1 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
3 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
4 ("enabled" attribute): Handle "isa" attribute.
5 (*movdi_internal): Substitute Yi and Yj constraint with x
6 and Ym and Yn constraint with y constraint. Update "isa"
7 attribute and set "preferred_for_speed" attribute from
8 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
9 (*movsi_internal): Ditto.
10 (*movdf_internal): Ditto.
11 (*movsf_internal): Ditto.
12 (*zero_extendsidi2): Ditto.
13 * config/i386/sse.md (vec_set<mode>_0): Ditto.
15 (*vec_extract<ssevecmodelower>_0): Ditto.
16 (*vec_extractv4si_0_zext_sse4): Ditto.
17 (vec_concatv2di): Ditto.
18 (*vec_dup<mode>): Ditto.
19 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
20 * config/i386/constraints.md (Yi): Remove.
25 2018-04-26 Nathan Sidwell <nathan@acm.org>
27 * dumpfile.c (dump_open): New.
28 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
29 (dump_finish): Detect stdio/stderr by value not name.
31 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
33 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
35 2018-04-26 Tom de Vries <tom@codesourcery.com>
38 * config/nvptx/nvptx.c (verify_neutering_jumps)
39 (verify_neutering_labels): New function
40 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
42 2018-04-26 Tom de Vries <tom@codesourcery.com>
45 * config/nvptx/nvptx.c (needs_neutering_p): New function.
46 (nvptx_single): Use needs_neutering_p to skip over insns that do not
49 2018-04-26 Richard Biener <rguenther@suse.de>
50 Tom de Vries <tom@codesourcery.com>
53 * lto-streamer-out.c (output_function): Fixup loops if required to match
54 discovery done in the reader.
56 2018-04-26 Richard Biener <rguenther@suse.de>
58 PR tree-optimization/85116
59 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
60 have a loop exit from the single latch predecessor. Remove
61 case of header with just condition.
62 (ch_base::copy_headers): Exclude infinite loops from any
64 (pass_ch::execute): Record exits.
66 2018-04-26 Richard Biener <rguenther@suse.de>
68 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
69 prologue cost vector and pass it to vect_get_load_cost.
70 (vect_get_peeling_costs_all_drs): Likewise.
71 (vect_peeling_hash_get_lowest_cost): Likewise.
72 (vect_enhance_data_refs_alignment): Likewise.
74 2018-04-26 Richard Biener <rguenther@suse.de>
77 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
78 checking of integer<->pointer conversions.
79 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
80 sign-/zero-extending pointer types.
81 (expand_omp_for_static_chunk): Likewise.
83 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
84 Jean Lee <xiaoyur347@gmail.com>
86 * config/mips/mips.c (mips_asan_shadow_offset): New function.
87 (TARGET_ASAN_SHADOW_OFFSET): Define.
88 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
89 true for -fsanitize=address.
91 2018-04-25 Mark Wielaard <mark@klomp.org>
93 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
96 2018-04-25 Jakub Jelinek <jakub@redhat.com>
98 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
99 than "alu", remove explicit "memory" and "imm_disp" attributes.
100 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
103 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
104 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
107 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
110 * config/i386/i386.c (ix86_expand_builtin): Change memory
111 operand to XI, extend p0 to Pmode.
112 * config/i386/i386.md: Change unspec volatile and operand
113 1 mode to XI, change operand 0 mode to P.
115 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
117 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
118 GET_MODE_MASK before any checking.
119 (nds32_can_use_bset_p): Likewise.
120 (nds32_can_use_btgl_p): Likewise.
122 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
124 * config/nds32/nds32-doubleword.md: New define_split pattern for
125 illegal register number.
127 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
129 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
131 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
133 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
135 2018-04-25 Richard Biener <rguenther@suse.de>
137 * lto-streamer.h (LTO_major_version): Bump to 8.
139 2018-04-25 Jakub Jelinek <jakub@redhat.com>
141 * BASE-VER: Set to 9.0.0.
143 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
145 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
146 in __abskf2 and __powikf2.
148 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
151 * config/aarch64/constraints.md (Usg, Usj): New constraints.
152 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
153 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
154 Use the above on operand 2. Reindent.
155 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
157 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
160 * common/config/i386/i386-common.c (ix86_handle_option): Don't
162 * config/i386/i386.opt (mcet): Removed.
163 * doc/install.texi: Remove -mcet documentation.
164 * doc/invoke.texi: Likewise.
166 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
169 * doc/install.texi: Remove -mcet from bootstrap-cet.
171 2018-04-24 Jakub Jelinek <jakub@redhat.com>
174 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
175 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
179 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
180 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
181 containing a CONST_VECTOR.
183 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
185 * doc/install.texi: Update newlib dependency for nvptx.
187 2018-04-24 Jakub Jelinek <jakub@redhat.com>
190 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
191 instead of INTVAL when shifting x left.
193 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
195 PR tree-optimization/85478
196 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
197 vect_grouped_store_supported for single element vectors.
199 2018-04-24 Richard Biener <rguenther@suse.de>
202 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
203 load cost increase to the case of non-constant step.
205 2018-04-24 Jakub Jelinek <jakub@redhat.com>
208 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
209 destination if any_malformed_asm.
211 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
214 * expr.c (store_field): In the bitfield case, if the value comes from
215 a function call and is returned in registers by means of a PARALLEL,
216 do not change the mode of the temporary unless BLKmode and VOIDmode.
218 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
220 PR rtl-optimization/85423
221 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
222 dependencies to debug insns when the previous insn is non-debug.
224 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
226 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
227 enums into a single definition.
228 (fls): Fix predicates and printing.
231 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
233 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
234 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
235 and short u6 immediate.
236 (check_if_valid_sleep_operand): Remove.
237 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
239 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
241 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
242 flag_always_save_lp condition.
243 * config/nds32/nds32.opt (malways-save-lp): New option.
245 2018-04-22 Shiva Chen <shiva0217@gmail.com>
247 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
248 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
249 * config/nds32/nds32.h
250 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
251 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
253 2018-04-22 Shiva Chen <shiva0217@gmail.com>
255 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
256 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
258 2018-04-22 Shiva Chen <shiva0217@gmail.com>
259 Chung-Ju Wu <jasonwucj@gmail.com>
261 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
263 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
264 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
266 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
268 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
270 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
272 * config/nds32/nds32-protos.h (nds32_data_alignment,
273 nds32_local_alignment): Declare.
274 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
275 nds32_local_alignment): New functions.
276 (TARGET_CONSTANT_ALIGNMENT): Define.
277 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
279 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
281 * config/nds32/nds32.c
282 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
283 (TARGET_MODES_TIEABLE_P): Likewise.
285 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
287 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
290 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
291 Chung-Ju Wu <jasonwucj@gmail.com>
293 * config/nds32/constants.md (unspec_volatile_element): Add enum values
294 for unaligned access.
295 * config/nds32/nds32-intrinsic.c: Implementation of expanding
297 * config/nds32/nds32-intrinsic.md: Likewise.
298 * config/nds32/nds32_intrinsic.h: Likewise.
299 * config/nds32/nds32.h (nds32_builtins): Likewise.
300 * config/nds32/nds32.opt (munaligned-access): New option.
301 * config/nds32/nds32.c (nds32_asm_file_start): Display
302 flag_unaligned_access status.
304 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
306 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
307 -mno-relax is present.
308 * config/riscv/linux.h (LINK_SPEC): Ditto.
310 2018-04-20 Martin Sebor <msebor@redhat.com>
313 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
315 (gimple_fold_builtin_stxcpy_chk): Same.
316 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
318 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
321 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
322 __powikf2 when long double is IEEE 128-bit.
324 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
326 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
327 step to make sure stack always aligned.
329 2018-04-20 Carl Love <cel@us.ibm.com>
332 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
335 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
336 Tom de Vries <tom@codesourcery.com>
339 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
340 Emit insns for calls too.
341 (nvptx_find_par): Always look for worker-level predecessor insn.
342 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
344 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
345 (nvptx_process_pars): Propagate frames for calls.
347 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
350 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
352 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
353 (ix86_handle_option): Don't handle OPT_mibt.
354 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
356 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
358 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
359 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
360 (ix86_target_macros): Define __CET__ with flag_cf_protection
362 * config/i386/i386.c (isa2_opts): Remove -mibt.
363 * config/i386/i386.h (TARGET_IBT): Removed.
364 (TARGET_IBT_P): Likewise.
365 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
366 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
367 * config/i386/i386.opt (mcet): Update help message.
370 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
371 -mcet as an alias for -mshstk.
373 2018-04-20 Richard Biener <rguenther@suse.de>
376 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
377 complexity by forcing a single use of the multiply operand.
379 2018-04-20 Martin Jambor <mjambor@suse.cz>
382 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
383 recursion dependency to only apply to non-clones.
385 2018-04-20 Martin Jambor <mjambor@suse.cz>
388 * ipa-cp.c (create_specialized_node): Check that clones of
389 self-recursive edges exist during IPA-CP.
391 2018-04-19 Toon Moene <toon@moene.org>
393 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
396 2018-04-19 Jakub Jelinek <jakub@redhat.com>
398 PR tree-optimization/85467
399 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
400 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
401 VECTOR_CST element to type.
403 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
406 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
407 * config/i386/i386.md (builtin_setjmp_setup): Removed.
408 (builtin_longjmp): Likewise.
409 (save_stack_nonlocal): New pattern.
410 (restore_stack_nonlocal): Likewise.
412 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
415 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
416 Replace ASM_OUTPUT_LABEL with fprintf.
418 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
421 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
422 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
423 * config/i386/i386-c.c (ix86_target_macros_internal): Also
424 define __IBT__ and __SHSTK__ for -fcf-protection.
425 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
427 (ix86_trampoline_init): Likewise.
428 (x86_output_mi_thunk): Likewise.
429 (ix86_notrack_prefixed_insn_p): Likewise.
430 (ix86_option_override_internal): Don't disallow -fcf-protection.
431 * config/i386/i386.md (rdssp<mode>): Also enable for
433 (incssp<mode>): Likewise.
434 (nop_endbr): Likewise.
435 * config/i386/i386.opt (mcet): Change help message to built-in
439 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
440 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
441 enable CET built-in functions.
443 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
445 * common/config/i386/i386-common.c
446 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
447 OPTION_MASK_ISA_MOVDIRI_UNSET,
448 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
449 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
450 * config.gcc (movdirintrin.h): New header.
451 * config/i386/cpuid.h (bit_MOVDIRI,
452 bit_MOVDIR64B): New bits.
453 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
455 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
456 (VOID, PVOID, PCVOID)): New function types.
457 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
458 __builtin_ia32_directstoreu_u64,
459 __builtin_ia32_movdir64b): New builtins.
460 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
461 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
463 (ix86_valid_target_attribute_inner_p): Ditto.
464 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
465 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
466 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
467 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
468 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
469 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
470 (movdiri<mode>, movdir64b_<mode>): New.
471 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
472 * config/i386/immintrin.h: Include movdirintrin.h.
473 * config/i386/movdirintrin.h: New file.
474 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
476 2018-04-19 Richard Biener <rguenther@suse.de>
479 * cfg.c (clear_bb_flags): When loop state says we have
480 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
482 2018-04-19 Richard Biener <rguenther@suse.de>
484 PR tree-optimization/84737
485 * tree-vect-data-refs.c (vect_copy_ref_info): New function
486 copying restrict info.
487 (vect_setup_realignment): Use it.
488 * tree-vectorizer.h (vect_copy_ref_info): Declare.
489 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
490 the first DR to all generated stores.
491 (vectorizable_load): Likewise for loads.
493 2018-04-19 Jakub Jelinek <jakub@redhat.com>
495 PR tree-optimization/85446
496 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
497 the integral and pointer types to have the same precision.
499 * doc/install.texi: Document --disable-cet being the default and
502 2018-04-18 Martin Liska <mliska@suse.cz>
504 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
507 2018-04-18 Martin Liska <mliska@suse.cz>
510 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
513 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
514 arguments if they are comparable.
516 2018-04-18 Martin Liska <mliska@suse.cz>
519 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
522 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
525 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
528 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
529 ENDBR after calling __morestack.
531 2018-04-18 David Malcolm <dmalcolm@redhat.com>
534 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
535 by using gcc_base_ver to generate a gcc_driver_version, and use
536 it when generating GCC_DRIVER_NAME.
537 * configure: Regenerate.
539 2018-04-18 Jakub Jelinek <jakub@redhat.com>
542 * config.gcc: Obsolete powerpc*-*-*spe*.
544 2018-04-17 Jakub Jelinek <jakub@redhat.com>
547 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
548 (stabstr_D): Change type of unum from unsigned int to
549 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
552 2018-04-17 Jim Wilson <jimw@sifive.com>
555 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
556 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
557 Set arg_pointer_offset after using pretend_args_size.
559 2018-04-17 Jakub Jelinek <jakub@redhat.com>
561 PR rtl-optimization/85431
562 * dse.c (record_store): Ignore zero width stores.
565 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
566 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
567 __builtin_stack_restore rather than after it.
568 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
569 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
570 argument instead of virtual_dynamic_stack_rtx.
572 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
574 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
576 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
577 Add note to error message to explain internal mapping of overloaded
578 built-in function name to non-overloaded built-in function name.
579 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
582 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
585 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
586 where the inputs overlap with the output.
588 2018-04-17 Jakub Jelinek <jakub@redhat.com>
590 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
591 (=v, v) alternative and explicit "memory" attribute.
592 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
593 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
595 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
596 "sselog1" type instead of "sselog".
597 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
598 "sselog". Remove explicit "memory" attribute.
599 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
600 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
602 (vec_extract_hi_v32hi): Merge all alternatives into one, use
603 "sselog1" type instead of "sselog". Remove explicit "memory"
605 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
606 use "sselog1" type instead of "sselog". Remove explicit "memory"
608 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
609 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
611 (vec_extract_hi_v64qi): Merge all alternatives into one, use
612 "sselog1" type instead of "sselog". Remove explicit "memory"
614 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
615 use "sselog1" type instead of "sselog". Remove explicit "memory"
619 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
622 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
625 2018-04-17 Martin Jambor <mjambor@suse.cz>
628 * ipa-cp.c (create_specialized_node): Call
629 expand_all_artificial_thunks if necessary.
631 2018-04-17 Martin Liska <mliska@suse.cz>
634 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
635 in message, remote space in between '_G' and '('.
637 2018-04-17 Jakub Jelinek <jakub@redhat.com>
640 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
641 avx512f_vmcmp<mode>3<round_saeonly_name>,
642 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
643 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
644 avx512f_rndscale<mode><round_saeonly_name>,
645 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
646 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
647 Use %<iptr>2 instead of %2 for -masm=intel.
648 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
649 avx512f_vcvttss2usi<round_saeonly_name>,
650 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
652 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
653 avx512f_vcvttsd2usi<round_saeonly_name>,
654 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
655 Use %q1 instead of %1 for -masm=intel.
656 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
657 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
658 of %3 for -masm=intel.
659 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
661 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
663 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
665 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
666 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
668 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
670 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
671 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
672 %g1 and one with %0 and %1.
673 (avx512er_vmrcp28<mode><round_saeonly_name>,
674 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
676 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
677 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
678 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
679 of %0 and %{%4%} for -masm=intel.
680 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
681 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
682 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
683 order of %0 and %{%5%}%{z%} for -masm=intel.
685 2018-04-17 Jan Hubicka <jh@suse.cz>
688 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
690 2018-04-17 Martin Liska <mliska@suse.cz>
693 * multiple_target.c (create_dispatcher_calls): Set apostrophes
694 for target_clone error message. Make default implementation
695 clone to be a local declaration.
696 (separate_attrs): Add new argument and check for an empty
698 (expand_target_clones): Handle it.
699 (ipa_target_clone): Make redirection just for target_clones
702 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
703 Tom de Vries <tom@codesourcery.com>
706 * omp-expand.c (expand_oacc_for): Add dummy false branch for
707 tiled basic blocks without omp continue statements.
709 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
712 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
713 vec_extract expression as having side effects to make sure it gets
716 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
719 * config/i386/i386.c (get_builtin_code_for_version): Check
722 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
725 * gcc/config.gcc: Support "skylake".
726 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
728 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
729 (processor_target_table): Add "skylake".
730 (ix86_option_override_internal): Add "skylake".
731 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
732 PROCESSOR_CANNONLAKE.
733 (get_builtin_code_for_version): Fix priority for
734 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
735 PROCESSOR_SKYLAKE-AVX512.
736 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
737 (processor_type): Add PROCESSOR_SKYLAKE.
739 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
740 Jason Merrill <jason@redhat.com>
743 * convert.c (convert_to_integer_1): Use direct recursion for
744 enumeral types and types with a precision less than the number
745 of bits in their mode.
747 2018-04-16 Julia Koval <julia.koval@intel.com>
750 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
751 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
753 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
756 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
757 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
758 and -mno-direct-move.
760 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
763 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
764 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
765 Ensure negative shifts result in {0}.
767 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
769 PR rtl-optimization/79916
770 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
771 regs (if any) to define how to gnerate SD moves when LRA is in
774 2018-04-13 Jakub Jelinek <jakub@redhat.com>
776 PR rtl-optimization/85393
777 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
778 * except.c (expand_dw2_landing_pad_for_region): Make static.
779 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
780 a label and unconditional jump to old_bb, rather than
781 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
784 PR rtl-optimization/85376
785 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
786 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
787 instead of a specific value.
