495 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
3496 regparm >= 3 with no arg reg available also for calls with
3497 flag_force_indirect_call. Pass decl to ix86_function_regparm.
3500 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
3502 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
3504 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
3507 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
3509 2018-03-05 Richard Biener <rguenther@suse.de>
3511 PR tree-optimization/84486
3512 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
3513 When inserting a __builtin_assume_aligned call set the LHS
3514 SSA name alignment info accordingly.
3516 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
3518 PR tree-optimization/84114
3519 * config/aarch64/aarch64.c (aarch64_reassociation_width)
3520 Avoid reassociation of FLOAT_MODE addition.
3522 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
3524 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
3525 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
3526 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
3527 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
3528 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
3529 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
3530 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
3532 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
3533 __builtin_ia32_wbinvd): New builtins.
3534 (SPECIAL_ARGS2): New.
3535 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
3536 (SPECIAL_ARGS2): New.
3537 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
3538 (ix86_valid_target_attribute_inner_p): Ditto.
3539 (ix86_init_mmx_sse_builtins): Add special_args2.
3540 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
3541 TARGET_WBNOINVD_P): New.
3542 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
3543 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
3544 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
3545 * config/i386/immintrin.h (_wbinvd): New intrinsic.
3546 * config/i386/pconfigintrin.h: New file.
3547 * config/i386/wbnoinvdintrin.h: Ditto.
3548 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
3549 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
3551 2018-03-05 Richard Biener <rguenther@suse.de>
3553 PR tree-optimization/84670
3554 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3556 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3557 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3558 (compute_antic_aux): Only assert the number of values in ANTIC_IN
3559 doesn't grow if all successors (recursively) were visited at least
3562 2018-03-05 Richard Biener <rguenther@suse.de>
3564 PR tree-optimization/84650
3565 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
3566 if executed in the loop pipeline.
3568 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
3570 * doc/configfiles.texi (Configuration Files): Move info about
3571 conditionalizing $target-protos.h to...
3572 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
3573 differs from $target-protos.h.
3575 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
3576 Chung-Ju Wu <jasonwucj@gmail.com>
3578 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
3579 * config/nds32/nds32-multiple.md (setmemsi): Define.
3580 * config/nds32/nds32-memory-manipulation.c
3581 (nds32_gen_dup_4_byte_to_word_value): New.
3582 (emit_setmem_word_loop): New.
3583 (emit_setmem_byte_loop): New.
3584 (nds32_expand_setmem_loop): New.
3585 (nds32_expand_setmem_loop_v3m): New.
3586 (nds32_expand_setmem_unroll): New.
3587 (nds32_expand_setmem): New.
3589 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3590 Chung-Ju Wu <jasonwucj@gmail.com>
3592 * config/nds32/nds32-memory-manipulation.c
3593 (nds32_emit_load_store): New.
3594 (nds32_emit_post_inc_load_store): New.
3595 (nds32_emit_mem_move): New.
3596 (nds32_emit_mem_move_block): New.
3597 (nds32_expand_movmemsi_loop_unknown_size): New.
3598 (nds32_expand_movmemsi_loop_known_size): New.
3599 (nds32_expand_movmemsi_loop): New.
3600 (nds32_expand_movmemsi_unroll): New.
3601 (nds32_expand_movmemqi): Rename ...
3602 (nds32_expand_movmemsi): ... to this.
3603 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
3604 (movmemsi): ... to this.
3605 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
3606 (nds32_expand_movmemsi): ... to this.
3608 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3609 Monk Chiang <sh.chiang04@gmail.com>
3610 Chung-Ju Wu <jasonwucj@gmail.com>
3612 * config/nds32/nds32-protos.h
3613 (nds32_expand_load_multiple): New arguments.
3614 (nds32_expand_store_multiple): Ditto.
3615 (nds32_valid_multiple_load_store): Rename ...
3616 (nds32_valid_multiple_load_store_p): ... to this.
3617 * config/nds32/nds32-memory-manipulation.c
3618 (nds32_expand_load_multiple): Refine implementation.
3619 (nds32_expand_store_multiple): Ditto.
3620 * config/nds32/nds32-multiple.md
3621 (load_multiple): Update nds32_expand_load_multiple interface.
3622 (store_multiple): Update nds32_expand_store_multiple interface.
3623 * config/nds32/nds32-predicates.c
3624 (nds32_valid_multiple_load_store): Rename ...
3625 (nds32_valid_multiple_load_store_p): ... to this and refine
3627 * config/nds32/predicates.md
3628 (nds32_load_multiple_and_update_address_operation): New predicate.
3629 (nds32_store_multiple_and_update_address_operation): New predicate.
3631 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3632 Chung-Ju Wu <jasonwucj@gmail.com>
3634 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
3635 (combo): New attribute.
3636 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
3638 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
3640 * config/nds32/nds32.opt: Change -mcmodel= default value.
3642 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
3643 Monk Chiang <sh.chiang04@gmail.com>
3644 Chung-Ju Wu <jasonwucj@gmail.com>
3646 * config/nds32/constants.md (unspec_element): New enum.
3647 * config/nds32/constraints.md (Umw): New constraint.
3648 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
3649 * config/nds32/nds32-intrinsic.md: Likewise.
3650 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
3651 (nds32_valid_smw_lwm_base_p): New.
3652 (nds32_output_smw_single_word): New.
3653 (nds32_output_lmw_single_word): New.
3654 (nds32_expand_unaligned_load): New.
3655 (nds32_expand_unaligned_store): New.
3656 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
3657 (nds32_output_smw_single_word): Declare.
3658 (nds32_output_lmw_single_word): Declare.
3659 (nds32_expand_unaligned_load): Declare.
3660 (nds32_expand_unaligned_store): Declare.
3661 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
3662 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
3663 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
3664 NDS32_BUILTIN_UASTORE_DW.
3665 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
3668 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
3669 Kito Cheng <kito.cheng@gmail.com>
3670 Chung-Ju Wu <jasonwucj@gmail.com>
3672 * config/nds32/nds32-intrinsic.c
3673 (nds32_expand_builtin_null_ftype_reg): Delete.
3674 (nds32_expand_builtin_reg_ftype_imm): Ditto.
3675 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
3676 (nds32_read_argument): New.
3677 (nds32_legitimize_target): Ditto.
3678 (nds32_legitimize_argument): Ditto.
3679 (nds32_check_constant_argument): Ditto.
3680 (nds32_expand_unop_builtin): Ditto.
3681 (nds32_expand_unopimm_builtin): Ditto.
3682 (nds32_expand_binop_builtin): Ditto.
3683 (nds32_builtin_decl_impl): Ditto.
3684 (builtin_description): Ditto.
3685 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
3686 (nds32_init_builtins_impl): Ditto.
3687 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
3688 (nds32_builtin_decl): New.
3689 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
3690 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
3692 2018-03-02 Jeff Law <law@redhat.com>
3694 * reorg.c (stop_search_p): Handle DEBUG_INSN.
3695 (redundant_insn, fill_simple_delay_slots): Likewise.
3696 (fill_slots_from_thread): Likewise.
3697 * resource.c (mark_referenced_resources): Likewise.
3698 (mark_set_resources, find_dead_or_set_registers): Likewise.
3700 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3702 * substring-locations.h (format_warning_va): Formatting fix for
3704 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
3706 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
3707 * substring-locations.c: Include intl.h.
3708 (format_warning_va): Turned into small wrapper around
3709 format_warning_n_va, renamed to ...
3710 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
3711 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
3713 (format_warning_at_substring_n): New function.
3714 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
3715 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
3716 format_warning_at_substring with just a shorter name instead of
3717 const function pointer.
3718 (fmtwarn_n): New function.
3719 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
3720 appropriate, get rid of all the fmtstr temporaries, move conditionals
3721 with G_() wrapped string literals directly into fmtwarn arguments,
3722 cast dir.len to (int), formatting fixes.
3724 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
3726 * doc/invoke.texi: Remove "Cilk Plus" references.
3728 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3729 Richard Biener <rguenther@suse.de>
3732 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
3733 for error or warning attributes if CALL_FROM_THUNK_P is set.
3736 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3739 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
3740 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
3743 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
3744 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
3746 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
3747 instead of -1U in last predictors element's probability member.
3749 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3752 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3753 arguments if they are comparable.
3755 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3757 PR tree-optimization/84634
3758 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
3759 masks and masked_loop_p with a single loop_masks, making sure it's
3760 null for bb vectorization.
3762 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3764 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3765 (vect_analyze_data_ref_access): Use loop->safe_len rather than
3766 loop->force_vectorize to check whether there is no alias.
3768 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3771 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
3773 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
3775 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
3776 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
3777 instead of a loop around prev_real_insn.
3778 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
3782 * config/i386/i386.c (ix86_print_operand): Use conditional
3783 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
3786 2018-03-02 Richard Biener <rguenther@suse.de>
3788 PR tree-optimization/84427
3789 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
3790 (bitmap_set_subtract_values): Rewrite to handle multiple
3793 (prune_clobbered_mems): Likewise.
3794 (phi_translate): Take edge instead of pred/phiblock.
3795 (phi_translate_1): Likewise.
3796 (phi_translate_set): Likewise. Insert all translated
3797 exprs for a value into the set, keeping possibly multiple
3798 expressions per value.
3799 (compute_antic_aux): Adjust for phi_translate changes.
3800 When intersecting union the expressions and prune those
3801 not in the final value set, keeping possibly multiple
3802 expressions per value. Do not use value-insertion
3803 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
3804 all expressions. Add verification that the value-sets
3805 only shrink during iteration.
3806 (compute_partial_antic_aux): Adjust for the phi_translate changes.
3807 (do_pre_regular_insertion): Likewise.
3808 (do_pre_partial_partial_insertion): Likewise.
3810 2018-03-02 Richard Biener <rguenther@suse.de>
3813 * config/darwin.c (saved_debug_info_level): New static global.
3814 (darwin_asm_lto_start): Disable debug info generation for LTO out.
3815 (darwin_asm_lto_end): Restore debug info generation settings.
3817 2018-03-01 Martin Liska <mliska@suse.cz>
3820 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
3823 2018-03-01 Richard Biener <rguenther@suse.de>
3826 * dwarf2out.c (gen_variable_die): Properly handle late VLA
3827 type annotation with LTO when debug was disabled at compile-time.
3829 2018-03-01 Matthew Fortune <mfortune@gmail.com>
3831 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
3833 (mips_final_postscan_insn): Likewise.
3835 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
3837 PR rtl-optimization/84528
3838 * alias.c (init_alias_target): Add commentary.
3839 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
3840 a unique base value if the frame pointer is not eliminated
3841 to the stack pointer.
3843 2018-03-01 Tom de Vries <tom@codesourcery.com>
3845 PR rtl-optimization/83327
3846 * lra-int.h (hard_regs_spilled_into): Declare.
3847 * lra.c (hard_regs_spilled_into): Define.
3848 (init_reg_info): Init hard_regs_spilled_into.
3849 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
3850 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
3851 (process_bb_lives): Handle hard_regs_spilled_into.
3852 (lra_create_live_ranges_1): Before doing liveness propagation, clear
3853 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
3855 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
3857 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
3858 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
3859 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
3860 * config/rs6000/aix72.h: New file.
3862 2018-02-28 Jakub Jelinek <jakub@redhat.com>
3864 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
3865 instead of warning_at with conditional singular and plural messages
3869 * stor-layout.c (update_alignment_for_field): For
3870 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
3871 && !DECL_PACKED (field), do the alignment update, just use
3872 only desired_align instead of MAX (type_align, desired_align)
3874 (place_field): Don't do known_align < desired_align handling
3875 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
3876 is non-NULL, instead do it after rli->prev_field handling and
3877 only if not within a bitfield word. For DECL_PACKED (field)
3878 use type_align of BITS_PER_UNIT.
3880 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
3882 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
3883 superfluous parentheses and trailing spaces.
3885 2018-02-28 Richard Biener <rguenther@suse.de>
3887 PR tree-optimization/84584
3888 * graphite-scop-detection.c (scop_detection::add_scop): Discard
3889 SCoPs with fake exit edge.
3891 2018-02-28 Martin Liska <mliska@suse.cz>
3894 * timevar.c (timer::print): Fix format to properly print 100%
3897 2018-02-28 Richard Biener <rguenther@suse.de>
3900 * genmatch.c (capture_info::walk_match): Do not mark
3901 captured expressions without operands as expr_p given
3902 they act more like predicates and should be subject to
3903 "lost tail" side-effect preserving.
3905 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
3907 PR rtl-optimization/81611
3908 * auto-inc-dec.c (attempt_change): Move dead note from
3909 mem_insn if it's the next use of regno
3910 (find_address): Take address use of reg holding
3911 non-incremented value. Add parm to limit search to the named
3913 (merge_in_block): Attempt to use a mem insn that is the next
3914 use of the original regno.
3916 2018-02-27 Martin Sebor <msebor@redhat.com>
3919 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
3920 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
3922 2018-02-27 Martin Sebor <msebor@redhat.com>
3924 PR translation/84207
3925 * diagnostic-core.h (warning_n, error_n, inform_n): Change
3926 n argument to unsigned HOST_WIDE_INT.
3927 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
3928 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
3929 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
3930 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
3932 2018-02-27 Richard Biener <rguenther@suse.de>
3934 PR tree-optimization/84512
3935 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
3936 Do not use the estimate returned from record_stmt_cost for
3937 the scalar iteration cost but sum properly using add_stmt_cost.
3939 2018-02-27 Richard Biener <rguenther@suse.de>
3941 PR tree-optimization/84466
3942 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3943 Adjust last change to less strictly validate use operands.
3945 2018-02-27 Martin Liska <mliska@suse.cz>
3947 PR gcov-profile/84548
3948 * gcov.c (process_file): Allow partial overlap and consider it
3949 also as group functions.
3950 (output_lines): Properly calculate range of lines for a group.
3952 2018-02-27 Martin Liska <mliska@suse.cz>
3954 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
3955 'ggc' suffixes. Change first column width.
3956 (timer::print): Fix formatting of the column.
3958 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
3960 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
3961 preserve inline entry blocks for the sake of debug inline
3962 entry point markers alone.
3963 (remove_unused_locals): Suggest in comments a better place to
3964 force the preservation of inline entry blocks that are
3965 otherwise unused, but do not preserve them.
3967 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3969 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
3971 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3974 * config/i386/constraints.md (Bs): Replace
3975 ix86_indirect_branch_register with
3976 TARGET_INDIRECT_BRANCH_REGISTER.
3978 * config/i386/i386.md (indirect_jump): Likewise.
3979 (tablejump): Likewise.
3980 (*sibcall_memory): Likewise.
3981 (*sibcall_value_memory): Likewise.
3982 Peepholes of indirect call and jump via memory: Likewise.
3983 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
3984 (*sibcall_value_GOT_32): Likewise.
3985 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3986 (GOT_memory_operand): Likewise.
3987 (call_insn_operand): Likewise.
3988 (sibcall_insn_operand): Likewise.
3989 (GOT32_symbol_operand): Likewise.
3990 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
3992 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3994 PR rtl-optimization/83496
3995 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
3996 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
3997 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
3998 redundant insn, if any.
3999 (relax_delay_slots): Likewise.
4000 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
4002 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
4004 PR tree-optimization/83965
4005 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
4006 that grouped statements are part of a reduction chain. Return
4007 true if the statement is not marked as a reduction itself but
4009 (vect_recog_dot_prod_pattern): Don't check whether the statement
4010 is part of a group here.
4011 (vect_recog_sad_pattern): Likewise.
4012 (vect_recog_widen_sum_pattern): Likewise.
4014 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4017 * final.c (rest_of_clean_state): Also look for calls inside sequences.
4019 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4022 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
4024 (ix86_output_indirect_function_return): New prototype.
4025 (ix86_split_simple_return_pop_internal): Likewise.
4026 * config/i386/i386.c (indirect_return_via_cx): New.
4027 (indirect_return_via_cx_bnd): Likewise.
4028 (indirect_thunk_name): Handle return va CX_REG.
4029 (output_indirect_thunk_function): Create alias for
4030 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
4031 (ix86_output_indirect_jmp): Remove the bool argument.
4032 (ix86_output_indirect_function_return): New function.
4033 (ix86_split_simple_return_pop_internal): Likewise.
4034 * config/i386/i386.md (*indirect_jump): Don't pass false
4035 to ix86_output_indirect_jmp.
4036 (*tablejump_1): Likewise.
4037 (simple_return_pop_internal): Change it to define_insn_and_split.
4038 Call ix86_split_simple_return_pop_internal to split it for
4040 (simple_return_indirect_internal): Call
4041 ix86_output_indirect_function_return instead of
4042 ix86_output_indirect_jmp.
4044 2018-02-26 Jakub Jelinek <jakub@redhat.com>
4047 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
4048 memset and value initialization afterwards.
4050 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
4052 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
4054 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4057 * common/config/aarch64/aarch64-common.c
4058 (aarch_option_optimization_table[]): Switch
4059 off fomit-frame-pointer
4061 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4062 Chung-Ju Wu <jasonwucj@gmail.com>
4064 * config/nds32/nds32-multiple.md (load_multiple): Disallow
4066 (store_multiple): Ditto.
4068 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4070 * config.gcc: Add --with-cpu support for nds32 target.
4071 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
4072 * config/nds32/nds32.opt: Add -mcpu= option.
4074 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
4076 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
4077 isel=yes): Warn for these deprecated options.
4079 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
4081 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
4082 ISA_2_5_MASKS_EMBEDDED.
4084 2018-02-23 Jakub Jelinek <jakub@redhat.com>
4086 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
4087 p->max as pointers rather than using iterative_hash_expr.
4089 2018-02-23 Carl Love <cel@us.ibm.com>
4091 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
4092 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
4094 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
4095 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
4096 P8V_BUILTIN_VEC_VUNSIGNED2.
4098 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
4101 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
4102 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
4104 * lra-constraints.c (curr_insn_transform): Set up
4105 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
4107 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
4108 * lra-eliminations.c (spill_pseudos): Ditto.
4109 (process_insn_for_elimination): Ditto.
4110 * lra-lives.c (reg_early_clobber_p): Use the new macros.
4111 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
4112 LRA_NON_CLOBBERED_ALT.
4114 2018-02-22 Martin Sebor <msebor@redhat.com>
4116 PR tree-optimization/84480
4117 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
4118 to maybe_diag_stxncpy_trunc. Call it.
4119 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
4120 from gimple_fold_builtin_strcpy. Print inlining stack.
4121 (handle_builtin_stxncpy): Print inlining stack.
4122 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
4124 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
4127 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
4128 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
4129 and -fcheck-pointer-bounds are used together.
4130 (indirect_thunk_prefix): New enum.
4131 (indirect_thunk_need_prefix): New function.
4132 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
4133 "_nt" instead of "_bnd" for NOTRACK prefix.
4134 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
4135 (output_indirect_thunk_function): Likewise.
4137 (ix86_code_end): Update output_indirect_thunk_function calls.
4138 (ix86_output_indirect_branch_via_reg): Replace
4139 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
4140 (ix86_output_indirect_branch_via_push): Likewise.
4141 (ix86_output_function_return): Likewise.
4142 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
4143 incompatible with -fcf-protection=branch and
4144 -fcheck-pointer-bounds.
4146 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4149 * config/aarch64/aarch64.c (aarch64_print_address_internal):
4150 Change gcc_assert call to output_operand_lossage.
4152 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4154 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
4156 2018-02-22 DJ Delorie <dj@redhat.com>
4157 Sebastian Perta <sebastian.perta@renesas.com>
4158 Oleg Endo <olegendo@gcc.gnu.org>
4160 * config/rx/rx.c (rx_rtx_costs): New function.
4161 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
4163 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
4165 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
4167 2018-02-22 Martin Liska <mliska@suse.cz>
4170 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
4171 Add "native" as a possible value.
4173 2018-02-22 Martin Liska <mliska@suse.cz>
4176 * config/i386/i386.c (ix86_option_override_internal):
4177 Add "native" as a possible value for -march and -mtune.
4179 2018-02-22 Jakub Jelinek <jakub@redhat.com>
4182 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
4183 to all type variants.
4185 PR tree-optimization/84503
4186 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
4187 width as info->bitpos + info->bitsize - start.
4188 (merged_store_group::merge_overlapping): Simplify width computation.
4189 (check_no_overlap): New function.
4190 (imm_store_chain_info::try_coalesce_bswap): Compute expected
4191 start + width and last_order of the group, fail if check_no_overlap
4193 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
4194 to group if check_no_overlap fails.
4196 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4198 * config/rs6000/altivec.md: Delete contraint arguments to
4199 define_expand, define_split, and define_peephole2, and in
4200 define_insn_and_split if always unused.
4201 * config/rs6000/darwin.md: Ditto.
4202 * config/rs6000/dfp.md: Ditto.
4203 * config/rs6000/rs6000.md: Ditto.
4204 * config/rs6000/sync.md: Ditto.
4205 * config/rs6000/vector.md: Ditto.
4206 * config/rs6000/vsx.md: Ditto.
4208 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4210 * config/rs6000/altivec.md: Write output control strings as braced
4211 blocks instead of double-quoted strings.
4212 * config/rs6000/darwin.md: Ditto.
4213 * config/rs6000/rs6000.md: Ditto.
4214 * config/rs6000/vector.md: Ditto.
4215 * config/rs6000/vsx.md: Ditto.
4217 2018-02-21 Jason Merrill <jason@redhat.com>
4219 PR c++/84314 - ICE with templates and fastcall attribute.
4220 * attribs.c (build_type_attribute_qual_variant): Remove assert.
4222 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4224 * ipa-cp.c (determine_versionability): Fix comment typos.
4226 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4229 * ipa-cp.c (determine_versionability): Do not version functions caling
4232 2018-02-21 Martin Liska <mliska@suse.cz>
4235 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
4236 Add "native" as a possible value.
4237 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
4238 the macro when native cpu detection is available.
4240 2018-02-21 Martin Liska <mliska@suse.cz>
4243 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
4244 Add "native" as a possible value.
4245 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
4246 when native cpu detection is available.
4248 2018-02-21 Jakub Jelinek <jakub@redhat.com>
4249 Martin Sebor <msebor@redhat.com>
4251 PR tree-optimization/84478
4252 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
4254 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
4255 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
4256 support which is conservatively correct, for 2 only stay conservative
4257 for maxlen. Formatting and comment capitalization fixes. Add STRICT
4258 argument to the 2 argument get_range_strlen, adjust 6 arg
4259 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
4261 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
4262 (gimple_fold_builtin_strlen): Pass true as last argument to
4265 2018-02-20 Martin Sebor <msebor@redhat.com>
4268 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
4269 (builtin_memref::set_base_and_offset): Same. Handle inner references.
4270 (builtin_memref::builtin_memref): Factor out parts into
4271 set_base_and_offset and call it.
4273 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4276 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
4277 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
4278 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
4279 search at the associated MODE_INT.
4281 2018-02-20 Jeff Law <law@redhat.com>
4284 PR tree-optimization/81592
4286 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
4287 for range data rather than using global data.
4288 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
4289 range data rather than using global data.
4290 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
4291 pass it to children as needed.
4292 (struct directive::fmtresult): Similarly.
4293 (struct directive::set_width): Similarly.
4294 (struct directive::set_precision): Similarly.
4295 (format_integer, format_directive, parse_directive): Similarly.
4296 (format_none): Accept unnamed vr_values parameter.
4297 (format_percent, format_floating, format_character): Similarly.
4298 (format_string, format_plain): Similarly.
4299 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
4300 the EVRP range analyzer for range data rather than using global data.
4301 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
4302 gimple-ssa-evrp-analyze.h
4303 (class sprintf_dom_walker): Add after_dom_children member function.
4304 Add evrp_range_analyzer member.
4305 (sprintf_dom_walker::before_dom_children): Call into the EVRP
4306 range analyzer as needed.
4307 (sprintf_dom_walker::after_dom_children): New member function.
4308 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
4310 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
4311 (evrp_range_analyzer::pop_to_marker): Likewise.
4313 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4315 PR tree-optimization/84419
4316 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
4317 with the required type if its current type is compatible but
4320 2018-02-20 Jakub Jelinek <jakub@redhat.com>
4323 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
4324 after vectorization.
4326 2018-02-20 Martin Liska <mliska@suse.cz>
4329 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
4330 possible values if we don't have a hint.
4332 2018-02-20 Martin Liska <mliska@suse.cz>
4336 * final.c (shorten_branches): Build align_tab array with one
4338 * opts.c (finish_options): Add alignment option limit check.
4339 (MAX_CODE_ALIGN): Likewise.
4340 (MAX_CODE_ALIGN_VALUE): Likewise.
4341 * doc/invoke.texi: Document maximum allowed option value for
4342 all -falign-* options.
4344 2018-02-19 Jakub Jelinek <jakub@redhat.com>
4347 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
4348 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
4349 * var-tracking.c (emit_note_insn_var_location): Remove all references
4350 to NOTE_INSN_CALL_ARG_LOCATION.
4351 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
4352 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
4353 Use copy_rtx_if_shared.
4354 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
4355 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
4356 (dwarf2out_var_location): Remove handling of
4357 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
4359 * final.c (final_scan_insn): Remove all references to
4360 NOTE_INSN_CALL_ARG_LOCATION.
4361 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
4362 before dumping final insns.
4363 * except.c (emit_note_eh_region_end): Remove all references to
4364 NOTE_INSN_CALL_ARG_LOCATION.
4365 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
4366 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
4367 * config/arc/arc.c (hwloop_optimize): Likewise.
4368 * config/arm/arm.c (create_fix_barrier): Likewise.
4369 * config/s390/s390.c (s390_chunkify_start): Likewise.
4370 * config/sh/sh.c (find_barrier): Likewise.
4371 * config/i386/i386.c (rest_of_insert_endbranch,
4372 ix86_seh_fixup_eh_fallthru): Likewise.
4373 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
4374 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
4375 * config/frv/frv.c (frv_function_prologue): Likewise.
4376 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
4378 (note_outside_basic_block_p): Remove all references to
4379 NOTE_INSN_CALL_ARG_LOCATION.
4380 * gengtype.c (adjust_field_rtx_def): Likewise.
4381 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
4383 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
4384 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
4387 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
4390 PR tree-optimization/84452
4391 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
4392 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
4395 2018-02-19 Martin Liska <mliska@suse.cz>
4398 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
4400 2018-02-19 Martin Liska <mliska@suse.cz>
4401 Richard Sandiford <richard.sandiford@linaro.org>
4403 PR tree-optimization/82491
4404 * gimple-fold.c (get_base_constructor): Make earlier bail out
4407 2018-02-19 Carl Love <cel@us.ibm.com>
4409 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
4410 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
4412 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
4413 P8V_BUILTIN_VEC_NEG.
4415 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
4417 * config/rl78/rl78.md (movdf): New define expand.
4419 2018-02-19 Martin Liska <mliska@suse.cz>
4422 * doc/invoke.texi: Fix typo.
4423 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
4425 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
4427 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
4428 handle rs6000_single_float and rs6000_double_float specially for
4431 2018-02-16 Jeff Law <law@redhat.com>
4433 * config/rx/rx.c (add_pop_cfi_notes): New function.;
4436 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4439 * ipa-inline.c (inline_small_functions): Fix a typo.
4441 2018-02-16 Nathan Sidwell <nathan@acm.org>
4443 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
4445 2018-02-16 Carl Love <cel@us.ibm.com>
4447 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
4448 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
4449 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
4450 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
4451 expansion to P8V_BUILTIN_VEC_FLOAT2.
4453 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
4455 PR rtl-optimization/70023
4456 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
4457 src_regno into account.
4459 2018-02-16 Carl Love <cel@us.ibm.com>
4461 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
4462 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
4463 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
4464 * config/rs6000/rs6000.c: Remove case statements for
4465 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
4466 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
4467 and P9V_BUILTIN_VEC_VINSERT4B.
4468 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
4469 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
4470 * config/rs6000/vsx.md:
4471 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
4474 2018-02-16 Carl Love <cel@us.ibm.com>
4476 * config/rs6000/altivec.h: Add builtin names vec_extract4b
4478 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
4480 * config/rs6000/rs6000-c.c: Add the definitions for
4481 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
4482 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
4483 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
4484 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
4485 definition for insert4b and define insn *insert3b_internal.
4486 * doc/extend.texi: Add documentation for vec_extract4b.
4488 2018-02-16 Nathan Sidwell <nathan@acm.org>
4490 * doc/extend.texi (Backwards Compatibility): Mention friend
4491 injection. Note for-scope is deprecated.
4492 * doc/invoke.texi (-ffriend-injection): Deprecate.
4494 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
4496 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
4497 that moved to I2, also allow destinations that are a paradoxical
4498 subreg (instead of a normal reg).
4500 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
4503 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
4506 2018-02-16 Richard Biener <rguenther@suse.de>
4508 PR tree-optimization/84037
4509 PR tree-optimization/84016
4511 * config/i386/i386.c (ix86_builtin_vectorization_cost):
4512 Adjust vec_construct for the fact we need additional higher latency
4513 128bit inserts for AVX256 and AVX512 vector builds.
4514 (ix86_add_stmt_cost): Scale vector construction cost for
4517 2018-02-16 Richard Biener <rguenther@suse.de>
4519 PR tree-optimization/84417
4520 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
4521 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
4522 (non_rewritable_lvalue_p): Likewise, use poly-ints.
4524 2018-02-16 Martin Liska <mliska@suse.cz>
4527 * internal-fn.def (ASAN_CHECK): Set proper flags.
4528 (ASAN_MARK): Likewise.
4530 2018-02-16 Julia Koval <julia.koval@intel.com>
4532 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
4533 from PTA_CANNONLAKE.
4535 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4538 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
4539 Use ++iter rather than iter++ for std::list iterators.
4540 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
4541 defer deleting them until all nodes in the forest are processed. Do
4542 free even leaf nodes. Change to_process into auto_vec.
4545 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
4546 * vec.h (vec_default_construct): Use memset instead of placement new
4547 if BROKEN_VALUE_INITIALIZATION is defined.
4548 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
4549 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
4552 PR rtl-optimization/83723
4553 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
4554 * lra.c (lra_substitute_pseudo): Likewise. If true, use
4555 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
4557 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
4559 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
4561 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
4563 PR rtl-optimization/81443
4564 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
4565 from inner REGs to paradoxical SUBREGs.
4567 2018-02-16 Richard Biener <rguenther@suse.de>
4569 PR tree-optimization/84399
4570 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4571 For operands we can analyze at their definition make sure we can
4572 analyze them at each use as well.
4574 2018-02-16 Richard Biener <rguenther@suse.de>
4576 PR tree-optimization/84190
4577 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
4578 volatile accesses if the decl isn't volatile.
4580 2018-02-15 Jason Merrill <jason@redhat.com>
4582 PR c++/84314 - ICE with templates and fastcall attribute.
4583 * attribs.c (build_type_attribute_qual_variant): Don't clobber
4584 TYPE_CANONICAL on an existing type.
4586 2018-02-15 Jakub Jelinek <jakub@redhat.com>
4588 PR tree-optimization/84383
4589 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
4590 dstoff nor call operand_equal_p if dstbase is NULL.