789 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
790 Bin Cheng <bin.cheng@arm.com>
792 PR tree-optimization/82965
793 PR tree-optimization/83991
794 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
795 by_profile_only parameter.
796 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
797 information if the loop was predicted to iterate too many times.
798 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
800 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
803 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
806 2018-04-13 Martin Liska <mliska@suse.cz>
807 Jakub Jelinek <jakub@redhat.com>
810 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
811 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
812 * builtins.c (expand_builtin_memory_copy_args): Use
813 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
814 handle dest_addr == pc_rtx.
816 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
819 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
820 asked to not generate direct moves.
821 (fix_trunc<mode>si2_stfiwx): Similar.
822 (fix_trunc<mode>si2_internal): Similar.
824 2018-04-12 Jakub Jelinek <jakub@redhat.com>
827 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
828 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
829 lookup if dest in some wider mode is known to be const0_rtx and
830 if so, record permanent equivalence for it to be ZERO_EXTEND of
831 the narrower mode destination.
833 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
835 * lto-streamer-out.c (output_function): Revert 259346.
836 * omp-expand.c (expand_oacc_for): Likewise.
838 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
840 PR rtl-optimization/85354
841 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
842 * sel-sched.c (sel_global_init): ... here.
844 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
847 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
848 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
849 mode for PE-COFF targets.
850 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
851 (i386_pe_asm_lto_end): Likewise.
852 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
853 (TARGET_ASM_LTO_END): Likewise.
854 * config/i386/winnt.c (saved_debug_info_level): New static variable.
855 (i386_pe_asm_lto_start): New function.
856 (i386_pe_asm_lto_end): Likewise.
858 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
859 Richard Biener <rguenther@suse.de>
862 * lto-streamer-out.c (output_function): Fix CFG loop state before
864 * omp-expand.c (expand_oacc_for): Handle calls to internal
865 functions like regular functions.
867 2018-04-12 Richard Biener <rguenther@suse.de>
870 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
871 for the early LTO debug to properly generate references to it
872 during DIE emission. Do not re-use that for the skeleton for
874 (dwarf2out_early_finish): Likewise.
876 2018-04-12 Jakub Jelinek <jakub@redhat.com>
880 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
881 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
882 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
883 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
884 and output is a reg, avoid creating invalid lowpart subreg, but
885 instead split into a 512-bit move. Don't split if not AVX512VL,
886 input is xmm16+ reg and output is a mem.
887 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
888 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
889 xmm16+ reg and output is a mem.
891 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
893 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
894 also for flag_dwarf2_cfi_asm.
896 2018-04-12 Jakub Jelinek <jakub@redhat.com>
898 PR rtl-optimization/85342
899 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
900 a bool scalar var inside of the loop instead. Don't try to update
901 recog_data.operand after failed apply_change_group.
903 2018-04-12 Tom de Vries <tom@codesourcery.com>
906 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
907 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
908 array with flexible array member as array without given dimension.
909 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
910 argument for undefined param to true.
912 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
915 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
916 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
917 from PowerPC section.
918 * config/rs6000/sysv4.opt (mcall-): Improve help text.
919 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
920 help text that is too long.
921 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
922 help text that is too long.
923 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
924 help text that is too long.
926 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
928 * config/alpha/alpha.md (stack_probe_internal): Rename
929 from "probe_stack". Update all callers.
931 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
933 PR rtl-optimization/84566
934 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
935 sched_macro_fuse_insns.
937 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
940 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
941 (compute_block_dependences): ... from here.
943 2018-04-11 Jakub Jelinek <jakub@redhat.com>
945 PR tree-optimization/85331
946 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
947 from int to HOST_WIDE_INT.
949 2018-04-11 Martin Jambor <mjambor@suse.cz>
952 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
953 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
954 not the same as the source val.
955 (cgraph_edge_brings_value_p): New parameter.
956 (gather_edges_for_value): Pass destination value to
957 cgraph_edge_brings_value_p.
958 (perhaps_add_new_callers): Likewise.
959 (get_info_about_necessary_edges): Likewise and exclude values brought
960 only by self-recursive edges.
961 (create_specialized_node): Redirect only clones of self-calling edges.
962 (+self_recursive_pass_through_p): New function.
963 (find_more_scalar_values_for_callers_subset): Use it.
964 (find_aggregate_values_for_callers_subset): Likewise.
965 (known_aggs_to_agg_replacement_list): Removed.
966 (decide_whether_version_node): Re-calculate known constants for all
967 remaining context clones.
969 2018-04-11 Richard Biener <rguenther@suse.de>
972 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
973 from early DWARF output.
974 (dwarf2out_early_finish): Output line info unconditionally into
975 early DWARF and add reference to it.
977 2018-04-11 Jakub Jelinek <jakub@redhat.com>
980 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
981 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
982 other than V2DFmode using iptr mode attribute.
983 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
985 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
987 PR rtl-optimization/84659
988 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
990 2018-04-11 Jakub Jelinek <jakub@redhat.com>
993 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
995 (output_loc_list): Pass address of a dummy size variable even in the
996 locview handling loop.
997 (index_location_lists): Add comment on why skip_loc_list_entry can't
1000 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
1003 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
1006 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
1009 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1010 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
1011 and -mstring-compare-inline-limit.
1013 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1016 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
1017 for stack clash protection in a register whenever we need it to be in
1020 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1022 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
1023 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
1025 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1028 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
1030 (mlong-double-): Ditto.
1031 * config/rs6000/sysv4.opt (msdata=): Ditto.
1032 (mtls-size=): Ditto.
1034 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
1036 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1037 erroneous entries for
1038 "vector int vec_ldl (int, long int *)", and
1039 "vector unsigned int vec_ldl (int, unsigned long int *)".
1040 Add comments and entries for
1041 "vector bool char vec_ldl (int, bool char *)",
1042 "vector bool short vec_ldl (int, bool short *)",
1043 "vector bool int vec_ldl (int, bool int *)",
1044 "vector bool long long vec_ldl (int, bool long long *)",
1045 "vector pixel vec_ldl (int, pixel *)",
1046 "vector long long vec_ldl (int, long long *)",
1047 "vector unsigned long long vec_ldl (int, unsigned long long *)".
1048 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
1049 type tree bool_long_long_type_node and correct definition of
1050 bool_V2DI_type_node to make reference to this new type tree.
1051 (rs6000_mangle_type): Replace erroneous reference to
1052 bool_long_type_node with bool_long_long_type_node.
1053 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
1054 comments to emphasize sign distinctions for char and int types and
1055 replace RS6000_BTI_bool_long constant with
1056 RS6000_BTI_bool_long_long constant. Also add comment to restrict
1057 use of RS6000_BTI_pixel.
1058 (bool_long_type_node): Remove this macro definition.
1059 (bool_long_long_type_node): New macro definition
1061 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1063 PR rtl-optimization/85300
1064 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
1065 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
1066 simplify_unary_operation fails.
1068 2018-04-10 Martin Liska <mliska@suse.cz>
1070 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
1071 cgraph_edge and ipa_ref.
1073 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1077 * config/i386/sse.md
1078 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
1079 computation of the VEC_MERGE selector from mask.
1080 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
1081 Fix decoding of the VEC_MERGE selector into mask.
1083 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1085 PR tree-optimization/85286
1086 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
1088 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1090 * final.c (final_1): Set insn_last_address as well as
1091 insn_current_address.
1093 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1096 * explow.c (emit_stack_probe): Call validize_mem on memory location
1097 before passing it to gen_probe_stack. Create address operand and
1098 legitimize it for the probe_stack_address case.
1100 2018-04-09 Jan Hubicka <jh@suse.cz>
1103 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
1104 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
1105 * tree.c (free_lang_data_in_type): Fix handling of binfos;
1107 (free_lang_data): Rebuild type inheritance graph.
1109 2018-04-09 Martin Sebor <msebor@redhat.com>
1111 * invoke.texi (-finline-small-functions): Mention other optimization
1113 (-findirect-inlining, -fpartial-inlining): Same.
1114 (-finline-functions-called-once): Same.
1115 (-freorder-blocks-and-partition): Same.
1117 2018-04-09 Jan Hubicka <jh@suse.cz>
1120 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
1121 jumps; choose last target that matches the criteria (i.e.
1122 no partition changes for non-crossing jumps).
1123 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
1124 support for redirecting crossing jumps to non-crossing.
1126 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
1128 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
1129 also for naked functions.
1131 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
1133 * config/arc/arc.md (add_shift): New pattern.
1134 (add_shift2): Likewise.
1135 (sub_shift): Likewise.
1136 (sub_shift_cmp0_noout): Likewise.
1137 (compare_si_ashiftsi): Likewise.
1138 (xbfu_cmp0_noout): New combine pattern.
1139 (xbfu_cmp0"): Likewise.
1140 (movsi_set_cc_insn): Place the predicable variant first.
1141 (commutative_binary_cmp0_noout): Remove clobber.
1142 (commutative_binary_cmp0): New pattern.
1143 (noncommutative_binary_cmp0): Likewise.
1144 (noncommutative_binary_cmp0_noout): Likewise.
1145 (noncommutative_binary_comparison_result_used): Removed.
1146 (rsub_cmp0): New pattern.
1147 (rsub_cmp0_noout): Likewise.
1148 (extzvsi): Changed, keep only meaningful variants.
1149 (SQH, SEZ): New iterators.
1150 (SQH_postfix): New mode attribute.
1151 (SEZ_prefix): New code attribute.
1152 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
1153 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
1154 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
1156 (noncommutative_operator): Check the availability of barrel
1159 2018-04-09 Richard Biener <rguenther@suse.de>
1161 PR tree-optimization/85284
1162 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
1163 Only use the niter constraining form of simple_iv when the exit
1166 2018-04-09 Tom de Vries <tom@codesourcery.com>
1169 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1170 (define_expand "*memory_barrier"): New define_expand.
1171 (define_insn "memory_barrier"): New insn.
1173 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1175 PR rtl-optimization/80463
1176 PR rtl-optimization/83972
1177 PR rtl-optimization/83480
1179 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
1180 correct producer for the insn.
1181 (tidy_control_flow): Fixup seqnos in case of debug insns.
1183 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1185 PR rtl-optimization/83913
1187 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
1188 different sched-times when merging exprs.
1190 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1192 PR rtl-optimization/83962
1194 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
1195 tidy_fallthru_edge and tidy_control_flow.
1197 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1199 PR rtl-optimization/83530
1201 * sel-sched.c (force_next_insn): New global variable.
1202 (remove_insn_for_debug): When force_next_insn is true, also leave only
1203 next insn in the ready list.
1204 (sel_sched_region): When the region wasn't scheduled, make another pass
1205 over it with force_next_insn set to 1.
1207 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1209 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
1211 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1212 for interrupt control.
1213 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1214 functions for interrupt control.
1215 * config/nds32/nds32-intrinsic.md: Likewise.
1216 * config/nds32/nds32_intrinsic.h: Likewise.
1217 * config/nds32/nds32.h (nds32_builtins): Likewise.
1219 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1221 * config/nds32/nds32.c (nds32_init_machine_status,
1222 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
1223 strict_aligned_p field.
1224 (nds32_expand_to_rtl_hook): New function.
1225 (TARGET_EXPAND_TO_RTL_HOOK): Define.
1226 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
1228 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1229 Chung-Ju Wu <jasonwucj@gmail.com>
1231 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
1232 * config/nds32/nds32-n7.md: New file.
1233 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
1234 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
1236 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
1237 * config/nds32/nds32.md (pipeline_model): Add n7.
1238 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
1239 * config/nds32/pipelines.md: Include n7 settings.
1241 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1242 Chung-Ju Wu <jasonwucj@gmail.com>
1244 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
1245 * config/nds32/nds32-e8.md: New file.
1246 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
1247 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
1249 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
1250 * config/nds32/nds32.md (pipeline_model): Add e8.
1251 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
1252 * config/nds32/pipelines.md: Include e8 settings.
1254 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1255 Chung-Ju Wu <jasonwucj@gmail.com>
1257 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
1258 * config/nds32/nds32-n8.md: New file.
1259 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
1260 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
1262 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
1263 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
1264 * config/nds32/nds32.md (pipeline_model): Add n8.
1265 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
1266 * config/nds32/pipelines.md: Include n8 settings.
1268 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1269 Chung-Ju Wu <jasonwucj@gmail.com>
1271 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
1272 * config/nds32/nds32-n9-2r1w.md: New file.
1273 * config/nds32/nds32-n9-3r2w.md: New file.
1274 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
1275 nds32_register_ports): New or modify for cpu n9.
1276 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
1278 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
1279 * config/nds32/nds32-utils.c: New file.
1280 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
1281 TARGET_MUL_SLOW): Define.
1282 * config/nds32/nds32.md (pipeline_model): New attribute.
1283 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
1284 New options that support cpu n9.
1285 * config/nds32/pipelines.md: Include n9 settings.
1286 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
1288 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1290 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
1291 information if necessary.
1292 (output_cond_branch_compare_zero): Likewise.
1293 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
1294 (nds32_target_alignment): Refine for alignment.
1295 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
1296 (FUNCTION_BOUNDARY): Modify.
1297 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
1299 * config/nds32/nds32.opt (malways-align, malign-functions): New.
1301 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1303 * config/nds32/constants.md (unspec_volatile_element): Add values for
1304 TLB operation and data prefetch.
1305 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1306 functions for TLB operation and data prefetch.
1307 * config/nds32/nds32-intrinsic.md: Likewise.
1308 * config/nds32/nds32_intrinsic.h: Likewise.
1309 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
1310 (nds32_print_operand): Likewise.
1311 * config/nds32/nds32.h (nds32_builtins): Likewise.
1313 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
1314 Andrew Pinski <pinsika@gcc.gnu.org>
1317 * match.pd: Use constant_boolean_node of correct type instead of
1318 boolean_true_node or boolean_false_node for simplifying
1319 pointer comparisons to zero.
1321 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1323 PR tree-optimization/80021
1324 * tree.c (verify_type_variant): Make error call in verify_variant_match
1325 translatable and remove final full stop.
1327 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1329 * config/nds32/constants.md (unspec_volatile_element): Add
1330 UNSPEC_VOLATILE_EH_RETURN.
1331 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
1332 nds32_output_stack_pop): Support dwarf exception handling process.
1333 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
1334 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
1335 exception handling process.
1336 (nds32_compute_stack_frame): Likewise.
1337 (nds32_return_addr_rtx): Likewise.
1338 (nds32_initial_elimination_offset): Likewise.
1339 (nds32_expand_prologue): Likewise.
1340 (nds32_expand_epilogue): Likewise.
1341 (nds32_dynamic_chain_address): New function.
1342 * config/nds32/nds32.h (machine_function): Add fields for dwarf
1344 (DYNAMIC_CHAIN_ADDRESS): Define.
1345 (EH_RETURN_DATA_REGNO): Define.
1346 (EH_RETURN_STACKADJ_RTX): Define.
1347 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
1348 patterns for dwarf exception handling.
1350 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1352 * config/nds32/nds32.h: Clean up obsolete macros.
1354 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1356 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1357 Add enum values for particular instructions.
1358 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1359 particular intrinsic functions.
1360 * config/nds32/nds32-intrinsic.md: Likewise.
1361 * config/nds32/nds32_intrinsic.h: Likewise.
1362 * config/nds32/nds32.h (nds32_builtins): Likewise.
1363 * config/nds32/nds32.md (type): Add pbsad and pbsada.
1364 (btst, ave): New patterns for particular instructions.
1366 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1368 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1369 Add enum values for atomic load/store and memory sync.
1370 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
1372 * config/nds32/nds32-intrinsic.md: Likewise.
1373 * config/nds32/nds32_intrinsic.h: Likewise.
1374 * config/nds32/nds32.h (nds32_builtins): Likewise.
1376 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1378 PR tree-optimization/85257
1379 * fold-const.c (native_encode_vector): If not all elts could fit
1380 and off is -1, return 0 rather than offset.
1381 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
1382 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
1383 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
1384 adjust buffer in native_interpret_expr call.
1386 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1388 * config/nds32/constants.md (unspec_volatile_element): Add cache
1389 control enum values.
1390 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
1391 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
1392 * config/nds32/nds32.c (nds32_cctl_names): New.
1393 (nds32_print_operand): Handle cache control register names.
1394 * config/nds32/nds32.h (nds32_builtins): New enum values.
1395 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
1397 * config/nds32/nds32.md (type): Add mmu.
1398 * config/nds32/pipelines.md (simple_insn): Add mmu.
1400 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1402 * config/nds32/nds32.md (type): Remove call.
1403 * config/nds32/pipelines.md (simple_insn): Likewise.
1405 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1407 * config/nds32/constants.md (unspec_volatile_element): Add
1408 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
1409 UNSPEC_VOLATILE_FMFCFG.
1410 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
1411 description for fmfcfg and fmfcsr.
1412 (bdesc_1arg): Add fmtcsr.
1413 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
1414 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
1415 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
1416 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
1417 unspec_fmfcfg): New patterns.
1418 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
1419 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
1420 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
1421 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
1422 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
1423 __nds32__fmfcfg): Define.
1425 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1427 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
1428 intrinsic register names.
1429 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
1430 intrinsic register enum values and macros.
1432 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1434 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
1435 for load/store addressing form.
1436 (nds32_print_operand_address): Likewise.
1438 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
1441 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
1442 based on LABEL_REF. Remove useless assertion.
1443 (pic_address_needs_scratch): Fix formatting.
1444 (sparc_legitimize_pic_address): Minor tweaks.
1445 (sparc_delegitimize_address): Adjust assertion accordingly.
1446 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
1447 into symbolic_operand.
1448 (movsi_high_pic_label_ref): Likewise.
1449 (movsi_lo_sum_pic_label_ref): Likewise.
1450 (movdi_pic_label_ref): Likewise.
1451 (movdi_high_pic_label_ref): Likewise.
1452 (movdi_lo_sum_pic_label_ref): Likewise.
1454 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
1456 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
1457 custom LIB_SPEC setup.
1459 2018-04-06 Ruslan Bukin <br@bsdpad.com>
1460 Kito Cheng <kito.cheng@gmail.com>
1462 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
1463 * config/riscv/freebsd.h: New.
1465 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1467 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
1468 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
1471 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1472 Kito Cheng <kito.cheng@gmail.com>
1474 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
1475 nds32_output_call, nds32_symbol_binds_local_p): New functions.
1476 * config/nds32/nds32-protos.h (nds32_output_call,
1477 nds32_output_return): Declare.
1478 * config/nds32/nds32.md: Refine all the call and return patterns.
1480 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1483 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
1484 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
1486 PR rtl-optimization/84872
1487 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
1488 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
1491 2018-04-06 Tamar Christina <tamar.christina@arm.com>
1493 * expr.c (copy_blkmode_to_reg): Revert 254862.
1494 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
1496 2018-04-06 Richard Biener <rguenther@suse.de>
1499 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
1500 after seeing a component reference with an adjacent field. Treat
1501 refs to arrays at struct end of external decls similar to
1502 refs to unconstrained commons.
1504 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1507 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
1508 look through SAVE_EXPRs with non-side-effects argument. Adjust
1510 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
1513 2018-04-06 Richard Biener <rguenther@suse.de>
1516 * alias.c (find_base_term): New wrapper around find_base_term
1517 unwinding CSELIB_VAL_PTR changes.
1518 (find_base_term): Do not restore CSELIB_VAL_PTR during the
1521 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1523 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
1525 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
1526 constant definitions.
1527 ("nop"): lr 0,0 -> nopr r0
1528 ("nop_lr0", "nop_lr1"): New insn definitions.
1530 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1532 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
1533 NDS32_V3PUSH_AVAILABLE_P macro.
1535 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
1536 Chung-Ju Wu <jasonwucj@gmail.com>
1538 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
1539 (nds32*-*-*): Add float and fpu_config into supported_defaults.
1540 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
1541 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
1542 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
1543 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
1544 * config/nds32/constraints.md: New constraints and checking for hard
1545 float configuration.
1546 * config/nds32/iterators.md: New mode iterator and attribute for hard
1547 float configuration.
1548 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
1550 * config/nds32/nds32-fpu.md: New file.
1551 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
1552 deal with hard float code generation.
1553 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
1555 (abi_type, float_reg_number): New enum type.
1556 * config/nds32/nds32-predicates.c: New predicates for hard float.
1557 * config/nds32/nds32-protos.h: Declare functions for hard float.
1558 * config/nds32/nds32.c: Implementation for hard float configuration.
1559 * config/nds32/nds32.h: Definitions for hard float configuration.
1560 * config/nds32/nds32.md: Include hard float machine description and
1561 modify patterns for hard float configuration.
1562 * config/nds32/nds32.opt: New options for hard float configuration.
1563 * config/nds32/predicates.md: New predicates for hard float
1566 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1568 * common/config/nds32/nds32-common.c
1569 (nds32_option_optimization_table): Enable -mreleax-hint by default.
1571 2018-04-05 Jakub Jelinek <jakub@redhat.com>
1574 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
1575 CONSTRUCTOR_ELT (ctor, ...)->value.
1577 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
1580 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
1582 2018-04-05 Tom de Vries <tom@codesourcery.com>
1585 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
1588 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1589 Kito Cheng <kito.cheng@gmail.com>
1591 * config/nds32/constraints.md (U33): Fine-tune checking condition.
1592 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
1593 * config/nds32/nds32.h (nds32_16bit_address_type): Add
1594 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
1596 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1597 Kito Cheng <kito.cheng@gmail.com>
1599 * config/nds32/constraints.md (Ufe): New memory constraint.
1600 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
1601 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
1602 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
1604 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
1605 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
1607 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1609 * config/nds32/nds32.md: Use optimize_size in the condition for
1610 alu-shift instructions.
1612 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1614 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
1616 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1618 * config/nds32/nds32.md (negsi2): Refine pattern.
1620 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1621 Chung-Ju Wu <jasonwucj@gmail.com>
1623 * config/nds32/iterators.md (shift_rotate): New code iterator.
1624 (shift): New code attribute.
1625 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
1626 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
1627 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
1628 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
1629 bit-wise operations.
1630 (andsi3, *andsi3): Ditto.
1631 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
1632 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
1633 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
1634 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
1635 nds32_ior_operand, nds32_xor_operand): New predicates.
1637 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1639 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
1640 (addsi3, subsi3): ... this.
1642 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1644 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
1646 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1648 * config/nds32/nds32.md: Adjust indention.
1650 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1652 * config/nds32/nds32.md (feature): New attribute.
1654 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1656 * config/nds32/nds32.md (subtype): New attribute.
1658 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
1661 * config/arm/arm-builtins.c (arm_expand_builtin): Change
1662 expansion to perform a bitwise AND of the argument followed by a
1663 boolean negation of the result.
1665 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
1667 PR rtl-optimization/84878
1668 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
1669 the basic block. Assert the use reference is not artificial and that
1670 it has an associated insn.
1672 2018-04-04 Michael Matz <matz@suse.de>
1674 * builtins.c (compute_objsize): Pass correct operand
1675 to array_at_struct_end_p.
1677 2018-04-04 Richard Biener <rguenther@suse.de>
1680 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
1681 from contexts for DINFO_LEVEL_TERSE and below.
1683 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1685 * config/nds32/nds32-doubleword.md (move_<mode>): Require
1686 resiter_operand condition.
1687 * config/nds32/nds32.md (*move<mode>): Ditto.
1689 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1690 Monk Chiang <sh.chiang04@gmail.com>
1692 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
1694 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1696 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
1698 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1699 Kito Cheng <kito.cheng@gmail.com>
1701 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
1702 nds32_cond_code_str, output_cond_branch,
1703 output_cond_branch_compare_zero, nds32_expand_cbranch,
1704 nds32_expand_cstore, nds32_expand_movcc,
1705 nds32_output_cbranchsi4_equality_zero,
1706 nds32_output_cbranchsi4_equality_reg,
1707 nds32_output_cbranchsi4_equality_reg_or_const_int,
1708 nds32_output_cbranchsi4_greater_less_zero: New functions.
1709 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
1710 nds32_expand_cstore, nds32_expand_movcc,
1711 nds32_output_cbranchsi4_equality_zero,
1712 nds32_output_cbranchsi4_equality_reg,
1713 nds32_output_cbranchsi4_equality_reg_or_const_int,
1714 nds32_output_cbranchsi4_greater_less_zero): Declare.
1715 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
1716 nds32_rimm11s_operand): New predicates.
1717 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
1718 * config/nds32/nds32.md: Rewrite all the branch and conditional move
1721 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1723 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
1724 * config/nds32/nds32.md: Ditto.
1725 * config/nds32/pipelines.md: Ditto.
1727 2018-04-04 Richard Biener <rguenther@suse.de>
1729 PR tree-optimization/85168
1730 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
1731 propagating abnormals.
1733 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1735 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
1737 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1738 Kito Cheng <kito.cheng@gmail.com>
1740 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
1741 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
1742 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
1743 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
1744 * config/nds32/nds32.md (sibcall_internal): New.
1745 (sibcall_register): Remove.
1746 (sibcall_immediate): Remove.
1747 (sibcall_value_internal): New.
1748 (sibcall_value_register): Remove.
1749 (sibcall_value_immediate): Remove.
1750 * config/nds32/predicates.md (nds32_general_register_operand): New.
1751 (nds32_call_address_operand): New.
1753 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1755 PR rtl-optimization/85167
1756 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
1757 bb_defs if *split_p, instead preinitialize it to NULL.
1759 PR tree-optimization/85156
1760 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
1761 evaluating the argument multiple times.
1763 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
1765 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
1767 (_mm_cvtpd_ps): Likewise.
1768 (_mm_cvttpd_epi32): Likewise.
1769 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
1770 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
1771 vector, pixel, and bool following altivec.h include.
1773 2018-04-03 Martin Sebor <msebor@redhat.com>
1775 * doc/extend.texi (Common Function Attributes): Clarify.
1776 (const attribute): Likewise.
1777 (pure attribute): Likewise.
1779 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1782 * config/i386/i386.c (ix86_expand_vector_set): Use
1783 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
1785 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
1787 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
1788 instructions when changing rounding bits to preserve precision bits
1789 in the x87 control word.
1791 2018-04-03 Martin Liska <mliska@suse.cz>
1793 PR tree-optimization/82491
1794 * rtl.h (strip_offset_and_add): Replace += suboffset with
1795 poly_uint64 () + suboffset.
1797 2018-03-29 Martin Liska <mliska@suse.cz>
1798 Martin Jambor <mjambor@suse.cz>
1801 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
1802 param_type is not an integral or pointer type.
1804 2018-04-03 Richard Biener <rguenther@suse.de>
1806 * sese.h (recompute_all_dominators): Remove.
1808 2018-04-02 Martin Sebor <msebor@redhat.com>
1810 * doc/invoke.texi (-Wrestrict): Fix typos.
1812 2018-04-02 Jim Wilson <jimw@sifive.com>
1814 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
1815 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
1816 (<optab>di3, <optab>si3_extend): Likewise.
1817 (<optab>si3_mask, <optab>si3_mask_1): New.
1818 (<optab>di3_mask, <optab>di3_mask_1): New.
1819 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
1820 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
1821 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
1823 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
1825 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
1828 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
1830 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
1831 (nds32_canonicalize_comparison): New function.
1833 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1834 Kito Cheng <kito.cheng@gmail.com>
1835 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1837 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
1838 * config/nds32/constants.md (unspec_volatile_element): Add
1839 UNSPEC_VOLATILE_RELAX_GROUP.
1840 * config/nds32/nds32-relax-opt.c: New file.
1841 * config/nds32/nds32-predicates.c
1842 (nds32_symbol_load_store_p): New function.
1843 * config/nds32/nds32-protos.h
1844 (nds32_symbol_load_store_p): Declare function.
1845 (make_pass_nds32_relax_opt): Declare new rtl pass function.
1846 * config/nds32/nds32.c
1847 (nds32_register_pass): New function to register pass.
1848 (nds32_register_passes): New function to register passes.
1849 * config/nds32/nds32.md (relax_group): New pattern.
1850 * config/nds32/nds32.opt (mrelax-hint): New option.
1851 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
1853 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
1855 * config/nds32/t-nds32: Modify files dependency.
1857 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1859 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
1860 (PROFILE_HOOK): Define its implementation.
1862 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1864 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
1865 type and 32-bit size.
1867 2018-04-01 Jakub Jelinek <jakub@redhat.com>
1870 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
1871 (V_128_256): New mode iterator.
1872 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
1873 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
1874 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
1876 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
1877 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
1879 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
1882 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
1883 NaN inputs correctly.
1885 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
1888 * config/rs6000/vsx.md (??r): New mode attribute.
1889 (*vsx_mov<mode>_64bit): Use it.
1890 (*vsx_mov<mode>_32bit): Likewise.
1892 2018-03-30 Martin Sebor <msebor@redhat.com>
1894 PR tree-optimization/84818
1895 * builtins.c (check_access): Use warning_n.
1897 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1900 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
1902 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
1905 2018-03-30 Julia Koval <julia.koval@intel.com>
1908 * x86-tune.def (movx, partial_reg_dependency): Enable for
1911 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
1914 * lra-constraints.c (process_alt_operands): Move setting
1915 this_alternative_matches below.
1917 2018-03-29 Martin Liska <mliska@suse.cz>
1920 * doc/invoke.texi: Document how LTO works with debug info.
1921 Describe auto-load support of binutils. Mention 'x86-64'
1922 as valid option value of -march option.
1924 2018-03-29 Jakub Jelinek <jakub@redhat.com>
1926 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
1929 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
1930 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
1931 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
1934 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1937 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
1938 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
1939 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
1940 for RS6000_BTM_POWERPC64.
1941 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
1942 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
1943 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
1948 2018-03-28 Carl Love <cel@us.ibm.com>
1951 2017-09-27 Carl Love <cel@us.ibm.com>
1953 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
1954 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
1955 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
1958 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1960 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
1961 instead of __vector bool.
1962 (_mm_max_pu8): Likewise.
1963 (_mm_min_pi16): Likewise.
1965 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1968 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
1969 (DIVWEUO): Likewise.
1971 (DIVDEUO): Likewise.
1972 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
1973 DIVWEUO and DIVDEUO.
1974 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
1975 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
1976 (div_extend): Likewise.
1977 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
1979 (__builtin_divweuo): Likewise.
1980 (__builtin_divdeo): Likewise.
1981 (__builtin_divdeuo): Likewise.
1983 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1986 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
1987 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
1989 PR tree-optimization/82004
1990 * gimple-match-head.c (optimize_pow_to_exp): New function.
1991 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
1992 Don't fold to exp if optimize_pow_to_exp is false.
1994 2018-03-28 Martin Liska <mliska@suse.cz>
1997 * calls.c (initialize_argument_information): Fix trailing space.
1998 * common.opt: Fix typo and provide better explanation for
1999 -fsanitize-coverage option.
2000 * config/i386/i386.opt: Fix typo.
2002 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2003 Martin Liska <mliska@suse.cz>
2006 * gimplify.c (asan_poison_variable): Don't do the check for
2007 gimplify_omp_ctxp here.
2008 (gimplify_decl_expr): Do it here.
2009 (gimplify_target_expr): Likewise.
2011 2018-03-28 Martin Liska <mliska@suse.cz>
2014 * config/i386/i386.c (ix86_function_arg_advance): Do not call
2015 chkp_type_bounds_count if MPX is not enabled.
2017 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
2019 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
2021 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
2024 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
2025 function to create the function decl for complex long double
2026 multiply and divide for -mabi=ieeelongdouble.
2027 (init_float128_ieee): Call it.
2029 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
2032 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
2033 -fcf-protection=branch -mibt.
2034 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
2036 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2039 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
2041 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
2044 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
2045 extern array declarations.
2047 2018-03-27 Richard Biener <rguenther@suse.de>
2050 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
2051 explicit single_use checks.
2053 2018-03-27 Richard Biener <rguenther@suse.de>
2055 PR tree-optimization/85082
2056 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
2059 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2061 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
2062 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
2063 Turn on fasynchronous-unwind-tables and funwind-tables.
2065 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
2068 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
2069 (*bmi_blsr_<mode>_ccz): Ditto.
2071 2018-03-26 Tom de Vries <tom@codesourcery.com>
2073 PR tree-optimization/85063
2074 * omp-general.c (offloading_function_p): New function. Factor out
2076 * omp-offload.c (pass_omp_target_link::gate): ... here.
2077 * omp-general.h (offloading_function_p): Declare.
2078 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
2079 with attribute omp declare target for offloading functions.
2081 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
2083 PR tree-optimization/84005
2084 * tree-data-ref.h (get_base_for_alignment): Declare.
2085 * tree-data-ref.c (get_base_for_alignment_1): New function.
2086 (get_base_for_alignment): Likewise.
2087 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
2088 get_base_for_alignment to find a suitable base object, instead
2089 of always using drb->base_address.
2091 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2094 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
2095 known size by default.
2097 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
2100 * lra-constraints.c (process_alt_operands): Don't match BLKmode
2101 and non BLKmode operands.
2103 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2106 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
2107 Clean up attributes.
2109 2018-03-23 Richard Biener <rguenther@suse.de>
2112 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
2113 we are going to emit early debug for LTO.
2115 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2118 * function.c (match_asm_constraints_1): Don't optimize if input
2119 doesn't satisfy general_operand predicate for output's mode.
2122 * alias.c (write_dependence_p): Don't require for x_canonicalized
2123 non-VOIDmode if x has VOIDmode.
2126 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
2127 just don't try to optimize it rather than assert it never happens.
2129 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2131 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
2132 macro expansions for definition of ST_INTERNAL_<mode> and
2133 LD_INTERNAL_<mode> builtins.
2134 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
2136 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
2138 (altivec_expand_st_builtin): Likewise.
2139 (altivec_expand_builtin): Remove calls to deleted functions.
2140 (rs6000_address_for_altivec): Delete this function.
2141 * config/rs6000/vector.md: Remove expands for
2142 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
2144 2018-03-22 Sudakshina Das <sudi.das@arm.com>
2147 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
2148 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
2149 re-computing once computed.
2150 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
2151 (arm_init_machine_status): Initialize
2152 machine->static_chain_stack_bytes.
2154 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2157 * doc/extend.texi: Add four new prototypes for vec_ld.
2158 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
2159 definitions for more logical presentation.
2160 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
2161 entries for V1TI variants of __builtin_altivec_ld builtin.
2162 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
2163 handling of V1TI variant of LVX icode pattern.
2164 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
2165 (rs6000_gimple_fold_builtin): Likewise.
2166 (altivec_init_builtins): Add code to define
2167 __builtin_altivec_lvx_v1ti function.
2169 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2172 * function.c (match_asm_constraints_1): Don't do the optimization
2173 if input isn't a REG, SUBREG, MEM or constant.