4592 PR tree-optimization/84334
4593 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
4594 also a CONSTANT_CLASS_P, punt.
4596 2018-02-14 Jim Wilson <jimw@sifive.com>
4598 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
4599 first SMALL_OPERAND check. New local min_second_step. Move assert
4600 to where locals are set. Add TARGET_RVC support.
4601 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
4603 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
4605 * doc/invoke.texi: Correct -Wformat-overflow code sample.
4607 2018-02-14 Martin Sebor <msebor@redhat.com>
4609 PR tree-optimization/83698
4610 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
4611 arrays constrain the offset range to their bounds.
4612 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
4613 (builtin_access::overlap): Avoid setting the size of overlap if it's
4615 (maybe_diag_overlap): Also consider arrays when deciding what values
4616 of offsets to include in diagnostics.
4618 2018-02-14 Martin Sebor <msebor@redhat.com>
4621 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
4622 that correspond to the kind of a declaration.
4624 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
4627 * config/pa/pa.md: Load address of PIC label using the linkage table
4628 if the label is nonlocal.
4630 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
4632 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
4633 warning message if user requests -maltivec=be.
4634 * doc/invoke.texi: Document deprecation of -maltivec=be.
4636 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
4639 * config/rs6000/rs6000-c.c: Update definitions for
4640 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
4641 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
4643 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4646 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
4647 add _get_ssp intrinsics. Remove argument from
4648 __builtin_ia32_rdssp[d|q].
4649 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
4650 * config/i386/i386-builtin.def: Remove argument from
4651 __builtin_ia32_rdssp[d|q].
4652 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
4653 ix86_expand_special_args_builtin for _rdssp[d|q].
4654 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
4655 Clear register before usage.
4656 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
4657 Add documentation for new _get_ssp and _inc_ssp intrinsics.
4659 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
4661 PR tree-optimization/84357
4662 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
4663 operand 1 of an ARRAY_REF too.
4665 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
4668 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
4669 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4671 (set_of_reg): New struct.
4672 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
4673 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
4674 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4676 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
4677 Split into bitclr, bitset, bitinvert patterns if appropriate.
4678 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
4679 use rx_fuse_in_memory_bitop.
4680 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
4681 to named insn, correct maximum insn length.
4683 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
4686 * machmode.def: Define a complex mode for PARTIAL_INT.
4687 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
4689 * doc/rtl.texi: Document CSPImode.
4690 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
4692 (msp430_hard_regno_nregs_with_padding): Likewise.
4694 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
4697 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
4699 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
4701 PR rtl-optimization/84169
4702 * combine.c (try_combine): New variable split_i2i3. Set it to true if
4703 we generated a parallel as new i3 and we split that to new i2 and i3
4704 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
4705 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
4706 those to i2, not i1. Partially rewrite this scan code.
4708 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4711 * stor-layout.c (place_field): For variable length fields, adjust
4712 offset_align afterwards not just based on the field's alignment,
4713 but also on the size.
4716 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
4717 of exps and logs in the use_exp2 case.
4719 2018-02-13 Jeff Law <law@redhat.com>
4721 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
4724 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
4727 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
4731 * common.opt (gas-loc-support, gas-locview-support): New.
4732 (ginline-points, ginternal-reset-location-views): New.
4733 * doc/invoke.texi: Document them. Use @itemx where intended.
4734 (gvariable-location-views): Adjust.
4735 * target.def (reset_location_view): New.
4736 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
4737 (TARGET_RESET_LOCATION_VIEW): New.
4738 * doc/tm.texi: Rebuilt.
4739 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
4740 (dwarf2out_default_as_locview_support): New.
4741 (output_asm_line_debug_info): Use option variables.
4742 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
4743 (output_loc_list): Likewise.
4744 (add_high_low_attributes): Check option variables.
4745 Don't output entry view attribute in strict mode.
4746 (gen_inlined_subroutine_die): Check option variables.
4747 (dwarf2out_inline_entry): Likewise.
4748 (init_sections_and_labels): Likewise.
4749 (dwarf2out_early_finish): Likewise.
4750 (maybe_reset_location_view): New, from...
4751 (dwarf2out_var_location): ... here. Call it.
4752 * debug.h (dwarf2out_default_as_loc_support): Declare.
4753 (dwarf2out_default_as_locview_support): Declare.
4754 * hooks.c (hook_int_rtx_insn_0): New.
4755 * hooks.h (hook_int_rtx_insn_0): Declare.
4756 * toplev.c (process_options): Take -gas-loc-support and
4757 -gas-locview-support from dwarf2out. Enable
4758 -gvariable-location-views by default only with locview
4759 assembler support. Enable -ginternal-reset-location-views by
4760 default only if the target defines the corresponding hook.
4761 Enable -ginline-points by default if location views are
4762 enabled; force it disabled if statement frontiers are
4764 * tree-inline.c (expand_call_inline): Check option variables.
4765 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
4767 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
4769 PR tree-optimization/84321
4770 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
4771 handling. Also check whether the anti-range contains any values
4772 that satisfy the mask; switch to a VR_RANGE if not.
4774 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
4777 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
4779 2018-02-13 Martin Jambor <mjambor@suse.cz>
4782 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
4783 of call statements, also set location of a load to a temporary.
4785 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
4787 * config/rl78/rl78.c (add_vector_labels): New function.
4788 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
4789 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
4790 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
4791 which checks that no arguments are passed.
4792 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
4793 * doc/extend.texi: Documentation for the new attribute.
4795 2018-02-13 Andreas Schwab <schwab@suse.de>
4797 * config/riscv/linux.h (CPP_SPEC): Define.
4799 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4802 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
4803 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
4804 OPTION_MASK_ISA_AES as first argument to def_builtin_const
4805 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
4806 instead of OPTION_MASK_ISA_PCLMUL as first argument to
4807 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
4808 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
4809 temporarily for AES and PCLMUL builtins.
4811 PR tree-optimization/84339
4812 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
4813 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
4817 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
4818 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
4819 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
4821 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
4823 * omp-simd-clone.h: New file.
4824 * omp-simd-clone.c: Include omp-simd-clone.h.
4825 (expand_simd_clones): No longer static.
4826 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
4827 cgraph.h and omp-simd-clone.h.
4828 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
4829 (vect_recog_widen_shift_pattern): Formatting fix.
4830 (vect_pattern_recog_1): Don't check optab for calls.
4833 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
4834 operands[2] into a REG before using gen_lowpart on it.
4836 2018-02-12 Jeff Law <law@redhat.com>
4839 * config/sh/sh.c (find_barrier): Consider a sibling call
4842 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
4843 successfully back substituting a reg.
4845 2018-02-12 Richard Biener <rguenther@suse.de>
4847 PR tree-optimization/84037
4848 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
4849 parameter, move visited init to caller.
4850 (vect_slp_analyze_operations): Separate cost from validity
4851 check, initialize visited once for all instances.
4852 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
4854 * tree-vect-stmts.c (vect_model_simple_cost): Make early
4856 (vect_model_promotion_demotion_cost): Likewise.
4857 (vectorizable_bswap): Guard cost modeling with !slp_node
4858 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
4860 (vectorizable_call): Likewise.
4861 (vectorizable_conversion): Likewise.
4862 (vectorizable_assignment): Likewise.
4863 (vectorizable_shift): Likewise.
4864 (vectorizable_operation): Likewise.
4865 (vectorizable_store): Likewise.
4866 (vectorizable_load): Likewise.
4867 (vectorizable_condition): Likewise.
4868 (vectorizable_comparison): Likewise.
4870 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
4873 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
4874 (ASAN_MARK): Fix fnspec to account for return value, change pointer
4875 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
4877 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4880 * params.def (inline-min-speedup): Increase from 8 to 15.
4881 (max-inline-insns-auto): Decrease from 40 to 30.
4882 * ipa-split.c (consider_split): Add some buffer for function to
4883 be considered inlining candidate.
4884 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
4887 2018-02-12 Richard Biener <rguenther@suse.de>
4889 PR tree-optimization/84037
4890 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
4891 matched stmts if we cannot swap the non-matched ones.
4893 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4895 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
4896 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
4897 _mm_maskz_scalef_round_ss): New intrinsics.
4898 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
4899 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
4900 __builtin_ia32_scalefss_round): Remove.
4901 (__builtin_ia32_scalefsd_mask_round,
4902 __builtin_ia32_scalefss_mask_round): New intrinsics.
4903 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
4904 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
4905 ((match_operand:VF_128 2 "<round_nimm_predicate>"
4906 "<round_constraint>")): Changed to ...
4907 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
4908 "<round_scalar_constraint>")): ... this.
4909 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
4910 %0, %1, %2<round_op3>}"): Changed to ...
4911 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
4912 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
4913 %2<round_scalar_mask_op3>}"): ... this.
4914 * config/i386/subst.md (round_scalar_nimm_predicate): New.
4916 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4918 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
4919 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
4920 (_mm_maskz_sqrt_round_ss): New intrinsics.
4921 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
4922 (__builtin_ia32_sqrtsd_mask_round)
4923 (__builtin_ia32_sqrtss_mask_round): New builtins.
4924 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
4925 (__builtin_ia32_sqrtss_round): Remove.
4926 (__builtin_ia32_sqrtsd_mask_round)
4927 (__builtin_ia32_sqrtss_mask_round): New builtins.
4928 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
4929 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
4930 ((match_operand:VF_128 1 "vector_operand"
4931 "xBm,<round_constraint>")): Changed to ...
4932 ((match_operand:VF_128 1 "vector_operand"
4933 "xBm,<round_scalar_constraint>")): ... this.
4934 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
4935 %0, %2, %<iptr>1<round_op3>}): Changed to ...
4936 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
4937 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
4938 %<iptr>1<round_scalar_mask_op3>}): ... this.
4939 ((set_attr "prefix" "<round_prefix>")): Changed to ...
4940 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
4942 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
4945 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
4946 Cast vec_cmpeq result to correct type.
4947 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
4948 Cast vec_cmpgt result to correct type.
4950 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
4952 * final.c (final_scan_insn_1): Renamed from...
4953 (final_scan_insn): ... this. New wrapper, to recover
4954 seen from the outermost call in recursive ones.
4955 * config/sparc/sparc.c (output_return): Drop seen from call.
4956 (output_sibcall): Likewise.
4957 * config/visium/visium.c (output_branch): Likewise.
4959 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
4961 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
4964 2018-02-10 Alan Modra <amodra@gmail.com>
4967 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
4968 Specify LR as an input.
4970 2018-02-10 Jakub Jelinek <jakub@redhat.com>
4973 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
4974 remove_member_access_dummy_vars): New functions.
4975 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
4976 lower_omp_1, execute_lower_omp): Use them.
4978 PR rtl-optimization/84308
4979 * shrink-wrap.c (spread_components): Release todo vector.
4981 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
4983 PR rtl-optimization/57193
4984 * ira-color.c (struct allocno_color_data): Add member
4985 conflict_allocno_hard_prefs.
4986 (update_conflict_allocno_hard_prefs): New.
4987 (bucket_allocno_compare_func): Add a preference based on
4988 conflict_allocno_hard_prefs.
4989 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
4990 (color_allocnos): Remove a dead code. Initiate
4991 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
4993 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4996 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
4997 constraint from =wa to wa. Avoid a subreg on the output operand,
4998 instead use a pseudo and subreg it in a move.
4999 (p9_xxbrd_<mode>): Changed to ...
5000 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
5001 (p9_xxbrd_v2df): New expander.
5002 (p9_xxbrw_<mode>): Changed to ...
5003 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
5004 (p9_xxbrw_v4sf): New expander.
5006 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5008 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
5010 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
5013 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
5014 multiply in 32-bit mode.
5015 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
5016 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
5019 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5021 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
5022 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
5023 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
5024 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
5026 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
5029 * dwarf2out.c (is_trivial_indirect_ref): New function.
5030 (dwarf2out_late_global_decl): Do not generate a location
5031 attribute for variables that have a non-trivial DECL_VALUE_EXPR
5032 and that are not defined in the current unit.
5034 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5036 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
5037 instead of a libcall for UNORDERED.
5039 2018-02-09 Tamar Christina <tamar.christina@arm.com>
5042 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
5043 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
5045 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5048 * config/s390/s390.c (s390_set_current_function): Invoke
5049 s390_indirect_branch_settings also if fndecl didn't change.
5051 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5053 * config/rs6000/rs6000.md (blockage): Set length to zero.
5055 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5057 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
5059 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5062 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
5063 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
5067 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
5068 PARALLEL incoming that failed vt_get_decl_and_offset check.
5071 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
5072 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
5074 (get_variable_section): For decls in named .bss* sections pass true as
5075 second argument to bss_initializer_p.
5077 2018-02-09 Marek Polacek <polacek@redhat.com>
5078 Jakub Jelinek <jakub@redhat.com>
5081 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
5082 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
5083 Sync some changes from cxx_fold_indirect_ref.
5085 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5087 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
5089 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
5090 (BLOCK_INLINE_ENTRY_LABEL): New.
5091 (dwarf2out_var_location): Disregard inline entry markers.
5092 (inline_entry_data): New struct.
5093 (inline_entry_data_hasher): New hashtable type.
5094 (inline_entry_data_hasher::hash): New.
5095 (inline_entry_data_hasher::equal): New.
5096 (inline_entry_data_table): New variable.
5097 (add_high_low_attributes): Add DW_AT_entry_pc and
5098 DW_AT_GNU_entry_view attributes if a pending entry is found
5099 in inline_entry_data_table. Add old entry_pc attribute only
5100 if debug nonbinding markers are disabled.
5101 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
5102 markers are enabled.
5103 (block_within_block_p, dwarf2out_inline_entry): New.
5104 (dwarf2out_finish): Check that no entries remained in
5105 inline_entry_data_table.
5106 * final.c (reemit_insn_block_notes): Handle inline entry notes.
5107 (final_scan_insn, notice_source_line): Likewise.
5108 (rest_of_clean_state): Skip inline entry markers.
5109 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
5111 * gimple.c (gimple_build_debug_inline_entry): New.
5112 * gimple.h (enum gimple_debug_subcode): Add
5113 GIMPLE_DEBUG_INLINE_ENTRY.
5114 (gimple_build_debug_inline_entry): Declare.
5115 (gimple_debug_inline_entry_p): New.
5116 (gimple_debug_nonbind_marker_p): Adjust.
5117 * insn-notes.def (INLINE_ENTRY): New.
5118 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
5119 inline entry marker notes.
5120 (print_insn): Likewise.
5121 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
5122 (INSN_DEBUG_MARKER_KIND): Likewise.
5123 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
5124 * tree-inline.c (expand_call_inline): Build and insert
5125 debug_inline_entry stmt.
5126 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
5127 inline entry blocks early, if nonbind markers are enabled.
5128 (dump_scope_block): Dump fragment info.
5129 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
5130 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
5131 (gimple_build_debug_inline_entry): New.
5132 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
5133 Enable/disable inline entry points too.
5134 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
5135 (DEBUG_INSN): Describe inline entry markers.
5137 * common.opt (gvariable-location-views): New.
5138 (gvariable-location-views=incompat5): New.
5139 * config.in: Rebuilt.
5140 * configure: Rebuilt.
5141 * configure.ac: Test assembler for view support.
5142 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
5143 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
5144 * dwarf2out.c (var_loc_view): New typedef.
5145 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
5146 (dwarf2out_locviews_in_attribute): New.
5147 (dwarf2out_locviews_in_loclist): New.
5148 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
5149 (enum dw_line_info_opcode): Add LI_adv_address.
5150 (struct dw_line_info_table): Add view.
5151 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
5152 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
5153 (zero_view_p): New variable.
5154 (ZERO_VIEW_P): New macro.
5155 (output_asm_line_debug_info): New.
5156 (struct var_loc_node): Add view.
5157 (add_AT_view_list, AT_loc_list): New.
5158 (add_var_loc_to_decl): Add view param. Test it against last.
5159 (new_loc_list): Add view params. Record them.
5160 (AT_loc_list_ptr): Handle loc and view lists.
5161 (view_list_to_loc_list_val_node): New.
5162 (print_dw_val): Handle dw_val_class_view_list.
5163 (size_of_die): Likewise.
5164 (value_format): Likewise.
5165 (loc_list_has_views): New.
5166 (gen_llsym): Set vl_symbol too.
5167 (maybe_gen_llsym, skip_loc_list_entry): New.
5168 (dwarf2out_maybe_output_loclist_view_pair): New.
5169 (output_loc_list): Output view list or entries too.
5170 (output_view_list_offset): New.
5171 (output_die): Handle dw_val_class_view_list.
5172 (output_dwarf_version): New.
5173 (output_compilation_unit_header): Use it.
5174 (output_skeleton_debug_sections): Likewise.
5175 (output_rnglists, output_line_info): Likewise.
5176 (output_pubnames, output_aranges): Update version comments.
5177 (output_one_line_info_table): Output view numbers in asm comments.
5178 (dw_loc_list): Determine current endview, pass it to new_loc_list.
5179 Call maybe_gen_llsym.
5180 (loc_list_from_tree_1): Adjust.
5181 (add_AT_location_description): Create view list attribute if
5182 needed, check it's absent otherwise.
5183 (convert_cfa_to_fb_loc_list): Adjust.
5184 (maybe_emit_file): Call output_asm_line_debug_info for test.
5185 (dwarf2out_var_location): Reset views as needed. Precompute
5186 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
5187 attribute. Set view.
5188 (new_line_info_table): Reset next view.
5189 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
5190 (dwarf2out_source_line): Likewise. Output view resets and labels to
5191 the assembler, or select appropriate line info opcodes.
5192 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
5193 (optimize_string_length): Catch it. Adjust.
5194 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
5195 dw_val_class_view_list, and remove it if no longer needed.
5196 (hash_loc_list): Hash view numbers.
5197 (loc_list_hasher::equal): Compare them.
5198 (optimize_location_lists): Check whether a view list symbol is
5199 needed, and whether the locview attribute is present, and
5200 whether they match. Remove the locview attribute if no longer
5202 (index_location_lists): Call skip_loc_list_entry for test.
5203 (dwarf2out_finish): Call output_asm_line_debug_info for test.
5204 Use output_dwarf_version.
5205 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
5206 (struct dw_val_node): Add val_view_list.
5207 * final.c (SEEN_NEXT_VIEW): New.
5208 (set_next_view_needed): New.
5209 (clear_next_view_needed): New.
5210 (maybe_output_next_view): New.
5211 (final_start_function): Rename to...
5212 (final_start_function_1): ... this. Take pointer to FIRST,
5213 add SEEN parameter. Emit param bindings in the initial view.
5214 (final_start_function): Reintroduce SEEN-less interface.
5215 (final): Rename to...
5216 (final_1): ... this. Take SEEN parameter. Output final pending
5217 next view at the end.
5218 (final): Reintroduce seen-less interface.
5219 (final_scan_insn): Output pending next view before switching
5220 sections or ending a block. Mark the next view as needed when
5221 outputting variable locations. Notify debug backend of section
5222 changes, and of location view changes.
5223 (rest_of_handle_final): Adjust.
5224 * toplev.c (process_options): Autodetect value for debug variable
5225 location views option. Warn on incompat5 without -gdwarf-5.
5226 * doc/invoke.texi (gvariable-location-views): New.
5227 (gvariable-location-views=incompat5): New.
5228 (gno-variable-location-views): New.
5230 2018-02-08 David Malcolm <dmalcolm@redhat.com>
5232 PR tree-optimization/84136
5233 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
5234 that the result of find_edge is non-NULL.
5236 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5239 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
5240 storing integer register in SImode. Fix cost of 256 and 512
5241 byte aligned SSE register store.
5243 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5245 * config/i386/i386.c (ix86_multiplication_cost): Fix
5246 multiplication cost for TARGET_AVX512DQ.
5248 2018-02-08 Marek Polacek <polacek@redhat.com>
5250 PR tree-optimization/84238
5251 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
5254 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5256 PR tree-optimization/84265
5257 * tree-vect-stmts.c (vectorizable_store): Don't treat
5258 VMAT_CONTIGUOUS accesses as grouped.
5259 (vectorizable_load): Likewise.
5261 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5263 PR tree-optimization/81635
5264 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
5265 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
5266 (test_round_for_mask): New functions.
5267 (wide_int_cc_tests): Call test_round_for_mask.
5268 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
5269 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
5270 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
5271 range returned by get_range_info.
5273 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5276 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
5277 * symtab.c: Include builtins.h
5278 (symtab_node::output_to_lto_symbol_table_p): Move here
5279 from lto-streamer-out.c:output_symbol_p.
5280 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
5281 (output_symbol_p): Move all logic to symtab.c
5282 (produce_symtab): Update.
5284 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5286 * config/s390/s390-opts.h (enum indirect_branch): Define.
5287 * config/s390/s390-protos.h (s390_return_addr_from_memory)
5288 (s390_indirect_branch_via_thunk)
5289 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
5290 (enum s390_indirect_branch_type): Define.
5291 * config/s390/s390.c (struct s390_frame_layout, struct
5292 machine_function): Remove.
5293 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
5294 (indirect_branch_table_label_no, indirect_branch_table_name):
5296 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
5297 (enum s390_indirect_branch_option): Define.
5298 (s390_return_addr_from_memory): New function.
5299 (s390_handle_string_attribute): New function.
5300 (s390_attribute_table): Add new attribute handler.
5301 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
5302 (s390_indirect_branch_via_thunk): New function.
5303 (s390_indirect_branch_via_inline_thunk): New function.
5304 (s390_function_ok_for_sibcall): When jumping via thunk disallow
5305 sibling call optimization for non z10 compiles.
5306 (s390_emit_call): Force indirect branch target to be a single
5307 register. Add r1 clobber for non-z10 compiles.
5308 (s390_emit_epilogue): Emit return jump via return_use expander.
5309 (s390_reorg): Handle JUMP_INSNs as execute targets.
5310 (s390_option_override_internal): Perform validity checks for the
5311 new command line options.
5312 (s390_indirect_branch_attrvalue): New function.
5313 (s390_indirect_branch_settings): New function.
5314 (s390_set_current_function): Invoke s390_indirect_branch_settings.
5315 (s390_output_indirect_thunk_function): New function.
5316 (s390_code_end): Implement target hook.
5317 (s390_case_values_threshold): Implement target hook.
5318 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
5320 * config/s390/s390.h (struct s390_frame_layout)
5321 (struct machine_function): Move here from s390.c.
5322 (TARGET_INDIRECT_BRANCH_NOBP_RET)
5323 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
5324 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
5325 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
5326 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
5327 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
5328 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
5329 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
5330 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
5331 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
5332 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
5333 (mnemonic attribute): Add values which aren't recognized
5335 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
5336 pattern for branch conversion. Fix mnemonic attribute.
5337 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
5338 indirect branch via thunk if requested.
5339 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
5340 ("*indirect_jump"): Disable for branch conversion using out of
5342 ("indirect_jump_via_thunk<mode>_z10")
5343 ("indirect_jump_via_thunk<mode>")
5344 ("indirect_jump_via_inlinethunk<mode>_z10")
5345 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
5346 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
5347 ("casesi_jump_via_inlinethunk<mode>_z10")
5348 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
5349 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
5350 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
5351 ("*indirect2_jump"): Disable for branch conversion.
5352 ("casesi_jump"): Turn into expander and expand patterns for branch
5354 ("return_use"): New expander.
5355 ("*return"): Emit return via thunk and rename it to ...
5356 ("*return<mode>"): ... this one.
5357 * config/s390/s390.opt: Add new options and and enum for the
5360 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5362 * lra-constraints.c (match_reload): Unconditionally use
5363 gen_lowpart_SUBREG, rather than selecting between that
5364 and equivalent gen_rtx_SUBREG code.
5366 2018-02-08 Richard Biener <rguenther@suse.de>
5368 PR tree-optimization/84233
5369 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
5370 changed flag instead of boguously re-using phi_inserted.
5372 2018-02-08 Martin Jambor <mjambor@suse.cz>
5374 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
5375 static local variables.
5377 2018-02-08 Richard Biener <rguenther@suse.de>
5379 PR tree-optimization/84278
5380 * tree-vect-stmts.c (vectorizable_store): When looking for
5381 smaller vector types to perform grouped strided loads/stores
5382 make sure the mode is supported by the target.
5383 (vectorizable_load): Likewise.
5385 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5387 * config/aarch64/aarch64.c (aarch64_components_for_bb):
5388 Increase LDP/STP opportunities by adding adjacent callee-saves.
5390 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5392 PR rtl-optimization/84068
5393 PR rtl-optimization/83459
5394 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
5396 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
5398 PR tree-optimization/84224
5399 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
5400 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
5403 2018-02-07 Iain Sandoe <iain@codesourcery.com>
5406 * config/rs6000/altivec.md (*restore_world): Remove LR use.
5407 * config/rs6000/predicates.md (restore_world_operation): Adjust op
5408 count, remove one USE.
5410 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
5412 * doc/install.texi (Configuration): Document the
5413 --with-long-double-format={ibm,ieee} PowerPC configuration
5417 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
5418 Convert from define_expand to be define_insn_and_split. Rework
5419 float/double/_Float128 conversions to QI/HI/SImode to work with
5420 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
5421 conversions to QI/HImode types did a store and then a load to
5422 truncate the value. For conversions to VSX registers, don't split
5423 the insn, instead emit the code directly. Use the code iterator
5424 any_fix to combine signed and unsigned conversions.
5425 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
5426 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
5427 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5428 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
5429 (fix_<mode>di2_hw): Likewise.
5430 (fixuns_<mode>di2_hw): Likewise.
5431 (fix_<mode>si2_hw): Likewise.
5432 (fixuns_<mode>si2_hw): Likewise.
5433 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
5434 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5435 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
5436 fix<uns>_trunc<SFDF:mode>si2_p8.
5437 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
5439 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
5440 (fix<uns>_<mode>_mem): Likewise.
5441 (fctiw<u>z_<mode>_mem): Likewise.
5442 (fix<uns>_<mode>_mem): Likewise.
5443 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
5444 the register allocator from doing a direct move to the GPRs to do
5445 a store, and instead use the ISA 3.0 store byte/half-word from
5446 vector register instruction. For IEEE 128-bit floating point,
5447 also optimize stores of 32-bit ints.
5448 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
5450 2018-02-07 Alan Hayward <alan.hayward@arm.com>
5452 * genextract.c (push_pathstr_operand): New function to support
5454 (walk_rtx): Call push_pathstr_operand.
5455 (print_path): Support [a-zA-Z].
5457 2018-02-07 Richard Biener <rguenther@suse.de>
5459 PR tree-optimization/84037
5460 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
5461 (cse_and_gimplify_to_preheader): Declare.
5462 (vect_get_place_in_interleaving_chain): Likewise.
5463 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5465 (_loop_vec_info::~_loop_vec_info): Delete it.
5466 (cse_and_gimplify_to_preheader): New function.
5467 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
5468 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
5469 (vectorizable_load): Likewise. For grouped stores always base
5470 the IV on the first element.
5471 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
5472 condition before gimplifying.
5474 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5476 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
5477 *DIV_EXPR and *MOD_EXPR.
5479 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
5482 * config/i386/i386.c (ix86_option_override_internal): Mask out
5483 the CF_SET bit when checking -fcf-protection.
5485 2018-02-07 Tom de Vries <tom@codesourcery.com>
5488 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
5491 2018-02-07 Richard Biener <rguenther@suse.de>
5493 PR tree-optimization/84204
5494 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
5497 PR tree-optimization/84205
5498 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
5499 special-case isl_ast_op_zdiv_r.
5501 PR tree-optimization/84223
5502 * graphite-scop-detection.c (gather_bbs::before_dom_children):
5503 Only add conditions from within the region.
5504 (gather_bbs::after_dom_children): Adjust.
5506 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
5509 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
5510 * config/avr/avr.md: Only post-reload split REG-REG moves if
5511 either register is GENERAL_REG_P.
5513 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5515 PR tree-optimization/84235
5516 * tree-ssa-scopedtables.c
5517 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
5518 if the subtraction is performed in floating point type where NaNs are
5519 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
5520 build 1. Formatting fix.
5522 2018-02-06 Jakub Jelinek <jakub@redhat.com>
5525 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
5526 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
5527 and skip it regardless of bb boundaries. Use CALL_P macro,
5528 don't test INSN_P (insn) together with CALL_P or JUMP_P check
5529 unnecessarily, formatting fix.
5531 2018-02-06 Michael Collison <michael.collison@arm.com>
5533 * config/arm/thumb2.md:
5534 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
5535 (*thumb_mov_notscc): Ditto.
5537 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
5540 * config/rs6000/rs6000.md (su code attribute): Use "u" for
5541 unsigned_fix, not "s".
5543 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5545 * configure.ac (gcc_fn_eh_frame_ro): New function.
5546 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
5547 correct .eh_frame permissions.
5548 * configure: Regenerate.
5550 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
5552 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
5555 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5557 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5558 Display warning message for -mno-speculate-indirect-jumps.
5560 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
5562 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
5564 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
5566 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
5568 PR tree-optimization/84225
5569 * tree-eh.c (find_trapping_overflow): Only call
5570 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
5572 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5575 * config/i386/i386.c: Reimplement the check of possible options
5576 -mibt/-mshstk conbination. Change error messages.
5577 * doc/invoke.texi: Fix a typo: remove extra '='.
5579 2018-02-06 Marek Polacek <polacek@redhat.com>
5581 PR tree-optimization/84228
5582 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
5584 2018-02-06 Tamar Christina <tamar.christina@arm.com>
5587 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
5588 emitted arch directives.
5589 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
5590 __ARM_FEATURE_COPROC before changing architectures.
5592 2018-02-06 Richard Biener <rguenther@suse.de>
5594 * config/i386/i386.c (print_reg): Fix typo.
5595 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
5597 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
5599 * configure: Regenerate.
5601 2018-02-05 Martin Sebor <msebor@redhat.com>
5603 PR tree-optimization/83369
5604 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
5607 2018-02-05 Martin Liska <mliska@suse.cz>
5609 * doc/invoke.texi: Cherry-pick upstream r323995.
5611 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
5613 * ira.c (ira_init_register_move_cost): Adjust comment.
5615 2018-02-05 Martin Liska <mliska@suse.cz>
5617 PR gcov-profile/84137
5618 * doc/gcov.texi: Fix typo in documentation.
5620 2018-02-05 Martin Liska <mliska@suse.cz>
5622 PR gcov-profile/83879
5623 * doc/gcov.texi: Document necessity of --dynamic-list-data when
5624 using dlopen functionality.
5626 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
5628 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
5629 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
5630 _mm_maskz_range_ss, _mm_mask_range_round_ss,
5631 _mm_maskz_range_round_ss): New intrinsics.
5632 (__builtin_ia32_rangesd128_round)
5633 (__builtin_ia32_rangess128_round): Remove.