2175 2018-03-22 Tom de Vries <tom@codesourcery.com>
2177 PR tree-optimization/84956
2178 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
2179 bb_has_abnormal_pred.
2181 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2184 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
2185 DECL_INITIAL (decl) to decl at the end.
2186 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
2189 2018-03-21 Joseph Myers <joseph@codesourcery.com>
2191 * doc/extend.texi (__builtin_tgmath): Document when complex
2192 integer types are treated as _Complex _Float64.
2194 2018-03-21 Tom de Vries <tom@codesourcery.com>
2196 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
2198 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2200 PR tree-optimization/84960
2201 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
2202 if it is ENTRY block, move them into single succ of ENTRY in that case.
2204 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
2206 PR tree-optimization/84811
2207 * poly-int.h (poly_span_traits): Remove the T3 parameter and
2208 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
2209 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
2210 (known_subrange_p): Update accordingly. Cast each value involved
2211 in the size comparison, rather than casting the result of the
2214 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2216 PR tree-optimization/84982
2217 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
2218 by flipping the least significant bit rather than all bits from
2219 bitpos to bitpos + bitsize - 1.
2221 2018-03-21 Nathan Sidwell <nathan@acm.org>
2223 * doc/extend.texi (Deprecated Features): Remove mention of
2224 long-deleted deprecations.
2226 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2229 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
2230 * configure: Regenerate.
2232 2018-03-21 Tom de Vries <tom@codesourcery.com>
2234 PR tree-optimization/83126
2235 * tree-parloops.c (num_phis): New function.
2236 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
2238 2018-03-21 Nathan Sidwell <nathan@acm.org>
2240 * doc/extend.texi (Deprecated Features): Update deprecated flags,
2241 mention anon-struct/union members and trailing attributes.
2243 2018-03-21 Bin Cheng <bin.cheng@arm.com>
2245 PR tree-optimization/84969
2246 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
2247 builtin memset partitions if they set different rhs values.
2249 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2251 PR rtl-optimization/84989
2252 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
2253 VEC_DUPLICATE with scalar result mode.
2255 2018-03-21 Martin Liska <mliska@suse.cz>
2258 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
2259 not intended return statement.
2261 2018-03-21 Martin Liska <mliska@suse.cz>
2264 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
2265 (chkp_find_bound_slots_1): Limit number of iterations.
2267 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
2270 * Minor grammar fixes for x86 options.
2272 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2275 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
2276 holding REG_CFA_RESTORE notes, instead turn them into a USE.
2278 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
2281 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
2282 (altivec_lvx_<mode>_1op): Likewise.
2283 (altivec_stvx_<mode>_2op): Likewise.
2284 (altivec_stvx_<mode>_1op): Likewise.
2285 (altivec_lvx_<VM2:mode>): New define_expand.
2286 (altivec_stvx_<VM2:mode>): Likewise.
2287 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
2288 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2289 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
2290 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2291 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
2292 (rs6000_gen_lvx): Likewise.
2293 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
2294 (altivec_expand_stv_builtin): Likewise.
2295 (altivec_expand_builtin): Likewise.
2296 * config/rs6000/vector.md: Likewise.
2298 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2301 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
2304 2018-03-20 Richard Biener <rguenther@suse.de>
2307 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
2308 sign-conversions as zero, fall back to standard scalar_stmt
2311 2018-03-20 Martin Liska <mliska@suse.cz>
2314 * predict.c (rebuild_frequencies): Handle case when we have
2315 PROFILE_ABSENT, but flag_guess_branch_prob is false.
2317 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2320 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
2321 flag_section_anchors.
2322 * varasm.c (use_blocks_for_decl_p): Remove hack for
2323 dw2_force_const_mem.
2326 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
2328 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
2329 be created, use lowpart_subreg of operands[0] rather than operands[0]
2331 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
2332 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
2333 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
2334 and n constraint instead of aarch64_shift_imm_di and Usd.
2335 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
2336 (*aarch64_<optab>_reg_minus<mode>3): ... this.
2338 2018-03-20 Sudakshina Das <sudi.das@arm.com>
2341 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
2342 to favor GPR over NEON registers.
2343 (<shift>di3_neon): Likewise.
2345 2018-03-20 Tom de Vries <tom@codesourcery.com>
2348 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
2349 (nvptx_process_pars): Emit bar.sync asap and alap.
2351 2018-03-20 Tom de Vries <tom@codesourcery.com>
2354 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
2355 seen_label if seen_label is already set.
2357 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2360 * config/i386/i386.c (fold_builtin_cpu): For features above 31
2361 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
2362 Use 1U instead of 1. Formatting fixes.
2365 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
2366 instead of TREE_TYPE (s1) for the return value.
2368 2018-03-19 Jakub Jelinek <jakub@redhat.com>
2370 PR tree-optimization/84946
2371 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
2372 bitsize + bitsize in poly_uint64 rather than poly_int64.
2375 * dwarf2asm.c: Include fold-const.c.
2376 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
2377 of decl rather than decl itself.
2379 PR rtl-optimization/84643
2380 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
2382 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
2385 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
2386 calling assemble_variable.
2388 2018-03-19 Sudakshina Das <sudi.das@arm.com>
2391 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
2392 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
2394 2018-03-19 Jim Wilson <jimw@sifive.com>
2397 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
2398 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
2399 (riscv_first_stack_step): Likewise.
2400 (riscv_option_override): Use STACK_BOUNDARY instead of
2402 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
2404 (BIGGEST_ALIGNMENT): Set to 128.
2405 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
2406 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
2409 2018-03-19 Richard Biener <rguenther@suse.de>
2411 PR tree-optimization/84933
2412 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
2413 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
2415 2018-03-19 Richard Biener <rguenther@suse.de>
2417 PR tree-optimization/84859
2418 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
2419 (cond_if_else_store_replacement): Perform sinking operation on
2420 single-store BBs regardless of MAX_STORES_TO_SINK setting.
2421 Generalize what a BB with a single eligible store is.
2423 2018-03-19 Richard Biener <rguenther@suse.de>
2425 PR tree-optimization/84929
2426 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
2427 chrec_is_positive against non-chrec arg.
2429 2018-03-19 Tamar Christina <tamar.christina@arm.com>
2432 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
2434 2018-03-18 Martin Liska <mliska@suse.cz>
2436 PR rtl-optimization/84635
2437 * regrename.c (build_def_use): Use matches_mode only when
2440 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
2442 PR tree-optimization/84913
2443 * tree-vect-loop.c (vectorizable_reduction): Don't try to
2444 vectorize chains of COND_EXPRs.
2446 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2448 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
2450 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2452 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
2454 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2456 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
2458 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2459 Kito Cheng <kito.cheng@gmail.com>
2461 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
2462 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
2463 (nds32_adjust_reg_alloc_order): New function.
2464 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
2466 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
2468 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
2469 nds32_print_operand, nds32_print_operand_address): Use
2470 HOST_WIDE_INT_PRINT_DEC instead.
2472 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2474 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
2476 2018-03-17 Jakub Jelinek <jakub@redhat.com>
2479 * config/i386/i386.c (initial_ix86_tune_features,
2480 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
2482 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
2483 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
2484 rather than 1u << ix86_tune. Formatting fix.
2485 (ix86_option_override_internal): Change ix86_arch_mask from
2486 unsigned int to unsigned HOST_WIDE_INT, initialize to
2487 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
2488 (ix86_function_specific_restore): Likewise.
2490 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2493 * postreload.c (reload_combine_recognize_pattern): Perform
2494 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
2495 truncate_int_for_mode the result for the destination's mode.
2498 * hsa-gen.c (mem_type_for_type): Fix comment typo.
2499 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
2501 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2504 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
2507 * lra-assigns.c (lra_split_hard_reg_for): Don't use
2508 regno_allocno_class_array and sorted_pseudos.
2509 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
2510 insns where regno is used.
2512 2018-03-16 Martin Liska <mliska@suse.cz>
2515 * multiple_target.c (create_dispatcher_calls): Redirect
2516 reference in the symbol table.
2518 2018-03-16 Martin Liska <mliska@suse.cz>
2521 * multiple_target.c (create_dispatcher_calls): Redirect also
2524 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2528 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
2529 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
2530 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
2532 2018-03-16 Julia Koval <julia.koval@intel.com>
2534 * doc/invoke.texi (Skylake Server): Add CLWB.
2535 Cannonlake): Remove CLWB.
2537 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2539 PR tree-optimization/84841
2540 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
2542 (FLOAT_ONE_CONST_TYPE): Define.
2543 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
2544 (sort_by_operand_rank): Put entries with higher constant_type last
2545 rather than first to match comments.
2547 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
2549 * config/nios2/nios2.md (movsi_internal): Fix thinko in
2552 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2555 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
2556 check and use address of target always.
2558 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
2561 * config/i386/i386.c (indirect_thunk_needed): Update comments.
2562 (indirect_thunk_bnd_needed): Likewise.
2563 (indirect_thunks_used): Likewise.
2564 (indirect_thunks_bnd_used): Likewise.
2565 (indirect_return_needed): New.
2566 (indirect_return_bnd_needed): Likewise.
2567 (output_indirect_thunk_function): Add a bool argument for
2569 (output_indirect_thunk_function): Don't generate alias for
2570 function return thunk.
2571 (ix86_code_end): Call output_indirect_thunk_function to generate
2572 function return thunks.
2573 (ix86_output_function_return): Set indirect_return_bnd_needed
2574 and indirect_return_needed instead of indirect_thunk_bnd_needed
2575 and indirect_thunk_needed.
2577 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
2579 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
2580 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
2581 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
2583 2018-03-15 David Malcolm <dmalcolm@redhat.com>
2584 Paul Hua <paul.hua.gm@gmail.com>
2587 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
2589 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
2591 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
2592 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
2595 2018-03-15 Tamar Christina <tamar.christina@arm.com>
2598 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
2599 instead of GET_MODE_SIZE when comparing Units.
2601 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2604 * varasm.c (hash_section): Return an unchangeble hash value
2605 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
2606 Return !aarch64_can_use_per_function_literal_pools_p ().
2608 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2611 * optabs.c (emit_conditional_move): Pass address of cmode's copy
2612 rather than address of cmode as last argument to prepare_cmp_insn.
2614 2018-03-15 Julia Koval <julia.koval@intel.com>
2616 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
2617 F_AVX512VNNI, F_AVX512BITALG): New.
2619 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
2622 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
2623 insn for floating-point loads and stores.
2625 2018-03-14 Carl Love <cel@us.ibm.com>
2627 * config/rs6000/rs6000-c.c: Add macro definitions for
2628 ALTIVEC_BUILTIN_VEC_PERMXOR.
2629 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
2630 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
2631 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
2632 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
2634 * config/doc/extend.texi: Add prototypes for vec_permxor.
2636 2018-03-14 David Malcolm <dmalcolm@redhat.com>
2639 * diagnostic-show-locus.c (class layout_point): Convert m_line
2640 from int to linenum_type.
2641 (line_span::comparator): Use linenum "compare" function when
2642 comparing line numbers.
2643 (test_line_span): New function.
2644 (layout_range::contains_point): Convert param "row" from int to
2646 (layout_range::intersects_line_p): Likewise.
2647 (layout::will_show_line_p): Likewise.
2648 (layout::print_source_line): Likewise.
2649 (layout::should_print_annotation_line_p): Likewise.
2650 (layout::print_annotation_line): Likewise.
2651 (layout::print_leading_fixits): Likewise.
2652 (layout::annotation_line_showed_range_p): Likewise.
2653 (struct line_corrections): Likewise for field m_row.
2654 (line_corrections::line_corrections): Likewise for param "row".
2655 (layout::print_trailing_fixits): Likewise.
2656 (layout::get_state_at_point): Likewise.
2657 (layout::get_x_bound_for_row): Likewise.
2658 (layout::print_line): Likewise.
2659 (diagnostic_show_locus): Likewise for locals "last_line" and
2661 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
2662 * input.c (selftest::test_linenum_comparisons): New function.
2663 (selftest::input_c_tests): Call it.
2664 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
2665 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
2666 * selftest.h (ASSERT_GT): New macro.
2667 (ASSERT_GT_AT): New macro.
2668 (ASSERT_LT): New macro.
2669 (ASSERT_LT_AT): New macro.
2671 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
2673 PR rtl-optimization/84780
2674 * combine.c (distribute_links): Don't make a link based on pc_rtx.
2676 2018-03-14 Martin Liska <mliska@suse.cz>
2678 * tree.c (record_node_allocation_statistics): Use
2679 get_stats_node_kind.
2680 (get_stats_node_kind): New function extracted from
2681 record_node_allocation_statistics.
2682 (free_node): Use get_stats_node_kind.
2684 2018-03-14 Richard Biener <rguenther@suse.de>
2686 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
2687 that the value-set of ANTIC_IN doesn't grow.
2690 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2692 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2693 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2695 2018-03-14 Julia Koval <julia.koval@intel.com>
2697 * config.gcc (icelake-client, icelake-server): New.
2699 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
2700 (initial_ix86_arch_features): Ditto.
2701 (PTA_SKYLAKE): Add SGX.
2702 (PTA_ICELAKE): Remove.
2703 (PTA_ICELAKE_CLIENT): New.
2704 (PTA_ICELAKE_SERVER): New.
2705 (ix86_option_override_internal): Split up icelake on icelake client and
2707 (get_builtin_code_for_version): Ditto.
2708 (fold_builtin_cpu): Ditto.
2709 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
2710 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
2711 * config/i386/i386.h (processor_type): Ditto.
2712 * doc/invoke.texi: Ditto.
2714 2018-03-14 Jakub Jelinek <jakub@redhat.com>
2717 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
2718 INTEGER_CST offset, add it together with bitpos / 8 and
2719 sign extend based on POINTER_SIZE.
2723 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
2726 * config/i386/constraints.md (Yc): New register constraint.
2727 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
2728 Use Yc constraint for alternative 2 of operand 0. Remove
2729 preferred_for_speed attribute.
2731 2018-03-14 Richard Biener <rguenther@suse.de>
2733 PR tree-optimization/84830
2734 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
2735 with the old one to avoid oscillations.
2737 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
2740 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
2742 (assign_by_spills): Return a flag of reload assignment failure.
2743 Do not process the reload assignment failures. Do not spill other
2744 reload pseudos if they has the same reg class. Update n if
2746 (lra_assign): Add a return arg. Set up from the result of
2747 assign_by_spills call.
2748 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2749 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2750 usage_insns if it is not NULL.
2751 (spill_hard_reg_in_range): New function.
2752 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2753 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2754 function prototypes.
2755 (lra_assign): Change prototype.
2756 * lra.c (lra): Add code to deal with fails by splitting hard reg
2759 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
2761 * config/riscv/riscv.opt (mrelax): New option.
2762 * config/riscv/riscv.c (riscv_file_start): Emit ".option
2763 "norelax" when riscv_mrelax is disabled.
2764 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
2766 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2769 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
2770 reassociation for int modes.
2772 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2774 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
2775 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
2777 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
2778 * config/aarch64/aarch64-sve.md
2779 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
2780 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
2781 (*extend<mode><Vwide>2): Rename to...
2782 (aarch64_sve_extend<mode><Vwide>2): ...this.
2783 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
2784 renaming the old pattern to...
2785 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
2787 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
2788 define_expand, renaming the old pattern to...
2789 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
2790 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
2791 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
2792 account when deciding which SVE instruction the optab should use.
2793 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
2795 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2797 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
2798 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
2799 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
2800 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
2801 (tlsdesc_small_<mode>): Turn a define_expand and use
2802 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
2803 (tlsdesc_small_advsimd_<mode>): ...this.
2804 (tlsdesc_small_sve_<mode>): New pattern.
2806 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2808 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
2809 (UNSPEC_UMUL_HIGHPART): New constants.
2810 (MUL_HIGHPART): New int iteraor.
2811 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
2812 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
2814 (*<su>mul<mode>3_highpart): New define_insn.
2816 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2819 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2822 2018-03-13 Martin Liska <mliska@suse.cz>
2825 * (sem_item_optimizer::sem_item_optimizer): Initialize new
2827 (sem_item_optimizer::~sem_item_optimizer): Release it.
2828 (sem_item_optimizer::merge_classes): Register variable aliases.
2829 (sem_item_optimizer::fixup_pt_set): New function.
2830 (sem_item_optimizer::fixup_points_to_sets): Likewise.
2831 * ipa-icf.h: Declare new variables and functions.
2833 2018-03-13 Jakub Jelinek <jakub@redhat.com>
2836 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
2837 integer_pow2p@2 and test integer_pow2p in condition.
2838 (A < 0 ? C : 0): Similarly for @1.
2841 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
2842 characters starting at p contain '\0' character, don't look beyond
2846 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
2847 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
2850 * reg-stack.c (change_stack): Change update_end var from int to
2851 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
2852 also call set_block_for_insn on the newly added insns and rescan.
2855 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
2856 on the last operand.
2859 * tree.c (stabilize_reference_1): Return save_expr (e) for
2860 STATEMENT_LIST even if it doesn't have side-effects.
2862 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
2864 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
2866 2018-03-12 Renlin Li <renlin.li@arm.com>
2868 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
2869 aarch64_output_scalar_simd_mov_immediate.
2871 2018-03-12 Martin Sebor <msebor@redhat.com>
2873 PR tree-optimization/83456
2874 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
2875 for perfectly overlapping calls to memcpy.
2876 (gimple_fold_builtin_memory_chk): Same.
2877 (gimple_fold_builtin_strcpy): Handle no-warning.
2878 (gimple_fold_builtin_stxcpy_chk): Same.