5634 (__builtin_ia32_rangesd128_mask_round,
5635 __builtin_ia32_rangess128_mask_round): New builtins.
5636 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
5637 __builtin_ia32_rangess128_round): Remove.
5638 (__builtin_ia32_rangesd128_mask_round,
5639 __builtin_ia32_rangess128_mask_round): New builtins.
5640 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
5641 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
5642 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
5643 "<round_saeonly_constraint>")): Changed to ...
5644 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
5645 "<round_saeonly_scalar_constraint>")): ... this.
5646 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
5647 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
5648 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
5649 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
5650 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
5652 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
5654 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
5656 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
5657 Remove all values except native, 8540 and 8548.
5659 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
5661 * config/i386/i386.c (ix86_output_function_return): Pass
5662 INVALID_REGNUM, instead of -1, as invalid register number to
5663 indirect_thunk_name and output_indirect_thunk.
5665 2018-02-02 Julia Koval <julia.koval@intel.com>
5667 * config.gcc: Add -march=icelake.
5668 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
5669 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
5670 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
5671 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
5672 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
5673 (processor_target_table): Add icelake.
5674 (ix86_option_override_internal): Handle new PTAs.
5675 (get_builtin_code_for_version): Handle icelake.
5676 (M_INTEL_COREI7_ICELAKE): New.
5677 (fold_builtin_cpu): Handle icelake.
5678 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
5679 * doc/invoke.texi: Add -march=icelake.
5681 2018-02-02 Julia Koval <julia.koval@intel.com>
5683 * config/i386/i386.c (ix86_option_override_internal): Change flags type
5684 to wide_int_bitmask.
5685 * wide-int-bitmask.h: New.
5687 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5690 * config/i386/i386.md: Replace Pmode with word_mode in
5691 builtin_setjmp_setup and builtin_longjmp to support x32.
5693 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
5697 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
5699 (rs6000_supported_cpu_names): New static variable.
5700 (linux_cpu_translation_table): Likewise.
5701 (elf_platform) <cpu>: Define new static variable and use it.
5702 Translate kernel AT_PLATFORM name to canonical name if needed.
5703 Error if platform name is unknown.
5705 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
5708 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
5710 2018-02-01 Jeff Law <law@redhat.com>
5713 * config/i386/i386.c (release_scratch_register_on_entry): Add new
5714 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
5715 the scratch if RELEASE_VIA_POP is false.
5716 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
5717 If we have to save a temporary register, decrement SIZE appropriately.
5718 Pass new arguments to release_scratch_register_on_entry.
5719 (ix86_adjust_stack_and_probe): Likewise.
5720 (ix86_emit_probe_stack_range): Pass new arguments to
5721 release_scratch_register_on_entry.
5723 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
5725 PR rtl-optimization/84157
5726 * combine.c (change_zero_ext): Use REG_P predicate in
5727 front of HARD_REGISTER_P predicate.
5729 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
5731 * config/avr/avr.c (avr_option_override): Move disabling of
5732 -fdelete-null-pointer-checks to...
5733 * common/config/avr/avr-common.c (avr_option_optimization_table):
5736 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5738 PR tree-optimization/81635
5739 * tree-data-ref.c (split_constant_offset_1): For types that
5740 wrap on overflow, try to use range info to prove that wrapping
5743 2018-02-01 Renlin Li <renlin.li@arm.com>
5746 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
5748 (aarch64_register_move_cost): Likewise.
5749 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
5751 (REG_CLASS_NAMES): Likewise.
5752 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
5753 TAILCALL_ADDR_REGS. Remove IP registers.
5754 * config/aarch64/aarch64.md (Ucs): Update register constraint.
5756 2018-02-01 Richard Biener <rguenther@suse.de>
5758 * domwalk.h (dom_walker::dom_walker): Add additional constructor
5759 for specifying RPO order and allow NULL for that.
5760 * domwalk.c (dom_walker::dom_walker): Likewise.
5761 (dom_walker::walk): Handle NULL RPO order.
5762 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
5764 (rewrite_update_dom_walker): Likewise.
5765 (mark_def_dom_walker): Likewise.
5767 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5769 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
5770 (aarch64_maybe_expand_sve_subreg_move): Declare.
5771 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
5772 * config/aarch64/predicates.md (aarch64_any_register_operand): New
5774 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
5775 that are semantically a reverse operation.
5776 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
5777 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
5778 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
5780 (aarch64_can_change_mode_class): For big-endian, forbid changes
5781 between two SVE modes if they have different element sizes.
5783 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5785 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
5786 the TImode handling for big-endian targets.
5788 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5790 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
5791 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
5793 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
5794 Remove BSWAP handing for big-endian targets and use the form of
5795 LD1RQ appropariate for the mode.
5797 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5799 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
5800 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
5803 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5806 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
5807 check for operands that need to go through aarch64_sve_reload_be.
5809 2018-02-01 Jakub Jelinek <jakub@redhat.com>
5811 PR tree-optimization/81661
5812 PR tree-optimization/84117
5813 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
5814 * tree-eh.c: Include gimplify.h.
5815 (find_trapping_overflow, replace_trapping_overflow,
5816 rewrite_to_non_trapping_overflow): New functions.
5817 * tree-vect-loop.c: Include tree-eh.h.
5818 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
5819 * tree-data-ref.c: Include tree-eh.h.
5820 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
5822 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
5824 PR rtl-optimization/84123
5825 * combine.c (change_zero_ext): Check if hard register satisfies
5826 can_change_dest_mode before calling gen_lowpart_SUBREG.
5828 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
5831 * ira.c (ira_init_register_move_cost): Remove assert.
5833 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5835 PR rtl-optimization/84071
5836 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
5837 * doc/tm.texi: Regenerate.
5839 2018-01-31 Richard Biener <rguenther@suse.de>
5841 PR tree-optimization/84132
5842 * tree-data-ref.c (analyze_miv_subscript): Properly
5843 check whether evolution_function_is_affine_multivariate_p
5844 before calling gcd_of_steps_may_divide_p.
5846 2018-01-31 Julia Koval <julia.koval@intel.com>
5849 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
5850 * config/i386/i386.md (rdpid_rex64) New.
5851 (rdpid): Make 32bit only.
5853 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
5856 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
5857 an IDENTIFIER_NODE for FUNCTION_TYPE's.
5859 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5862 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5864 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5866 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5868 PR rtl-optimization/84071
5869 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
5870 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
5872 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5874 * config/arc/arc.c (arc_handle_aux_attribute): New function.
5875 (arc_attribute_table): Add 'aux' attribute.
5876 (arc_in_small_data_p): Consider aux like variables.
5877 (arc_is_aux_reg_p): New function.
5878 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
5879 (arc_get_aux_arg): New function.
5880 (prepare_move_operands): Handle aux-register access.
5881 (arc_handle_aux_attribute): New function.
5882 * doc/extend.texi (ARC Variable attributes): Add subsection.
5884 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5886 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
5887 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
5888 (arc_attribute_table): Add 'uncached' attribute.
5889 (arc_print_operand): Print '.di' flag for uncached memory
5891 (arc_in_small_data_p): Do not consider for small data the uncached
5893 (arc_is_uncached_mem_p): New function.
5894 * config/arc/predicates.md (compact_store_memory_operand): Check
5895 for uncached memory accesses.
5896 (nonvol_nonimm_operand): Likewise.
5897 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
5899 2018-01-31 Jakub Jelinek <jakub@redhat.com>
5902 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
5903 falign-loops=): Add Optimization flag.
5905 2018-01-30 Jeff Law <law@redhat.com>
5908 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
5909 INT_REGISTERS_SAVED. Check it prior to calling
5910 get_scratch_register_on_entry.
5911 (ix86_adjust_stack_and_probe): Similarly.
5912 (ix86_emit_probe_stack_range): Similarly.
5913 (ix86_expand_prologue): Corresponding changes.
5915 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5918 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
5919 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
5921 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
5924 * lra-constraints.c (curr_insn_transform): Process AND in the
5927 2018-01-30 Jakub Jelinek <jakub@redhat.com>
5929 PR rtl-optimization/83986
5930 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
5931 dependence against last_pending_memory_flush in addition to
5934 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
5936 PR tree-optimization/81611
5937 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
5940 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5943 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
5946 2018-01-30 Richard Biener <rguenther@suse.de>
5947 Jakub Jelinek <jakub@redhat.com>
5949 PR tree-optimization/84111
5950 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
5951 inner loops added during recursion, as they don't have up-to-date
5954 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5957 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
5958 (can_inline_edge_by_limits_p): ... here.
5959 (can_early_inline_edge_p, check_callers,
5960 update_caller_keys, update_callee_keys, recursive_inlining,
5961 add_new_edges_to_heap, speculation_useful_p,
5962 inline_small_functions,
5963 inline_small_functions, flatten_function,
5964 inline_to_all_callers_1): Update.
5966 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5968 * profile-count.c (profile_count::combine_with_ipa_count): Handle
5971 2018-01-30 Richard Biener <rguenther@suse.de>
5973 PR tree-optimization/83008
5974 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
5975 invariant and constant vector uses in stmts when they need
5978 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5981 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
5982 * configure: Regenerate.
5984 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5986 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
5988 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
5989 Use gen_rtx_REG rather than gen_lowpart.
5991 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5993 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
5994 rather than 0 when creating partial subregs.
5996 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5998 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
6001 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
6004 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
6005 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
6006 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
6007 flags. This restores the settings used before the 2017-07-24.
6008 Turning off pre increment/decrement/modify allows IVOPTS to
6009 optimize DF/SF loops where the index is an int.
6011 2018-01-29 Richard Biener <rguenther@suse.de>
6012 Kelvin Nilsen <kelvin@gcc.gnu.org>
6015 * tree-vect-stmts.c (vectorizable_call): Don't call
6016 targetm.vectorize_builtin_md_vectorized_function if callee is
6019 2018-01-22 Carl Love <cel@us.ibm.com>
6021 * doc/extend.tex: Fix typo in second arg in
6022 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
6024 2018-01-29 Richard Biener <rguenther@suse.de>
6026 PR tree-optimization/84086
6027 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
6028 (flush_ssaname_freelist): When SSA names were released reset
6029 the SCEV hash table.
6031 2018-01-29 Richard Biener <rguenther@suse.de>
6033 PR tree-optimization/84057
6034 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
6035 removed paths when removing edges.
6037 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
6039 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
6040 -mfunction-return=@var{choice}.
6042 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
6045 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
6047 (layout::print_source_line): Likewise.
6048 (test_get_line_width_without_trailing_whitespace): Add test cases.
6050 2018-01-27 Jakub Jelinek <jakub@redhat.com>
6053 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
6056 2018-01-26 Jim Wilson <jimw@sifive.com>
6058 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
6060 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
6063 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6065 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
6066 and CMP + SUB-immediate -> SUBS.
6068 2018-01-26 Martin Sebor <msebor@redhat.com>
6070 PR tree-optimization/83896
6071 * tree-ssa-strlen.c (get_string_len): Rename...
6072 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
6073 Avoid assuming length is constant.
6074 (handle_char_store): Use HOST_WIDE_INT for string length.
6076 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
6079 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
6080 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
6082 2018-01-26 Richard Biener <rguenther@suse.de>
6084 PR rtl-optimization/84003
6085 * dse.c (record_store): Only record redundant stores when
6086 the earlier store aliases at least all accesses the later one does.
6088 2018-01-26 Jakub Jelinek <jakub@redhat.com>
6090 PR rtl-optimization/83985
6091 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
6092 REG_CFA_RESTORE insns.
6093 (delete_unmarked_insns): Don't ignore separate shrink wrapping
6094 REG_CFA_RESTORE insns here.
6097 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
6098 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
6100 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6102 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
6103 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
6104 (arc_init): Likewise.
6105 (arc_override_options): Likewise.
6106 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
6108 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
6110 * config/arc/arc.h (TARGET_DBNZ): Define.
6111 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
6112 properly set the tune attribute.
6113 (dbnz): Use TARGET_DBNZ guard.
6114 * config/arc/arc.opt (mtune): Add core3 option.
6116 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6118 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
6119 recognize new pic like addresses.
6120 (arc_delegitimize_address): Clean up.
6122 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6124 * config/arc/arc-arches.def: Option mrf16 valid for all
6126 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
6127 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
6128 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
6129 * config/arc/arc-tables.opt: Regenerate.
6130 * config/arc/arc.c (arc_conditional_register_usage): Handle
6131 reduced register file case.
6132 (arc_file_start): Set must have build attributes.
6133 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
6135 * config/arc/arc.opt (mrf16): Add new option.
6136 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
6137 * config/arc/genmultilib.awk: Handle new mrf16 option.
6138 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
6139 * config/arc/t-multilib: Regenerate.
6140 * doc/invoke.texi (ARC Options): Document mrf16 option.
6142 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6144 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
6145 * config/arc/arc.c (arc_handle_secure_attribute): New function.
6146 (arc_attribute_table): Add 'secure_call' attribute.
6147 (arc_print_operand): Print secure call operand.
6148 (arc_function_ok_for_sibcall): Don't optimize tail calls when
6150 (arc_is_secure_call_p): New function. * config/arc/arc.md
6151 (call_i): Add support for sjli instruction.
6152 (call_value_i): Likewise.
6153 * config/arc/constraints.md (Csc): New constraint.
6155 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6156 John Eric Martin <John.Martin@emmicro-us.com>
6158 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
6159 * config/arc/arc.c (_arc_jli_section): New struct.
6160 (arc_jli_section): New type.
6161 (rc_jli_sections): New static variable.
6162 (arc_handle_jli_attribute): New function.
6163 (arc_attribute_table): Add jli_always and jli_fixed attribute.
6164 (arc_file_end): New function.
6165 (TARGET_ASM_FILE_END): Define.
6166 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
6167 (arc_add_jli_section): New function.
6168 (jli_call_scan): Likewise.
6169 (arc_reorg): Call jli_call_scan.
6170 (arc_output_addsi): Remove 'S' from printing asm operand.
6171 (arc_is_jli_call_p): New function.
6172 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
6174 (movhi_insn): Likewise.
6175 (movsi_insn): Likewise.
6176 (movsi_set_cc_insn): Likewise.
6177 (loadqi_update): Likewise.
6178 (load_zeroextendqisi_update): Likewise.
6179 (load_signextendqisi_update): Likewise.
6180 (loadhi_update): Likewise.
6181 (load_zeroextendhisi_update): Likewise.
6182 (load_signextendhisi_update): Likewise.
6183 (loadsi_update): Likewise.
6184 (loadsf_update): Likewise.
6185 (movsicc_insn): Likewise.
6186 (bset_insn): Likewise.
6187 (bxor_insn): Likewise.
6188 (bclr_insn): Likewise.
6189 (bmsk_insn): Likewise.
6190 (bicsi3_insn): Likewise.
6191 (cmpsi_cc_c_insn): Likewise.
6192 (movsi_ne): Likewise.
6193 (movsi_cond_exec): Likewise.
6194 (clrsbsi2): Likewise.
6205 (call_i): Remove 'S' asm letter, add jli instruction.
6206 (call_value_i): Likewise.
6207 * config/arc/arc.op (mjli-always): New option.
6208 * config/arc/constraints.md (Cji): New constraint.
6209 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
6211 (subsf3_fpx): Likewise.
6212 (mulsf3_fpx): Likewise.
6213 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
6215 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
6217 * doc/invoke.texi (ARC): Document mjli-always option.
6219 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
6221 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
6222 avoid addition with 0 and use incw and decw where possible.
6224 2018-01-26 Richard Biener <rguenther@suse.de>
6226 PR tree-optimization/81082
6227 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
6228 association if it requires casting to unsigned.
6229 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
6230 from fold_plusminus_mult_expr to catch important cases late when
6231 range info is available.
6233 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6235 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
6236 * configure.ac (hidden_linkonce): New test.
6237 * configure: Regenerate.
6238 * config.in: Regenerate.
6240 2018-01-26 Julia Koval <julia.koval@intel.com>
6242 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
6243 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
6244 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
6245 _mm_mask_bitshuffle_epi64_mask): Fix type.
6246 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
6247 USI_FTYPE_V4DI_V4DI_USI): Remove.
6248 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
6249 __builtin_ia32_vpshufbitqmb256_mask,
6250 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
6251 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
6252 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
6254 2018-01-26 Alan Modra <amodra@gmail.com>
6257 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
6258 UNSPEC_VBPERMQ. Sort other unspecs.
6260 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
6262 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
6264 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
6267 * predict.c (drop_profile): Do not push/pop cfun; update also
6269 (handle_missing_profiles): Fix logic looking for zero profiles.
6271 2018-01-25 Jakub Jelinek <jakub@redhat.com>
6274 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
6275 on functions with #pragma omp declare simd or functions with simd
6277 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
6278 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
6279 Remove trailing \n from warning_at calls.
6281 2018-01-25 Tom de Vries <tom@codesourcery.com>
6284 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6285 for neutered workers.
6287 2018-01-24 Joseph Myers <joseph@codesourcery.com>
6290 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
6291 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
6293 2018-01-24 Jeff Law <law@redhat.com>
6296 * i386.c (get_probe_interval): Move to earlier point.
6297 (ix86_compute_frame_layout): If -fstack-clash-protection and
6298 the frame is larger than the probe interval, then use pushes
6299 to save registers rather than reg->mem moves.
6300 (ix86_expand_prologue): Remove conditional for int_registers_saved
6303 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
6306 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
6307 min/max for never referenced object.
6309 2018-01-24 Jakub Jelinek <jakub@redhat.com>
6312 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
6314 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
6315 attributes from DECL_ATTRIBUTES (decl) without affecting
6316 DECL_ATTRIBUTES (current_function_decl).
6317 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
6318 functions with non-NULL DECL_ABSTRACT_ORIGIN.
6320 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
6322 PR tree-optimization/83979
6323 * fold-const.c (fold_comparison): Use constant_boolean_node
6324 instead of boolean_{true,false}_node.
6326 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
6328 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
6331 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6333 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6334 Simplify the clause that sets the length attribute.
6335 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6336 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
6337 clause that sets the length attribute.
6338 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6340 2018-01-24 Tom de Vries <tom@codesourcery.com>
6343 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
6344 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
6345 Add strict parameter.
6346 (prevent_branch_around_nothing): Insert dummy insn between branch to
6347 label and label with no ptx insn inbetween.
6348 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
6350 2018-01-24 Tom de Vries <tom@codesourcery.com>
6353 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6354 for neutered threads in warp.
6355 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
6357 2018-01-24 Richard Biener <rguenther@suse.de>
6359 PR tree-optimization/83176
6360 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
6363 2018-01-24 Richard Biener <rguenther@suse.de>
6365 PR tree-optimization/82819
6366 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
6367 code generating pluses that are no-ops in the target precision.
6369 2018-01-24 Richard Biener <rguenther@suse.de>
6372 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
6374 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6376 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
6377 to merge probabilities.
6378 * predict.c (probably_never_executed): Also mark as cold functions
6379 with global 0 profile and guessed local profile.
6380 * profile-count.c (profile_probability::combine_with_count): New
6382 * profile-count.h (profile_probability::operator*,
6383 profile_probability::operator*=, profile_probability::operator/,
6384 profile_probability::operator/=): Reduce precision to adjusted
6385 and set value to guessed on contradictory divisions.
6386 (profile_probability::combine_with_freq): Remove.
6387 (profile_probability::combine_wiht_count): Declare.
6388 (profile_count::force_nonzero):: Set to adjusted.
6389 (profile_count::probability_in):: Set quality to adjusted.
6390 * tree-ssa-tail-merge.c (replace_block_by): Use
6393 2018-01-23 Andrew Waterman <andrew@sifive.com>
6394 Jim Wilson <jimw@sifive.com>
6396 * config/riscv/riscv.c (riscv_stack_boundary): New.
6397 (riscv_option_override): Set riscv_stack_boundary. Handle
6398 riscv_preferred_stack_boundary_arg.
6399 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
6400 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
6401 (STACK_BOUNDARY): Set to riscv_stack_boundary.
6402 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
6403 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
6404 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
6406 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
6409 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
6410 of struct ix86_frame.
6411 (ix86_expand_epilogue): Likewise. Add a local variable for
6412 the reg_save_offset field in struct ix86_frame.
6414 2018-01-23 Bin Cheng <bin.cheng@arm.com>
6416 PR tree-optimization/82604
6417 * tree-loop-distribution.c (enum partition_kind): New enum item
6418 PKIND_PARTIAL_MEMSET.
6419 (partition_builtin_p): Support above new enum item.
6420 (generate_code_for_partition): Ditto.
6421 (compute_access_range): Differentiate cases that equality can be
6422 proven at all loops, the innermost loops or no loops.
6423 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
6424 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
6425 (finalize_partitions, distribute_loop): Don't fuse partition of
6426 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
6427 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
6430 2018-01-23 Martin Liska <mliska@suse.cz>
6432 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
6433 order to ignore the predictor.
6434 (PRED_POLYMORPHIC_CALL): Likewise.
6435 (PRED_RECURSIVE_CALL): Likewise.
6437 2018-01-23 Martin Liska <mliska@suse.cz>
6439 * tree-profile.c (tree_profiling): Print function header to
6440 aware reader which function we are working on.
6441 * value-prof.c (gimple_find_values_to_profile): Do not print
6442 not interesting value histograms.
6444 2018-01-23 Martin Liska <mliska@suse.cz>
6446 * profile-count.h (enum profile_quality): Add
6447 profile_uninitialized as the first value. Do not number values
6448 as they are zero based.
6449 (profile_count::verify): Update sanity check.
6450 (profile_probability::verify): Likewise.
6452 2018-01-23 Nathan Sidwell <nathan@acm.org>
6454 * doc/invoke.texi (ffor-scope): Deprecate.
6456 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6458 PR tree-optimization/83510
6459 * domwalk.c (set_all_edges_as_executable): New function.
6460 (dom_walker::dom_walker): Convert bool param
6461 "skip_unreachable_blocks" to enum reachability. Move setup of
6462 edge flags to set_all_edges_as_executable and only do it when
6463 reachability is REACHABLE_BLOCKS.
6464 * domwalk.h (enum dom_walker::reachability): New enum.
6465 (dom_walker::dom_walker): Convert bool param
6466 "skip_unreachable_blocks" to enum reachability.
6467 (set_all_edges_as_executable): New decl.
6468 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
6469 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
6471 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
6472 but converting true to REACHABLE_BLOCKS.
6473 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
6475 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
6476 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
6477 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
6479 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
6480 if check_all_array_refs will be called.
6482 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6484 * tree.c (selftest::test_location_wrappers): Add more test
6487 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6489 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
6490 (selftest::test_bit_in_range): Likewise.
6492 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6495 * doc/sourcebuild.texi (vect_float): Say that the selector
6496 only describes the situation when -funsafe-math-optimizations is on.
6497 (vect_float_strict): Document.
6499 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6501 PR tree-optimization/83965
6502 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
6503 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
6504 instead of checking only for a reduction.
6505 (vect_recog_widen_sum_pattern): Likewise.
6507 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6509 * predict.c (probably_never_executed): Only use precise profile info.
6510 (compute_function_frequency): Skip after inlining hack since we now
6511 have quality checking.
6513 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6515 * profile-count.h (profile_probability::very_unlikely,
6516 profile_probability::unlikely, profile_probability::even): Set
6517 precision to guessed.
6519 2018-01-23 Richard Biener <rguenther@suse.de>
6521 PR tree-optimization/83963
6522 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
6523 Properly terminate dominator walk when crossing the exit edge not
6524 when visiting its source block.
6526 2018-01-23 Jakub Jelinek <jakub@redhat.com>
6529 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
6530 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
6532 2018-01-22 Jakub Jelinek <jakub@redhat.com>
6534 PR tree-optimization/83957
6535 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
6536 semicolon after for body surrounded by braces.
6538 PR tree-optimization/83081
6539 * profile-count.h (profile_probability::split): New method.
6540 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
6541 Use profile_probability::split.
6542 (do_compare_rtx_and_jump): Fix adjustment of probabilities
6543 when splitting a single conditional jump into 2.
6545 2018-01-22 David Malcolm <dmalcolm@redhat.com>
6547 PR tree-optimization/69452
6548 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
6551 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6553 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
6554 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
6555 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
6557 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6559 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
6560 * config/rl78/rl78.md (movdi): New define_expand.
6561 * config/rl78/rl78.c (rl78_split_movdi): New function.
6563 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
6566 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
6568 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
6569 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
6570 128-bit to produce an UNSPEC move to get the double word with the
6571 signbit and then a shift directly to do signbit.
6572 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
6573 implementation with a new version that just does either a direct
6574 move or a regular move. Move memory interface to separate insns.
6575 Move insns so they are next to the expander.
6576 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
6577 with signbit move. Split big and little endian case.
6578 (signbit<mode>2_dm_mem_le): Likewise.
6579 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
6580 (signbit<mode>2_dm2): Likewise.
6582 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6584 * config/rl78/rl78.md (anddi3): New define_expand.
6586 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6588 * config/rl78/rl78.md (umindi3): New define_expand.
6590 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6592 * config/rl78/rl78.md (smindi3): New define_expand.
6594 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6596 * config/rl78/rl78.md (smaxdi3): New define_expand.
6598 2018-01-22 Carl Love <cel@us.ibm.com>
6600 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
6601 LVX_V1TI): Add macro expansion.
6602 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
6603 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
6604 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
6605 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
6606 Change check to determine if the instruction is a byte reversing
6607 entry. Fix typo in comment.
6608 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
6609 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
6610 Add def_builtin calls for new builtins.
6611 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
6612 Add define_insn expansion.
6614 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6616 * config/rl78/rl78.md (umaxdi3): New define_expand.
6618 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6620 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
6621 for non-QImode registers.
6623 2018-01-22 Richard Biener <rguenther@suse.de>
6625 PR tree-optimization/83963
6626 * graphite-scop-detection.c (scop_detection::get_sese): Delay
6627 including the loop exit block.
6628 (scop_detection::merge_sese): Likewise.
6629 (scop_detection::add_scop): Do it here instead.
6631 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6633 * doc/sourcebuild.texi (arm_softfloat): Document.
6635 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
6638 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
6639 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
6640 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
6642 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6643 David Edelsohn <dje.gcc@gmail.com>
6646 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6647 Change "crset eq" to "crset 2".
6648 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6649 (*call_indirect_aix<mode>_nospec): Likewise.
6650 (*call_value_indirect_aix<mode>_nospec): Likewise.
6651 (*call_indirect_elfv2<mode>_nospec): Likewise.
6652 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
6653 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
6654 change assembly output from . to $.
6655 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6656 (indirect_jump<mode>_nospec): Change assembly output from . to $.
6657 (*tablejump<mode>_internal1_nospec): Likewise.
6659 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
6662 * config/sh/sh_optimize_sett_clrt.cc:
6663 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
6665 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
6667 PR tree-optimization/83940
6668 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
6669 offset_dt to vect_constant_def rather than vect_unknown_def_type.
6670 (vect_check_load_store_mask): Add a mask_dt_out parameter and
6671 use it to pass back the definition type.
6672 (vect_check_store_rhs): Likewise rhs_dt_out.
6673 (vect_build_gather_load_calls): Add a mask_dt argument and use
6674 it instead of a call to vect_is_simple_use.
6675 (vectorizable_store): Update calls to vect_check_load_store_mask
6676 and vect_check_store_rhs. Use the dt returned by the latter instead
6677 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
6678 instead of calls to vect_is_simple_use. Pass the scalar rather
6679 than the vector operand to vect_is_simple_use when handling
6680 second and subsequent copies of an rhs value.
6681 (vectorizable_load): Update calls to vect_check_load_store_mask
6682 and vect_build_gather_load_calls. Use the cached mask_dt and
6683 gs_info.offset_dt instead of calls to vect_is_simple_use.
6685 2018-01-20 Jakub Jelinek <jakub@redhat.com>
6688 * tree-emutls.c: Include gimplify.h.
6689 (lower_emutls_2): New function.
6690 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
6691 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
6692 it before further processing.
6695 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
6696 UINTVAL (trueop1) instead of INTVAL (op1).
6698 2018-01-19 Jakub Jelinek <jakub@redhat.com>
6702 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
6703 INCOMING_FRAME_SP_OFFSET if not defined.
6704 (scan_trace): Add ENTRY argument. If true and
6705 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
6706 emit a note to adjust the CFA offset.
6707 (create_cfi_notes): Adjust scan_trace callers.
6708 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
6709 INCOMING_FRAME_SP_OFFSET in the CIE.
6710 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
6711 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
6713 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
6714 * doc/tm.texi: Regenerated.
6716 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6718 PR rtl-optimization/83147
6719 * lra-constraints.c (remove_inheritance_pseudos): Use
6720 lra_substitute_pseudo_within_insn.
6722 2018-01-19 Tom de Vries <tom@codesourcery.com>
6723 Cesar Philippidis <cesar@codesourcery.com>
6726 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
6728 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
6731 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
6732 spaces for function labels.
6734 2018-01-19 Martin Liska <mliska@suse.cz>
6736 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
6737 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
6738 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
6739 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
6740 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
6741 (PRED_CONST_RETURN): Change from 69 to 65.
6742 (PRED_NULL_RETURN): Change from 91 to 71.
6743 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
6744 (PRED_LOOP_GUARD): Change from 66 to 73.
6746 2018-01-19 Martin Liska <mliska@suse.cz>
6748 * predict.c (predict_insn_def): Add new assert.
6749 (struct branch_predictor): Change type to signed integer.
6750 (test_prediction_value_range): Amend test to cover
6752 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
6753 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
6754 (PRED_LOOP_ITERATIONS_MAX): Likewise.
6755 (PRED_LOOP_IV_COMPARE): Likewise.
6756 * predict.h (PROB_UNINITIALIZED): Define new constant.
6758 2018-01-19 Martin Liska <mliska@suse.cz>
6760 * predict.c (dump_prediction): Add new format for
6761 analyze_brprob.py script which is enabled with -details
6763 * profile-count.h (precise_p): New function.
6765 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6767 PR tree-optimization/83922
6768 * tree-vect-loop.c (vect_verify_full_masking): Return false if
6769 there are no statements that need masking.
6770 (vect_active_double_reduction_p): New function.
6771 (vect_analyze_loop_operations): Use it when handling phis that
6772 are not in the loop header.
6774 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6776 PR tree-optimization/83914
6777 * tree-vect-loop.c (vectorizable_induction): Don't convert
6778 init_expr or apply the peeling adjustment for inductions
6779 that are nested within the vectorized loop.
6781 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6783 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
6786 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6790 * function.h (gimplify_parameters): Add gimple_seq * argument.
6791 * function.c: Include gimple.h and options.h.
6792 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
6793 for the added local temporaries if needed.
6794 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
6795 if there are any parameter cleanups, wrap whole body into a
6796 try/finally with the cleanups.
6798 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
6801 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
6802 Use GET_MODE_CLASS for scalar floating point.
6804 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6808 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
6809 Fix call of call_cgraph_insertion_hooks.
6811 2018-01-18 Martin Sebor <msebor@redhat.com>
6813 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
6815 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6818 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
6821 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
6824 * common.opt: (-ffile-prefix-map): New option.
6825 * opts.c (common_handle_option): Defer it.
6826 * opts-global.c (handle_common_deferred_options): Handle it.
6827 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
6828 * file-prefix-map.h: New file.
6829 (remap_debug_filename, add_debug_prefix_map): ...here.
6830 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
6831 * final.c (debug_prefix_map, add_debug_prefix_map
6832 remap_debug_filename): Move to...
6833 * file-prefix-map.c: New file.
6834 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
6835 generalize, get rid of alloca(), use strrchr() instead of strchr().
6836 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
6837 Implement in terms of add_prefix_map().
6838 (remap_macro_filename, remap_debug_filename): Implement in term of
6840 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
6841 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
6842 * dbxout.c: Include file-prefix-map.h.
6843 * varasm.c: Likewise.
6844 * vmsdbgout.c: Likewise.
6845 * xcoffout.c: Likewise.
6846 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
6847 * doc/cppopts.texi (-fmacro-prefix-map): Document.
6848 * doc/invoke.texi (-ffile-prefix-map): Document.
6849 (-fdebug-prefix-map): Update description.
6851 2018-01-18 Martin Liska <mliska@suse.cz>
6853 * config/i386/i386.c (indirect_thunk_name): Document that also
6855 (output_indirect_thunk): Document why both instructions
6856 (pause and lfence) are generated.
6858 2018-01-18 Richard Biener <rguenther@suse.de>
6860 PR tree-optimization/83887
6861 * graphite-scop-detection.c
6862 (scop_detection::get_nearest_dom_with_single_entry): Remove.
6863 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
6864 (scop_detection::merge_sese): Re-implement with a flood-fill
6865 algorithm that properly finds a SESE region if it exists.
6867 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6870 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
6871 pointer_diff optimizations use view_convert instead of convert.
6873 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6875 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6876 Generate different code for -mno-speculate-indirect-jumps.
6877 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6878 (*call_indirect_aix<mode>): Disable for
6879 -mno-speculate-indirect-jumps.
6880 (*call_indirect_aix<mode>_nospec): New define_insn.
6881 (*call_value_indirect_aix<mode>): Disable for
6882 -mno-speculate-indirect-jumps.
6883 (*call_value_indirect_aix<mode>_nospec): New define_insn.
6884 (*sibcall_nonlocal_sysv<mode>): Generate different code for
6885 -mno-speculate-indirect-jumps.
6886 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6888 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6890 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
6891 long double type, set the flags for noting the default long double
6892 type, even if we don't pass or return a long double type.
6894 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
6897 * ipa-inline.c (flatten_function): Do not overwrite final inlining
6900 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
6902 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
6903 support for merge[hl].
6904 (fold_mergehl_helper): New helper function.
6905 (tree-vector-builder.h): New #include for tree_vector_builder usage.
6906 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
6907 (altivec_vmrglw_direct): Add xxmrglw insn.
6909 2018-01-17 Andrew Waterman <andrew@sifive.com>
6911 * config/riscv/riscv.c (riscv_conditional_register_usage): If
6912 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
6914 2018-01-17 David Malcolm <dmalcolm@redhat.com>
6917 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
6918 call the lto_location_cache before reading the
6919 DECL_SOURCE_LOCATION of the types.
6921 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
6922 Richard Sandiford <richard.sandiford@linaro.org>
6924 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
6925 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
6926 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
6927 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
6928 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
6930 * config/aarch64/constraints.md (aarch64_movti_operand):
6932 * config/aarch64/predicates.md (Uti): Add new constraint.
6934 2018-01-17 Carl Love <cel@us.ibm.com>
6936 * config/rs6000/vsx.md (define_expand xl_len_r,
6937 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
6938 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
6940 (define_expand, define_insn): Move the shift left from the
6941 define_insn to the define_expand for lxvl and stxvl instructions.
6942 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
6943 and XL_LEN_R definitions to PURE.
6945 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
6947 * config/i386/i386.c (indirect_thunk_name): Declare regno
6948 as unsigned int. Compare regno with INVALID_REGNUM.
6949 (output_indirect_thunk): Ditto.
6950 (output_indirect_thunk_function): Ditto.
6951 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
6952 in the call to output_indirect_thunk_function.
6954 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
6957 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
6958 rather than the size of inner_type to determine the stack slot size
6959 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
6961 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
6964 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
6967 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
6969 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
6970 endian Linux systems to optionally enable multilibs for selecting
6971 the long double type if the user configured an explicit type.
6972 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
6973 have no long double multilibs if not defined.
6974 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
6975 warn if the user used -mabi={ieee,ibm}longdouble and we built
6976 multilibs for long double.
6977 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
6978 appropriate multilib option.
6979 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
6981 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
6982 for building long double multilibs.
6983 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
6985 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
6987 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
6990 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
6992 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
6995 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
6998 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
7001 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7003 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
7004 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
7006 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
7008 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
7009 different rtl trees depending on TARGET_64BIT.
7010 (rs6000_gen_lvx): Likewise.
7012 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7014 * config/visium/visium.md (nop): Tweak comment.
7015 (hazard_nop): Likewise.
7017 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7019 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
7020 -mspeculate-indirect-jumps.
7021 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
7022 for -mno-speculate-indirect-jumps.
7023 (*call_indirect_elfv2<mode>_nospec): New define_insn.
7024 (*call_value_indirect_elfv2<mode>): Disable for
7025 -mno-speculate-indirect-jumps.
7026 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
7027 (indirect_jump): Emit different RTL for
7028 -mno-speculate-indirect-jumps.
7029 (*indirect_jump<mode>): Disable for
7030 -mno-speculate-indirect-jumps.
7031 (*indirect_jump<mode>_nospec): New define_insn.
7032 (tablejump): Emit different RTL for
7033 -mno-speculate-indirect-jumps.
7034 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
7035 (tablejumpsi_nospec): New define_expand.
7036 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
7037 (tablejumpdi_nospec): New define_expand.
7038 (*tablejump<mode>_internal1): Disable for
7039 -mno-speculate-indirect-jumps.
7040 (*tablejump<mode>_internal1_nospec): New define_insn.
7041 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
7044 2018-01-16 Artyom Skrobov tyomitch@gmail.com
7046 * caller-save.c (insert_save): Drop unnecessary parameter. All
7049 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7050 Richard Biener <rguenth@suse.de>
7053 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
7054 return early, inline manually is_gimple_sizepos. Make sure if we
7055 call gimplify_expr we don't end up with a gimple constant.
7056 * tree.c (variably_modified_type_p): Don't return true for
7057 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
7058 * gimplify.h (is_gimple_sizepos): Remove.
7060 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7062 PR tree-optimization/83857
7063 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
7064 vectorizable_live_operation for pure SLP statements.
7065 (vectorizable_live_operation): Handle PHIs.
7067 2018-01-16 Richard Biener <rguenther@suse.de>
7069 PR tree-optimization/83867
7070 * tree-vect-stmts.c (vect_transform_stmt): Precompute
7071 nested_in_vect_loop_p since the scalar stmt may get invalidated.
7073 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7076 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
7077 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
7078 If off is not INTEGER_CST, issue a may not be aligned warning
7079 rather than isn't aligned. Use isn%'t rather than isn't.
7080 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
7082 <case MULT_EXPR>: Improve the case when bottom and one of the
7083 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
7084 operand, in that case check if the other operand is multiple of
7085 bottom divided by the INTEGER_CST operand.
7087 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7090 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
7091 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
7092 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
7093 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
7094 * config/pa/pa.c (pa_function_arg_advance): Likewise.
7095 (pa_function_arg, pa_arg_partial_bytes): Likewise.
7096 (pa_function_arg_size): New function.
7098 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7100 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
7101 in a separate statement.
7103 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7105 PR tree-optimization/83847
7106 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
7107 group gathers and scatters.
7109 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7111 PR rtl-optimization/86620
7112 * params.def (max-sched-ready-insns): Bump minimum value to 1.
7114 PR rtl-optimization/83213
7115 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
7116 to last if both are JUMP_INSNs.
7118 PR tree-optimization/83843
7119 * gimple-ssa-store-merging.c
7120 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
7121 store_immediate_info for bswap/nop orig_stores.
7123 2018-01-15 Andrew Waterman <andrew@sifive.com>
7125 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
7127 <UDIV>: Increase cost if !TARGET_DIV.
7129 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
7131 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
7132 (define_attr "cr_logical_3op"): New.
7133 (cceq_ior_compare): Adjust.
7134 (cceq_ior_compare_complement): Adjust.
7135 (*cceq_rev_compare): Adjust.
7136 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
7137 (is_cracked_insn): Adjust.
7138 (insn_must_be_first_in_group): Adjust.
7139 * config/rs6000/40x.md: Adjust.
7140 * config/rs6000/440.md: Adjust.
7141 * config/rs6000/476.md: Adjust.
7142 * config/rs6000/601.md: Adjust.
7143 * config/rs6000/603.md: Adjust.
7144 * config/rs6000/6xx.md: Adjust.
7145 * config/rs6000/7450.md: Adjust.
7146 * config/rs6000/7xx.md: Adjust.
7147 * config/rs6000/8540.md: Adjust.
7148 * config/rs6000/cell.md: Adjust.
7149 * config/rs6000/e300c2c3.md: Adjust.
7150 * config/rs6000/e500mc.md: Adjust.
7151 * config/rs6000/e500mc64.md: Adjust.
7152 * config/rs6000/e5500.md: Adjust.
7153 * config/rs6000/e6500.md: Adjust.
7154 * config/rs6000/mpc.md: Adjust.
7155 * config/rs6000/power4.md: Adjust.
7156 * config/rs6000/power5.md: Adjust.
7157 * config/rs6000/power6.md: Adjust.
7158 * config/rs6000/power7.md: Adjust.
7159 * config/rs6000/power8.md: Adjust.
7160 * config/rs6000/power9.md: Adjust.
7161 * config/rs6000/rs64.md: Adjust.
7162 * config/rs6000/titan.md: Adjust.
7164 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7166 * config/i386/predicates.md (indirect_branch_operand): Rewrite
7167 ix86_indirect_branch_register logic.
7169 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7171 * config/i386/constraints.md (Bs): Update
7172 ix86_indirect_branch_register check. Don't check
7173 ix86_indirect_branch_register with GOT_memory_operand.
7175 * config/i386/predicates.md (GOT_memory_operand): Don't check
7176 ix86_indirect_branch_register here.
7177 (GOT32_symbol_operand): Likewise.
7179 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7181 * config/i386/predicates.md (constant_call_address_operand):
7182 Rewrite ix86_indirect_branch_register logic.
7183 (sibcall_insn_operand): Likewise.
7185 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7187 * config/i386/constraints.md (Bs): Replace
7188 ix86_indirect_branch_thunk_register with
7189 ix86_indirect_branch_register.
7191 * config/i386/i386.md (indirect_jump): Likewise.
7192 (tablejump): Likewise.
7193 (*sibcall_memory): Likewise.
7194 (*sibcall_value_memory): Likewise.
7195 Peepholes of indirect call and jump via memory: Likewise.
7196 * config/i386/i386.opt: Likewise.
7197 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7198 (GOT_memory_operand): Likewise.
7199 (call_insn_operand): Likewise.
7200 (sibcall_insn_operand): Likewise.
7201 (GOT32_symbol_operand): Likewise.
7203 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7206 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
7207 type rather than type addr's type points to.
7208 (expand_omp_atomic_mutex): Likewise.
7209 (expand_omp_atomic): Likewise.
7211 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7214 * config/i386/i386.c (output_indirect_thunk_function): Use
7215 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
7216 for __x86_return_thunk.
7218 2018-01-15 Richard Biener <rguenther@suse.de>
7221 * expmed.c (extract_bit_field_1): Fix typo.
7223 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7226 * config/arm/iterators.md (VF): New mode iterator.
7227 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
7228 Remove integer-related logic from pattern.
7229 (neon_vabd<mode>_3): Likewise.
7231 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7234 * common.opt (fstrict-overflow): No longer an alias.
7235 (fwrapv-pointer): New option.
7236 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
7237 also for pointer types based on flag_wrapv_pointer.
7238 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
7239 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
7240 opts->x_flag_wrapv got set.
7241 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
7242 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
7243 POINTER_TYPE_OVERFLOW_UNDEFINED.
7244 * match.pd: Likewise in address comparison pattern.
7245 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
7247 2018-01-15 Richard Biener <rguenther@suse.de>
7250 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
7251 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
7252 Reset type names to their identifier if their TYPE_DECL doesn't
7253 have linkage (and thus is used for ODR and devirt).
7254 (save_debug_info_for_decl): Remove.
7255 (save_debug_info_for_type): Likewise.
7256 (add_tree_to_fld_list): Adjust.
7257 * tree-pretty-print.c (dump_generic_node): Make dumping of
7258 type names more robust.
7260 2018-01-15 Richard Biener <rguenther@suse.de>
7262 * BASE-VER: Bump to 8.0.1.
7264 2018-01-14 Martin Sebor <msebor@redhat.com>
7267 * builtins.c (check_access): Avoid warning when the no-warning bit
7270 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
7272 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
7273 * ira-color (allocno_hard_regs_compare): Likewise.
7275 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
7278 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
7279 Use .pushsection/.popsection.
7281 2018-01-14 Martin Sebor <msebor@redhat.com>
7284 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
7286 2018-01-14 Jakub Jelinek <jakub@redhat.com>
7288 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
7289 entry from extra_headers.
7290 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
7291 extra_headers, make the list bitwise identical to the i?86-*-* one.
7293 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7295 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
7296 -mcmodel=large with -mindirect-branch=thunk,
7297 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
7298 -mfunction-return=thunk-extern.
7299 * doc/invoke.texi: Document -mcmodel=large is incompatible with
7300 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
7301 -mfunction-return=thunk and -mfunction-return=thunk-extern.
7303 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7305 * config/i386/i386.c (print_reg): Print the name of the full
7306 integer register without '%'.
7307 (ix86_print_operand): Handle 'V'.
7308 * doc/extend.texi: Document 'V' modifier.
7310 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7312 * config/i386/constraints.md (Bs): Disallow memory operand for
7313 -mindirect-branch-register.
7315 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7316 (GOT_memory_operand): Likewise.
7317 (call_insn_operand): Likewise.
7318 (sibcall_insn_operand): Likewise.
7319 (GOT32_symbol_operand): Likewise.
7320 * config/i386/i386.md (indirect_jump): Call convert_memory_address
7321 for -mindirect-branch-register.
7322 (tablejump): Likewise.
7323 (*sibcall_memory): Likewise.
7324 (*sibcall_value_memory): Likewise.
7325 Disallow peepholes of indirect call and jump via memory for
7326 -mindirect-branch-register.
7327 (*call_pop): Replace m with Bw.
7328 (*call_value_pop): Likewise.
7329 (*sibcall_pop_memory): Replace m with Bs.
7330 * config/i386/i386.opt (mindirect-branch-register): New option.
7331 * doc/invoke.texi: Document -mindirect-branch-register option.
7333 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7335 * config/i386/i386-protos.h (ix86_output_function_return): New.
7336 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
7337 set function_return_type.
7338 (indirect_thunk_name): Add ret_p to indicate thunk for function
7340 (output_indirect_thunk_function): Pass false to
7341 indirect_thunk_name.
7342 (ix86_output_indirect_branch_via_reg): Likewise.
7343 (ix86_output_indirect_branch_via_push): Likewise.
7344 (output_indirect_thunk_function): Create alias for function
7345 return thunk if regno < 0.
7346 (ix86_output_function_return): New function.
7347 (ix86_handle_fndecl_attribute): Handle function_return.
7348 (ix86_attribute_table): Add function_return.
7349 * config/i386/i386.h (machine_function): Add
7350 function_return_type.
7351 * config/i386/i386.md (simple_return_internal): Use
7352 ix86_output_function_return.
7353 (simple_return_internal_long): Likewise.
7354 * config/i386/i386.opt (mfunction-return=): New option.
7355 (indirect_branch): Mention -mfunction-return=.
7356 * doc/extend.texi: Document function_return function attribute.
7357 * doc/invoke.texi: Document -mfunction-return= option.
7359 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7361 * config/i386/i386-opts.h (indirect_branch): New.
7362 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
7363 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
7364 with local indirect jump when converting indirect call and jump.
7365 (ix86_set_indirect_branch_type): New.
7366 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
7367 (indirectlabelno): New.
7368 (indirect_thunk_needed): Likewise.
7369 (indirect_thunk_bnd_needed): Likewise.
7370 (indirect_thunks_used): Likewise.
7371 (indirect_thunks_bnd_used): Likewise.
7372 (INDIRECT_LABEL): Likewise.
7373 (indirect_thunk_name): Likewise.
7374 (output_indirect_thunk): Likewise.
7375 (output_indirect_thunk_function): Likewise.
7376 (ix86_output_indirect_branch_via_reg): Likewise.
7377 (ix86_output_indirect_branch_via_push): Likewise.
7378 (ix86_output_indirect_branch): Likewise.
7379 (ix86_output_indirect_jmp): Likewise.
7380 (ix86_code_end): Call output_indirect_thunk_function if needed.
7381 (ix86_output_call_insn): Call ix86_output_indirect_branch if
7383 (ix86_handle_fndecl_attribute): Handle indirect_branch.
7384 (ix86_attribute_table): Add indirect_branch.
7385 * config/i386/i386.h (machine_function): Add indirect_branch_type
7386 and has_local_indirect_jump.
7387 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
7389 (tablejump): Likewise.
7390 (*indirect_jump): Use ix86_output_indirect_jmp.
7391 (*tablejump_1): Likewise.
7392 (simple_return_indirect_internal): Likewise.
7393 * config/i386/i386.opt (mindirect-branch=): New option.
7394 (indirect_branch): New.
7397 (thunk-inline): Likewise.
7398 (thunk-extern): Likewise.
7399 * doc/extend.texi: Document indirect_branch function attribute.
7400 * doc/invoke.texi: Document -mindirect-branch= option.
7402 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
7405 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
7407 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
7409 * ipa-inline.c (want_inline_small_function_p): Return false if
7410 inlining has already failed with CIF_FINAL_ERROR.
7411 (update_caller_keys): Call want_inline_small_function_p before
7413 (update_callee_keys): Likewise.
7415 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7417 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
7419 (rs6000_quadword_masked_address_p): Likewise.
7420 (quad_aligned_load_p): Likewise.
7421 (quad_aligned_store_p): Likewise.
7422 (const_load_sequence_p): Add comment to describe the outer-most loop.
7423 (mimic_memory_attributes_and_flags): New function.
7424 (rs6000_gen_stvx): Likewise.
7425 (replace_swapped_aligned_store): Likewise.
7426 (rs6000_gen_lvx): Likewise.
7427 (replace_swapped_aligned_load): Likewise.
7428 (replace_swapped_load_constant): Capitalize argument name in
7429 comment describing this function.
7430 (rs6000_analyze_swaps): Add a third pass to search for vector loads
7431 and stores that access quad-word aligned addresses and replace
7432 with stvx or lvx instructions when appropriate.
7433 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
7434 New function prototype.
7435 (rs6000_quadword_masked_address_p): Likewise.
7436 (rs6000_gen_lvx): Likewise.
7437 (rs6000_gen_stvx): Likewise.
7438 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
7439 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
7440 when memory address is aligned.
7441 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
7442 this split to select lvx instruction when memory address is aligned.
7443 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
7444 instruction when memory address is aligned.
7445 (*vsx_le_perm_load_v16qi): Likewise.
7446 (four unnamed splitters): Modify to select the stvx instruction
7447 when memory is aligned.
7449 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
7451 * predict.c (determine_unlikely_bbs): Handle correctly BBs
7452 which appears in the queue multiple times.
7454 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7455 Alan Hayward <alan.hayward@arm.com>
7456 David Sherwood <david.sherwood@arm.com>
7458 * tree-vectorizer.h (vec_lower_bound): New structure.
7459 (_loop_vec_info): Add check_nonzero and lower_bounds.
7460 (LOOP_VINFO_CHECK_NONZERO): New macro.
7461 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
7462 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
7463 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
7464 fields. Make seg_len the distance travelled, not including the
7466 (dr_direction_indicator): Declare.
7467 (dr_zero_step_indicator): Likewise.
7468 (dr_known_forward_stride_p): Likewise.
7469 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
7471 (runtime_alias_check_p): Allow runtime alias checks with
7473 (operator ==): Compare access_size and align.
7474 (prune_runtime_alias_test_list): Rework for new distinction between
7475 the access_size and seg_len.
7476 (create_intersect_range_checks_index): Likewise. Cope with polynomial
7478 (get_segment_min_max): New function.
7479 (create_intersect_range_checks): Use it.
7480 (dr_step_indicator): New function.
7481 (dr_direction_indicator): Likewise.
7482 (dr_zero_step_indicator): Likewise.
7483 (dr_known_forward_stride_p): Likewise.
7484 * tree-loop-distribution.c (data_ref_segment_size): Return
7485 DR_STEP * (niters - 1).
7486 (compute_alias_check_pairs): Update call to the dr_with_seg_len
7488 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
7489 (vect_preserves_scalar_order_p): New function, split out from...
7490 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
7491 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
7492 (vect_vfa_access_size): New function.
7493 (vect_vfa_align): Likewise.
7494 (vect_compile_time_alias): Take access_size_a and access_b arguments.
7495 (dump_lower_bound): New function.
7496 (vect_check_lower_bound): Likewise.
7497 (vect_small_gap_p): Likewise.
7498 (vectorizable_with_step_bound_p): Likewise.
7499 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
7500 depencies if the vectorization factor is 1. Convert the checks
7501 for nonzero steps into checks on the bounds of DR_STEP. Try using
7502 a bunds check for variable steps if the minimum required step is
7503 relatively small. Update calls to the dr_with_seg_len
7504 constructor and to vect_compile_time_alias.
7505 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
7507 (vect_loop_versioning): Call it.
7508 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
7510 (vect_estimate_min_profitable_iters): Account for any bounds checks.
7512 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7513 Alan Hayward <alan.hayward@arm.com>
7514 David Sherwood <david.sherwood@arm.com>
7516 * doc/sourcebuild.texi (vect_scatter_store): Document.
7517 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
7519 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
7521 * genopinit.c (main): Add supports_vec_scatter_store and
7522 supports_vec_scatter_store_cached to target_optabs.
7523 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
7524 IFN_MASK_SCATTER_STORE.
7525 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
7527 * internal-fn.h (internal_store_fn_p): Declare.
7528 (internal_fn_stored_value_index): Likewise.
7529 * internal-fn.c (scatter_store_direct): New macro.
7530 (expand_scatter_store_optab_fn): New function.
7531 (direct_scatter_store_optab_supported_p): New macro.
7532 (internal_store_fn_p): New function.
7533 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
7534 IFN_MASK_SCATTER_STORE.
7535 (internal_fn_mask_index): Likewise.
7536 (internal_fn_stored_value_index): New function.
7537 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
7539 * optabs-query.h (supports_vec_scatter_store_p): Declare.
7540 * optabs-query.c (supports_vec_scatter_store_p): New function.
7541 * tree-vectorizer.h (vect_get_store_rhs): Declare.
7542 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
7543 true for scatter stores.
7544 (vect_gather_scatter_fn_p): Handle scatter stores too.
7545 (vect_check_gather_scatter): Consider using scatter stores if
7546 supports_vec_scatter_store_p.
7547 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
7549 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7550 internal_fn_stored_value_index.
7551 (check_load_store_masking): Handle scatter stores too.
7552 (vect_get_store_rhs): Make public.
7553 (vectorizable_call): Use internal_store_fn_p.
7554 (vectorizable_store): Handle scatter store internal functions.
7555 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
7556 when deciding whether the end of the group has been reached.
7557 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
7558 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
7559 (mask_scatter_store<mode>): New insns.
7561 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7562 Alan Hayward <alan.hayward@arm.com>
7563 David Sherwood <david.sherwood@arm.com>
7565 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
7566 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
7567 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
7569 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
7570 Use vect_truncate_gather_scatter_offset if we can't treat the
7571 operation as a normal gather load or scatter store.
7572 (get_group_load_store_type): Take the gather_scatter_info
7573 as argument. Try using a gather load or scatter store for
7574 single-element groups.
7575 (get_load_store_type): Update calls to get_group_load_store_type
7576 and vect_use_strided_gather_scatters_p.
7578 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7579 Alan Hayward <alan.hayward@arm.com>
7580 David Sherwood <david.sherwood@arm.com>
7582 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
7583 optional tree argument.
7584 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
7586 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
7587 but continue to use the current value as a fallback.
7588 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
7589 to compare the updates.
7590 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
7591 (get_load_store_type): Use it when handling a strided access.
7592 (vect_get_strided_load_store_ops): New function.
7593 (vect_get_data_ptr_increment): Likewise.
7594 (vectorizable_load): Handle strided gather loads. Always pass
7595 a step to vect_create_data_ref_ptr and bump_vector_ptr.
7597 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7598 Alan Hayward <alan.hayward@arm.com>
7599 David Sherwood <david.sherwood@arm.com>
7601 * doc/md.texi (gather_load@var{m}): Document.
7602 (mask_gather_load@var{m}): Likewise.
7603 * genopinit.c (main): Add supports_vec_gather_load and
7604 supports_vec_gather_load_cached to target_optabs.
7605 * optabs-tree.c (init_tree_optimization_optabs): Use
7606 ggc_cleared_alloc to allocate target_optabs.
7607 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
7608 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
7610 * internal-fn.h (internal_load_fn_p): Declare.
7611 (internal_gather_scatter_fn_p): Likewise.
7612 (internal_fn_mask_index): Likewise.
7613 (internal_gather_scatter_fn_supported_p): Likewise.
7614 * internal-fn.c (gather_load_direct): New macro.
7615 (expand_gather_load_optab_fn): New function.
7616 (direct_gather_load_optab_supported_p): New macro.
7617 (direct_internal_fn_optab): New function.
7618 (internal_load_fn_p): Likewise.
7619 (internal_gather_scatter_fn_p): Likewise.
7620 (internal_fn_mask_index): Likewise.
7621 (internal_gather_scatter_fn_supported_p): Likewise.
7622 * optabs-query.c (supports_at_least_one_mode_p): New function.
7623 (supports_vec_gather_load_p): Likewise.
7624 * optabs-query.h (supports_vec_gather_load_p): Declare.
7625 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
7626 and memory_type field.
7627 (NUM_PATTERNS): Bump to 15.
7628 * tree-vect-data-refs.c: Include internal-fn.h.
7629 (vect_gather_scatter_fn_p): New function.
7630 (vect_describe_gather_scatter_call): Likewise.
7631 (vect_check_gather_scatter): Try using internal functions for
7632 gather loads. Recognize existing calls to a gather load function.
7633 (vect_analyze_data_refs): Consider using gather loads if
7634 supports_vec_gather_load_p.
7635 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
7636 (vect_get_gather_scatter_offset_type): Likewise.
7637 (vect_convert_mask_for_vectype): Likewise.
7638 (vect_add_conversion_to_patterm): Likewise.
7639 (vect_try_gather_scatter_pattern): Likewise.
7640 (vect_recog_gather_scatter_pattern): New pattern recognizer.
7641 (vect_vect_recog_func_ptrs): Add it.
7642 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7643 internal_fn_mask_index and internal_gather_scatter_fn_p.
7644 (check_load_store_masking): Take the gather_scatter_info as an
7645 argument and handle gather loads.
7646 (vect_get_gather_scatter_ops): New function.
7647 (vectorizable_call): Check internal_load_fn_p.
7648 (vectorizable_load): Likewise. Handle gather load internal
7650 (vectorizable_store): Update call to check_load_store_masking.
7651 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
7652 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
7653 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
7654 (aarch64_gather_scale_operand_d): New predicates.
7655 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
7656 (mask_gather_load<mode>): New insns.
7658 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7659 Alan Hayward <alan.hayward@arm.com>
7660 David Sherwood <david.sherwood@arm.com>
7662 * optabs.def (fold_left_plus_optab): New optab.
7663 * doc/md.texi (fold_left_plus_@var{m}): Document.
7664 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
7665 * internal-fn.c (fold_left_direct): Define.
7666 (expand_fold_left_optab_fn): Likewise.
7667 (direct_fold_left_optab_supported_p): Likewise.
7668 * fold-const-call.c (fold_const_fold_left): New function.
7669 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
7670 * tree-parloops.c (valid_reduction_p): New function.
7671 (gather_scalar_reductions): Use it.
7672 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
7673 (vect_finish_replace_stmt): Declare.
7674 * tree-vect-loop.c (fold_left_reduction_fn): New function.
7675 (needs_fold_left_reduction_p): New function, split out from...
7676 (vect_is_simple_reduction): ...here. Accept reductions that
7677 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
7678 (vect_force_simple_reduction): Also store the reduction type in
7679 the assignment's STMT_VINFO_REDUC_TYPE.
7680 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
7681 (merge_with_identity): New function.
7682 (vect_expand_fold_left): Likewise.
7683 (vectorize_fold_left_reduction): Likewise.
7684 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
7685 scalar phi in place for it. Check for target support and reject
7686 cases that would reassociate the operation. Defer the transform
7687 phase to vectorize_fold_left_reduction.
7688 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
7689 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
7690 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
7692 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7694 * tree-if-conv.c (predicate_mem_writes): Remove redundant
7695 call to ifc_temp_var.
7697 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7698 Alan Hayward <alan.hayward@arm.com>
7699 David Sherwood <david.sherwood@arm.com>
7701 * target.def (legitimize_address_displacement): Take the original
7702 offset as a poly_int.
7703 * targhooks.h (default_legitimize_address_displacement): Update
7705 * targhooks.c (default_legitimize_address_displacement): Likewise.
7706 * doc/tm.texi: Regenerate.
7707 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
7708 as an argument, moving assert of ad->disp == ad->disp_term to...
7709 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
7710 Try calling targetm.legitimize_address_displacement before expanding
7711 the address rather than afterwards, and adjust for the new interface.
7712 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
7713 Match the new hook interface. Handle SVE addresses.
7714 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
7717 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7719 * Makefile.in (OBJS): Add early-remat.o.
7720 * target.def (select_early_remat_modes): New hook.
7721 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
7722 * doc/tm.texi: Regenerate.
7723 * targhooks.h (default_select_early_remat_modes): Declare.
7724 * targhooks.c (default_select_early_remat_modes): New function.
7725 * timevar.def (TV_EARLY_REMAT): New timevar.
7726 * passes.def (pass_early_remat): New pass.
7727 * tree-pass.h (make_pass_early_remat): Declare.
7728 * early-remat.c: New file.
7729 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
7731 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
7733 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7734 Alan Hayward <alan.hayward@arm.com>
7735 David Sherwood <david.sherwood@arm.com>
7737 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
7738 vfm1 with a bound_epilog parameter.
7739 (vect_do_peeling): Update calls accordingly, and move the prologue
7740 call earlier in the function. Treat the base bound_epilog as 0 for
7741 fully-masked loops and retain vf - 1 for other loops. Add 1 to
7742 this base when peeling for gaps.