2879 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
2881 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2883 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
2884 parameter. Use it for SFmode.
2885 (rs6000_function_arg_advance_1): Adjust.
2886 (rs6000_function_arg): Adjust.
2887 (rs6000_gimplify_va_arg): Pass false for that new parameter.
2889 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2891 PR rtl-optimization/84169
2892 PR rtl-optimization/84780
2893 * combine.c (can_combine_p): Check for a 2-insn combination whether
2894 the destination register is used between the two insns, too.
2896 2018-03-12 Richard Biener <rguenther@suse.de>
2898 PR tree-optimization/84803
2899 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
2900 for refs DR analysis didn't process.
2902 2018-03-12 Richard Biener <rguenther@suse.de>
2904 PR tree-optimization/84777
2905 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
2906 force-vectorize loops ignore whether we are optimizing for size.
2908 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
2910 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
2911 (TARGET_MD_ASM_ADJUST): Define.
2913 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
2914 Kito Cheng <kito.cheng@gmail.com>
2915 Chung-Ju Wu <jasonwucj@gmail.com>
2917 * config/nds32/nds32.c (nds32_compute_stack_frame,
2918 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
2919 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
2920 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
2921 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
2922 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
2923 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
2924 * config/nds32/nds32.md (prologue, epilogue): Use macro
2925 NDS32_V3PUSH_AVAILABLE_P to do checking.
2927 2018-03-11 Jakub Jelinek <jakub@redhat.com>
2930 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
2931 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
2932 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
2933 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
2934 addition of most attributes on !orig_type_die or the attribute not
2935 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
2937 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2938 Chung-Ju Wu <jasonwucj@gmail.com>
2940 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
2942 * config/nds32/nds32.opt (mvh): New option.
2944 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2945 Chung-Ju Wu <jasonwucj@gmail.com>
2947 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
2949 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
2950 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
2953 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2954 Chung-Ju Wu <jasonwucj@gmail.com>
2956 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
2958 * config/nds32/nds32-multiple.md (strlensi): New pattern.
2959 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
2961 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
2962 Kito Cheng <kito.cheng@gmail.com>
2963 Chung-Ju Wu <jasonwucj@gmail.com>
2965 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
2966 UNSPEC_FFMISM and UNSPEC_FLMISM.
2967 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
2968 for ffb, ffmism and flmism.
2969 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
2970 (unspec_ffmism): Ditto.
2971 (unspec_flmism): Ditto.
2972 (nds32_expand_builtin_impl): Check if string extension is available.
2973 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
2974 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
2976 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
2979 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2982 * lra-assigns.c (assign_by_spills): Return a flag of reload
2983 assignment failure. Do not process the reload assignment
2984 failures. Do not spill other reload pseudos if they has the same
2986 (lra_assign): Add a return arg. Set up from the result of
2987 assign_by_spills call.
2988 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2989 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2990 usage_insns if it is not NULL.
2991 (spill_hard_reg_in_range): New function.
2992 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2993 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2994 function prototypes.
2995 (lra_assign): Change prototype.
2996 * lra.c (lra): Add code to deal with fails by splitting hard reg
2999 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
3002 * config/i386/i386.opt: Replace Enforcment with Enforcement.
3004 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
3007 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
3008 (dw_val_node): Add val_symbolic_view.
3009 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
3010 (symview_upper_bound): New.
3011 (new_line_info_table): Initialize symviews_since_reset.
3012 (dwarf2out_source_line): Count symviews_since_reset and set
3013 symview_upper_bound.
3014 (dw_val_equal_p): Handle symview.
3015 (add_AT_symview): New.
3016 (print_dw_val): Handle symview.
3017 (attr_checksum, attr_checksum_ordered): Likewise.
3018 (same_dw_val_p, size_of_die): Likewise.
3019 (value_format, output_die): Likewise.
3020 (add_high_low_attributes): Use add_AT_symview for entry_view.
3021 (dwarf2out_finish): Reset symview_upper_bound, clear
3024 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
3027 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
3028 Add strict argument and use it.
3029 (rs6000_split_multireg_move): Update for new strict argument.
3030 (mem_operand_gpr): Disallow all non-offsettable addresses.
3031 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
3033 2018-03-09 Jakub Jelinek <jakub@redhat.com>
3036 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
3037 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
3038 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
3041 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
3042 decl, use remap_type if we want to use the type.
3044 2018-03-09 Martin Sebor <msebor@redhat.com>
3046 PR tree-optimization/84526
3047 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3049 (builtin_access::generic_overlap): Be prepared to handle non-array
3052 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
3054 PR rtl-optimization/84682
3055 * lra-constraints.c (process_address_1): Check is_address flag
3056 for address constraints.
3057 (process_alt_operands): Likewise.
3058 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
3059 preprocess_constraints.
3060 * recog.h (preprocess_constraints): Add oploc parameter.
3062 * recog.c (preprocess_constraints): Test address_operand for
3063 CT_ADDRESS constraints.
3065 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3068 * lra-assigns.c (assign_by_spills): Return a flag of reload
3069 assignment failure. Do not process the reload assignment
3070 failures. Do not spill other reload pseudos if they has the same
3072 (lra_assign): Add a return arg. Set up from the result of
3073 assign_by_spills call.
3074 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3075 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3076 usage_insns if it is not NULL.
3077 (spill_hard_reg_in_range): New function.
3078 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3079 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3080 function prototypes.
3081 (lra_assign): Change prototype.
3082 * lra.c (lra): Add code to deal with fails by splitting hard reg
3085 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3088 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
3089 Accept complain bool parameter. Only emit errors if it is true.
3090 (arm_parse_cpu_option_name): Likewise.
3091 (arm_target_thumb_only): Adjust callers of the above.
3092 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
3093 prototype to take a default true bool parameter.
3094 (arm_parse_arch_option_name): Likewise.
3096 2018-03-09 David Malcolm <dmalcolm@redhat.com>
3097 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
3101 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
3102 * configure: Regenerate.
3103 * configure.ac ("linker --version-script option"): New.
3104 ("linker soname option"): New.
3106 2018-03-09 Richard Biener <rguenther@suse.de>
3108 PR tree-optimization/84775
3109 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
3110 immediate uses of predicate stmts and mark them modified.
3113 PR tree-optimization/84178
3114 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3116 (version_loop_for_if_conversion): Delay update_ssa call.
3117 (tree_if_conversion): Delay update_ssa until after predicate
3120 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
3123 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
3124 when the function accesses prior frames.
3126 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3129 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
3130 gen_llsym, otherwise call maybe_gen_llsym.
3133 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
3134 has ',' character inside of it.
3136 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3139 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
3140 as clobbering CC_REGNUM.
3142 2018-03-08 Richard Biener <rguenther@suse.de>
3145 * tree-scalar-evolution.c: Include tree-into-ssa.h.
3146 (follow_copies_to_constant): Do not follow SSA names registered
3149 2018-03-08 Richard Biener <rguenther@suse.de>
3151 PR tree-optimization/84178
3152 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3154 (version_loop_for_if_conversion): Delay update_ssa call.
3155 (tree_if_conversion): Delay update_ssa until after predicate
3158 2018-03-08 David Malcolm <dmalcolm@redhat.com>
3160 PR tree-optimization/84178
3161 * tree-if-conv.c (release_bb_predicate): Remove the
3162 the assertion that the stmts have NULL use_ops.
3163 Discard the statements, asserting that they haven't
3164 yet been added to a BB.
3166 2018-03-08 Richard Biener <rguenther@suse.de>
3168 PR tree-optimization/84746
3169 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
3170 (phi_translate): Pass in destination ANTIC_OUT set.
3171 (phi_translate_1): Likewise. For a simplified result lookup
3172 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
3173 (phi_translate_set): Adjust.
3174 (do_pre_regular_insertion): Likewise.
3175 (do_pre_partial_partial_insertion): Likewise.
3177 2018-03-08 Martin Liska <mliska@suse.cz>
3179 PR gcov-profile/84735
3180 * doc/gcov.texi: Document usage of profile files.
3181 * gcov-io.h: Document changes in the format.
3183 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
3187 * dwarf2out.c (struct dw_line_info_table): Update comments for
3189 (FORCE_RESET_NEXT_VIEW): New.
3190 (FORCE_RESETTING_VIEW_P): New.
3191 (RESETTING_VIEW_P): Check for -1 too.
3192 (ZERO_VIEW_P): Likewise.
3193 (new_line_info_table): Force-reset next view.
3194 (dwarf2out_begin_function): Likewise.
3195 (dwarf2out_source_line): Simplify zero_view_p initialization.
3196 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
3197 view directly. Omit view when omitting .loc at line 0.
3199 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3201 PR tree-optimization/84740
3202 * tree-switch-conversion.c (process_switch): Call build_constructors
3203 only if info.phi_count is non-zero.
3205 PR tree-optimization/84739
3206 * tree-tailcall.c (find_tail_calls): Check call arguments against
3207 DECL_ARGUMENTS (current_function_decl) rather than
3208 DECL_ARGUMENTS (func) when checking for tail recursion.
3210 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3212 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
3213 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
3214 Volker Reichelt's entry and add entries for people that perform
3215 GCC fuzzy testing and report numerous bugs.
3217 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
3220 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
3221 readonly data in sdata, if that is disabled.
3222 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
3223 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3224 -mreadonly-in-sdata option.
3226 2018-03-07 Martin Sebor <msebor@redhat.com>
3228 PR tree-optimization/84468
3229 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
3230 basic block when looking for nul assignment.
3232 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
3235 * except.h (output_function_exception_table): Adjust prototype.
3236 * except.c (output_function_exception_table): Remove FNNAME parameter
3237 and add SECTION parameter. Ouput one part of the table at a time.
3238 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
3239 the first part of the exception table and emit unwind directives.
3240 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
3241 (i386_pe_seh_cold_init): Likewise.
3242 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
3243 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
3244 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
3245 (ix86_output_call_insn): Emit a nop in one more case for SEH.
3246 * config/i386/winnt.c: Include except.h.
3247 (struct seh_frame_state): Add reg_offset, after_prologue and
3248 in_cold_section fields.
3249 (i386_pe_seh_end_prologue): Set seh->after_prologue.
3250 (i386_pe_seh_cold_init): New function.
3251 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
3252 to seh->in_cold_section.
3253 (seh_emit_push): Record the offset of the push.
3254 (seh_emit_save): Record the offet of the save.
3255 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
3256 Test seh->after_prologue to disregard the epilogue.
3257 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
3258 (i386_pe_end_cold_function): New function.
3260 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3263 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
3264 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
3267 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
3269 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
3270 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
3273 * multiple_target.c: Include tree-inline.h and intl.h.
3274 (expand_target_clones): Diagnose and fail if node->definition and
3275 !tree_versionable_function_p (node->decl).
3277 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
3279 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
3281 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
3282 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
3283 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
3285 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3288 * combine.c (try_combine): Use reg_or_subregno instead of handling
3289 just paradoxical SUBREGs and REGs.
3291 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
3293 * config/arc/arc.c (arc_finalize_pic): Remove function.
3294 (arc_must_save_register): We use single base PIC register, remove
3295 checks to save/restore the PIC register.
3296 (arc_expand_prologue): Likewise.
3297 * config/arc/arc-protos.h (arc_set_default_type_attributes):
3299 (arc_verify_short): Likewise.
3300 (arc_attr_type): Likewise.
3301 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
3302 (walk_stores): Likewise.
3303 (arc_address_cost): Make it static.
3304 (arc_verify_short): Likewise.
3305 (branch_dest): Likewise.
3306 (arc_attr_type): Likewise.
3307 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
3308 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
3309 (arc_final_prescan_insn): Remove inserting the nops due to
3310 hardware hazards. It is done in reorg step.
3311 (insn_length_variant_t): Remove.
3312 (insn_length_parameters_t): Likewise.
3313 (arc_insn_length_parameters): Likewise.
3314 (arc_get_insn_variants): Likewise.
3315 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
3317 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3320 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
3323 PR tree-optimization/84687
3324 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
3326 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
3328 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3330 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
3331 Rename to ppc_speculation_barrier.
3332 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
3333 __builtin_ppc_speculation_barrier.
3335 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3338 * combine.c (combine_simplify_rtx): Don't try to simplify if
3339 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
3342 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
3344 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
3345 to 32 bytes when compiling for POWER9.
3347 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3350 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
3351 regparm >= 3 with no arg reg available also for calls with
3352 flag_force_indirect_call. Pass decl to ix86_function_regparm.
3355 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
3357 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
3359 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
3362 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
3364 2018-03-05 Richard Biener <rguenther@suse.de>
3366 PR tree-optimization/84486
3367 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
3368 When inserting a __builtin_assume_aligned call set the LHS
3369 SSA name alignment info accordingly.
3371 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
3373 PR tree-optimization/84114
3374 * config/aarch64/aarch64.c (aarch64_reassociation_width)
3375 Avoid reassociation of FLOAT_MODE addition.
3377 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
3379 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
3380 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
3381 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
3382 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
3383 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
3384 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
3385 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
3387 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
3388 __builtin_ia32_wbinvd): New builtins.
3389 (SPECIAL_ARGS2): New.
3390 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
3391 (SPECIAL_ARGS2): New.
3392 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
3393 (ix86_valid_target_attribute_inner_p): Ditto.
3394 (ix86_init_mmx_sse_builtins): Add special_args2.
3395 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
3396 TARGET_WBNOINVD_P): New.
3397 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
3398 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
3399 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
3400 * config/i386/immintrin.h (_wbinvd): New intrinsic.
3401 * config/i386/pconfigintrin.h: New file.
3402 * config/i386/wbnoinvdintrin.h: Ditto.
3403 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
3404 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
3406 2018-03-05 Richard Biener <rguenther@suse.de>
3408 PR tree-optimization/84670
3409 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3411 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3412 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3413 (compute_antic_aux): Only assert the number of values in ANTIC_IN
3414 doesn't grow if all successors (recursively) were visited at least
3417 2018-03-05 Richard Biener <rguenther@suse.de>
3419 PR tree-optimization/84650
3420 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
3421 if executed in the loop pipeline.
3423 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
3425 * doc/configfiles.texi (Configuration Files): Move info about
3426 conditionalizing $target-protos.h to...
3427 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
3428 differs from $target-protos.h.
3430 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
3431 Chung-Ju Wu <jasonwucj@gmail.com>
3433 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
3434 * config/nds32/nds32-multiple.md (setmemsi): Define.
3435 * config/nds32/nds32-memory-manipulation.c
3436 (nds32_gen_dup_4_byte_to_word_value): New.
3437 (emit_setmem_word_loop): New.
3438 (emit_setmem_byte_loop): New.
3439 (nds32_expand_setmem_loop): New.
3440 (nds32_expand_setmem_loop_v3m): New.
3441 (nds32_expand_setmem_unroll): New.
3442 (nds32_expand_setmem): New.
3444 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3445 Chung-Ju Wu <jasonwucj@gmail.com>
3447 * config/nds32/nds32-memory-manipulation.c
3448 (nds32_emit_load_store): New.
3449 (nds32_emit_post_inc_load_store): New.
3450 (nds32_emit_mem_move): New.
3451 (nds32_emit_mem_move_block): New.
3452 (nds32_expand_movmemsi_loop_unknown_size): New.
3453 (nds32_expand_movmemsi_loop_known_size): New.
3454 (nds32_expand_movmemsi_loop): New.
3455 (nds32_expand_movmemsi_unroll): New.
3456 (nds32_expand_movmemqi): Rename ...
3457 (nds32_expand_movmemsi): ... to this.
3458 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
3459 (movmemsi): ... to this.
3460 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
3461 (nds32_expand_movmemsi): ... to this.
3463 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3464 Monk Chiang <sh.chiang04@gmail.com>
3465 Chung-Ju Wu <jasonwucj@gmail.com>
3467 * config/nds32/nds32-protos.h
3468 (nds32_expand_load_multiple): New arguments.
3469 (nds32_expand_store_multiple): Ditto.
3470 (nds32_valid_multiple_load_store): Rename ...
3471 (nds32_valid_multiple_load_store_p): ... to this.
3472 * config/nds32/nds32-memory-manipulation.c
3473 (nds32_expand_load_multiple): Refine implementation.
3474 (nds32_expand_store_multiple): Ditto.
3475 * config/nds32/nds32-multiple.md
3476 (load_multiple): Update nds32_expand_load_multiple interface.
3477 (store_multiple): Update nds32_expand_store_multiple interface.
3478 * config/nds32/nds32-predicates.c
3479 (nds32_valid_multiple_load_store): Rename ...
3480 (nds32_valid_multiple_load_store_p): ... to this and refine
3482 * config/nds32/predicates.md
3483 (nds32_load_multiple_and_update_address_operation): New predicate.
3484 (nds32_store_multiple_and_update_address_operation): New predicate.
3486 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3487 Chung-Ju Wu <jasonwucj@gmail.com>
3489 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
3490 (combo): New attribute.
3491 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
3493 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
3495 * config/nds32/nds32.opt: Change -mcmodel= default value.
3497 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
3498 Monk Chiang <sh.chiang04@gmail.com>
3499 Chung-Ju Wu <jasonwucj@gmail.com>
3501 * config/nds32/constants.md (unspec_element): New enum.
3502 * config/nds32/constraints.md (Umw): New constraint.
3503 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
3504 * config/nds32/nds32-intrinsic.md: Likewise.
3505 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
3506 (nds32_valid_smw_lwm_base_p): New.
3507 (nds32_output_smw_single_word): New.