7743 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
7744 with fully-masked loops.
7745 (vect_estimate_min_profitable_iters): Handle the single peeled
7746 iteration in that case.
7748 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7749 Alan Hayward <alan.hayward@arm.com>
7750 David Sherwood <david.sherwood@arm.com>
7752 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
7753 single-element interleaving even if the size is not a power of 2.
7754 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
7755 accesses for single-element interleaving if the group size is
7758 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7759 Alan Hayward <alan.hayward@arm.com>
7760 David Sherwood <david.sherwood@arm.com>
7762 * doc/md.texi (fold_extract_last_@var{m}): Document.
7763 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
7764 * optabs.def (fold_extract_last_optab): New optab.
7765 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
7766 * internal-fn.c (fold_extract_direct): New macro.
7767 (expand_fold_extract_optab_fn): Likewise.
7768 (direct_fold_extract_optab_supported_p): Likewise.
7769 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
7770 * tree-vect-loop.c (vect_model_reduction_cost): Handle
7771 EXTRACT_LAST_REDUCTION.
7772 (get_initial_def_for_reduction): Do not create an initial vector
7773 for EXTRACT_LAST_REDUCTION reductions.
7774 (vectorizable_reduction): Leave the scalar phi in place for
7775 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
7776 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
7777 epilogue code for EXTRACT_LAST_REDUCTION and defer the
7778 transform phase to vectorizable_condition.
7779 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
7781 (vect_finish_stmt_generation): ...here.
7782 (vect_finish_replace_stmt): New function.
7783 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
7784 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
7786 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
7788 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7789 Alan Hayward <alan.hayward@arm.com>
7790 David Sherwood <david.sherwood@arm.com>
7792 * doc/md.texi (extract_last_@var{m}): Document.
7793 * optabs.def (extract_last_optab): New optab.
7794 * internal-fn.def (EXTRACT_LAST): New internal function.
7795 * internal-fn.c (cond_unary_direct): New macro.
7796 (expand_cond_unary_optab_fn): Likewise.
7797 (direct_cond_unary_optab_supported_p): Likewise.
7798 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
7799 loops using EXTRACT_LAST.
7800 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
7801 (extract_last_<mode>): ...this optab.
7802 (vec_extract<mode><Vel>): Update accordingly.
7804 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7805 Alan Hayward <alan.hayward@arm.com>
7806 David Sherwood <david.sherwood@arm.com>
7808 * target.def (empty_mask_is_expensive): New hook.
7809 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
7810 * doc/tm.texi: Regenerate.
7811 * targhooks.h (default_empty_mask_is_expensive): Declare.
7812 * targhooks.c (default_empty_mask_is_expensive): New function.
7813 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
7814 if the target says that empty masks are expensive.
7815 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
7817 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
7819 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7820 Alan Hayward <alan.hayward@arm.com>
7821 David Sherwood <david.sherwood@arm.com>
7823 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
7824 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
7825 (vect_use_loop_mask_for_alignment_p): New function.
7826 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
7827 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
7828 niters_skip argument. Make sure that the first niters_skip elements
7829 of the first iteration are inactive.
7830 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
7831 Update call to vect_set_loop_masks_directly.
7832 (get_misalign_in_elems): New function, split out from...
7833 (vect_gen_prolog_loop_niters): ...here.
7834 (vect_update_init_of_dr): Take a code argument that specifies whether
7835 the adjustment should be added or subtracted.
7836 (vect_update_init_of_drs): Likewise.
7837 (vect_prepare_for_masked_peels): New function.
7838 (vect_do_peeling): Skip prologue peeling if we're using a mask
7839 instead. Update call to vect_update_inits_of_drs.
7840 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7842 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
7843 alignment. Do not include the number of peeled iterations in
7844 the minimum threshold in that case.
7845 (vectorizable_induction): Adjust the start value down by
7846 LOOP_VINFO_MASK_SKIP_NITERS iterations.
7847 (vect_transform_loop): Call vect_prepare_for_masked_peels.
7848 Take the number of skipped iterations into account when calculating
7850 * tree-vect-stmts.c (vect_gen_while_not): New function.
7852 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7853 Alan Hayward <alan.hayward@arm.com>
7854 David Sherwood <david.sherwood@arm.com>
7856 * doc/sourcebuild.texi (vect_fully_masked): Document.
7857 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
7859 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
7861 (vect_analyze_loop_2): ...here. Don't check the vectorization
7862 factor against the number of loop iterations if the loop is
7865 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7866 Alan Hayward <alan.hayward@arm.com>
7867 David Sherwood <david.sherwood@arm.com>
7869 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
7870 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
7871 (dump_groups): Update accordingly.
7872 (iv_use::mem_type): New member variable.
7873 (address_p): New function.
7874 (record_use): Add a mem_type argument and initialize the new
7876 (record_group_use): Add a mem_type argument. Use address_p.
7877 Remove obsolete null checks of base_object. Update call to record_use.
7878 (find_interesting_uses_op): Update call to record_group_use.
7879 (find_interesting_uses_cond): Likewise.
7880 (find_interesting_uses_address): Likewise.
7881 (get_mem_type_for_internal_fn): New function.
7882 (find_address_like_use): Likewise.
7883 (find_interesting_uses_stmt): Try find_address_like_use before
7884 calling find_interesting_uses_op.
7885 (addr_offset_valid_p): Use the iv mem_type field as the type
7886 of the addressed memory.
7887 (add_autoinc_candidates): Likewise.
7888 (get_address_cost): Likewise.
7889 (split_small_address_groups_p): Use address_p.
7890 (split_address_groups): Likewise.
7891 (add_iv_candidate_for_use): Likewise.
7892 (autoinc_possible_for_pair): Likewise.
7893 (rewrite_groups): Likewise.
7894 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
7895 (determine_group_iv_cost): Update after split of USE_ADDRESS.
7896 (get_alias_ptr_type_for_ptr_address): New function.
7897 (rewrite_use_address): Rewrite address uses in calls that were
7898 identified by find_address_like_use.
7900 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7901 Alan Hayward <alan.hayward@arm.com>
7902 David Sherwood <david.sherwood@arm.com>
7904 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
7906 * gimple-expr.h (is_gimple_addressable: Likewise.
7907 * gimple-expr.c (is_gimple_address): Likewise.
7908 * internal-fn.c (expand_call_mem_ref): New function.
7909 (expand_mask_load_optab_fn): Use it.
7910 (expand_mask_store_optab_fn): Likewise.
7912 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7913 Alan Hayward <alan.hayward@arm.com>
7914 David Sherwood <david.sherwood@arm.com>
7916 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
7917 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
7918 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
7919 (cond_umax@var{mode}): Document.
7920 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
7921 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
7922 (cond_umin_optab, cond_umax_optab): New optabs.
7923 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
7924 (COND_IOR, COND_XOR): New internal functions.
7925 * internal-fn.h (get_conditional_internal_fn): Declare.
7926 * internal-fn.c (cond_binary_direct): New macro.
7927 (expand_cond_binary_optab_fn): Likewise.
7928 (direct_cond_binary_optab_supported_p): Likewise.
7929 (get_conditional_internal_fn): New function.
7930 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
7931 Cope with reduction statements that are vectorized as calls rather
7933 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
7934 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
7935 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
7936 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
7937 (UNSPEC_COND_EOR): New unspecs.
7938 (optab): Add mappings for them.
7939 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
7940 (sve_int_op, sve_fp_op): New int attributes.
7942 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7943 Alan Hayward <alan.hayward@arm.com>
7944 David Sherwood <david.sherwood@arm.com>
7946 * optabs.def (while_ult_optab): New optab.
7947 * doc/md.texi (while_ult@var{m}@var{n}): Document.
7948 * internal-fn.def (WHILE_ULT): New internal function.
7949 * internal-fn.h (direct_internal_fn_supported_p): New override
7950 that takes two types as argument.
7951 * internal-fn.c (while_direct): New macro.
7952 (expand_while_optab_fn): New function.
7953 (convert_optab_supported_p): Likewise.
7954 (direct_while_optab_supported_p): New macro.
7955 * wide-int.h (wi::udiv_ceil): New function.
7956 * tree-vectorizer.h (rgroup_masks): New structure.
7957 (vec_loop_masks): New typedef.
7958 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
7960 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
7961 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
7962 (vect_max_vf): New function.
7963 (slpeel_make_loop_iterate_ntimes): Delete.
7964 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
7965 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
7966 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
7967 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
7968 internal-fn.h, stor-layout.h and optabs-query.h.
7969 (vect_set_loop_mask): New function.
7970 (add_preheader_seq): Likewise.
7971 (add_header_seq): Likewise.
7972 (interleave_supported_p): Likewise.
7973 (vect_maybe_permute_loop_masks): Likewise.
7974 (vect_set_loop_masks_directly): Likewise.
7975 (vect_set_loop_condition_masked): Likewise.
7976 (vect_set_loop_condition_unmasked): New function, split out from
7977 slpeel_make_loop_iterate_ntimes.
7978 (slpeel_make_loop_iterate_ntimes): Rename to..
7979 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
7980 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
7981 (vect_do_peeling): Update call accordingly.
7982 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
7984 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7985 mask_compare_type, can_fully_mask_p and fully_masked_p.
7986 (release_vec_loop_masks): New function.
7987 (_loop_vec_info): Use it to free the loop masks.
7988 (can_produce_all_loop_masks_p): New function.
7989 (vect_get_max_nscalars_per_iter): Likewise.
7990 (vect_verify_full_masking): Likewise.
7991 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
7992 retries, and free the mask rgroups before retrying. Check loop-wide
7993 reasons for disallowing fully-masked loops. Make the final decision
7994 about whether use a fully-masked loop or not.
7995 (vect_estimate_min_profitable_iters): Do not assume that peeling
7996 for the number of iterations will be needed for fully-masked loops.
7997 (vectorizable_reduction): Disable fully-masked loops.
7998 (vectorizable_live_operation): Likewise.
7999 (vect_halve_mask_nunits): New function.
8000 (vect_double_mask_nunits): Likewise.
8001 (vect_record_loop_mask): Likewise.
8002 (vect_get_loop_mask): Likewise.
8003 (vect_transform_loop): Handle the case in which the final loop
8004 iteration might handle a partial vector. Call vect_set_loop_condition
8005 instead of slpeel_make_loop_iterate_ntimes.
8006 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
8007 (check_load_store_masking): New function.
8008 (prepare_load_store_mask): Likewise.
8009 (vectorizable_store): Handle fully-masked loops.
8010 (vectorizable_load): Likewise.
8011 (supportable_widening_operation): Use vect_halve_mask_nunits for
8013 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
8014 (vect_gen_while): New function.
8015 * config/aarch64/aarch64.md (umax<mode>3): New expander.
8016 (aarch64_uqdec<mode>): New insn.
8018 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8019 Alan Hayward <alan.hayward@arm.com>
8020 David Sherwood <david.sherwood@arm.com>
8022 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
8023 (reduc_xor_scal_optab): New optabs.
8024 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
8025 (reduc_xor_scal_@var{m}): Document.
8026 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
8027 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
8029 * fold-const-call.c (fold_const_call): Handle them.
8030 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
8031 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
8032 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
8033 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
8034 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
8035 (UNSPEC_XORV): New unspecs.
8036 (optab): Add entries for them.
8037 (BITWISEV): New int iterator.
8038 (bit_reduc_op): New int attributes.
8040 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8041 Alan Hayward <alan.hayward@arm.com>
8042 David Sherwood <david.sherwood@arm.com>
8044 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
8045 * internal-fn.def (VEC_SHL_INSERT): New internal function.
8046 * optabs.def (vec_shl_insert_optab): New optab.
8047 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
8048 (duplicate_and_interleave): Likewise.
8049 * tree-vect-loop.c: Include internal-fn.h.
8050 (neutral_op_for_slp_reduction): New function, split out from
8051 get_initial_defs_for_reduction.
8052 (get_initial_def_for_reduction): Handle option 2 for variable-length
8053 vectors by loading the neutral value into a vector and then shifting
8054 the initial value into element 0.
8055 (get_initial_defs_for_reduction): Replace the code argument with
8056 the neutral value calculated by neutral_op_for_slp_reduction.
8057 Use gimple_build_vector for constant-length vectors.
8058 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
8059 but the first group_size elements have a neutral value.
8060 Use duplicate_and_interleave otherwise.
8061 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
8062 Update call to get_initial_defs_for_reduction. Handle SLP
8063 reductions for variable-length vectors by creating one vector
8064 result for each scalar result, with the elements associated
8065 with other scalar results stubbed out with the neutral value.
8066 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
8067 Require IFN_VEC_SHL_INSERT for double reductions on
8068 variable-length vectors, or SLP reductions that have
8069 a neutral value. Require can_duplicate_and_interleave_p
8070 support for variable-length unchained SLP reductions if there
8071 is no neutral value, such as for MIN/MAX reductions. Also require
8072 the number of vector elements to be a multiple of the number of
8073 SLP statements when doing variable-length unchained SLP reductions.
8074 Update call to vect_create_epilog_for_reduction.
8075 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
8076 and remove initial values.
8077 (duplicate_and_interleave): Make public.
8078 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
8079 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
8081 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8082 Alan Hayward <alan.hayward@arm.com>
8083 David Sherwood <david.sherwood@arm.com>
8085 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
8086 (can_duplicate_and_interleave_p): New function.
8087 (vect_get_and_check_slp_defs): Take the vector of statements
8088 rather than just the current one. Remove excess parentheses.
8089 Restriction rejectinon of vect_constant_def and vect_external_def
8090 for variable-length vectors to boolean types, or types for which
8091 can_duplicate_and_interleave_p is false.
8092 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
8093 (duplicate_and_interleave): New function.
8094 (vect_get_constant_vectors): Use gimple_build_vector for
8095 constant-length vectors and suitable variable-length constant
8096 vectors. Use duplicate_and_interleave for other variable-length
8097 vectors. Don't defer the update when inserting new statements.
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 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
8104 min_profitable_iters doesn't go negative.
8106 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8107 Alan Hayward <alan.hayward@arm.com>
8108 David Sherwood <david.sherwood@arm.com>
8110 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
8111 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
8112 * optabs.def (vec_mask_load_lanes_optab): New optab.
8113 (vec_mask_store_lanes_optab): Likewise.
8114 * internal-fn.def (MASK_LOAD_LANES): New internal function.
8115 (MASK_STORE_LANES): Likewise.
8116 * internal-fn.c (mask_load_lanes_direct): New macro.
8117 (mask_store_lanes_direct): Likewise.
8118 (expand_mask_load_optab_fn): Handle masked operations.
8119 (expand_mask_load_lanes_optab_fn): New macro.
8120 (expand_mask_store_optab_fn): Handle masked operations.
8121 (expand_mask_store_lanes_optab_fn): New macro.
8122 (direct_mask_load_lanes_optab_supported_p): Likewise.
8123 (direct_mask_store_lanes_optab_supported_p): Likewise.
8124 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
8126 (vect_load_lanes_supported): Likewise.
8127 * tree-vect-data-refs.c (strip_conversion): New function.
8128 (can_group_stmts_p): Likewise.
8129 (vect_analyze_data_ref_accesses): Use it instead of checking
8130 for a pair of assignments.
8131 (vect_store_lanes_supported): Take a masked_p parameter.
8132 (vect_load_lanes_supported): Likewise.
8133 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
8134 vect_store_lanes_supported and vect_load_lanes_supported.
8135 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8136 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
8137 parameter. Don't allow gaps for masked accesses.
8138 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
8139 and vect_load_lanes_supported.
8140 (get_load_store_type): Take a masked_p parameter and update
8141 call to get_group_load_store_type.
8142 (vectorizable_store): Update call to get_load_store_type.
8143 Handle IFN_MASK_STORE_LANES.
8144 (vectorizable_load): Update call to get_load_store_type.
8145 Handle IFN_MASK_LOAD_LANES.
8147 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8148 Alan Hayward <alan.hayward@arm.com>
8149 David Sherwood <david.sherwood@arm.com>
8151 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
8153 * config/aarch64/aarch64-protos.h
8154 (aarch64_sve_struct_memory_operand_p): Declare.
8155 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
8156 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
8157 (VPRED, vpred): Handle SVE structure modes.
8158 * config/aarch64/constraints.md (Utx): New constraint.
8159 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
8160 (aarch64_sve_struct_nonimmediate_operand): New predicates.
8161 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
8162 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
8163 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
8164 structure modes. Split into pieces after RA.
8165 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
8166 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
8168 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
8169 SVE structure modes.
8170 (aarch64_classify_address): Likewise.
8171 (sizetochar): Move earlier in file.
8172 (aarch64_print_operand): Handle SVE register lists.
8173 (aarch64_array_mode): New function.
8174 (aarch64_sve_struct_memory_operand_p): Likewise.
8175 (TARGET_ARRAY_MODE): Redefine.
8177 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8178 Alan Hayward <alan.hayward@arm.com>
8179 David Sherwood <david.sherwood@arm.com>
8181 * target.def (array_mode): New target hook.
8182 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
8183 * doc/tm.texi: Regenerate.
8184 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
8185 * hooks.c (hook_optmode_mode_uhwi_none): New function.
8186 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
8188 * stor-layout.c (mode_for_array): Likewise. Support polynomial
8191 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8192 Alan Hayward <alan.hayward@arm.com>
8193 David Sherwood <david.sherwood@arm.com>
8195 * fold-const.c (fold_binary_loc): Check the argument types
8196 rather than the result type when testing for a vector operation.
8198 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8200 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
8201 * doc/tm.texi: Regenerate.
8203 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8204 Alan Hayward <alan.hayward@arm.com>
8205 David Sherwood <david.sherwood@arm.com>
8207 * doc/invoke.texi (-msve-vector-bits=): Document new option.
8208 (sve): Document new AArch64 extension.
8209 * doc/md.texi (w): Extend the description of the AArch64
8210 constraint to include SVE vectors.
8211 (Upl, Upa): Document new AArch64 predicate constraints.
8212 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
8214 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
8215 (msve-vector-bits=): New option.
8216 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
8217 SVE when these are disabled.
8218 (sve): New extension.
8219 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
8220 modes. Adjust their number of units based on aarch64_sve_vg.
8221 (MAX_BITSIZE_MODE_ANY_MODE): Define.
8222 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
8223 aarch64_addr_query_type.
8224 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
8225 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
8226 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
8227 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
8228 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
8229 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
8230 (aarch64_simd_imm_zero_p): Delete.
8231 (aarch64_check_zero_based_sve_index_immediate): Declare.
8232 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8233 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8234 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8235 (aarch64_sve_float_mul_immediate_p): Likewise.
8236 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8238 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
8239 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
8240 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
8241 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
8242 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
8243 (aarch64_regmode_natural_size): Likewise.
8244 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
8245 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
8247 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
8248 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
8249 for VG and the SVE predicate registers.
8250 (V_ALIASES): Add a "z"-prefixed alias.
8251 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
8252 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
8253 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
8254 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
8255 (REG_CLASS_NAMES): Add entries for them.
8256 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
8257 and the predicate registers.
8258 (aarch64_sve_vg): Declare.
8259 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
8260 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
8261 (REGMODE_NATURAL_SIZE): Define.
8262 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
8264 * config/aarch64/aarch64.c: Include cfgrtl.h.
8265 (simd_immediate_info): Add a constructor for series vectors,
8266 and an associated step field.
8267 (aarch64_sve_vg): New variable.
8268 (aarch64_dbx_register_number): Handle VG and the predicate registers.
8269 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
8270 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
8271 (VEC_ANY_DATA, VEC_STRUCT): New constants.
8272 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
8273 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
8274 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
8275 (aarch64_get_mask_mode): New functions.
8276 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
8277 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
8278 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
8279 predicate modes and predicate registers. Explicitly restrict
8280 GPRs to modes of 16 bytes or smaller. Only allow FP registers
8281 to store a vector mode if it is recognized by
8282 aarch64_classify_vector_mode.
8283 (aarch64_regmode_natural_size): New function.
8284 (aarch64_hard_regno_caller_save_mode): Return the original mode
8286 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
8287 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
8288 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
8289 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
8291 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
8292 does not overlap dest if the function is frame-related. Handle
8294 (aarch64_split_add_offset): New function.
8295 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
8296 them aarch64_add_offset.
8297 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
8298 and update call to aarch64_sub_sp.
8299 (aarch64_add_cfa_expression): New function.
8300 (aarch64_expand_prologue): Pass extra temporary registers to the
8301 functions above. Handle the case in which we need to emit new
8302 DW_CFA_expressions for registers that were originally saved
8303 relative to the stack pointer, but now have to be expressed
8304 relative to the frame pointer.
8305 (aarch64_output_mi_thunk): Pass extra temporary registers to the
8307 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
8308 IP0 and IP1 values for SVE frames.
8309 (aarch64_expand_vec_series): New function.
8310 (aarch64_expand_sve_widened_duplicate): Likewise.
8311 (aarch64_expand_sve_const_vector): Likewise.
8312 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
8313 Handle SVE constants. Use emit_move_insn to move a force_const_mem
8314 into the register, rather than emitting a SET directly.
8315 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
8316 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
8317 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
8318 (offset_9bit_signed_scaled_p): New functions.
8319 (aarch64_replicate_bitmask_imm): New function.
8320 (aarch64_bitmask_imm): Use it.
8321 (aarch64_cannot_force_const_mem): Reject expressions involving
8322 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
8323 (aarch64_classify_index): Handle SVE indices, by requiring
8324 a plain register index with a scale that matches the element size.
8325 (aarch64_classify_address): Handle SVE addresses. Assert that
8326 the mode of the address is VOIDmode or an integer mode.
8327 Update call to aarch64_classify_symbol.
8328 (aarch64_classify_symbolic_expression): Update call to
8329 aarch64_classify_symbol.
8330 (aarch64_const_vec_all_in_range_p): New function.
8331 (aarch64_print_vector_float_operand): Likewise.
8332 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
8333 "vN" for FP registers with SVE modes. Handle (const ...) vectors
8334 and the FP immediates 1.0 and 0.5.
8335 (aarch64_print_address_internal): Handle SVE addresses.
8336 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
8337 (aarch64_regno_regclass): Handle predicate registers.
8338 (aarch64_secondary_reload): Handle big-endian reloads of SVE
8340 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
8341 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
8342 (aarch64_convert_sve_vector_bits): New function.
8343 (aarch64_override_options): Use it to handle -msve-vector-bits=.
8344 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8346 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
8347 Handle SVE vector and predicate modes. Accept VL-based constants
8348 that need only one temporary register, and VL offsets that require
8349 no temporary registers.
8350 (aarch64_conditional_register_usage): Mark the predicate registers
8351 as fixed if SVE isn't available.
8352 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
8353 Return true for SVE vector and predicate modes.
8354 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
8355 rather than an unsigned int. Handle SVE modes.
8356 (aarch64_preferred_simd_mode): Update call accordingly. Handle
8358 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
8360 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8361 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8362 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8363 (aarch64_sve_float_mul_immediate_p): New functions.
8364 (aarch64_sve_valid_immediate): New function.
8365 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
8366 Explicitly reject structure modes. Check for INDEX constants.
8367 Handle PTRUE and PFALSE constants.
8368 (aarch64_check_zero_based_sve_index_immediate): New function.
8369 (aarch64_simd_imm_zero_p): Delete.
8370 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
8371 vector modes. Accept constants in the range of CNT[BHWD].
8372 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
8373 ask for an Advanced SIMD mode.
8374 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
8375 (aarch64_simd_vector_alignment): Handle SVE predicates.
8376 (aarch64_vectorize_preferred_vector_alignment): New function.
8377 (aarch64_simd_vector_alignment_reachable): Use it instead of
8379 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
8380 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
8382 (MAX_VECT_LEN): Delete.
8383 (expand_vec_perm_d): Add a vec_flags field.
8384 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
8385 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
8386 (aarch64_evpc_ext): Don't apply a big-endian lane correction
8388 (aarch64_evpc_rev): Rename to...
8389 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
8390 (aarch64_evpc_rev_global): New function.
8391 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
8392 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
8394 (aarch64_evpc_sve_tbl): New function.
8395 (aarch64_expand_vec_perm_const_1): Update after rename of
8396 aarch64_evpc_rev. Handle SVE permutes too, trying
8397 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
8398 than aarch64_evpc_tbl.
8399 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
8400 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
8401 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
8402 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
8403 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
8404 (aarch64_expand_sve_vcond): New functions.
8405 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
8406 of aarch64_vector_mode_p.
8407 (aarch64_dwarf_poly_indeterminate_value): New function.
8408 (aarch64_compute_pressure_classes): Likewise.
8409 (aarch64_can_change_mode_class): Likewise.
8410 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
8411 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
8412 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
8413 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
8414 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
8415 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
8416 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
8417 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
8419 (Dn, Dl, Dr): Accept const as well as const_vector.
8420 (Dz): Likewise. Compare against CONST0_RTX.
8421 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
8422 of "vector" where appropriate.
8423 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
8424 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
8425 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
8426 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
8427 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
8428 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
8429 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
8430 (v_int_equiv): Extend to SVE modes.
8431 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
8433 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
8434 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
8435 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
8436 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
8437 (SVE_COND_FP_CMP): New int iterators.
8438 (perm_hilo): Handle the new unpack unspecs.
8439 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
8441 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
8442 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
8443 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
8444 (aarch64_equality_operator, aarch64_constant_vector_operand)
8445 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
8446 (aarch64_sve_nonimmediate_operand): Likewise.
8447 (aarch64_sve_general_operand): Likewise.
8448 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
8449 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
8450 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
8451 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
8452 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
8453 (aarch64_sve_float_arith_immediate): Likewise.
8454 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
8455 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
8456 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
8457 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
8458 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
8459 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
8460 (aarch64_sve_float_arith_operand): Likewise.
8461 (aarch64_sve_float_arith_with_sub_operand): Likewise.
8462 (aarch64_sve_float_mul_operand): Likewise.
8463 (aarch64_sve_vec_perm_operand): Likewise.
8464 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
8465 (aarch64_mov_operand): Accept const_poly_int and const_vector.
8466 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
8467 as well as const_vector.
8468 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
8469 in file. Use CONST0_RTX and CONSTM1_RTX.
8470 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
8471 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
8472 Use aarch64_simd_imm_zero.
8473 * config/aarch64/aarch64-sve.md: New file.
8474 * config/aarch64/aarch64.md: Include it.
8475 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
8476 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
8477 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
8478 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
8479 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
8480 (sve): New attribute.
8481 (enabled): Disable instructions with the sve attribute unless
8483 (movqi, movhi): Pass CONST_POLY_INT operaneds through
8484 aarch64_expand_mov_immediate.
8485 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
8486 CNT[BHSD] immediates.
8487 (movti): Split CONST_POLY_INT moves into two halves.
8488 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
8489 Split additions that need a temporary here if the destination
8490 is the stack pointer.
8491 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
8492 (*add<mode>3_poly_1): New instruction.
8493 (set_clobber_cc): New expander.
8495 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8497 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
8498 parameter and use it instead of GET_MODE_SIZE (innermode). Use
8499 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
8500 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
8501 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
8502 Change innermode from fixed_mode_size to machine_mode.
8503 (simplify_subreg): Update call accordingly. Handle a constant-sized
8504 subreg of a variable-length CONST_VECTOR.
8506 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8507 Alan Hayward <alan.hayward@arm.com>
8508 David Sherwood <david.sherwood@arm.com>
8510 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
8511 (add_offset_to_base): New function, split out from...
8512 (create_mem_ref): ...here. When handling a scale other than 1,
8513 check first whether the address is valid without the offset.
8514 Add it into the base if so, leaving the index and scale as-is.
8516 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8519 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
8520 fold_for_warn before checking if arg2 is INTEGER_CST.
8522 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
8524 * config/rs6000/predicates.md (load_multiple_operation): Delete.
8525 (store_multiple_operation): Delete.
8526 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
8527 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
8528 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
8529 guarded by TARGET_STRING.
8530 (rs6000_output_load_multiple): Delete.
8531 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
8532 OPTION_MASK_STRING / TARGET_STRING handling.
8533 (print_operand) <'N', 'O'>: Add comment that these are unused now.
8534 (const rs6000_opt_masks) <"string">: Change mask to 0.
8535 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
8536 (MASK_STRING): Delete.
8537 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
8539 (load_multiple): Delete.
8546 (store_multiple): Delete.
8553 (movmemsi_8reg): Delete.
8554 (corresponding unnamed define_insn): Delete.
8555 (movmemsi_6reg): Delete.
8556 (corresponding unnamed define_insn): Delete.
8557 (movmemsi_4reg): Delete.
8558 (corresponding unnamed define_insn): Delete.
8559 (movmemsi_2reg): Delete.
8560 (corresponding unnamed define_insn): Delete.
8561 (movmemsi_1reg): Delete.
8562 (corresponding unnamed define_insn): Delete.
8563 * config/rs6000/rs6000.opt (mno-string): New.
8564 (mstring): Replace by deprecation warning stub.
8565 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
8567 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8569 * regrename.c (regrename_do_replace): If replacing the same
8570 reg multiple times, try to reuse last created gen_raw_REG.
8573 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
8574 main to workaround a bug in GDB.
8576 2018-01-12 Tom de Vries <tom@codesourcery.com>
8579 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
8581 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
8583 PR rtl-optimization/80481
8584 * ira-color.c (get_cap_member): New function.
8585 (allocnos_conflict_by_live_ranges_p): Use it.
8586 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
8587 (setup_slot_coalesced_allocno_live_ranges): Ditto.
8589 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
8592 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
8593 (*saddl_se_1): Ditto.
8595 (*ssubl_se_1): Ditto.
8597 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8599 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
8600 rather than wi::to_widest for DR_INITs.
8601 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
8602 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
8603 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
8605 (vect_analyze_group_access_1): Note that here.
8607 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8609 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
8610 polynomial type sizes.
8612 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8614 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
8615 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
8616 (gimple_add_tmp_var): Likewise.
8618 2018-01-12 Martin Liska <mliska@suse.cz>
8620 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
8621 (gimple_alloc_sizes): Likewise.
8622 (dump_gimple_statistics): Use PRIu64 in printf format.
8623 * gimple.h: Change uint64_t to int.
8625 2018-01-12 Martin Liska <mliska@suse.cz>
8627 * tree-core.h: Use uint64_t instead of int.
8628 * tree.c (tree_node_counts): Likewise.
8629 (tree_node_sizes): Likewise.
8630 (dump_tree_statistics): Use PRIu64 in printf format.
8632 2018-01-12 Martin Liska <mliska@suse.cz>
8634 * Makefile.in: As qsort_chk is implemented in vec.c, add
8635 vec.o to linkage of gencfn-macros.
8636 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
8637 passing the info to record_node_allocation_statistics.
8638 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
8640 * ggc-common.c (struct ggc_usage): Add operator== and use
8641 it in operator< and compare function.
8642 * mem-stats.h (struct mem_usage): Likewise.
8643 * vec.c (struct vec_usage): Remove operator< and compare
8644 function. Can be simply inherited.
8646 2018-01-12 Martin Jambor <mjambor@suse.cz>
8649 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
8650 * tree-ssa-math-opts.c: Include domwalk.h.
8651 (convert_mult_to_fma_1): New function.
8652 (fma_transformation_info): New type.
8653 (fma_deferring_state): Likewise.
8654 (cancel_fma_deferring): New function.
8655 (result_of_phi): Likewise.
8656 (last_fma_candidate_feeds_initial_phi): Likewise.
8657 (convert_mult_to_fma): Added deferring logic, split actual
8658 transformation to convert_mult_to_fma_1.
8659 (math_opts_dom_walker): New type.
8660 (math_opts_dom_walker::after_dom_children): New method, body moved
8661 here from pass_optimize_widening_mul::execute, added deferring logic
8663 (pass_optimize_widening_mul::execute): Moved most of code to
8664 math_opts_dom_walker::after_dom_children.
8665 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
8666 * config/i386/i386.c (ix86_option_override_internal): Added
8667 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
8669 2018-01-12 Richard Biener <rguenther@suse.de>
8672 * dwarf2out.c (gen_variable_die): Do not reset old_die for
8673 inline instance vars.
8675 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
8678 * config/rx/rx.c (rx_is_restricted_memory_address):
8681 2018-01-12 Richard Biener <rguenther@suse.de>
8683 PR tree-optimization/80846
8684 * target.def (split_reduction): New target hook.
8685 * targhooks.c (default_split_reduction): New function.
8686 * targhooks.h (default_split_reduction): Declare.
8687 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
8688 target requests first reduce vectors by combining low and high
8690 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
8691 (get_vectype_for_scalar_type_and_size): Export.
8692 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
8693 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
8694 * doc/tm.texi: Regenerate.
8695 * config/i386/i386.c (ix86_split_reduction): Implement
8696 TARGET_VECTORIZE_SPLIT_REDUCTION.
8698 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8701 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
8702 in PIC mode except for TARGET_VXWORKS_RTP.
8703 * config/sparc/sparc.c: Include cfgrtl.h.
8704 (TARGET_INIT_PIC_REG): Define.
8705 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
8706 (sparc_pic_register_p): New predicate.
8707 (sparc_legitimate_address_p): Use it.
8708 (sparc_legitimize_pic_address): Likewise.
8709 (sparc_delegitimize_address): Likewise.
8710 (sparc_mode_dependent_address_p): Likewise.
8711 (gen_load_pcrel_sym): Remove 4th parameter.
8712 (load_got_register): Adjust call to above. Remove obsolete stuff.
8713 (sparc_expand_prologue): Do not call load_got_register here.
8714 (sparc_flat_expand_prologue): Likewise.
8715 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
8716 (sparc_use_pseudo_pic_reg): New function.
8717 (sparc_init_pic_reg): Likewise.
8718 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8719 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
8721 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
8723 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
8724 Add item for branch_cost.
8726 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8728 PR rtl-optimization/83565
8729 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
8730 not extend the result to a larger mode for rotate operations.
8731 (num_sign_bit_copies1): Likewise.
8733 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8736 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
8738 Use values-Xc.o for -pedantic.
8739 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
8741 2018-01-12 Martin Liska <mliska@suse.cz>
8744 * ipa-devirt.c (final_warning_record::grow_type_warnings):
8746 (possible_polymorphic_call_targets): Use it.
8747 (ipa_devirt): Likewise.
8749 2018-01-12 Martin Liska <mliska@suse.cz>
8751 * profile-count.h (enum profile_quality): Use 0 as invalid
8752 enum value of profile_quality.
8754 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
8756 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
8757 -mext-string options.
8759 2018-01-12 Richard Biener <rguenther@suse.de>
8761 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
8762 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
8763 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
8765 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
8767 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
8769 * configure.ac (--with-long-double-format): Add support for the
8770 configuration option to change the default long double format on
8772 * config.gcc (powerpc*-linux*-*): Likewise.
8773 * configure: Regenerate.
8774 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
8775 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
8776 used without modification.
8778 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8780 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
8781 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
8782 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
8783 MISC_BUILTIN_SPEC_BARRIER.
8784 (rs6000_init_builtins): Likewise.
8785 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
8787 (speculation_barrier): New define_insn.
8788 * doc/extend.texi: Document __builtin_speculation_barrier.
8790 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8793 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
8794 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
8795 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
8797 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
8798 integral modes instead of "ss" and "sd".
8799 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
8800 vectors with 32-bit and 64-bit elements.
8801 (vecdupssescalarmodesuffix): New mode attribute.
8802 (vec_dup<mode>): Use it.
8804 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
8807 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
8808 frame if argument is passed on stack.
8810 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8813 * ree.c (combine_reaching_defs): Optimize also
8814 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
8815 reg2=any_extend(exp); reg1=reg2;, formatting fix.
8817 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8820 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
8822 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8825 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
8826 after they are computed.
8828 2018-01-11 Bin Cheng <bin.cheng@arm.com>
8830 PR tree-optimization/83695
8831 * gimple-loop-linterchange.cc
8832 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
8833 reset cached scev information after interchange.
8834 (pass_linterchange::execute): Remove call to scev_reset_htab.
8836 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8838 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
8839 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
8840 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
8841 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
8842 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
8843 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
8844 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
8845 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
8846 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
8847 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
8848 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
8849 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
8850 (V_lane_reg): Likewise.
8851 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
8853 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
8854 (vfmal_lane_low<mode>_intrinsic,
8855 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
8856 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
8857 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
8858 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
8859 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
8860 vfmsl_lane_high<mode>_intrinsic): New define_insns.
8862 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8864 * config/arm/arm-cpus.in (fp16fml): New feature.
8865 (ALL_SIMD): Add fp16fml.
8866 (armv8.2-a): Add fp16fml as an option.
8867 (armv8.3-a): Likewise.
8868 (armv8.4-a): Add fp16fml as part of fp16.
8869 * config/arm/arm.h (TARGET_FP16FML): Define.
8870 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
8872 * config/arm/arm-modes.def (V2HF): Define.
8873 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
8874 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
8875 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
8876 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
8877 vfmsl_low, vfmsl_high): New set of builtins.
8878 * config/arm/iterators.md (PLUSMINUS): New code iterator.
8879 (vfml_op): New code attribute.
8880 (VFMLHALVES): New int iterator.
8881 (VFML, VFMLSEL): New mode attributes.
8882 (V_reg): Define mapping for V2HF.
8883 (V_hi, V_lo): New mode attributes.
8884 (VF_constraint): Likewise.
8885 (vfml_half, vfml_half_selector): New int attributes.
8886 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
8888 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
8889 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
8891 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
8892 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
8893 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
8894 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
8896 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
8897 Document new effective target and option set.
8899 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8901 * config/arm/arm-cpus.in (armv8_4): New feature.
8902 (ARMv8_4a): New fgroup.
8903 (armv8.4-a): New arch.
8904 * config/arm/arm-tables.opt: Regenerate.
8905 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
8906 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
8907 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
8908 Add matching rules for -march=armv8.4-a and extensions.
8909 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
8911 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
8914 * config/rx/rx.md (BW): New mode attribute.
8915 (sync_lock_test_and_setsi): Add mode suffix to insn output.
8917 2018-01-11 Richard Biener <rguenther@suse.de>
8919 PR tree-optimization/83435
8920 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
8921 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
8922 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
8924 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8925 Alan Hayward <alan.hayward@arm.com>
8926 David Sherwood <david.sherwood@arm.com>
8928 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
8930 (aarch64_classify_address): Initialize it. Track polynomial offsets.
8931 (aarch64_print_address_internal): Use it to check for a zero offset.
8933 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8934 Alan Hayward <alan.hayward@arm.com>
8935 David Sherwood <david.sherwood@arm.com>
8937 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
8938 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
8939 Return a poly_int64 rather than a HOST_WIDE_INT.
8940 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
8941 rather than a HOST_WIDE_INT.
8942 * config/aarch64/aarch64.h (aarch64_frame): Protect with
8943 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
8944 hard_fp_offset, frame_size, initial_adjust, callee_offset and
8945 final_offset from HOST_WIDE_INT to poly_int64.
8946 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
8947 to_constant when getting the number of units in an Advanced SIMD
8949 (aarch64_builtin_vectorized_function): Check for a constant number
8951 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
8953 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
8954 attribute instead of GET_MODE_NUNITS.
8955 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
8956 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
8957 GET_MODE_SIZE for fixed-size registers.
8958 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
8959 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
8960 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
8961 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
8962 (aarch64_print_operand, aarch64_print_address_internal)
8963 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
8964 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
8965 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
8966 Handle polynomial GET_MODE_SIZE.
8967 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
8968 wider than SImode without modification.
8969 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
8970 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
8971 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
8972 passing and returning SVE modes.
8973 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
8974 rather than GEN_INT.
8975 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
8976 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
8977 (aarch64_allocate_and_probe_stack_space): Likewise.
8978 (aarch64_layout_frame): Cope with polynomial offsets.
8979 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
8980 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
8982 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
8983 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
8984 poly_int64 rather than a HOST_WIDE_INT.
8985 (aarch64_get_separate_components, aarch64_process_components)
8986 (aarch64_expand_prologue, aarch64_expand_epilogue)
8987 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
8988 (aarch64_anchor_offset): New function, split out from...
8989 (aarch64_legitimize_address): ...here.
8990 (aarch64_builtin_vectorization_cost): Handle polynomial
8991 TYPE_VECTOR_SUBPARTS.
8992 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
8994 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
8995 number of elements from the PARALLEL rather than the mode.
8996 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
8997 rather than GET_MODE_BITSIZE.
8998 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
8999 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
9000 (aarch64_expand_vec_perm_const_1): Handle polynomial
9001 d->perm.length () and d->perm elements.
9002 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
9003 Apply to_constant to d->perm elements.
9004 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
9005 polynomial CONST_VECTOR_NUNITS.
9006 (aarch64_move_pointer): Take amount as a poly_int64 rather
9008 (aarch64_progress_pointer): Avoid temporary variable.
9009 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
9010 the mode attribute instead of GET_MODE.
9012 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9013 Alan Hayward <alan.hayward@arm.com>
9014 David Sherwood <david.sherwood@arm.com>
9016 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
9017 x exists before using it.
9018 (aarch64_add_constant_internal): Rename to...
9019 (aarch64_add_offset_1): ...this. Replace regnum with separate
9020 src and dest rtxes. Handle the case in which they're different,
9021 including when the offset is zero. Replace scratchreg with an rtx.
9022 Use 2 additions if there is no spare register into which we can
9023 move a 16-bit constant.
9024 (aarch64_add_constant): Delete.
9025 (aarch64_add_offset): Replace reg with separate src and dest
9026 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
9027 Use aarch64_add_offset_1.
9028 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
9029 an rtx rather than an int. Take the delta as a poly_int64
9030 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
9031 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
9032 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
9033 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
9034 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
9036 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
9037 aarch64_add_constant.
9039 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9041 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
9042 Use scalar_float_mode.
9044 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9046 * config/aarch64/aarch64-simd.md
9047 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
9048 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
9049 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
9050 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
9051 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
9052 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
9053 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
9054 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
9055 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
9056 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
9058 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9061 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
9062 targ_options->x_arm_arch_string is non NULL.
9064 2018-01-11 Tamar Christina <tamar.christina@arm.com>
9066 * config/aarch64/aarch64.h
9067 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
9069 2018-01-11 Sudakshina Das <sudi.das@arm.com>
9072 * expmed.c (emit_store_flag_force): Swap if const op0
9073 and change VOIDmode to mode of op0.
9075 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9077 PR rtl-optimization/83761
9078 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
9079 than bytes to mode_for_size.
9081 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9084 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
9085 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
9088 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9091 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
9092 when in layout mode.
9093 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
9094 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
9097 2018-01-10 Michael Collison <michael.collison@arm.com>
9099 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
9100 * config/aarch64/aarch64-option-extension.def: Add
9101 AARCH64_OPT_EXTENSION of 'fp16fml'.
9102 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9103 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
9104 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
9105 * config/aarch64/constraints.md (Ui7): New constraint.
9106 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
9107 (VFMLA_SEL_W): Ditto.
9110 (VFMLA16_LOW): New int iterator.
9111 (VFMLA16_HIGH): Ditto.
9112 (UNSPEC_FMLAL): New unspec.
9113 (UNSPEC_FMLSL): Ditto.
9114 (UNSPEC_FMLAL2): Ditto.
9115 (UNSPEC_FMLSL2): Ditto.
9116 (f16mac): New code attribute.
9117 * config/aarch64/aarch64-simd-builtins.def
9118 (aarch64_fmlal_lowv2sf): Ditto.
9119 (aarch64_fmlsl_lowv2sf): Ditto.
9120 (aarch64_fmlalq_lowv4sf): Ditto.
9121 (aarch64_fmlslq_lowv4sf): Ditto.
9122 (aarch64_fmlal_highv2sf): Ditto.
9123 (aarch64_fmlsl_highv2sf): Ditto.
9124 (aarch64_fmlalq_highv4sf): Ditto.
9125 (aarch64_fmlslq_highv4sf): Ditto.
9126 (aarch64_fmlal_lane_lowv2sf): Ditto.
9127 (aarch64_fmlsl_lane_lowv2sf): Ditto.
9128 (aarch64_fmlal_laneq_lowv2sf): Ditto.
9129 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
9130 (aarch64_fmlalq_lane_lowv4sf): Ditto.
9131 (aarch64_fmlsl_lane_lowv4sf): Ditto.
9132 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
9133 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
9134 (aarch64_fmlal_lane_highv2sf): Ditto.
9135 (aarch64_fmlsl_lane_highv2sf): Ditto.
9136 (aarch64_fmlal_laneq_highv2sf): Ditto.
9137 (aarch64_fmlsl_laneq_highv2sf): Ditto.
9138 (aarch64_fmlalq_lane_highv4sf): Ditto.
9139 (aarch64_fmlsl_lane_highv4sf): Ditto.
9140 (aarch64_fmlalq_laneq_highv4sf): Ditto.
9141 (aarch64_fmlsl_laneq_highv4sf): Ditto.
9142 * config/aarch64/aarch64-simd.md:
9143 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
9144 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9145 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
9146 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9147 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
9148 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
9149 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
9150 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
9151 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
9152 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
9153 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
9154 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
9155 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
9156 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
9157 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
9158 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
9159 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
9160 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
9161 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
9162 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
9163 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
9164 (vfmlsl_low_u32): Ditto.
9165 (vfmlalq_low_u32): Ditto.
9166 (vfmlslq_low_u32): Ditto.
9167 (vfmlal_high_u32): Ditto.
9168 (vfmlsl_high_u32): Ditto.
9169 (vfmlalq_high_u32): Ditto.
9170 (vfmlslq_high_u32): Ditto.
9171 (vfmlal_lane_low_u32): Ditto.
9172 (vfmlsl_lane_low_u32): Ditto.
9173 (vfmlal_laneq_low_u32): Ditto.
9174 (vfmlsl_laneq_low_u32): Ditto.
9175 (vfmlalq_lane_low_u32): Ditto.
9176 (vfmlslq_lane_low_u32): Ditto.
9177 (vfmlalq_laneq_low_u32): Ditto.
9178 (vfmlslq_laneq_low_u32): Ditto.
9179 (vfmlal_lane_high_u32): Ditto.
9180 (vfmlsl_lane_high_u32): Ditto.
9181 (vfmlal_laneq_high_u32): Ditto.
9182 (vfmlsl_laneq_high_u32): Ditto.
9183 (vfmlalq_lane_high_u32): Ditto.
9184 (vfmlslq_lane_high_u32): Ditto.
9185 (vfmlalq_laneq_high_u32): Ditto.
9186 (vfmlslq_laneq_high_u32): Ditto.
9187 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
9188 (AARCH64_FL_FOR_ARCH8_4): New.
9189 (AARCH64_ISA_F16FML): New ISA flag.
9190 (TARGET_F16FML): New feature flag for fp16fml.
9191 (doc/invoke.texi): Document new fp16fml option.
9193 2018-01-10 Michael Collison <michael.collison@arm.com>
9195 * config/aarch64/aarch64-builtins.c:
9196 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
9197 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9198 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
9199 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
9200 (AARCH64_ISA_SHA3): New ISA flag.
9201 (TARGET_SHA3): New feature flag for sha3.
9202 * config/aarch64/iterators.md (sha512_op): New int attribute.
9203 (CRYPTO_SHA512): New int iterator.
9204 (UNSPEC_SHA512H): New unspec.
9205 (UNSPEC_SHA512H2): Ditto.
9206 (UNSPEC_SHA512SU0): Ditto.
9207 (UNSPEC_SHA512SU1): Ditto.
9208 * config/aarch64/aarch64-simd-builtins.def
9209 (aarch64_crypto_sha512hqv2di): New builtin.
9210 (aarch64_crypto_sha512h2qv2di): Ditto.
9211 (aarch64_crypto_sha512su0qv2di): Ditto.
9212 (aarch64_crypto_sha512su1qv2di): Ditto.
9213 (aarch64_eor3qv8hi): Ditto.
9214 (aarch64_rax1qv2di): Ditto.
9215 (aarch64_xarqv2di): Ditto.
9216 (aarch64_bcaxqv8hi): Ditto.
9217 * config/aarch64/aarch64-simd.md:
9218 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
9219 (aarch64_crypto_sha512su0qv2di): Ditto.
9220 (aarch64_crypto_sha512su1qv2di): Ditto.
9221 (aarch64_eor3qv8hi): Ditto.
9222 (aarch64_rax1qv2di): Ditto.
9223 (aarch64_xarqv2di): Ditto.
9224 (aarch64_bcaxqv8hi): Ditto.
9225 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
9226 (vsha512h2q_u64): Ditto.
9227 (vsha512su0q_u64): Ditto.
9228 (vsha512su1q_u64): Ditto.
9229 (veor3q_u16): Ditto.
9230 (vrax1q_u64): Ditto.
9232 (vbcaxq_u16): Ditto.
9233 * config/arm/types.md (crypto_sha512): New type attribute.
9234 (crypto_sha3): Ditto.
9235 (doc/invoke.texi): Document new sha3 option.
9237 2018-01-10 Michael Collison <michael.collison@arm.com>
9239 * config/aarch64/aarch64-builtins.c:
9240 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
9241 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9242 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
9243 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
9244 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
9245 (AARCH64_ISA_SM4): New ISA flag.
9246 (TARGET_SM4): New feature flag for sm4.
9247 * config/aarch64/aarch64-simd-builtins.def
9248 (aarch64_sm3ss1qv4si): Ditto.
9249 (aarch64_sm3tt1aq4si): Ditto.
9250 (aarch64_sm3tt1bq4si): Ditto.
9251 (aarch64_sm3tt2aq4si): Ditto.
9252 (aarch64_sm3tt2bq4si): Ditto.
9253 (aarch64_sm3partw1qv4si): Ditto.
9254 (aarch64_sm3partw2qv4si): Ditto.
9255 (aarch64_sm4eqv4si): Ditto.
9256 (aarch64_sm4ekeyqv4si): Ditto.
9257 * config/aarch64/aarch64-simd.md:
9258 (aarch64_sm3ss1qv4si): Ditto.
9259 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
9260 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
9261 (aarch64_sm4eqv4si): Ditto.
9262 (aarch64_sm4ekeyqv4si): Ditto.
9263 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
9264 (sm3part_op): Ditto.
9265 (CRYPTO_SM3TT): Ditto.
9266 (CRYPTO_SM3PART): Ditto.
9267 (UNSPEC_SM3SS1): New unspec.
9268 (UNSPEC_SM3TT1A): Ditto.
9269 (UNSPEC_SM3TT1B): Ditto.
9270 (UNSPEC_SM3TT2A): Ditto.
9271 (UNSPEC_SM3TT2B): Ditto.
9272 (UNSPEC_SM3PARTW1): Ditto.
9273 (UNSPEC_SM3PARTW2): Ditto.
9274 (UNSPEC_SM4E): Ditto.
9275 (UNSPEC_SM4EKEY): Ditto.
9276 * config/aarch64/constraints.md (Ui2): New constraint.
9277 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
9278 * config/arm/types.md (crypto_sm3): New type attribute.
9279 (crypto_sm4): Ditto.
9280 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
9281 (vsm3tt1aq_u32): Ditto.
9282 (vsm3tt1bq_u32): Ditto.
9283 (vsm3tt2aq_u32): Ditto.
9284 (vsm3tt2bq_u32): Ditto.
9285 (vsm3partw1q_u32): Ditto.
9286 (vsm3partw2q_u32): Ditto.
9287 (vsm4eq_u32): Ditto.
9288 (vsm4ekeyq_u32): Ditto.
9289 (doc/invoke.texi): Document new sm4 option.
9291 2018-01-10 Michael Collison <michael.collison@arm.com>
9293 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
9294 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
9295 (AARCH64_FL_FOR_ARCH8_4): New.
9296 (AARCH64_FL_V8_4): New flag.
9297 (doc/invoke.texi): Document new armv8.4-a option.
9299 2018-01-10 Michael Collison <michael.collison@arm.com>
9301 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9302 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
9303 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
9304 * config/aarch64/aarch64-option-extension.def: Add
9305 AARCH64_OPT_EXTENSION of 'sha2'.
9306 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
9307 (crypto): Disable sha2 and aes if crypto disabled.
9308 (crypto): Enable aes and sha2 if enabled.
9309 (simd): Disable sha2 and aes if simd disabled.
9310 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
9312 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
9313 (TARGET_SHA2): New feature flag for sha2.
9314 (TARGET_AES): New feature flag for aes.
9315 * config/aarch64/aarch64-simd.md:
9316 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
9317 conditional on TARGET_AES.
9318 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
9319 (aarch64_crypto_sha1hsi): Make pattern conditional
9321 (aarch64_crypto_sha1hv4si): Ditto.
9322 (aarch64_be_crypto_sha1hv4si): Ditto.
9323 (aarch64_crypto_sha1su1v4si): Ditto.
9324 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
9325 (aarch64_crypto_sha1su0v4si): Ditto.
9326 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
9327 (aarch64_crypto_sha256su0v4si): Ditto.
9328 (aarch64_crypto_sha256su1v4si): Ditto.
9329 (doc/invoke.texi): Document new aes and sha2 options.
9331 2018-01-10 Martin Sebor <msebor@redhat.com>
9333 PR tree-optimization/83781
9334 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
9337 2018-01-11 Martin Sebor <msebor@gmail.com>
9338 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9340 PR tree-optimization/83501
9341 PR tree-optimization/81703
9343 * tree-ssa-strlen.c (get_string_cst): Rename...
9344 (get_string_len): ...to this. Handle global constants.
9345 (handle_char_store): Adjust.
9347 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
9348 Jim Wilson <jimw@sifive.com>
9350 * config/riscv/riscv-protos.h (riscv_output_return): New.
9351 * config/riscv/riscv.c (struct machine_function): New naked_p field.
9352 (riscv_attribute_table, riscv_output_return),
9353 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
9354 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
9355 (riscv_compute_frame_info): Only compute frame->mask if not a naked
9357 (riscv_expand_prologue): Add early return for naked function.
9358 (riscv_expand_epilogue): Likewise.
9359 (riscv_function_ok_for_sibcall): Return false for naked function.
9360 (riscv_set_current_function): New.
9361 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
9362 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
9363 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
9364 * doc/extend.texi (RISC-V Function Attributes): New.
9366 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
9368 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
9369 check for 128-bit long double before checking TCmode.
9370 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
9371 128-bit long doubles before checking TFmode or TCmode.
9372 (FLOAT128_IBM_P): Likewise.
9374 2018-01-10 Martin Sebor <msebor@redhat.com>
9376 PR tree-optimization/83671
9377 * builtins.c (c_strlen): Unconditionally return zero for the empty
9379 Use -Warray-bounds for warnings.
9380 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
9381 for non-constant array indices with COMPONENT_REF, arrays of
9382 arrays, and pointers to arrays.
9383 (gimple_fold_builtin_strlen): Determine and set length range for
9384 non-constant character arrays.
9386 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
9389 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
9392 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
9394 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
9396 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9399 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
9400 VECTOR_MEM_ALTIVEC_OR_VSX_P.
9401 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
9402 indexed_or_indirect_operand predicate.
9403 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
9404 (*vsx_le_perm_load_v8hi): Likewise.
9405 (*vsx_le_perm_load_v16qi): Likewise.
9406 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
9407 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
9408 (*vsx_le_perm_store_v8hi): Likewise.
9409 (*vsx_le_perm_store_v16qi): Likewise.
9410 (eight unnamed splitters): Likewise.
9412 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9414 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
9415 * config/rs6000/emmintrin.h: Likewise.
9416 * config/rs6000/mmintrin.h: Likewise.
9417 * config/rs6000/xmmintrin.h: Likewise.
9419 2018-01-10 David Malcolm <dmalcolm@redhat.com>
9422 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
9424 * tree.c (tree_nop_conversion): Return true for location wrapper
9426 (maybe_wrap_with_location): New function.
9427 (selftest::check_strip_nops): New function.
9428 (selftest::test_location_wrappers): New function.
9429 (selftest::tree_c_tests): Call it.
9430 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
9431 (maybe_wrap_with_location): New decl.
9432 (EXPR_LOCATION_WRAPPER_P): New macro.
9433 (location_wrapper_p): New inline function.
9434 (tree_strip_any_location_wrapper): New inline function.
9436 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
9439 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
9440 stack_realign_offset for the largest alignment of stack slot
9442 (ix86_find_max_used_stack_alignment): New function.
9443 (ix86_finalize_stack_frame_flags): Use it. Set
9444 max_used_stack_alignment if we don't realign stack.
9445 * config/i386/i386.h (machine_function): Add
9446 max_used_stack_alignment.
9448 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
9450 * config/arm/arm.opt (-mbranch-cost): New option.
9451 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
9454 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
9457 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
9458 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
9460 2018-01-10 Richard Biener <rguenther@suse.de>
9463 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
9464 early out so it also covers the case where we have a non-NULL
9467 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9469 PR tree-optimization/83753
9470 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
9471 for non-strided grouped accesses if the number of elements is 1.
9473 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9476 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
9477 * i386.h (TARGET_USE_GATHER): Define.
9478 * x86-tune.def (X86_TUNE_USE_GATHER): New.
9480 2018-01-10 Martin Liska <mliska@suse.cz>
9483 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
9484 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
9486 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
9487 CLEANUP_NO_PARTITIONING is not set.
9489 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9491 * doc/rtl.texi: Remove documentation of (const ...) wrappers
9492 for vectors, as a partial revert of r254296.
9493 * rtl.h (const_vec_p): Delete.
9494 (const_vec_duplicate_p): Don't test for vector CONSTs.
9495 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
9496 * expmed.c (make_tree): Likewise.
9499 * common.md (E, F): Use CONSTANT_P instead of checking for
9501 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9502 checking for CONST_VECTOR.
9504 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9507 * predict.c (force_edge_cold): Handle in more sane way edges
9510 2018-01-09 Carl Love <cel@us.ibm.com>
9512 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
9514 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
9515 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
9516 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
9517 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
9518 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
9519 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
9520 * config/rs6000/rs6000-protos.h: Add extern defition for
9521 rs6000_generate_float2_double_code.
9522 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
9524 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
9525 (float2_v2df): Add define_expand.
9527 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
9530 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
9531 op_mode in the force_to_mode call.
9533 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9535 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
9536 instead of checking each element individually.
9537 (aarch64_evpc_uzp): Likewise.
9538 (aarch64_evpc_zip): Likewise.
9539 (aarch64_evpc_ext): Likewise.
9540 (aarch64_evpc_rev): Likewise.
9541 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
9542 instead of checking each element individually. Return true without
9544 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
9545 whether all selected elements come from the same input, instead of
9546 checking each element individually. Remove calls to gen_rtx_REG,
9547 start_sequence and end_sequence and instead assert that no rtl is
9550 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9552 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
9553 order of HIGH and CONST checks.
9555 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9557 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
9558 if the destination isn't an SSA_NAME.
9560 2018-01-09 Richard Biener <rguenther@suse.de>
9562 PR tree-optimization/83668
9563 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
9565 (canonicalize_loop_form): ... here, renamed from ...
9566 (canonicalize_loop_closed_ssa_form): ... this and amended to
9567 swap successor edges for loop exit blocks to make us use
9568 the RPO order we need for initial schedule generation.
9570 2018-01-09 Joseph Myers <joseph@codesourcery.com>
9572 PR tree-optimization/64811
9573 * match.pd: When optimizing comparisons with Inf, avoid
9574 introducing or losing exceptions from comparisons with NaN.
9576 2018-01-09 Martin Liska <mliska@suse.cz>
9579 * asan.c (shadow_mem_size): Add gcc_assert.
9581 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
9583 Don't save registers in main().
9586 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
9587 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
9588 * config/avr/avr.c (avr_set_current_function): Don't error if
9589 naked, OS_task or OS_main are specified at the same time.
9590 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
9592 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
9594 * common/config/avr/avr-common.c (avr_option_optimization_table):
9595 Switch on -mmain-is-OS_task for optimizing compilations.
9597 2018-01-09 Richard Biener <rguenther@suse.de>
9599 PR tree-optimization/83572
9600 * graphite.c: Include cfganal.h.
9601 (graphite_transform_loops): Connect infinite loops to exit
9602 and remove fake edges at the end.
9604 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9606 * ipa-inline.c (edge_badness): Revert accidental checkin.
9608 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9611 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
9612 symbols; not inline clones.
9614 2018-01-09 Jakub Jelinek <jakub@redhat.com>
9617 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
9618 hard registers. Formatting fixes.
9620 PR preprocessor/83722
9621 * gcc.c (try_generate_repro): Pass
9622 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
9623 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
9626 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
9627 Kito Cheng <kito.cheng@gmail.com>
9629 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
9630 (riscv_leaf_function_p): Delete.
9631 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
9633 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9635 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
9637 (do_ifelse): New function.
9638 (do_isel): New function.
9639 (do_sub3): New function.
9640 (do_add3): New function.
9641 (do_load_mask_compare): New function.
9642 (do_overlap_load_compare): New function.
9643 (expand_compare_loop): New function.
9644 (expand_block_compare): Call expand_compare_loop() when appropriate.
9645 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
9647 (-mblock-compare-inline-loop-limit): New option.
9649 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9652 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
9653 Reverse order of second and third operands in first alternative.
9654 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
9655 of first and second elements in UNSPEC_VPERMR vector.
9656 (altivec_expand_vec_perm_le): Likewise.
9658 2018-01-08 Jeff Law <law@redhat.com>
9660 PR rtl-optimizatin/81308
9661 * tree-switch-conversion.c (cfg_altered): New file scoped static.
9662 (process_switch): If group_case_labels makes a change, then set
9664 (pass_convert_switch::execute): If a switch is converted, then
9665 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
9667 PR rtl-optimization/81308
9668 * recog.c (split_all_insns): Conditionally cleanup the CFG after
9671 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
9673 PR target/83663 - Revert r255946
9674 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
9675 generation for cases where splatting a value is not useful.