3508 (nds32_output_lmw_single_word): New.
3509 (nds32_expand_unaligned_load): New.
3510 (nds32_expand_unaligned_store): New.
3511 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
3512 (nds32_output_smw_single_word): Declare.
3513 (nds32_output_lmw_single_word): Declare.
3514 (nds32_expand_unaligned_load): Declare.
3515 (nds32_expand_unaligned_store): Declare.
3516 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
3517 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
3518 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
3519 NDS32_BUILTIN_UASTORE_DW.
3520 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
3523 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
3524 Kito Cheng <kito.cheng@gmail.com>
3525 Chung-Ju Wu <jasonwucj@gmail.com>
3527 * config/nds32/nds32-intrinsic.c
3528 (nds32_expand_builtin_null_ftype_reg): Delete.
3529 (nds32_expand_builtin_reg_ftype_imm): Ditto.
3530 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
3531 (nds32_read_argument): New.
3532 (nds32_legitimize_target): Ditto.
3533 (nds32_legitimize_argument): Ditto.
3534 (nds32_check_constant_argument): Ditto.
3535 (nds32_expand_unop_builtin): Ditto.
3536 (nds32_expand_unopimm_builtin): Ditto.
3537 (nds32_expand_binop_builtin): Ditto.
3538 (nds32_builtin_decl_impl): Ditto.
3539 (builtin_description): Ditto.
3540 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
3541 (nds32_init_builtins_impl): Ditto.
3542 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
3543 (nds32_builtin_decl): New.
3544 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
3545 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
3547 2018-03-02 Jeff Law <law@redhat.com>
3549 * reorg.c (stop_search_p): Handle DEBUG_INSN.
3550 (redundant_insn, fill_simple_delay_slots): Likewise.
3551 (fill_slots_from_thread): Likewise.
3552 * resource.c (mark_referenced_resources): Likewise.
3553 (mark_set_resources, find_dead_or_set_registers): Likewise.
3555 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3557 * substring-locations.h (format_warning_va): Formatting fix for
3559 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
3561 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
3562 * substring-locations.c: Include intl.h.
3563 (format_warning_va): Turned into small wrapper around
3564 format_warning_n_va, renamed to ...
3565 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
3566 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
3568 (format_warning_at_substring_n): New function.
3569 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
3570 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
3571 format_warning_at_substring with just a shorter name instead of
3572 const function pointer.
3573 (fmtwarn_n): New function.
3574 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
3575 appropriate, get rid of all the fmtstr temporaries, move conditionals
3576 with G_() wrapped string literals directly into fmtwarn arguments,
3577 cast dir.len to (int), formatting fixes.
3579 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
3581 * doc/invoke.texi: Remove "Cilk Plus" references.
3583 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3584 Richard Biener <rguenther@suse.de>
3587 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
3588 for error or warning attributes if CALL_FROM_THUNK_P is set.
3591 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3594 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
3595 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
3598 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
3599 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
3601 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
3602 instead of -1U in last predictors element's probability member.
3604 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3607 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3608 arguments if they are comparable.
3610 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3612 PR tree-optimization/84634
3613 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
3614 masks and masked_loop_p with a single loop_masks, making sure it's
3615 null for bb vectorization.
3617 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3619 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3620 (vect_analyze_data_ref_access): Use loop->safe_len rather than
3621 loop->force_vectorize to check whether there is no alias.
3623 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3626 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
3628 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
3630 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
3631 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
3632 instead of a loop around prev_real_insn.
3633 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
3637 * config/i386/i386.c (ix86_print_operand): Use conditional
3638 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
3641 2018-03-02 Richard Biener <rguenther@suse.de>
3643 PR tree-optimization/84427
3644 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
3645 (bitmap_set_subtract_values): Rewrite to handle multiple
3648 (prune_clobbered_mems): Likewise.
3649 (phi_translate): Take edge instead of pred/phiblock.
3650 (phi_translate_1): Likewise.
3651 (phi_translate_set): Likewise. Insert all translated
3652 exprs for a value into the set, keeping possibly multiple
3653 expressions per value.
3654 (compute_antic_aux): Adjust for phi_translate changes.
3655 When intersecting union the expressions and prune those
3656 not in the final value set, keeping possibly multiple
3657 expressions per value. Do not use value-insertion
3658 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
3659 all expressions. Add verification that the value-sets
3660 only shrink during iteration.
3661 (compute_partial_antic_aux): Adjust for the phi_translate changes.
3662 (do_pre_regular_insertion): Likewise.
3663 (do_pre_partial_partial_insertion): Likewise.
3665 2018-03-02 Richard Biener <rguenther@suse.de>
3668 * config/darwin.c (saved_debug_info_level): New static global.
3669 (darwin_asm_lto_start): Disable debug info generation for LTO out.
3670 (darwin_asm_lto_end): Restore debug info generation settings.
3672 2018-03-01 Martin Liska <mliska@suse.cz>
3675 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
3678 2018-03-01 Richard Biener <rguenther@suse.de>
3681 * dwarf2out.c (gen_variable_die): Properly handle late VLA
3682 type annotation with LTO when debug was disabled at compile-time.
3684 2018-03-01 Matthew Fortune <mfortune@gmail.com>
3686 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
3688 (mips_final_postscan_insn): Likewise.
3690 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
3692 PR rtl-optimization/84528
3693 * alias.c (init_alias_target): Add commentary.
3694 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
3695 a unique base value if the frame pointer is not eliminated
3696 to the stack pointer.
3698 2018-03-01 Tom de Vries <tom@codesourcery.com>
3700 PR rtl-optimization/83327
3701 * lra-int.h (hard_regs_spilled_into): Declare.
3702 * lra.c (hard_regs_spilled_into): Define.
3703 (init_reg_info): Init hard_regs_spilled_into.
3704 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
3705 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
3706 (process_bb_lives): Handle hard_regs_spilled_into.
3707 (lra_create_live_ranges_1): Before doing liveness propagation, clear
3708 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
3710 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
3712 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
3713 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
3714 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
3715 * config/rs6000/aix72.h: New file.
3717 2018-02-28 Jakub Jelinek <jakub@redhat.com>
3719 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
3720 instead of warning_at with conditional singular and plural messages
3724 * stor-layout.c (update_alignment_for_field): For
3725 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
3726 && !DECL_PACKED (field), do the alignment update, just use
3727 only desired_align instead of MAX (type_align, desired_align)
3729 (place_field): Don't do known_align < desired_align handling
3730 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
3731 is non-NULL, instead do it after rli->prev_field handling and
3732 only if not within a bitfield word. For DECL_PACKED (field)
3733 use type_align of BITS_PER_UNIT.
3735 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
3737 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
3738 superfluous parentheses and trailing spaces.
3740 2018-02-28 Richard Biener <rguenther@suse.de>
3742 PR tree-optimization/84584
3743 * graphite-scop-detection.c (scop_detection::add_scop): Discard
3744 SCoPs with fake exit edge.
3746 2018-02-28 Martin Liska <mliska@suse.cz>
3749 * timevar.c (timer::print): Fix format to properly print 100%
3752 2018-02-28 Richard Biener <rguenther@suse.de>
3755 * genmatch.c (capture_info::walk_match): Do not mark
3756 captured expressions without operands as expr_p given
3757 they act more like predicates and should be subject to
3758 "lost tail" side-effect preserving.
3760 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
3762 PR rtl-optimization/81611
3763 * auto-inc-dec.c (attempt_change): Move dead note from
3764 mem_insn if it's the next use of regno
3765 (find_address): Take address use of reg holding
3766 non-incremented value. Add parm to limit search to the named
3768 (merge_in_block): Attempt to use a mem insn that is the next
3769 use of the original regno.
3771 2018-02-27 Martin Sebor <msebor@redhat.com>
3774 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
3775 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
3777 2018-02-27 Martin Sebor <msebor@redhat.com>
3779 PR translation/84207
3780 * diagnostic-core.h (warning_n, error_n, inform_n): Change
3781 n argument to unsigned HOST_WIDE_INT.
3782 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
3783 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
3784 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
3785 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
3787 2018-02-27 Richard Biener <rguenther@suse.de>
3789 PR tree-optimization/84512
3790 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
3791 Do not use the estimate returned from record_stmt_cost for
3792 the scalar iteration cost but sum properly using add_stmt_cost.
3794 2018-02-27 Richard Biener <rguenther@suse.de>
3796 PR tree-optimization/84466
3797 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3798 Adjust last change to less strictly validate use operands.
3800 2018-02-27 Martin Liska <mliska@suse.cz>
3802 PR gcov-profile/84548
3803 * gcov.c (process_file): Allow partial overlap and consider it
3804 also as group functions.
3805 (output_lines): Properly calculate range of lines for a group.
3807 2018-02-27 Martin Liska <mliska@suse.cz>
3809 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
3810 'ggc' suffixes. Change first column width.
3811 (timer::print): Fix formatting of the column.
3813 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
3815 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
3816 preserve inline entry blocks for the sake of debug inline
3817 entry point markers alone.
3818 (remove_unused_locals): Suggest in comments a better place to
3819 force the preservation of inline entry blocks that are
3820 otherwise unused, but do not preserve them.
3822 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3824 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
3826 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3829 * config/i386/constraints.md (Bs): Replace
3830 ix86_indirect_branch_register with
3831 TARGET_INDIRECT_BRANCH_REGISTER.
3833 * config/i386/i386.md (indirect_jump): Likewise.
3834 (tablejump): Likewise.
3835 (*sibcall_memory): Likewise.
3836 (*sibcall_value_memory): Likewise.
3837 Peepholes of indirect call and jump via memory: Likewise.
3838 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
3839 (*sibcall_value_GOT_32): Likewise.
3840 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3841 (GOT_memory_operand): Likewise.
3842 (call_insn_operand): Likewise.
3843 (sibcall_insn_operand): Likewise.
3844 (GOT32_symbol_operand): Likewise.
3845 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
3847 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3849 PR rtl-optimization/83496
3850 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
3851 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
3852 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
3853 redundant insn, if any.
3854 (relax_delay_slots): Likewise.
3855 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
3857 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
3859 PR tree-optimization/83965
3860 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
3861 that grouped statements are part of a reduction chain. Return
3862 true if the statement is not marked as a reduction itself but
3864 (vect_recog_dot_prod_pattern): Don't check whether the statement
3865 is part of a group here.
3866 (vect_recog_sad_pattern): Likewise.
3867 (vect_recog_widen_sum_pattern): Likewise.
3869 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3872 * final.c (rest_of_clean_state): Also look for calls inside sequences.
3874 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3877 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
3879 (ix86_output_indirect_function_return): New prototype.
3880 (ix86_split_simple_return_pop_internal): Likewise.
3881 * config/i386/i386.c (indirect_return_via_cx): New.
3882 (indirect_return_via_cx_bnd): Likewise.
3883 (indirect_thunk_name): Handle return va CX_REG.
3884 (output_indirect_thunk_function): Create alias for
3885 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
3886 (ix86_output_indirect_jmp): Remove the bool argument.
3887 (ix86_output_indirect_function_return): New function.
3888 (ix86_split_simple_return_pop_internal): Likewise.
3889 * config/i386/i386.md (*indirect_jump): Don't pass false
3890 to ix86_output_indirect_jmp.
3891 (*tablejump_1): Likewise.
3892 (simple_return_pop_internal): Change it to define_insn_and_split.
3893 Call ix86_split_simple_return_pop_internal to split it for
3895 (simple_return_indirect_internal): Call
3896 ix86_output_indirect_function_return instead of
3897 ix86_output_indirect_jmp.
3899 2018-02-26 Jakub Jelinek <jakub@redhat.com>
3902 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
3903 memset and value initialization afterwards.
3905 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
3907 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
3909 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3912 * common/config/aarch64/aarch64-common.c
3913 (aarch_option_optimization_table[]): Switch
3914 off fomit-frame-pointer
3916 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3917 Chung-Ju Wu <jasonwucj@gmail.com>
3919 * config/nds32/nds32-multiple.md (load_multiple): Disallow
3921 (store_multiple): Ditto.
3923 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3925 * config.gcc: Add --with-cpu support for nds32 target.
3926 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
3927 * config/nds32/nds32.opt: Add -mcpu= option.
3929 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
3931 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
3932 isel=yes): Warn for these deprecated options.
3934 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
3936 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
3937 ISA_2_5_MASKS_EMBEDDED.
3939 2018-02-23 Jakub Jelinek <jakub@redhat.com>
3941 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
3942 p->max as pointers rather than using iterative_hash_expr.
3944 2018-02-23 Carl Love <cel@us.ibm.com>
3946 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
3947 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
3949 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
3950 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
3951 P8V_BUILTIN_VEC_VUNSIGNED2.
3953 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
3956 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
3957 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
3959 * lra-constraints.c (curr_insn_transform): Set up
3960 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
3962 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
3963 * lra-eliminations.c (spill_pseudos): Ditto.
3964 (process_insn_for_elimination): Ditto.
3965 * lra-lives.c (reg_early_clobber_p): Use the new macros.
3966 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
3967 LRA_NON_CLOBBERED_ALT.
3969 2018-02-22 Martin Sebor <msebor@redhat.com>
3971 PR tree-optimization/84480
3972 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
3973 to maybe_diag_stxncpy_trunc. Call it.
3974 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
3975 from gimple_fold_builtin_strcpy. Print inlining stack.
3976 (handle_builtin_stxncpy): Print inlining stack.
3977 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
3979 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
3982 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
3983 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
3984 and -fcheck-pointer-bounds are used together.
3985 (indirect_thunk_prefix): New enum.
3986 (indirect_thunk_need_prefix): New function.
3987 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
3988 "_nt" instead of "_bnd" for NOTRACK prefix.
3989 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
3990 (output_indirect_thunk_function): Likewise.
3992 (ix86_code_end): Update output_indirect_thunk_function calls.
3993 (ix86_output_indirect_branch_via_reg): Replace
3994 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
3995 (ix86_output_indirect_branch_via_push): Likewise.
3996 (ix86_output_function_return): Likewise.
3997 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
3998 incompatible with -fcf-protection=branch and
3999 -fcheck-pointer-bounds.
4001 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4004 * config/aarch64/aarch64.c (aarch64_print_address_internal):
4005 Change gcc_assert call to output_operand_lossage.
4007 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4009 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
4011 2018-02-22 DJ Delorie <dj@redhat.com>
4012 Sebastian Perta <sebastian.perta@renesas.com>
4013 Oleg Endo <olegendo@gcc.gnu.org>
4015 * config/rx/rx.c (rx_rtx_costs): New function.
4016 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
4018 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
4020 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
4022 2018-02-22 Martin Liska <mliska@suse.cz>
4025 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
4026 Add "native" as a possible value.
4028 2018-02-22 Martin Liska <mliska@suse.cz>
4031 * config/i386/i386.c (ix86_option_override_internal):
4032 Add "native" as a possible value for -march and -mtune.
4034 2018-02-22 Jakub Jelinek <jakub@redhat.com>
4037 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
4038 to all type variants.
4040 PR tree-optimization/84503
4041 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
4042 width as info->bitpos + info->bitsize - start.
4043 (merged_store_group::merge_overlapping): Simplify width computation.
4044 (check_no_overlap): New function.
4045 (imm_store_chain_info::try_coalesce_bswap): Compute expected
4046 start + width and last_order of the group, fail if check_no_overlap
4048 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
4049 to group if check_no_overlap fails.
4051 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4053 * config/rs6000/altivec.md: Delete contraint arguments to
4054 define_expand, define_split, and define_peephole2, and in
4055 define_insn_and_split if always unused.
4056 * config/rs6000/darwin.md: Ditto.
4057 * config/rs6000/dfp.md: Ditto.
4058 * config/rs6000/rs6000.md: Ditto.
4059 * config/rs6000/sync.md: Ditto.
4060 * config/rs6000/vector.md: Ditto.
4061 * config/rs6000/vsx.md: Ditto.
4063 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4065 * config/rs6000/altivec.md: Write output control strings as braced
4066 blocks instead of double-quoted strings.
4067 * config/rs6000/darwin.md: Ditto.
4068 * config/rs6000/rs6000.md: Ditto.
4069 * config/rs6000/vector.md: Ditto.
4070 * config/rs6000/vsx.md: Ditto.
4072 2018-02-21 Jason Merrill <jason@redhat.com>
4074 PR c++/84314 - ICE with templates and fastcall attribute.
4075 * attribs.c (build_type_attribute_qual_variant): Remove assert.
4077 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4079 * ipa-cp.c (determine_versionability): Fix comment typos.
4081 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4084 * ipa-cp.c (determine_versionability): Do not version functions caling
4087 2018-02-21 Martin Liska <mliska@suse.cz>
4090 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
4091 Add "native" as a possible value.
4092 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
4093 the macro when native cpu detection is available.
4095 2018-02-21 Martin Liska <mliska@suse.cz>
4098 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
4099 Add "native" as a possible value.
4100 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
4101 when native cpu detection is available.
4103 2018-02-21 Jakub Jelinek <jakub@redhat.com>
4104 Martin Sebor <msebor@redhat.com>
4106 PR tree-optimization/84478
4107 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
4109 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
4110 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
4111 support which is conservatively correct, for 2 only stay conservative
4112 for maxlen. Formatting and comment capitalization fixes. Add STRICT
4113 argument to the 2 argument get_range_strlen, adjust 6 arg
4114 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
4116 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
4117 (gimple_fold_builtin_strlen): Pass true as last argument to
4120 2018-02-20 Martin Sebor <msebor@redhat.com>
4123 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
4124 (builtin_memref::set_base_and_offset): Same. Handle inner references.
4125 (builtin_memref::builtin_memref): Factor out parts into
4126 set_base_and_offset and call it.
4128 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4131 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
4132 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
4133 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
4134 search at the associated MODE_INT.
4136 2018-02-20 Jeff Law <law@redhat.com>
4139 PR tree-optimization/81592
4141 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
4142 for range data rather than using global data.
4143 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
4144 range data rather than using global data.
4145 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
4146 pass it to children as needed.
4147 (struct directive::fmtresult): Similarly.
4148 (struct directive::set_width): Similarly.
4149 (struct directive::set_precision): Similarly.
4150 (format_integer, format_directive, parse_directive): Similarly.
4151 (format_none): Accept unnamed vr_values parameter.
4152 (format_percent, format_floating, format_character): Similarly.
4153 (format_string, format_plain): Similarly.
4154 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
4155 the EVRP range analyzer for range data rather than using global data.
4156 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
4157 gimple-ssa-evrp-analyze.h
4158 (class sprintf_dom_walker): Add after_dom_children member function.
4159 Add evrp_range_analyzer member.
4160 (sprintf_dom_walker::before_dom_children): Call into the EVRP
4161 range analyzer as needed.
4162 (sprintf_dom_walker::after_dom_children): New member function.
4163 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
4165 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
4166 (evrp_range_analyzer::pop_to_marker): Likewise.
4168 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4170 PR tree-optimization/84419
4171 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
4172 with the required type if its current type is compatible but
4175 2018-02-20 Jakub Jelinek <jakub@redhat.com>
4178 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
4179 after vectorization.
4181 2018-02-20 Martin Liska <mliska@suse.cz>
4184 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
4185 possible values if we don't have a hint.
4187 2018-02-20 Martin Liska <mliska@suse.cz>
4191 * final.c (shorten_branches): Build align_tab array with one
4193 * opts.c (finish_options): Add alignment option limit check.
4194 (MAX_CODE_ALIGN): Likewise.
4195 (MAX_CODE_ALIGN_VALUE): Likewise.
4196 * doc/invoke.texi: Document maximum allowed option value for
4197 all -falign-* options.
4199 2018-02-19 Jakub Jelinek <jakub@redhat.com>
4202 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
4203 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
4204 * var-tracking.c (emit_note_insn_var_location): Remove all references
4205 to NOTE_INSN_CALL_ARG_LOCATION.
4206 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
4207 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
4208 Use copy_rtx_if_shared.
4209 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
4210 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
4211 (dwarf2out_var_location): Remove handling of
4212 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
4214 * final.c (final_scan_insn): Remove all references to
4215 NOTE_INSN_CALL_ARG_LOCATION.
4216 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
4217 before dumping final insns.
4218 * except.c (emit_note_eh_region_end): Remove all references to
4219 NOTE_INSN_CALL_ARG_LOCATION.
4220 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
4221 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
4222 * config/arc/arc.c (hwloop_optimize): Likewise.
4223 * config/arm/arm.c (create_fix_barrier): Likewise.
4224 * config/s390/s390.c (s390_chunkify_start): Likewise.
4225 * config/sh/sh.c (find_barrier): Likewise.
4226 * config/i386/i386.c (rest_of_insert_endbranch,
4227 ix86_seh_fixup_eh_fallthru): Likewise.
4228 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
4229 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
4230 * config/frv/frv.c (frv_function_prologue): Likewise.
4231 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
4233 (note_outside_basic_block_p): Remove all references to
4234 NOTE_INSN_CALL_ARG_LOCATION.
4235 * gengtype.c (adjust_field_rtx_def): Likewise.
4236 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
4238 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
4239 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
4242 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
4245 PR tree-optimization/84452
4246 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
4247 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
4250 2018-02-19 Martin Liska <mliska@suse.cz>
4253 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
4255 2018-02-19 Martin Liska <mliska@suse.cz>
4256 Richard Sandiford <richard.sandiford@linaro.org>
4258 PR tree-optimization/82491
4259 * gimple-fold.c (get_base_constructor): Make earlier bail out
4262 2018-02-19 Carl Love <cel@us.ibm.com>
4264 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
4265 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
4267 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
4268 P8V_BUILTIN_VEC_NEG.
4270 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
4272 * config/rl78/rl78.md (movdf): New define expand.
4274 2018-02-19 Martin Liska <mliska@suse.cz>
4277 * doc/invoke.texi: Fix typo.
4278 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
4280 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
4282 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
4283 handle rs6000_single_float and rs6000_double_float specially for
4286 2018-02-16 Jeff Law <law@redhat.com>
4288 * config/rx/rx.c (add_pop_cfi_notes): New function.;
4291 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4294 * ipa-inline.c (inline_small_functions): Fix a typo.
4296 2018-02-16 Nathan Sidwell <nathan@acm.org>
4298 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
4300 2018-02-16 Carl Love <cel@us.ibm.com>
4302 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
4303 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
4304 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
4305 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
4306 expansion to P8V_BUILTIN_VEC_FLOAT2.
4308 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
4310 PR rtl-optimization/70023
4311 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
4312 src_regno into account.
4314 2018-02-16 Carl Love <cel@us.ibm.com>
4316 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
4317 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
4318 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
4319 * config/rs6000/rs6000.c: Remove case statements for
4320 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
4321 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
4322 and P9V_BUILTIN_VEC_VINSERT4B.
4323 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
4324 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
4325 * config/rs6000/vsx.md:
4326 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
4329 2018-02-16 Carl Love <cel@us.ibm.com>
4331 * config/rs6000/altivec.h: Add builtin names vec_extract4b
4333 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
4335 * config/rs6000/rs6000-c.c: Add the definitions for
4336 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
4337 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
4338 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
4339 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
4340 definition for insert4b and define insn *insert3b_internal.
4341 * doc/extend.texi: Add documentation for vec_extract4b.
4343 2018-02-16 Nathan Sidwell <nathan@acm.org>
4345 * doc/extend.texi (Backwards Compatibility): Mention friend
4346 injection. Note for-scope is deprecated.
4347 * doc/invoke.texi (-ffriend-injection): Deprecate.
4349 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
4351 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
4352 that moved to I2, also allow destinations that are a paradoxical
4353 subreg (instead of a normal reg).
4355 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
4358 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
4361 2018-02-16 Richard Biener <rguenther@suse.de>
4363 PR tree-optimization/84037
4364 PR tree-optimization/84016
4366 * config/i386/i386.c (ix86_builtin_vectorization_cost):
4367 Adjust vec_construct for the fact we need additional higher latency
4368 128bit inserts for AVX256 and AVX512 vector builds.
4369 (ix86_add_stmt_cost): Scale vector construction cost for
4372 2018-02-16 Richard Biener <rguenther@suse.de>
4374 PR tree-optimization/84417
4375 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
4376 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
4377 (non_rewritable_lvalue_p): Likewise, use poly-ints.
4379 2018-02-16 Martin Liska <mliska@suse.cz>
4382 * internal-fn.def (ASAN_CHECK): Set proper flags.
4383 (ASAN_MARK): Likewise.
4385 2018-02-16 Julia Koval <julia.koval@intel.com>
4387 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
4388 from PTA_CANNONLAKE.
4390 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4393 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
4394 Use ++iter rather than iter++ for std::list iterators.
4395 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
4396 defer deleting them until all nodes in the forest are processed. Do
4397 free even leaf nodes. Change to_process into auto_vec.
4400 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
4401 * vec.h (vec_default_construct): Use memset instead of placement new
4402 if BROKEN_VALUE_INITIALIZATION is defined.
4403 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
4404 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
4407 PR rtl-optimization/83723
4408 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
4409 * lra.c (lra_substitute_pseudo): Likewise. If true, use
4410 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
4412 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
4414 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
4416 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
4418 PR rtl-optimization/81443
4419 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
4420 from inner REGs to paradoxical SUBREGs.
4422 2018-02-16 Richard Biener <rguenther@suse.de>
4424 PR tree-optimization/84399
4425 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4426 For operands we can analyze at their definition make sure we can
4427 analyze them at each use as well.
4429 2018-02-16 Richard Biener <rguenther@suse.de>
4431 PR tree-optimization/84190
4432 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
4433 volatile accesses if the decl isn't volatile.
4435 2018-02-15 Jason Merrill <jason@redhat.com>
4437 PR c++/84314 - ICE with templates and fastcall attribute.
4438 * attribs.c (build_type_attribute_qual_variant): Don't clobber
4439 TYPE_CANONICAL on an existing type.
4441 2018-02-15 Jakub Jelinek <jakub@redhat.com>
4443 PR tree-optimization/84383
4444 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
4445 dstoff nor call operand_equal_p if dstbase is NULL.
4447 PR tree-optimization/84334
4448 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
4449 also a CONSTANT_CLASS_P, punt.
4451 2018-02-14 Jim Wilson <jimw@sifive.com>
4453 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
4454 first SMALL_OPERAND check. New local min_second_step. Move assert
4455 to where locals are set. Add TARGET_RVC support.
4456 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
4458 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
4460 * doc/invoke.texi: Correct -Wformat-overflow code sample.
4462 2018-02-14 Martin Sebor <msebor@redhat.com>
4464 PR tree-optimization/83698
4465 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
4466 arrays constrain the offset range to their bounds.
4467 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
4468 (builtin_access::overlap): Avoid setting the size of overlap if it's
4470 (maybe_diag_overlap): Also consider arrays when deciding what values
4471 of offsets to include in diagnostics.
4473 2018-02-14 Martin Sebor <msebor@redhat.com>
4476 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
4477 that correspond to the kind of a declaration.
4479 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
4482 * config/pa/pa.md: Load address of PIC label using the linkage table
4483 if the label is nonlocal.
4485 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
4487 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
4488 warning message if user requests -maltivec=be.
4489 * doc/invoke.texi: Document deprecation of -maltivec=be.
4491 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
4494 * config/rs6000/rs6000-c.c: Update definitions for
4495 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
4496 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
4498 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4501 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
4502 add _get_ssp intrinsics. Remove argument from
4503 __builtin_ia32_rdssp[d|q].
4504 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
4505 * config/i386/i386-builtin.def: Remove argument from
4506 __builtin_ia32_rdssp[d|q].
4507 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
4508 ix86_expand_special_args_builtin for _rdssp[d|q].
4509 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
4510 Clear register before usage.
4511 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
4512 Add documentation for new _get_ssp and _inc_ssp intrinsics.
4514 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
4516 PR tree-optimization/84357
4517 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
4518 operand 1 of an ARRAY_REF too.
4520 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
4523 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
4524 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4526 (set_of_reg): New struct.
4527 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
4528 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
4529 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4531 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
4532 Split into bitclr, bitset, bitinvert patterns if appropriate.
4533 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
4534 use rx_fuse_in_memory_bitop.
4535 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
4536 to named insn, correct maximum insn length.
4538 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
4541 * machmode.def: Define a complex mode for PARTIAL_INT.
4542 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
4544 * doc/rtl.texi: Document CSPImode.
4545 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
4547 (msp430_hard_regno_nregs_with_padding): Likewise.
4549 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
4552 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
4554 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
4556 PR rtl-optimization/84169
4557 * combine.c (try_combine): New variable split_i2i3. Set it to true if
4558 we generated a parallel as new i3 and we split that to new i2 and i3
4559 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
4560 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
4561 those to i2, not i1. Partially rewrite this scan code.
4563 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4566 * stor-layout.c (place_field): For variable length fields, adjust
4567 offset_align afterwards not just based on the field's alignment,
4568 but also on the size.
4571 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
4572 of exps and logs in the use_exp2 case.
4574 2018-02-13 Jeff Law <law@redhat.com>
4576 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
4579 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
4582 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
4586 * common.opt (gas-loc-support, gas-locview-support): New.
4587 (ginline-points, ginternal-reset-location-views): New.
4588 * doc/invoke.texi: Document them. Use @itemx where intended.
4589 (gvariable-location-views): Adjust.
4590 * target.def (reset_location_view): New.
4591 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
4592 (TARGET_RESET_LOCATION_VIEW): New.
4593 * doc/tm.texi: Rebuilt.
4594 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
4595 (dwarf2out_default_as_locview_support): New.
4596 (output_asm_line_debug_info): Use option variables.
4597 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
4598 (output_loc_list): Likewise.
4599 (add_high_low_attributes): Check option variables.
4600 Don't output entry view attribute in strict mode.
4601 (gen_inlined_subroutine_die): Check option variables.
4602 (dwarf2out_inline_entry): Likewise.
4603 (init_sections_and_labels): Likewise.
4604 (dwarf2out_early_finish): Likewise.
4605 (maybe_reset_location_view): New, from...
4606 (dwarf2out_var_location): ... here. Call it.
4607 * debug.h (dwarf2out_default_as_loc_support): Declare.
4608 (dwarf2out_default_as_locview_support): Declare.
4609 * hooks.c (hook_int_rtx_insn_0): New.
4610 * hooks.h (hook_int_rtx_insn_0): Declare.
4611 * toplev.c (process_options): Take -gas-loc-support and
4612 -gas-locview-support from dwarf2out. Enable
4613 -gvariable-location-views by default only with locview
4614 assembler support. Enable -ginternal-reset-location-views by
4615 default only if the target defines the corresponding hook.
4616 Enable -ginline-points by default if location views are
4617 enabled; force it disabled if statement frontiers are
4619 * tree-inline.c (expand_call_inline): Check option variables.
4620 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
4622 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
4624 PR tree-optimization/84321
4625 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
4626 handling. Also check whether the anti-range contains any values
4627 that satisfy the mask; switch to a VR_RANGE if not.
4629 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
4632 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
4634 2018-02-13 Martin Jambor <mjambor@suse.cz>
4637 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
4638 of call statements, also set location of a load to a temporary.
4640 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
4642 * config/rl78/rl78.c (add_vector_labels): New function.
4643 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
4644 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
4645 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
4646 which checks that no arguments are passed.
4647 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
4648 * doc/extend.texi: Documentation for the new attribute.
4650 2018-02-13 Andreas Schwab <schwab@suse.de>
4652 * config/riscv/linux.h (CPP_SPEC): Define.
4654 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4657 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
4658 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
4659 OPTION_MASK_ISA_AES as first argument to def_builtin_const
4660 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
4661 instead of OPTION_MASK_ISA_PCLMUL as first argument to
4662 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
4663 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
4664 temporarily for AES and PCLMUL builtins.
4666 PR tree-optimization/84339
4667 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
4668 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
4672 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
4673 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
4674 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
4676 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
4678 * omp-simd-clone.h: New file.
4679 * omp-simd-clone.c: Include omp-simd-clone.h.
4680 (expand_simd_clones): No longer static.
4681 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
4682 cgraph.h and omp-simd-clone.h.
4683 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
4684 (vect_recog_widen_shift_pattern): Formatting fix.
4685 (vect_pattern_recog_1): Don't check optab for calls.
4688 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
4689 operands[2] into a REG before using gen_lowpart on it.
4691 2018-02-12 Jeff Law <law@redhat.com>
4694 * config/sh/sh.c (find_barrier): Consider a sibling call
4697 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
4698 successfully back substituting a reg.
4700 2018-02-12 Richard Biener <rguenther@suse.de>
4702 PR tree-optimization/84037
4703 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
4704 parameter, move visited init to caller.
4705 (vect_slp_analyze_operations): Separate cost from validity
4706 check, initialize visited once for all instances.
4707 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
4709 * tree-vect-stmts.c (vect_model_simple_cost): Make early
4711 (vect_model_promotion_demotion_cost): Likewise.
4712 (vectorizable_bswap): Guard cost modeling with !slp_node
4713 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
4715 (vectorizable_call): Likewise.
4716 (vectorizable_conversion): Likewise.
4717 (vectorizable_assignment): Likewise.
4718 (vectorizable_shift): Likewise.
4719 (vectorizable_operation): Likewise.
4720 (vectorizable_store): Likewise.
4721 (vectorizable_load): Likewise.
4722 (vectorizable_condition): Likewise.
4723 (vectorizable_comparison): Likewise.
4725 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
4728 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
4729 (ASAN_MARK): Fix fnspec to account for return value, change pointer
4730 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
4732 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4735 * params.def (inline-min-speedup): Increase from 8 to 15.
4736 (max-inline-insns-auto): Decrease from 40 to 30.
4737 * ipa-split.c (consider_split): Add some buffer for function to
4738 be considered inlining candidate.
4739 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
4742 2018-02-12 Richard Biener <rguenther@suse.de>
4744 PR tree-optimization/84037
4745 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
4746 matched stmts if we cannot swap the non-matched ones.
4748 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4750 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
4751 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
4752 _mm_maskz_scalef_round_ss): New intrinsics.
4753 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
4754 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
4755 __builtin_ia32_scalefss_round): Remove.
4756 (__builtin_ia32_scalefsd_mask_round,
4757 __builtin_ia32_scalefss_mask_round): New intrinsics.
4758 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
4759 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
4760 ((match_operand:VF_128 2 "<round_nimm_predicate>"
4761 "<round_constraint>")): Changed to ...
4762 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
4763 "<round_scalar_constraint>")): ... this.
4764 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
4765 %0, %1, %2<round_op3>}"): Changed to ...
4766 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
4767 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
4768 %2<round_scalar_mask_op3>}"): ... this.
4769 * config/i386/subst.md (round_scalar_nimm_predicate): New.
4771 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4773 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
4774 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
4775 (_mm_maskz_sqrt_round_ss): New intrinsics.