9676 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
9677 across a vec_duplicate and a paradoxical subreg forming a vector
9678 mode to a vec_concat.
9680 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9682 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
9683 -march=armv8.3-a variants.
9684 * config/arm/t-multilib: Likewise.
9685 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
9687 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9689 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
9691 (cceq_ior_compare_complement): Give it a name so I can use it, and
9692 change boolean_or_operator predicate to boolean_operator so it can
9693 be used to generate a crand.
9694 (eqne): New code iterator.
9695 (bd/bd_neg): New code_attrs.
9696 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
9697 a single define_insn.
9698 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
9699 decrement (bdnzt/bdnzf/bdzt/bdzf).
9700 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
9701 with the new names of the branch decrement patterns, and added the
9702 names of the branch decrement conditional patterns.
9704 2018-01-08 Richard Biener <rguenther@suse.de>
9706 PR tree-optimization/83563
9707 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
9710 2018-01-08 Richard Biener <rguenther@suse.de>
9713 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
9715 2018-01-08 Richard Biener <rguenther@suse.de>
9717 PR tree-optimization/83685
9718 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
9719 references to abnormals.
9721 2018-01-08 Richard Biener <rguenther@suse.de>
9724 * dwarf2out.c (output_indirect_strings): Handle empty
9725 skeleton_debug_str_hash.
9726 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
9728 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9730 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
9731 (emit_store_direct): Likewise.
9732 (arc_trampoline_adjust_address): Likewise.
9733 (arc_asm_trampoline_template): New function.
9734 (arc_initialize_trampoline): Use asm_trampoline_template.
9735 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
9736 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
9737 * config/arc/arc.md (flush_icache): Delete pattern.
9739 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9741 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
9742 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
9745 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9748 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
9749 by not USED_FOR_TARGET.
9750 (make_pass_resolve_sw_modes): Likewise.
9752 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9754 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
9757 2018-01-08 Richard Biener <rguenther@suse.de>
9760 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
9762 2018-01-08 Richard Biener <rguenther@suse.de>
9765 * match.pd ((t * 2) / 2) -> t): Add missing :c.
9767 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
9770 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
9771 basic blocks with a small number of successors.
9772 (convert_control_dep_chain_into_preds): Improve handling of
9774 (dump_predicates): Split apart into...
9775 (dump_pred_chain): ...here...
9776 (dump_pred_info): ...and here.
9777 (can_one_predicate_be_invalidated_p): Add debugging printfs.
9778 (can_chain_union_be_invalidated_p): Improve check for invalidation
9780 (uninit_uses_cannot_happen): Avoid unnecessary if
9781 convert_control_dep_chain_into_preds yielded nothing.
9783 2018-01-06 Martin Sebor <msebor@redhat.com>
9785 PR tree-optimization/83640
9786 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
9787 subtracting negative offset from size.
9788 (builtin_access::overlap): Adjust offset bounds of the access to fall
9789 within the size of the object if possible.
9791 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
9793 PR rtl-optimization/83699
9794 * expmed.c (extract_bit_field_1): Restrict the vector usage of
9795 extract_bit_field_as_subreg to cases in which the extracted
9796 value is also a vector.
9798 * lra-constraints.c (process_alt_operands): Test for the equivalence
9799 substitutions when detecting a possible reload cycle.
9801 2018-01-06 Jakub Jelinek <jakub@redhat.com>
9804 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
9805 by default if flag_selective_schedling{,2}. Formatting fixes.
9807 PR rtl-optimization/83682
9808 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
9809 if it has non-VECTOR_MODE element mode.
9810 (vec_duplicate_p): Likewise.
9813 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
9814 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
9816 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9819 * config/i386/i386-builtin.def
9820 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
9821 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
9822 Require also OPTION_MASK_ISA_AVX512F in addition to
9823 OPTION_MASK_ISA_GFNI.
9824 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
9825 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
9826 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
9827 to OPTION_MASK_ISA_GFNI.
9828 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
9829 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
9830 OPTION_MASK_ISA_AVX512BW.
9831 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
9832 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
9833 addition to OPTION_MASK_ISA_GFNI.
9834 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
9835 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
9836 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
9837 to OPTION_MASK_ISA_GFNI.
9838 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
9839 a requirement for all ISAs rather than any of them with a few
9841 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
9843 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
9844 bitmasks to be enabled with 3 exceptions, instead of requiring any
9845 enabled ISA with lots of exceptions.
9846 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
9847 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
9848 Change avx512bw in isa attribute to avx512f.
9849 * config/i386/sgxintrin.h: Add license boilerplate.
9850 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
9851 to __AVX512F__ and __AVX512VL to __AVX512VL__.
9852 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
9853 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
9855 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
9856 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
9857 temporarily sse2 rather than sse if not enabled already.
9860 * config/i386/sse.md (VI248_VLBW): Rename to ...
9861 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
9862 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
9863 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
9864 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
9865 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
9866 mode iterator instead of VI248_VLBW.
9868 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
9870 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
9871 (record_modified): Skip clobbers; add debug output.
9872 (param_change_prob): Use sreal frequencies.
9874 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9876 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
9877 punt for user-aligned variables.
9879 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9881 * tree-chrec.c (chrec_contains_symbols): Return true for
9884 2018-01-05 Sudakshina Das <sudi.das@arm.com>
9887 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
9888 of (x|y) == x for BICS pattern.
9890 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9892 PR tree-optimization/83605
9893 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
9894 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
9897 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
9899 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
9900 * config/epiphany/rtems.h: New file.
9902 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9903 Uros Bizjak <ubizjak@gmail.com>
9906 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
9907 QIreg_operand instead of register_operand predicate.
9908 * config/i386/i386.c (ix86_rop_should_change_byte_p,
9909 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
9910 comments instead of -fmitigate[-_]rop.
9912 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9915 * cgraphunit.c (symbol_table::compile): Switch to text_section
9916 before calling assembly_start debug hook.
9917 * run-rtl-passes.c (run_rtl_passes): Likewise.
9920 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9922 * tree-vrp.c (extract_range_from_binary_expr_1): Check
9923 range_int_cst_p rather than !symbolic_range_p before calling
9924 extract_range_from_multiplicative_op_1.
9926 2018-01-04 Jeff Law <law@redhat.com>
9928 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
9929 redundant test in assertion.
9931 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9933 * doc/rtl.texi: Document machine_mode wrapper classes.
9935 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9937 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
9940 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9942 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
9943 the VEC_PERM_EXPR fold to fail.
9945 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9948 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
9949 to switched_sections.
9951 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9954 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
9957 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
9960 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
9961 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
9963 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9966 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
9967 is BLKmode and bitpos not zero or mode change is needed.
9969 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9972 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
9975 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
9978 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
9979 instead of MULT rtx. Update all corresponding splitters.
9981 (*ssub<modesuffix>): Ditto.
9983 (*cmp_sadd_di): Update split patterns.
9984 (*cmp_sadd_si): Ditto.
9985 (*cmp_sadd_sidi): Ditto.
9986 (*cmp_ssub_di): Ditto.
9987 (*cmp_ssub_si): Ditto.
9988 (*cmp_ssub_sidi): Ditto.
9989 * config/alpha/predicates.md (const23_operand): New predicate.
9990 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
9991 Look for ASHIFT, not MULT inner operand.
9992 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
9994 2018-01-04 Martin Liska <mliska@suse.cz>
9996 PR gcov-profile/83669
9997 * gcov.c (output_intermediate_file): Add version to intermediate
9999 * doc/gcov.texi: Document new field 'version' in intermediate
10000 file format. Fix location of '-k' option of gcov command.
10002 2018-01-04 Martin Liska <mliska@suse.cz>
10005 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
10007 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10009 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
10011 2018-01-03 Martin Sebor <msebor@redhat.com>
10013 PR tree-optimization/83655
10014 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
10015 checking calls with invalid arguments.
10017 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10019 * tree-vect-stmts.c (vect_get_store_rhs): New function.
10020 (vectorizable_mask_load_store): Delete.
10021 (vectorizable_call): Return false for masked loads and stores.
10022 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
10023 instead of gimple_assign_rhs1.
10024 (vectorizable_load): Handle IFN_MASK_LOAD.
10025 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
10027 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10029 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
10031 (vectorizable_mask_load_store): ...here.
10032 (vectorizable_load): ...and here.
10034 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10036 * tree-vect-stmts.c (vect_build_all_ones_mask)
10037 (vect_build_zero_merge_argument): New functions, split out from...
10038 (vectorizable_load): ...here.
10040 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10042 * tree-vect-stmts.c (vect_check_store_rhs): New function,
10044 (vectorizable_mask_load_store): ...here.
10045 (vectorizable_store): ...and here.
10047 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10049 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
10051 (vectorizable_mask_load_store): ...here.
10053 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10055 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
10056 (vect_model_store_cost): Take a vec_load_store_type instead of a
10058 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
10059 (vect_model_store_cost): Take a vec_load_store_type instead of a
10061 (vectorizable_mask_load_store): Update accordingly.
10062 (vectorizable_store): Likewise.
10063 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
10065 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10067 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
10068 IFN_MASK_LOAD calls here rather than...
10069 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
10071 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10072 Alan Hayward <alan.hayward@arm.com>
10073 David Sherwood <david.sherwood@arm.com>
10075 * expmed.c (extract_bit_field_1): For vector extracts,
10076 fall back to extract_bit_field_as_subreg if vec_extract
10079 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10080 Alan Hayward <alan.hayward@arm.com>
10081 David Sherwood <david.sherwood@arm.com>
10083 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
10084 they are variable or constant sized.
10085 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
10086 slots for constant-sized data.
10088 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10089 Alan Hayward <alan.hayward@arm.com>
10090 David Sherwood <david.sherwood@arm.com>
10092 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
10093 handling COND_EXPRs with boolean comparisons, try to find a better
10094 basis for the mask type than the boolean itself.
10096 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10098 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
10099 is calculated and how it can be overridden.
10100 * genmodes.c (max_bitsize_mode_any_mode): New variable.
10101 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
10103 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
10106 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10107 Alan Hayward <alan.hayward@arm.com>
10108 David Sherwood <david.sherwood@arm.com>
10110 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
10111 Remove the mode argument.
10112 (aarch64_simd_valid_immediate): Remove the mode and inverse
10114 * config/aarch64/iterators.md (bitsize): New iterator.
10115 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
10116 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
10117 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
10118 aarch64_simd_valid_immediate.
10119 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
10120 (aarch64_reg_or_bic_imm): Likewise.
10121 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
10122 with an insn_type enum and msl with a modifier_type enum.
10123 Replace element_width with a scalar_mode. Change the shift
10124 to unsigned int. Add constructors for scalar_float_mode and
10125 scalar_int_mode elements.
10126 (aarch64_vect_float_const_representable_p): Delete.
10127 (aarch64_can_const_movi_rtx_p)
10128 (aarch64_simd_scalar_immediate_valid_for_move)
10129 (aarch64_simd_make_constant): Update call to
10130 aarch64_simd_valid_immediate.
10131 (aarch64_advsimd_valid_immediate_hs): New function.
10132 (aarch64_advsimd_valid_immediate): Likewise.
10133 (aarch64_simd_valid_immediate): Remove mode and inverse
10134 arguments. Rewrite to use the above. Use const_vec_duplicate_p
10135 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
10136 and aarch64_float_const_representable_p on the result.
10137 (aarch64_output_simd_mov_immediate): Remove mode argument.
10138 Update call to aarch64_simd_valid_immediate and use of
10139 simd_immediate_info.
10140 (aarch64_output_scalar_simd_mov_immediate): Update call
10143 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10144 Alan Hayward <alan.hayward@arm.com>
10145 David Sherwood <david.sherwood@arm.com>
10147 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
10148 (mode_nunits): Likewise CONST_MODE_NUNITS.
10149 * machmode.def (ADJUST_NUNITS): Document.
10150 * genmodes.c (mode_data::need_nunits_adj): New field.
10151 (blank_mode): Update accordingly.
10152 (adj_nunits): New variable.
10153 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
10155 (emit_mode_size_inline): Set need_bytesize_adj for all modes
10156 listed in adj_nunits.
10157 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
10158 listed in adj_nunits. Don't emit case statements for such modes.
10159 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
10160 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
10161 nothing if adj_nunits is nonnull.
10162 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
10163 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
10164 (emit_mode_fbit): Update use of print_maybe_const_decl.
10165 (emit_move_size): Likewise. Treat the array as non-const
10167 (emit_mode_adjustments): Handle adj_nunits.
10169 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10171 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
10172 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
10173 (VECTOR_MODES): Use it.
10174 (make_vector_modes): Take the prefix as an argument.
10176 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10177 Alan Hayward <alan.hayward@arm.com>
10178 David Sherwood <david.sherwood@arm.com>
10180 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
10181 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
10182 for MODE_VECTOR_BOOL.
10183 * machmode.def (VECTOR_BOOL_MODE): Document.
10184 * genmodes.c (VECTOR_BOOL_MODE): New macro.
10185 (make_vector_bool_mode): New function.
10186 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
10188 * lto-streamer-in.c (lto_input_mode_table): Likewise.
10189 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
10191 * stor-layout.c (int_mode_for_mode): Likewise.
10192 * tree.c (build_vector_type_for_mode): Likewise.
10193 * varasm.c (output_constant_pool_2): Likewise.
10194 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
10195 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
10196 for MODE_VECTOR_BOOL.
10197 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
10198 of mode class checks.
10199 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
10200 instead of a list of mode class checks.
10201 (expand_vector_scalar_condition): Likewise.
10202 (type_for_widest_vector_mode): Handle BImode as an inner mode.
10204 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10205 Alan Hayward <alan.hayward@arm.com>
10206 David Sherwood <david.sherwood@arm.com>
10208 * machmode.h (mode_size): Change from unsigned short to
10210 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
10211 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10212 or if measurement_type is not polynomial.
10213 (fixed_size_mode::includes_p): Check for constant-sized modes.
10214 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
10215 return a poly_uint16 rather than an unsigned short.
10216 (emit_mode_size): Change the type of mode_size from unsigned short
10217 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
10218 (emit_mode_adjustments): Cope with polynomial vector sizes.
10219 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10221 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10223 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
10224 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
10225 * caller-save.c (setup_save_areas): Likewise.
10226 (replace_reg_with_saved_mem): Likewise.
10227 * calls.c (emit_library_call_value_1): Likewise.
10228 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
10229 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
10230 (gen_lowpart_for_combine): Likewise.
10231 * convert.c (convert_to_integer_1): Likewise.
10232 * cse.c (equiv_constant, cse_insn): Likewise.
10233 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
10234 (cselib_subst_to_values): Likewise.
10235 * dce.c (word_dce_process_block): Likewise.
10236 * df-problems.c (df_word_lr_mark_ref): Likewise.
10237 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
10238 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
10239 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
10240 (rtl_for_decl_location): Likewise.
10241 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
10242 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
10243 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
10244 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
10245 (expand_expr_real_1): Likewise.
10246 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
10247 (pad_below): Likewise.
10248 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10249 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
10250 * ira.c (get_subreg_tracking_sizes): Likewise.
10251 * ira-build.c (ira_create_allocno_objects): Likewise.
10252 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
10253 (ira_sort_regnos_for_alter_reg): Likewise.
10254 * ira-costs.c (record_operand_costs): Likewise.
10255 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
10256 (resolve_simple_move): Likewise.
10257 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
10258 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
10259 (lra_constraints): Likewise.
10260 (CONST_POOL_OK_P): Reject variable-sized modes.
10261 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
10262 (add_pseudo_to_slot, lra_spill): Likewise.
10263 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10264 * optabs-query.c (get_best_extraction_insn): Likewise.
10265 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10266 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
10267 (expand_mult_highpart, valid_multiword_target_p): Likewise.
10268 * recog.c (offsettable_address_addr_space_p): Likewise.
10269 * regcprop.c (maybe_mode_change): Likewise.
10270 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
10271 * regrename.c (build_def_use): Likewise.
10272 * regstat.c (dump_reg_info): Likewise.
10273 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
10274 (find_reloads, find_reloads_subreg_address): Likewise.
10275 * reload1.c (eliminate_regs_1): Likewise.
10276 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
10277 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
10278 (simplify_binary_operation_1, simplify_subreg): Likewise.
10279 * targhooks.c (default_function_arg_padding): Likewise.
10280 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
10281 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
10282 (verify_gimple_assign_ternary): Likewise.
10283 * tree-inline.c (estimate_move_cost): Likewise.
10284 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10285 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
10286 (get_address_cost_ainc): Likewise.
10287 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
10288 (vect_supportable_dr_alignment): Likewise.
10289 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10290 (vectorizable_reduction): Likewise.
10291 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
10292 (vectorizable_operation, vectorizable_load): Likewise.
10293 * tree.c (build_same_sized_truth_vector_type): Likewise.
10294 * valtrack.c (cleanup_auto_inc_dec): Likewise.
10295 * var-tracking.c (emit_note_insn_var_location): Likewise.
10296 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
10297 (ADDR_VEC_ALIGN): Likewise.
10299 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10300 Alan Hayward <alan.hayward@arm.com>
10301 David Sherwood <david.sherwood@arm.com>
10303 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
10305 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10306 or if measurement_type is polynomial.
10307 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
10308 * combine.c (make_extraction): Likewise.
10309 * dse.c (find_shift_sequence): Likewise.
10310 * dwarf2out.c (mem_loc_descriptor): Likewise.
10311 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
10312 (extract_bit_field, extract_low_bits): Likewise.
10313 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
10314 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
10315 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
10316 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
10317 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10318 * reload.c (find_reloads): Likewise.
10319 * reload1.c (alter_reg): Likewise.
10320 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
10321 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
10322 * tree-if-conv.c (predicate_mem_writes): Likewise.
10323 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
10324 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
10325 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
10326 * valtrack.c (dead_debug_insert_temp): Likewise.
10327 * varasm.c (mergeable_constant_section): Likewise.
10328 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
10330 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10331 Alan Hayward <alan.hayward@arm.com>
10332 David Sherwood <david.sherwood@arm.com>
10334 * expr.c (expand_assignment): Cope with polynomial mode sizes
10335 when assigning to a CONCAT.
10337 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10338 Alan Hayward <alan.hayward@arm.com>
10339 David Sherwood <david.sherwood@arm.com>
10341 * machmode.h (mode_precision): Change from unsigned short to
10343 (mode_to_precision): Return a poly_uint16 rather than an unsigned
10345 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
10346 or if measurement_type is not polynomial.
10347 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
10348 in which the mode is already known to be a scalar_int_mode.
10349 * genmodes.c (emit_mode_precision): Change the type of mode_precision
10350 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
10352 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10353 for GET_MODE_PRECISION.
10354 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10355 for GET_MODE_PRECISION.
10356 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
10358 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
10359 (expand_field_assignment, make_extraction): Likewise.
10360 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
10361 (get_last_value): Likewise.
10362 * convert.c (convert_to_integer_1): Likewise.
10363 * cse.c (cse_insn): Likewise.
10364 * expr.c (expand_expr_real_1): Likewise.
10365 * lra-constraints.c (simplify_operand_subreg): Likewise.
10366 * optabs-query.c (can_atomic_load_p): Likewise.
10367 * optabs.c (expand_atomic_load): Likewise.
10368 (expand_atomic_store): Likewise.
10369 * ree.c (combine_reaching_defs): Likewise.
10370 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
10371 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
10372 * tree.h (type_has_mode_precision_p): Likewise.
10373 * ubsan.c (instrument_si_overflow): 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.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
10380 polynomial numbers of units.
10381 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
10382 (valid_vector_subparts_p): New function.
10383 (build_vector_type): Remove temporary shim and take the number
10384 of units as a poly_uint64 rather than an int.
10385 (build_opaque_vector_type): Take the number of units as a
10386 poly_uint64 rather than an int.
10387 * tree.c (build_vector_from_ctor): Handle polynomial
10388 TYPE_VECTOR_SUBPARTS.
10389 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
10390 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
10391 (build_vector_from_val): If the number of units is variable,
10392 use build_vec_duplicate_cst for constant operands and
10393 VEC_DUPLICATE_EXPR otherwise.
10394 (make_vector_type): Remove temporary is_constant ().
10395 (build_vector_type, build_opaque_vector_type): Take the number of
10396 units as a poly_uint64 rather than an int.
10397 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
10399 * cfgexpand.c (expand_debug_expr): Likewise.
10400 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
10401 (store_constructor, expand_expr_real_1): Likewise.
10402 (const_scalar_mask_from_tree): Likewise.
10403 * fold-const-call.c (fold_const_reduction): Likewise.
10404 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
10405 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
10406 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
10407 (fold_relational_const): Likewise.
10408 (native_interpret_vector): Likewise. Change the size from an
10409 int to an unsigned int.
10410 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
10411 TYPE_VECTOR_SUBPARTS.
10412 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
10413 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
10414 duplicating a non-constant operand into a variable-length vector.
10415 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
10416 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
10417 * ipa-icf.c (sem_variable::equals): Likewise.
10418 * match.pd: Likewise.
10419 * omp-simd-clone.c (simd_clone_subparts): Likewise.
10420 * print-tree.c (print_node): Likewise.
10421 * stor-layout.c (layout_type): Likewise.
10422 * targhooks.c (default_builtin_vectorization_cost): Likewise.
10423 * tree-cfg.c (verify_gimple_comparison): Likewise.
10424 (verify_gimple_assign_binary): Likewise.
10425 (verify_gimple_assign_ternary): Likewise.
10426 (verify_gimple_assign_single): Likewise.
10427 * tree-pretty-print.c (dump_generic_node): Likewise.
10428 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10429 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
10430 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
10431 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
10432 (vect_shift_permute_load_chain): Likewise.
10433 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
10434 (expand_vector_condition, optimize_vector_constructor): Likewise.
10435 (lower_vec_perm, get_compute_type): Likewise.
10436 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10437 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
10438 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
10439 (vect_recog_mask_conversion_pattern): Likewise.
10440 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
10441 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
10442 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10443 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
10444 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
10445 (vectorizable_shift, vectorizable_operation, vectorizable_store)
10446 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
10447 (supportable_widening_operation): Likewise.
10448 (supportable_narrowing_operation): Likewise.
10449 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
10451 * varasm.c (output_constant): Likewise.
10453 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10454 Alan Hayward <alan.hayward@arm.com>
10455 David Sherwood <david.sherwood@arm.com>
10457 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
10458 so that both the length == 3 and length != 3 cases set up their
10459 own permute vectors. Add comments explaining why we know the
10460 number of elements is constant.
10461 (vect_permute_load_chain): Likewise.
10463 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10464 Alan Hayward <alan.hayward@arm.com>
10465 David Sherwood <david.sherwood@arm.com>
10467 * machmode.h (mode_nunits): Change from unsigned char to
10469 (ONLY_FIXED_SIZE_MODES): New macro.
10470 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
10471 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
10472 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
10474 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
10475 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
10476 or if measurement_type is not polynomial.
10477 * genmodes.c (ZERO_COEFFS): New macro.
10478 (emit_mode_nunits_inline): Make mode_nunits_inline return a
10480 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
10481 Use ZERO_COEFFS when emitting initializers.
10482 * data-streamer.h (bp_pack_poly_value): New function.
10483 (bp_unpack_poly_value): Likewise.
10484 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10485 for GET_MODE_NUNITS.
10486 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10487 for GET_MODE_NUNITS.
10488 * tree.c (make_vector_type): Remove temporary shim and make
10489 the real function take the number of units as a poly_uint64
10490 rather than an int.
10491 (build_vector_type_for_mode): Handle polynomial nunits.
10492 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
10493 * emit-rtl.c (const_vec_series_p_1): Likewise.
10494 (gen_rtx_CONST_VECTOR): Likewise.
10495 * fold-const.c (test_vec_duplicate_folding): Likewise.
10496 * genrecog.c (validate_pattern): Likewise.
10497 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
10498 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10499 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
10500 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
10501 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
10502 * rtlanal.c (subreg_get_info): Likewise.
10503 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10504 (vect_grouped_load_supported): Likewise.
10505 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
10506 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
10507 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
10508 (simplify_const_unary_operation, simplify_binary_operation_1)
10509 (simplify_const_binary_operation, simplify_ternary_operation)
10510 (test_vector_ops_duplicate, test_vector_ops): Likewise.
10511 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
10512 instead of CONST_VECTOR_NUNITS.
10513 * varasm.c (output_constant_pool_2): Likewise.
10514 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
10515 explicit-encoded elements in the XVEC for variable-length vectors.
10517 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10519 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
10521 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10522 Alan Hayward <alan.hayward@arm.com>
10523 David Sherwood <david.sherwood@arm.com>
10525 * coretypes.h (fixed_size_mode): Declare.
10526 (fixed_size_mode_pod): New typedef.
10527 * builtins.h (target_builtins::x_apply_args_mode)
10528 (target_builtins::x_apply_result_mode): Change type to
10529 fixed_size_mode_pod.
10530 * builtins.c (apply_args_size, apply_result_size, result_vector)
10531 (expand_builtin_apply_args_1, expand_builtin_apply)
10532 (expand_builtin_return): Update accordingly.
10534 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10536 * cse.c (hash_rtx_cb): Hash only the encoded elements.
10537 * cselib.c (cselib_hash_rtx): Likewise.
10538 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
10539 CONST_VECTOR encoding.
10541 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10542 Jeff Law <law@redhat.com>
10545 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
10546 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
10547 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
10548 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
10551 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
10552 explicitly probe *sp in a noreturn function if there were any callee
10553 register saves or frame pointer is needed.
10555 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10558 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
10559 BLKmode for ternary, binary or unary expressions.
10562 * var-tracking.c (delete_vta_debug_insn): New inline function.
10563 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
10564 insns from get_insns () to NULL instead of each bb separately.
10565 Use delete_vta_debug_insn. No longer static.
10566 (vt_debug_insns_local, variable_tracking_main_1): Adjust
10567 delete_vta_debug_insns callers.
10568 * rtl.h (delete_vta_debug_insns): Declare.
10569 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
10570 instead of variable_tracking_main.
10572 2018-01-03 Martin Sebor <msebor@redhat.com>
10574 PR tree-optimization/83603
10575 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
10576 arguments past the endof the argument list in functions declared
10577 without a prototype.
10578 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
10579 Avoid checking when arguments are null.
10581 2018-01-03 Martin Sebor <msebor@redhat.com>
10584 * doc/extend.texi (attribute const): Fix a typo.
10585 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
10586 issuing -Wsuggest-attribute for void functions.
10588 2018-01-03 Martin Sebor <msebor@redhat.com>
10590 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
10591 offset_int::from instead of wide_int::to_shwi.
10592 (maybe_diag_overlap): Remove assertion.
10593 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
10594 * gimple-ssa-sprintf.c (format_directive): Same.
10595 (parse_directive): Same.
10596 (sprintf_dom_walker::compute_format_length): Same.
10597 (try_substitute_return_value): Same.
10599 2018-01-03 Jeff Law <law@redhat.com>
10601 PR middle-end/83654
10602 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
10603 non-constant residual for zero at runtime and avoid probing in
10604 that case. Reorganize code for trailing problem to mirror handling
10607 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10609 PR tree-optimization/83501
10610 * tree-ssa-strlen.c (get_string_cst): New.
10611 (handle_char_store): Call get_string_cst.
10613 2018-01-03 Martin Liska <mliska@suse.cz>
10615 PR tree-optimization/83593
10616 * tree-ssa-strlen.c: Include tree-cfg.h.
10617 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
10618 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
10619 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
10621 (strlen_dom_walker::before_dom_children): Call
10622 gimple_purge_dead_eh_edges. Dump tranformation with details
10624 (strlen_dom_walker::before_dom_children): Update call by adding
10625 new argument cleanup_eh.
10626 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
10628 2018-01-03 Martin Liska <mliska@suse.cz>
10631 * cif-code.def (VARIADIC_THUNK): New enum value.
10632 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
10635 2018-01-03 Jan Beulich <jbeulich@suse.com>
10637 * sse.md (mov<mode>_internal): Tighten condition for when to use
10638 vmovdqu<ssescalarsize> for TI and OI modes.
10640 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10642 Update copyright years.
10644 2018-01-03 Martin Liska <mliska@suse.cz>
10647 * ipa-visibility.c (function_and_variable_visibility): Skip
10648 functions with noipa attribure.
10650 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10652 * gcc.c (process_command): Update copyright notice dates.
10653 * gcov-dump.c (print_version): Ditto.
10654 * gcov.c (print_version): Ditto.
10655 * gcov-tool.c (print_version): Ditto.
10656 * gengtype.c (create_file): Ditto.
10657 * doc/cpp.texi: Bump @copying's copyright year.
10658 * doc/cppinternals.texi: Ditto.
10659 * doc/gcc.texi: Ditto.
10660 * doc/gccint.texi: Ditto.
10661 * doc/gcov.texi: Ditto.
10662 * doc/install.texi: Ditto.
10663 * doc/invoke.texi: Ditto.
10665 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10667 * vector-builder.h (vector_builder::m_full_nelts): Change from
10668 unsigned int to poly_uint64.
10669 (vector_builder::full_nelts): Update prototype accordingly.
10670 (vector_builder::new_vector): Likewise.
10671 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
10672 (vector_builder::operator ==): Likewise.
10673 (vector_builder::finalize): Likewise.
10674 * int-vector-builder.h (int_vector_builder::int_vector_builder):
10675 Take the number of elements as a poly_uint64 rather than an
10677 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
10678 from unsigned int to poly_uint64.
10679 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
10680 (vec_perm_indices::new_vector): Likewise.
10681 (vec_perm_indices::length): Likewise.
10682 (vec_perm_indices::nelts_per_input): Likewise.
10683 (vec_perm_indices::input_nelts): Likewise.
10684 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
10685 number of elements per input as a poly_uint64 rather than an
10686 unsigned int. Use the original encoding for variable-length
10687 vectors, rather than clamping each individual element.
10688 For the second and subsequent elements in each pattern,
10689 clamp the step and base before clamping their sum.
10690 (vec_perm_indices::series_p): Handle polynomial element counts.
10691 (vec_perm_indices::all_in_range_p): Likewise.
10692 (vec_perm_indices_to_tree): Likewise.
10693 (vec_perm_indices_to_rtx): Likewise.
10694 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
10695 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
10696 (tree_vector_builder::new_binary_operation): Handle polynomial
10697 element counts. Return false if we need to know the number
10698 of elements at compile time.
10699 * fold-const.c (fold_vec_perm): Punt if the number of elements
10700 isn't known at compile time.
10702 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10704 * vec-perm-indices.h (vec_perm_builder): Change element type
10705 from HOST_WIDE_INT to poly_int64.
10706 (vec_perm_indices::element_type): Update accordingly.
10707 (vec_perm_indices::clamp): Handle polynomial element_types.
10708 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10709 (vec_perm_indices::all_in_range_p): Likewise.
10710 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
10712 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
10713 polynomial vec_perm_indices element types.
10714 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
10715 * fold-const.c (fold_vec_perm): Likewise.
10716 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
10717 * tree-vect-generic.c (lower_vec_perm): Likewise.
10718 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10719 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
10720 element type to HOST_WIDE_INT.
10722 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10723 Alan Hayward <alan.hayward@arm.com>
10724 David Sherwood <david.sherwood@arm.com>
10726 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
10727 rather than an int. Use plus_constant.
10728 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
10729 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
10731 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10732 Alan Hayward <alan.hayward@arm.com>
10733 David Sherwood <david.sherwood@arm.com>
10735 * calls.c (emit_call_1, expand_call): Change struct_value_size from
10736 a HOST_WIDE_INT to a poly_int64.
10738 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10739 Alan Hayward <alan.hayward@arm.com>
10740 David Sherwood <david.sherwood@arm.com>
10742 * calls.c (load_register_parameters): Cope with polynomial
10743 mode sizes. Require a constant size for BLKmode parameters
10744 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
10745 forces a parameter to be padded at the lsb end in order to
10746 fill a complete number of words, require the parameter size
10747 to be ordered wrt UNITS_PER_WORD.
10749 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10750 Alan Hayward <alan.hayward@arm.com>
10751 David Sherwood <david.sherwood@arm.com>
10753 * reload1.c (spill_stack_slot_width): Change element type
10754 from unsigned int to poly_uint64_pod.
10755 (alter_reg): Treat mode sizes as polynomial.
10757 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10758 Alan Hayward <alan.hayward@arm.com>
10759 David Sherwood <david.sherwood@arm.com>
10761 * reload.c (complex_word_subreg_p): New function.
10762 (reload_inner_reg_of_subreg, push_reload): Use it.
10764 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10765 Alan Hayward <alan.hayward@arm.com>
10766 David Sherwood <david.sherwood@arm.com>
10768 * lra-constraints.c (process_alt_operands): Reject matched
10769 operands whose sizes aren't ordered.
10770 (match_reload): Refer to this check here.
10772 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10773 Alan Hayward <alan.hayward@arm.com>
10774 David Sherwood <david.sherwood@arm.com>
10776 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
10777 that the mode size is in the set {1, 2, 4, 8, 16}.
10779 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10780 Alan Hayward <alan.hayward@arm.com>
10781 David Sherwood <david.sherwood@arm.com>
10783 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
10784 Use plus_constant instead of gen_rtx_PLUS.
10786 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10787 Alan Hayward <alan.hayward@arm.com>
10788 David Sherwood <david.sherwood@arm.com>
10790 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
10791 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
10792 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
10793 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
10794 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
10795 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
10796 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
10797 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
10798 * config/i386/i386.c (ix86_push_rounding): ...this new function.
10799 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
10801 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
10802 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
10803 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
10804 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
10805 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
10806 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
10807 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
10808 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
10809 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
10810 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
10812 * expr.c (emit_move_resolve_push): Treat the input and result
10813 of PUSH_ROUNDING as a poly_int64.
10814 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
10815 (emit_push_insn): Likewise.
10816 * lra-eliminations.c (mark_not_eliminable): Likewise.
10817 * recog.c (push_operand): Likewise.
10818 * reload1.c (elimination_effects): Likewise.
10819 * rtlanal.c (nonzero_bits1): Likewise.
10820 * calls.c (store_one_arg): Likewise. Require the padding to be
10821 known at compile time.
10823 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10824 Alan Hayward <alan.hayward@arm.com>
10825 David Sherwood <david.sherwood@arm.com>
10827 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
10828 Use plus_constant instead of gen_rtx_PLUS.
10830 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10831 Alan Hayward <alan.hayward@arm.com>
10832 David Sherwood <david.sherwood@arm.com>
10834 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
10835 rather than an int.
10837 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10838 Alan Hayward <alan.hayward@arm.com>
10839 David Sherwood <david.sherwood@arm.com>
10841 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
10842 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
10843 via stack temporaries. Treat the mode size as polynomial too.
10845 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10846 Alan Hayward <alan.hayward@arm.com>
10847 David Sherwood <david.sherwood@arm.com>
10849 * expr.c (expand_expr_real_2): When handling conversions involving
10850 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
10851 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
10852 as a poly_uint64 too.
10854 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10855 Alan Hayward <alan.hayward@arm.com>
10856 David Sherwood <david.sherwood@arm.com>
10858 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
10860 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10861 Alan Hayward <alan.hayward@arm.com>
10862 David Sherwood <david.sherwood@arm.com>
10864 * combine.c (can_change_dest_mode): Handle polynomial
10865 REGMODE_NATURAL_SIZE.
10866 * expmed.c (store_bit_field_1): Likewise.
10867 * expr.c (store_constructor): Likewise.
10868 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
10869 and polynomial REGMODE_NATURAL_SIZE.
10870 (gen_lowpart_common): Likewise.
10871 * reginfo.c (record_subregs_of_mode): Likewise.
10872 * rtlanal.c (read_modify_subreg_p): Likewise.
10874 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10875 Alan Hayward <alan.hayward@arm.com>
10876 David Sherwood <david.sherwood@arm.com>
10878 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
10879 numbers of elements.
10881 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10882 Alan Hayward <alan.hayward@arm.com>
10883 David Sherwood <david.sherwood@arm.com>
10885 * match.pd: Cope with polynomial numbers of vector elements.
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 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
10892 in a POINTER_PLUS_EXPR.
10894 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10895 Alan Hayward <alan.hayward@arm.com>
10896 David Sherwood <david.sherwood@arm.com>
10898 * omp-simd-clone.c (simd_clone_subparts): New function.
10899 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
10900 (ipa_simd_modify_function_body): Likewise.
10902 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10903 Alan Hayward <alan.hayward@arm.com>
10904 David Sherwood <david.sherwood@arm.com>
10906 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
10907 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
10908 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
10909 (expand_vector_condition, vector_element): Likewise.
10910 (subparts_gt): New function.
10911 (get_compute_type): Use subparts_gt.
10912 (count_type_subparts): Delete.
10913 (expand_vector_operations_1): Use subparts_gt instead of
10914 count_type_subparts.
10916 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10917 Alan Hayward <alan.hayward@arm.com>
10918 David Sherwood <david.sherwood@arm.com>
10920 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
10921 (vect_compile_time_alias): ...this new function. Do the calculation
10922 on poly_ints rather than trees.
10923 (vect_prune_runtime_alias_test_list): Update call accordingly.
10925 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10926 Alan Hayward <alan.hayward@arm.com>
10927 David Sherwood <david.sherwood@arm.com>
10929 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
10931 (vect_schedule_slp_instance): Likewise.
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-slp.c (vect_get_and_check_slp_defs): Reject
10938 constant and extern definitions for variable-length vectors.
10939 (vect_get_constant_vectors): Note that the number of units
10940 is known to be constant.
10942 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10943 Alan Hayward <alan.hayward@arm.com>
10944 David Sherwood <david.sherwood@arm.com>
10946 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
10947 of units as polynomial. Choose between WIDE and NARROW based
10950 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10951 Alan Hayward <alan.hayward@arm.com>
10952 David Sherwood <david.sherwood@arm.com>
10954 * tree-vect-stmts.c (simd_clone_subparts): New function.
10955 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
10957 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10958 Alan Hayward <alan.hayward@arm.com>
10959 David Sherwood <david.sherwood@arm.com>
10961 * tree-vect-stmts.c (vectorizable_call): Treat the number of
10962 vectors as polynomial. Use build_index_vector for
10963 IFN_GOMP_SIMD_LANE.
10965 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10966 Alan Hayward <alan.hayward@arm.com>
10967 David Sherwood <david.sherwood@arm.com>
10969 * tree-vect-stmts.c (get_load_store_type): Treat the number of
10970 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
10971 for variable-length vectors.
10972 (vectorizable_mask_load_store): Treat the number of units as
10973 polynomial, asserting that it is constant if the condition has
10974 already been enforced.
10975 (vectorizable_store, vectorizable_load): Likewise.
10977 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10978 Alan Hayward <alan.hayward@arm.com>
10979 David Sherwood <david.sherwood@arm.com>
10981 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
10982 of units as polynomial. Punt if we can't tell at compile time
10983 which vector contains the final result.
10985 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10986 Alan Hayward <alan.hayward@arm.com>
10987 David Sherwood <david.sherwood@arm.com>
10989 * tree-vect-loop.c (vectorizable_induction): Treat the number
10990 of units as polynomial. Punt on SLP inductions. Use an integer
10991 VEC_SERIES_EXPR for variable-length integer reductions. Use a
10992 cast of such a series for variable-length floating-point
10995 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10996 Alan Hayward <alan.hayward@arm.com>
10997 David Sherwood <david.sherwood@arm.com>
10999 * tree.h (build_index_vector): Declare.
11000 * tree.c (build_index_vector): New function.
11001 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
11002 of units as polynomial, forcibly converting it to a constant if
11003 vectorizable_reduction has already enforced the condition.
11004 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
11005 to create a {1,2,3,...} vector.
11006 (vectorizable_reduction): Treat the number of units as polynomial.
11007 Choose vectype_in based on the largest scalar element size rather
11008 than the smallest number of units. Enforce the restrictions
11011 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11012 Alan Hayward <alan.hayward@arm.com>
11013 David Sherwood <david.sherwood@arm.com>
11015 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
11016 number of units as polynomial.
11018 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11019 Alan Hayward <alan.hayward@arm.com>
11020 David Sherwood <david.sherwood@arm.com>
11022 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
11023 * target.def (autovectorize_vector_sizes): Return the vector sizes
11024 by pointer, using vector_sizes rather than a bitmask.
11025 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
11026 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
11027 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
11029 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
11030 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
11031 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
11032 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
11033 * omp-general.c (omp_max_vf): Likewise.
11034 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11035 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
11036 * tree-vect-loop.c (vect_analyze_loop): Likewise.
11037 * tree-vect-slp.c (vect_slp_bb): Likewise.
11038 * doc/tm.texi: Regenerate.
11039 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
11041 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
11042 the vector size as a poly_uint64 rather than an unsigned int.
11043 (current_vector_size): Change from an unsigned int to a poly_uint64.
11044 (get_vectype_for_scalar_type): Update accordingly.
11045 * tree.h (build_truth_vector_type): Take the size and number of
11046 units as a poly_uint64 rather than an unsigned int.
11047 (build_vector_type): Add a temporary overload that takes
11048 the number of units as a poly_uint64 rather than an unsigned int.
11049 * tree.c (make_vector_type): Likewise.
11050 (build_truth_vector_type): Take the number of units as a poly_uint64
11051 rather than an unsigned int.
11053 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11054 Alan Hayward <alan.hayward@arm.com>
11055 David Sherwood <david.sherwood@arm.com>
11057 * target.def (get_mask_mode): Take the number of units and length
11058 as poly_uint64s rather than unsigned ints.
11059 * targhooks.h (default_get_mask_mode): Update accordingly.
11060 * targhooks.c (default_get_mask_mode): Likewise.
11061 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
11062 * doc/tm.texi: Regenerate.
11064 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11065 Alan Hayward <alan.hayward@arm.com>
11066 David Sherwood <david.sherwood@arm.com>
11068 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
11069 * omp-general.c (omp_max_vf): Likewise.
11070 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
11071 (expand_omp_simd): Handle polynomial safelen.
11072 * omp-low.c (omplow_simd_context): Add a default constructor.
11073 (omplow_simd_context::max_vf): Change from int to poly_uint64.
11074 (lower_rec_simd_input_clauses): Update accordingly.
11075 (lower_rec_input_clauses): Likewise.
11077 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11078 Alan Hayward <alan.hayward@arm.com>
11079 David Sherwood <david.sherwood@arm.com>
11081 * tree-vectorizer.h (vect_nunits_for_cost): New function.
11082 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
11083 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
11084 (vect_analyze_slp_cost): Likewise.
11085 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
11086 (vect_model_load_cost): Likewise.
11088 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11089 Alan Hayward <alan.hayward@arm.com>
11090 David Sherwood <david.sherwood@arm.com>
11092 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
11093 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
11094 from an unsigned int * to a poly_uint64_pod *.
11095 (calculate_unrolling_factor): New function.
11096 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
11098 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11099 Alan Hayward <alan.hayward@arm.com>
11100 David Sherwood <david.sherwood@arm.com>
11102 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
11103 from an unsigned int to a poly_uint64.
11104 (_loop_vec_info::slp_unrolling_factor): Likewise.
11105 (_loop_vec_info::vectorization_factor): Change from an int
11107 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
11108 (vect_get_num_vectors): New function.
11109 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
11110 (vect_get_num_copies): Use vect_get_num_vectors.
11111 (vect_analyze_data_ref_dependences): Change max_vf from an int *
11112 to an unsigned int *.
11113 (vect_analyze_data_refs): Change min_vf from an int * to a
11115 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11116 than an unsigned HOST_WIDE_INT.
11117 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
11118 (vect_analyze_data_ref_dependence): Change max_vf from an int *
11119 to an unsigned int *.
11120 (vect_analyze_data_ref_dependences): Likewise.
11121 (vect_compute_data_ref_alignment): Handle polynomial vf.
11122 (vect_enhance_data_refs_alignment): Likewise.
11123 (vect_prune_runtime_alias_test_list): Likewise.
11124 (vect_shift_permute_load_chain): Likewise.
11125 (vect_supportable_dr_alignment): Likewise.
11126 (dependence_distance_ge_vf): Take the vectorization factor as a
11127 poly_uint64 rather than an unsigned HOST_WIDE_INT.
11128 (vect_analyze_data_refs): Change min_vf from an int * to a
11130 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
11131 vfm1 as a poly_uint64 rather than an int. Make the same change
11132 for the returned bound_scalar.
11133 (vect_gen_vector_loop_niters): Handle polynomial vf.
11134 (vect_do_peeling): Likewise. Update call to
11135 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
11136 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
11138 * tree-vect-loop.c (vect_determine_vectorization_factor)
11139 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
11140 (vect_get_known_peeling_cost): Likewise.
11141 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
11142 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
11143 (vect_transform_loop): Likewise. Use the lowest possible VF when
11144 updating the upper bounds of the loop.
11145 (vect_min_worthwhile_factor): Make static. Return an unsigned int
11146 rather than an int.
11147 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
11148 polynomial unroll factors.
11149 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
11150 (vect_make_slp_decision): Likewise.
11151 (vect_supported_load_permutation_p): Likewise, and polynomial
11153 (vect_analyze_slp_cost): Handle polynomial vf.
11154 (vect_slp_analyze_node_operations): Likewise.
11155 (vect_slp_analyze_bb_1): Likewise.
11156 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11157 than an unsigned HOST_WIDE_INT.
11158 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
11159 (vectorizable_load): Handle polynomial vf.
11160 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
11162 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
11164 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11165 Alan Hayward <alan.hayward@arm.com>
11166 David Sherwood <david.sherwood@arm.com>
11168 * match.pd: Handle bit operations involving three constants
11169 and try to fold one pair.
11171 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11173 * tree-vect-loop-manip.c: Include gimple-fold.h.
11174 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
11175 niters_maybe_zero parameters. Handle other cases besides a step of 1.
11176 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
11177 Add a path that uses a step of VF instead of 1, but disable it
11179 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
11180 and niters_no_overflow parameters. Update calls to
11181 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
11182 Create a new SSA name if the latter choses to use a ste other
11183 than zero, and return it via niters_vector_mult_vf_var.
11184 * tree-vect-loop.c (vect_transform_loop): Update calls to
11185 vect_do_peeling, vect_gen_vector_loop_niters and
11186 slpeel_make_loop_iterate_ntimes.
11187 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
11188 (vect_gen_vector_loop_niters): Update declarations after above changes.
11190 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
11192 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
11193 128-bit round to integer instructions.
11194 (ceil<mode>2): Likewise.
11195 (btrunc<mode>2): Likewise.
11196 (round<mode>2): Likewise.
11198 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11200 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
11201 unaligned VSX load/store on P8/P9.
11202 (expand_block_clear): Allow the use of unaligned VSX
11203 load/store on P8/P9.
11205 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11207 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
11209 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
11210 swap associated with both a load and a store.
11212 2018-01-02 Andrew Waterman <andrew@sifive.com>
11214 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
11215 * config/riscv/riscv.md (clear_cache): Use it.
11217 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
11219 * web.c: Remove out-of-date comment.
11221 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11223 * expr.c (fixup_args_size_notes): Check that any existing
11224 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
11225 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
11226 (emit_single_push_insn): ...here.
11228 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11230 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
11231 (const_vector_encoded_nelts): New function.
11232 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
11233 (const_vector_int_elt, const_vector_elt): Declare.
11234 * emit-rtl.c (const_vector_int_elt_1): New function.
11235 (const_vector_elt): Likewise.
11236 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
11237 of CONST_VECTOR_ELT.
11239 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11241 * expr.c: Include rtx-vector-builder.h.
11242 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
11243 directly on the tree encoding.
11244 (const_vector_from_tree): Likewise.
11245 * optabs.c: Include rtx-vector-builder.h.
11246 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
11247 sequence of "u" values.
11248 * vec-perm-indices.c: Include rtx-vector-builder.h.
11249 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
11250 directly on the vec_perm_indices encoding.
11252 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11254 * doc/rtl.texi (const_vector): Describe new encoding scheme.
11255 * Makefile.in (OBJS): Add rtx-vector-builder.o.
11256 * rtx-vector-builder.h: New file.
11257 * rtx-vector-builder.c: Likewise.
11258 * rtl.h (rtx_def::u2): Add a const_vector field.
11259 (CONST_VECTOR_NPATTERNS): New macro.
11260 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
11261 (CONST_VECTOR_DUPLICATE_P): Likewise.
11262 (CONST_VECTOR_STEPPED_P): Likewise.
11263 (CONST_VECTOR_ENCODED_ELT): Likewise.
11264 (const_vec_duplicate_p): Check for a duplicated vector encoding.
11265 (unwrap_const_vec_duplicate): Likewise.
11266 (const_vec_series_p): Check for a non-duplicated vector encoding.
11267 Say that the function only returns true for integer vectors.
11268 * emit-rtl.c: Include rtx-vector-builder.h.
11269 (gen_const_vec_duplicate_1): Delete.
11270 (gen_const_vector): Call gen_const_vec_duplicate instead of
11271 gen_const_vec_duplicate_1.
11272 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
11273 (gen_const_vec_duplicate): Use rtx_vector_builder.
11274 (gen_const_vec_series): Likewise.
11275 (gen_rtx_CONST_VECTOR): Likewise.
11276 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
11277 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11278 Build a new vector rather than modifying a CONST_VECTOR in-place.
11279 (handle_special_swappables): Update call accordingly.
11280 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
11281 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11282 Build a new vector rather than modifying a CONST_VECTOR in-place.
11283 (handle_special_swappables): Update call accordingly.
11285 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11287 * simplify-rtx.c (simplify_const_binary_operation): Use
11288 CONST_VECTOR_ELT instead of XVECEXP.
11290 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11292 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
11293 the selector elements to be different from the data elements
11294 if the selector is a VECTOR_CST.
11295 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
11296 ssizetype for the selector.
11298 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11300 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
11301 before testing each element individually.
11302 * tree-vect-generic.c (lower_vec_perm): Likewise.
11304 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11306 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
11307 * selftest-run-tests.c (selftest::run_tests): Call it.
11308 * vector-builder.h (vector_builder::operator ==): New function.
11309 (vector_builder::operator !=): Likewise.
11310 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
11311 (vec_perm_indices::all_from_input_p): New function.
11312 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11313 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
11314 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
11315 instead of reading the VECTOR_CST directly. Detect whether both
11316 vector inputs are the same before constructing the vec_perm_indices,
11317 and update the number of inputs argument accordingly. Use the
11318 utility functions added above. Only construct sel2 if we need to.
11320 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11322 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
11323 the broadcast of the low byte.
11324 (expand_mult_highpart): Use an explicit encoding for the permutes.
11325 * optabs-query.c (can_mult_highpart_p): Likewise.
11326 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
11327 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11328 (vectorizable_bswap): Likewise.
11329 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
11330 explicit encoding for the power-of-2 permutes.
11331 (vect_permute_store_chain): Likewise.
11332 (vect_grouped_load_supported): Likewise.
11333 (vect_permute_load_chain): Likewise.
11335 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11337 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
11338 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
11339 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
11340 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11341 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
11342 (vect_gen_perm_mask_any): Likewise.
11344 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11346 * int-vector-builder.h: New file.
11347 * vec-perm-indices.h: Include int-vector-builder.h.
11348 (vec_perm_indices): Redefine as an int_vector_builder.
11349 (auto_vec_perm_indices): Delete.
11350 (vec_perm_builder): Redefine as a stand-alone class.
11351 (vec_perm_indices::vec_perm_indices): New function.
11352 (vec_perm_indices::clamp): Likewise.
11353 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
11354 (vec_perm_indices::new_vector): New function.
11355 (vec_perm_indices::new_expanded_vector): Update for new
11356 vec_perm_indices class.
11357 (vec_perm_indices::rotate_inputs): New function.
11358 (vec_perm_indices::all_in_range_p): Operate directly on the
11359 encoded form, without computing elided elements.
11360 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
11361 encoding. Update for new vec_perm_indices class.
11362 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
11363 the given vec_perm_builder.
11364 (expand_vec_perm_var): Update vec_perm_builder constructor.
11365 (expand_mult_highpart): Use vec_perm_builder instead of
11366 auto_vec_perm_indices.
11367 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
11368 vec_perm_indices instead of auto_vec_perm_indices. Use a single
11369 or double series encoding as appropriate.
11370 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
11371 vec_perm_indices instead of auto_vec_perm_indices.
11372 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11373 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11374 (vect_permute_store_chain): Likewise.
11375 (vect_grouped_load_supported): Likewise.
11376 (vect_permute_load_chain): Likewise.
11377 (vect_shift_permute_load_chain): Likewise.
11378 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11379 (vect_transform_slp_perm_load): Likewise.
11380 (vect_schedule_slp_instance): Likewise.
11381 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11382 (vectorizable_mask_load_store): Likewise.
11383 (vectorizable_bswap): Likewise.
11384 (vectorizable_store): Likewise.
11385 (vectorizable_load): Likewise.
11386 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
11387 vec_perm_indices instead of auto_vec_perm_indices. Use
11388 tree_to_vec_perm_builder to read the vector from a tree.
11389 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
11390 vec_perm_builder instead of a vec_perm_indices.
11391 (have_whole_vector_shift): Use vec_perm_builder and
11392 vec_perm_indices instead of auto_vec_perm_indices. Leave the
11393 truncation to calc_vec_perm_mask_for_shift.
11394 (vect_create_epilog_for_reduction): Likewise.
11395 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
11396 from auto_vec_perm_indices to vec_perm_indices.
11397 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11398 instead of changing individual elements.
11399 (aarch64_vectorize_vec_perm_const): Use new_vector to install
11400 the vector in d.perm.
11401 * config/arm/arm.c (expand_vec_perm_d::perm): Change
11402 from auto_vec_perm_indices to vec_perm_indices.
11403 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11404 instead of changing individual elements.
11405 (arm_vectorize_vec_perm_const): Use new_vector to install
11406 the vector in d.perm.
11407 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
11408 Update vec_perm_builder constructor.
11409 (rs6000_expand_interleave): Likewise.
11410 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
11411 (rs6000_expand_interleave): Likewise.
11413 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11415 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
11416 to qimode could truncate the indices.
11417 * optabs.c (expand_vec_perm_var): Likewise.
11419 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11421 * Makefile.in (OBJS): Add vec-perm-indices.o.
11422 * vec-perm-indices.h: New file.
11423 * vec-perm-indices.c: Likewise.
11424 * target.h (vec_perm_indices): Replace with a forward class
11426 (auto_vec_perm_indices): Move to vec-perm-indices.h.
11427 * optabs.h: Include vec-perm-indices.h.
11428 (expand_vec_perm): Delete.
11429 (selector_fits_mode_p, expand_vec_perm_var): Declare.
11430 (expand_vec_perm_const): Declare.
11431 * target.def (vec_perm_const_ok): Replace with...
11432 (vec_perm_const): ...this new hook.
11433 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
11434 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
11435 * doc/tm.texi: Regenerate.
11436 * optabs.def (vec_perm_const): Delete.
11437 * doc/md.texi (vec_perm_const): Likewise.
11438 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
11439 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
11440 expand_vec_perm for constant permutation vectors. Assert that
11441 the mode of variable permutation vectors is the integer equivalent
11442 of the mode that is being permuted.
11443 * optabs-query.h (selector_fits_mode_p): Declare.
11444 * optabs-query.c: Include vec-perm-indices.h.
11445 (selector_fits_mode_p): New function.
11446 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
11447 is defined, instead of checking whether the vec_perm_const_optab
11448 exists. Use targetm.vectorize.vec_perm_const instead of
11449 targetm.vectorize.vec_perm_const_ok. Check whether the indices
11450 fit in the vector mode before using a variable permute.
11451 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
11452 vec_perm_indices instead of an rtx.
11453 (expand_vec_perm): Replace with...
11454 (expand_vec_perm_const): ...this new function. Take the selector
11455 as a vec_perm_indices rather than an rtx. Also take the mode of
11456 the selector. Update call to shift_amt_for_vec_perm_mask.
11457 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
11458 Use vec_perm_indices::new_expanded_vector to expand the original
11459 selector into bytes. Check whether the indices fit in the vector
11460 mode before using a variable permute.
11461 (expand_vec_perm_var): Make global.
11462 (expand_mult_highpart): Use expand_vec_perm_const.
11463 * fold-const.c: Includes vec-perm-indices.h.
11464 * tree-ssa-forwprop.c: Likewise.
11465 * tree-vect-data-refs.c: Likewise.
11466 * tree-vect-generic.c: Likewise.
11467 * tree-vect-loop.c: Likewise.
11468 * tree-vect-slp.c: Likewise.
11469 * tree-vect-stmts.c: Likewise.
11470 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
11472 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
11473 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
11474 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
11475 (aarch64_vectorize_vec_perm_const): ...this new function.
11476 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11477 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11478 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
11479 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
11480 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11481 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11482 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
11484 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
11485 check for NEON modes.
11486 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
11487 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
11488 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
11489 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
11491 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
11492 the old VEC_PERM_CONST conditions.
11493 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
11494 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
11495 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
11496 (ia64_vectorize_vec_perm_const_ok): Merge into...
11497 (ia64_vectorize_vec_perm_const): ...this new function.
11498 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
11499 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
11500 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
11501 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
11502 * config/mips/mips.c (mips_expand_vec_perm_const)
11503 (mips_vectorize_vec_perm_const_ok): Merge into...
11504 (mips_vectorize_vec_perm_const): ...this new function.
11505 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
11506 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
11507 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
11508 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
11509 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
11510 (rs6000_expand_vec_perm_const): Delete.
11511 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
11513 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11514 (altivec_expand_vec_perm_const_le): Take each operand individually.
11515 Operate on constant selectors rather than rtxes.
11516 (altivec_expand_vec_perm_const): Likewise. Update call to
11517 altivec_expand_vec_perm_const_le.
11518 (rs6000_expand_vec_perm_const): Delete.
11519 (rs6000_vectorize_vec_perm_const_ok): Delete.
11520 (rs6000_vectorize_vec_perm_const): New function.
11521 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11522 an element count and rtx array.
11523 (rs6000_expand_extract_even): Update call accordingly.
11524 (rs6000_expand_interleave): Likewise.
11525 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
11526 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
11527 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
11528 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
11529 (rs6000_expand_vec_perm_const): Delete.
11530 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11531 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11532 (altivec_expand_vec_perm_const_le): Take each operand individually.
11533 Operate on constant selectors rather than rtxes.
11534 (altivec_expand_vec_perm_const): Likewise. Update call to
11535 altivec_expand_vec_perm_const_le.
11536 (rs6000_expand_vec_perm_const): Delete.
11537 (rs6000_vectorize_vec_perm_const_ok): Delete.
11538 (rs6000_vectorize_vec_perm_const): New function. Remove stray
11539 reference to the SPE evmerge intructions.
11540 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11541 an element count and rtx array.
11542 (rs6000_expand_extract_even): Update call accordingly.
11543 (rs6000_expand_interleave): Likewise.
11544 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
11545 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
11547 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11549 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11551 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
11552 vector mode and that that mode matches the mode of the data
11554 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
11555 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
11556 directly using expand_vec_perm_1 when forcing selectors into
11558 (expand_vec_perm_var): New function, split out from expand_vec_perm.
11560 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11562 * optabs-query.h (can_vec_perm_p): Delete.
11563 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
11564 * optabs-query.c (can_vec_perm_p): Split into...
11565 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
11566 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
11567 particular selector is valid.
11568 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11569 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11570 (vect_grouped_load_supported): Likewise.
11571 (vect_shift_permute_load_chain): Likewise.
11572 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11573 (vect_transform_slp_perm_load): Likewise.
11574 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11575 (vectorizable_bswap): Likewise.
11576 (vect_gen_perm_mask_checked): Likewise.
11577 * fold-const.c (fold_ternary_loc): Likewise. Don't take
11578 implementations of variable permutation vectors into account
11579 when deciding which selector to use.
11580 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
11581 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
11582 with a false third argument.
11583 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
11584 to test whether the constant selector is valid and can_vec_perm_var_p
11585 to test whether a variable selector is valid.
11587 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11589 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
11590 * optabs-query.c (can_vec_perm_p): Likewise.
11591 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
11592 instead of vec_perm_indices.
11593 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
11594 (vect_gen_perm_mask_checked): Likewise,
11595 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
11596 (vect_gen_perm_mask_checked): Likewise,
11598 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11600 * optabs-query.h (qimode_for_vec_perm): Declare.
11601 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
11602 (qimode_for_vec_perm): ...this new function.
11603 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
11605 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11607 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
11608 does not have a conditional at the top.
11610 2018-01-02 Richard Biener <rguenther@suse.de>
11612 * ipa-inline.c (big_speedup_p): Fix expression.
11614 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11617 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
11620 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11624 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
11625 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
11626 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
11627 cond_taken_branch_cost 3->4.
11629 2018-01-01 Jakub Jelinek <jakub@redhat.com>
11631 PR tree-optimization/83581
11632 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
11633 TODO_cleanup_cfg if any changes have been made.
11635 PR middle-end/83608
11636 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
11637 convert_modes if target mode has the right side, but different mode
11640 PR middle-end/83609
11641 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
11642 last argument when extracting from CONCAT. If either from_real or
11643 from_imag is NULL, use expansion through memory. If result is not
11644 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
11645 the parts directly to inner mode, if even that fails, use expansion
11648 PR middle-end/83623
11649 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
11650 check for bswap in mode rather than HImode and use that in expand_unop
11653 Copyright (C) 2018 Free Software Foundation, Inc.
11655 Copying and distribution of this file, with or without modification,
11656 are permitted in any medium without royalty provided the copyright
11657 notice and this notice are preserved.