4776 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
4777 (__builtin_ia32_sqrtsd_mask_round)
4778 (__builtin_ia32_sqrtss_mask_round): New builtins.
4779 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
4780 (__builtin_ia32_sqrtss_round): Remove.
4781 (__builtin_ia32_sqrtsd_mask_round)
4782 (__builtin_ia32_sqrtss_mask_round): New builtins.
4783 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
4784 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
4785 ((match_operand:VF_128 1 "vector_operand"
4786 "xBm,<round_constraint>")): Changed to ...
4787 ((match_operand:VF_128 1 "vector_operand"
4788 "xBm,<round_scalar_constraint>")): ... this.
4789 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
4790 %0, %2, %<iptr>1<round_op3>}): Changed to ...
4791 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
4792 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
4793 %<iptr>1<round_scalar_mask_op3>}): ... this.
4794 ((set_attr "prefix" "<round_prefix>")): Changed to ...
4795 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
4797 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
4800 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
4801 Cast vec_cmpeq result to correct type.
4802 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
4803 Cast vec_cmpgt result to correct type.
4805 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
4807 * final.c (final_scan_insn_1): Renamed from...
4808 (final_scan_insn): ... this. New wrapper, to recover
4809 seen from the outermost call in recursive ones.
4810 * config/sparc/sparc.c (output_return): Drop seen from call.
4811 (output_sibcall): Likewise.
4812 * config/visium/visium.c (output_branch): Likewise.
4814 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
4816 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
4819 2018-02-10 Alan Modra <amodra@gmail.com>
4822 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
4823 Specify LR as an input.
4825 2018-02-10 Jakub Jelinek <jakub@redhat.com>
4828 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
4829 remove_member_access_dummy_vars): New functions.
4830 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
4831 lower_omp_1, execute_lower_omp): Use them.
4833 PR rtl-optimization/84308
4834 * shrink-wrap.c (spread_components): Release todo vector.
4836 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
4838 PR rtl-optimization/57193
4839 * ira-color.c (struct allocno_color_data): Add member
4840 conflict_allocno_hard_prefs.
4841 (update_conflict_allocno_hard_prefs): New.
4842 (bucket_allocno_compare_func): Add a preference based on
4843 conflict_allocno_hard_prefs.
4844 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
4845 (color_allocnos): Remove a dead code. Initiate
4846 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
4848 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4851 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
4852 constraint from =wa to wa. Avoid a subreg on the output operand,
4853 instead use a pseudo and subreg it in a move.
4854 (p9_xxbrd_<mode>): Changed to ...
4855 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
4856 (p9_xxbrd_v2df): New expander.
4857 (p9_xxbrw_<mode>): Changed to ...
4858 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
4859 (p9_xxbrw_v4sf): New expander.
4861 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4863 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
4865 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
4868 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
4869 multiply in 32-bit mode.
4870 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
4871 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
4874 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4876 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
4877 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
4878 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
4879 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
4881 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
4884 * dwarf2out.c (is_trivial_indirect_ref): New function.
4885 (dwarf2out_late_global_decl): Do not generate a location
4886 attribute for variables that have a non-trivial DECL_VALUE_EXPR
4887 and that are not defined in the current unit.
4889 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4891 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
4892 instead of a libcall for UNORDERED.
4894 2018-02-09 Tamar Christina <tamar.christina@arm.com>
4897 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
4898 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
4900 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4903 * config/s390/s390.c (s390_set_current_function): Invoke
4904 s390_indirect_branch_settings also if fndecl didn't change.
4906 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4908 * config/rs6000/rs6000.md (blockage): Set length to zero.
4910 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4912 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
4914 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4917 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
4918 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
4922 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
4923 PARALLEL incoming that failed vt_get_decl_and_offset check.
4926 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
4927 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
4929 (get_variable_section): For decls in named .bss* sections pass true as
4930 second argument to bss_initializer_p.
4932 2018-02-09 Marek Polacek <polacek@redhat.com>
4933 Jakub Jelinek <jakub@redhat.com>
4936 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
4937 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
4938 Sync some changes from cxx_fold_indirect_ref.
4940 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4942 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
4944 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
4945 (BLOCK_INLINE_ENTRY_LABEL): New.
4946 (dwarf2out_var_location): Disregard inline entry markers.
4947 (inline_entry_data): New struct.
4948 (inline_entry_data_hasher): New hashtable type.
4949 (inline_entry_data_hasher::hash): New.
4950 (inline_entry_data_hasher::equal): New.
4951 (inline_entry_data_table): New variable.
4952 (add_high_low_attributes): Add DW_AT_entry_pc and
4953 DW_AT_GNU_entry_view attributes if a pending entry is found
4954 in inline_entry_data_table. Add old entry_pc attribute only
4955 if debug nonbinding markers are disabled.
4956 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
4957 markers are enabled.
4958 (block_within_block_p, dwarf2out_inline_entry): New.
4959 (dwarf2out_finish): Check that no entries remained in
4960 inline_entry_data_table.
4961 * final.c (reemit_insn_block_notes): Handle inline entry notes.
4962 (final_scan_insn, notice_source_line): Likewise.
4963 (rest_of_clean_state): Skip inline entry markers.
4964 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
4966 * gimple.c (gimple_build_debug_inline_entry): New.
4967 * gimple.h (enum gimple_debug_subcode): Add
4968 GIMPLE_DEBUG_INLINE_ENTRY.
4969 (gimple_build_debug_inline_entry): Declare.
4970 (gimple_debug_inline_entry_p): New.
4971 (gimple_debug_nonbind_marker_p): Adjust.
4972 * insn-notes.def (INLINE_ENTRY): New.
4973 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
4974 inline entry marker notes.
4975 (print_insn): Likewise.
4976 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
4977 (INSN_DEBUG_MARKER_KIND): Likewise.
4978 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
4979 * tree-inline.c (expand_call_inline): Build and insert
4980 debug_inline_entry stmt.
4981 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
4982 inline entry blocks early, if nonbind markers are enabled.
4983 (dump_scope_block): Dump fragment info.
4984 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
4985 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
4986 (gimple_build_debug_inline_entry): New.
4987 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
4988 Enable/disable inline entry points too.
4989 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
4990 (DEBUG_INSN): Describe inline entry markers.
4992 * common.opt (gvariable-location-views): New.
4993 (gvariable-location-views=incompat5): New.
4994 * config.in: Rebuilt.
4995 * configure: Rebuilt.
4996 * configure.ac: Test assembler for view support.
4997 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
4998 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
4999 * dwarf2out.c (var_loc_view): New typedef.
5000 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
5001 (dwarf2out_locviews_in_attribute): New.
5002 (dwarf2out_locviews_in_loclist): New.
5003 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
5004 (enum dw_line_info_opcode): Add LI_adv_address.
5005 (struct dw_line_info_table): Add view.
5006 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
5007 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
5008 (zero_view_p): New variable.
5009 (ZERO_VIEW_P): New macro.
5010 (output_asm_line_debug_info): New.
5011 (struct var_loc_node): Add view.
5012 (add_AT_view_list, AT_loc_list): New.
5013 (add_var_loc_to_decl): Add view param. Test it against last.
5014 (new_loc_list): Add view params. Record them.
5015 (AT_loc_list_ptr): Handle loc and view lists.
5016 (view_list_to_loc_list_val_node): New.
5017 (print_dw_val): Handle dw_val_class_view_list.
5018 (size_of_die): Likewise.
5019 (value_format): Likewise.
5020 (loc_list_has_views): New.
5021 (gen_llsym): Set vl_symbol too.
5022 (maybe_gen_llsym, skip_loc_list_entry): New.
5023 (dwarf2out_maybe_output_loclist_view_pair): New.
5024 (output_loc_list): Output view list or entries too.
5025 (output_view_list_offset): New.
5026 (output_die): Handle dw_val_class_view_list.
5027 (output_dwarf_version): New.
5028 (output_compilation_unit_header): Use it.
5029 (output_skeleton_debug_sections): Likewise.
5030 (output_rnglists, output_line_info): Likewise.
5031 (output_pubnames, output_aranges): Update version comments.
5032 (output_one_line_info_table): Output view numbers in asm comments.
5033 (dw_loc_list): Determine current endview, pass it to new_loc_list.
5034 Call maybe_gen_llsym.
5035 (loc_list_from_tree_1): Adjust.
5036 (add_AT_location_description): Create view list attribute if
5037 needed, check it's absent otherwise.
5038 (convert_cfa_to_fb_loc_list): Adjust.
5039 (maybe_emit_file): Call output_asm_line_debug_info for test.
5040 (dwarf2out_var_location): Reset views as needed. Precompute
5041 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
5042 attribute. Set view.
5043 (new_line_info_table): Reset next view.
5044 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
5045 (dwarf2out_source_line): Likewise. Output view resets and labels to
5046 the assembler, or select appropriate line info opcodes.
5047 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
5048 (optimize_string_length): Catch it. Adjust.
5049 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
5050 dw_val_class_view_list, and remove it if no longer needed.
5051 (hash_loc_list): Hash view numbers.
5052 (loc_list_hasher::equal): Compare them.
5053 (optimize_location_lists): Check whether a view list symbol is
5054 needed, and whether the locview attribute is present, and
5055 whether they match. Remove the locview attribute if no longer
5057 (index_location_lists): Call skip_loc_list_entry for test.
5058 (dwarf2out_finish): Call output_asm_line_debug_info for test.
5059 Use output_dwarf_version.
5060 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
5061 (struct dw_val_node): Add val_view_list.
5062 * final.c (SEEN_NEXT_VIEW): New.
5063 (set_next_view_needed): New.
5064 (clear_next_view_needed): New.
5065 (maybe_output_next_view): New.
5066 (final_start_function): Rename to...
5067 (final_start_function_1): ... this. Take pointer to FIRST,
5068 add SEEN parameter. Emit param bindings in the initial view.
5069 (final_start_function): Reintroduce SEEN-less interface.
5070 (final): Rename to...
5071 (final_1): ... this. Take SEEN parameter. Output final pending
5072 next view at the end.
5073 (final): Reintroduce seen-less interface.
5074 (final_scan_insn): Output pending next view before switching
5075 sections or ending a block. Mark the next view as needed when
5076 outputting variable locations. Notify debug backend of section
5077 changes, and of location view changes.
5078 (rest_of_handle_final): Adjust.
5079 * toplev.c (process_options): Autodetect value for debug variable
5080 location views option. Warn on incompat5 without -gdwarf-5.
5081 * doc/invoke.texi (gvariable-location-views): New.
5082 (gvariable-location-views=incompat5): New.
5083 (gno-variable-location-views): New.
5085 2018-02-08 David Malcolm <dmalcolm@redhat.com>
5087 PR tree-optimization/84136
5088 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
5089 that the result of find_edge is non-NULL.
5091 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5094 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
5095 storing integer register in SImode. Fix cost of 256 and 512
5096 byte aligned SSE register store.
5098 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5100 * config/i386/i386.c (ix86_multiplication_cost): Fix
5101 multiplication cost for TARGET_AVX512DQ.
5103 2018-02-08 Marek Polacek <polacek@redhat.com>
5105 PR tree-optimization/84238
5106 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
5109 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5111 PR tree-optimization/84265
5112 * tree-vect-stmts.c (vectorizable_store): Don't treat
5113 VMAT_CONTIGUOUS accesses as grouped.
5114 (vectorizable_load): Likewise.
5116 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5118 PR tree-optimization/81635
5119 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
5120 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
5121 (test_round_for_mask): New functions.
5122 (wide_int_cc_tests): Call test_round_for_mask.
5123 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
5124 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
5125 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
5126 range returned by get_range_info.
5128 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5131 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
5132 * symtab.c: Include builtins.h
5133 (symtab_node::output_to_lto_symbol_table_p): Move here
5134 from lto-streamer-out.c:output_symbol_p.
5135 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
5136 (output_symbol_p): Move all logic to symtab.c
5137 (produce_symtab): Update.
5139 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5141 * config/s390/s390-opts.h (enum indirect_branch): Define.
5142 * config/s390/s390-protos.h (s390_return_addr_from_memory)
5143 (s390_indirect_branch_via_thunk)
5144 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
5145 (enum s390_indirect_branch_type): Define.
5146 * config/s390/s390.c (struct s390_frame_layout, struct
5147 machine_function): Remove.
5148 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
5149 (indirect_branch_table_label_no, indirect_branch_table_name):
5151 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
5152 (enum s390_indirect_branch_option): Define.
5153 (s390_return_addr_from_memory): New function.
5154 (s390_handle_string_attribute): New function.
5155 (s390_attribute_table): Add new attribute handler.
5156 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
5157 (s390_indirect_branch_via_thunk): New function.
5158 (s390_indirect_branch_via_inline_thunk): New function.
5159 (s390_function_ok_for_sibcall): When jumping via thunk disallow
5160 sibling call optimization for non z10 compiles.
5161 (s390_emit_call): Force indirect branch target to be a single
5162 register. Add r1 clobber for non-z10 compiles.
5163 (s390_emit_epilogue): Emit return jump via return_use expander.
5164 (s390_reorg): Handle JUMP_INSNs as execute targets.
5165 (s390_option_override_internal): Perform validity checks for the
5166 new command line options.
5167 (s390_indirect_branch_attrvalue): New function.
5168 (s390_indirect_branch_settings): New function.
5169 (s390_set_current_function): Invoke s390_indirect_branch_settings.
5170 (s390_output_indirect_thunk_function): New function.
5171 (s390_code_end): Implement target hook.
5172 (s390_case_values_threshold): Implement target hook.
5173 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
5175 * config/s390/s390.h (struct s390_frame_layout)
5176 (struct machine_function): Move here from s390.c.
5177 (TARGET_INDIRECT_BRANCH_NOBP_RET)
5178 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
5179 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
5180 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
5181 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
5182 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
5183 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
5184 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
5185 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
5186 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
5187 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
5188 (mnemonic attribute): Add values which aren't recognized
5190 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
5191 pattern for branch conversion. Fix mnemonic attribute.
5192 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
5193 indirect branch via thunk if requested.
5194 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
5195 ("*indirect_jump"): Disable for branch conversion using out of
5197 ("indirect_jump_via_thunk<mode>_z10")
5198 ("indirect_jump_via_thunk<mode>")
5199 ("indirect_jump_via_inlinethunk<mode>_z10")
5200 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
5201 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
5202 ("casesi_jump_via_inlinethunk<mode>_z10")
5203 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
5204 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
5205 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
5206 ("*indirect2_jump"): Disable for branch conversion.
5207 ("casesi_jump"): Turn into expander and expand patterns for branch
5209 ("return_use"): New expander.
5210 ("*return"): Emit return via thunk and rename it to ...
5211 ("*return<mode>"): ... this one.
5212 * config/s390/s390.opt: Add new options and and enum for the
5215 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5217 * lra-constraints.c (match_reload): Unconditionally use
5218 gen_lowpart_SUBREG, rather than selecting between that
5219 and equivalent gen_rtx_SUBREG code.
5221 2018-02-08 Richard Biener <rguenther@suse.de>
5223 PR tree-optimization/84233
5224 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
5225 changed flag instead of boguously re-using phi_inserted.
5227 2018-02-08 Martin Jambor <mjambor@suse.cz>
5229 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
5230 static local variables.
5232 2018-02-08 Richard Biener <rguenther@suse.de>
5234 PR tree-optimization/84278
5235 * tree-vect-stmts.c (vectorizable_store): When looking for
5236 smaller vector types to perform grouped strided loads/stores
5237 make sure the mode is supported by the target.
5238 (vectorizable_load): Likewise.
5240 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5242 * config/aarch64/aarch64.c (aarch64_components_for_bb):
5243 Increase LDP/STP opportunities by adding adjacent callee-saves.
5245 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5247 PR rtl-optimization/84068
5248 PR rtl-optimization/83459
5249 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
5251 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
5253 PR tree-optimization/84224
5254 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
5255 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
5258 2018-02-07 Iain Sandoe <iain@codesourcery.com>
5261 * config/rs6000/altivec.md (*restore_world): Remove LR use.
5262 * config/rs6000/predicates.md (restore_world_operation): Adjust op
5263 count, remove one USE.
5265 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
5267 * doc/install.texi (Configuration): Document the
5268 --with-long-double-format={ibm,ieee} PowerPC configuration
5272 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
5273 Convert from define_expand to be define_insn_and_split. Rework
5274 float/double/_Float128 conversions to QI/HI/SImode to work with
5275 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
5276 conversions to QI/HImode types did a store and then a load to
5277 truncate the value. For conversions to VSX registers, don't split
5278 the insn, instead emit the code directly. Use the code iterator
5279 any_fix to combine signed and unsigned conversions.
5280 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
5281 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
5282 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5283 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
5284 (fix_<mode>di2_hw): Likewise.
5285 (fixuns_<mode>di2_hw): Likewise.
5286 (fix_<mode>si2_hw): Likewise.
5287 (fixuns_<mode>si2_hw): Likewise.
5288 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
5289 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5290 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
5291 fix<uns>_trunc<SFDF:mode>si2_p8.
5292 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
5294 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
5295 (fix<uns>_<mode>_mem): Likewise.
5296 (fctiw<u>z_<mode>_mem): Likewise.
5297 (fix<uns>_<mode>_mem): Likewise.
5298 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
5299 the register allocator from doing a direct move to the GPRs to do
5300 a store, and instead use the ISA 3.0 store byte/half-word from
5301 vector register instruction. For IEEE 128-bit floating point,
5302 also optimize stores of 32-bit ints.
5303 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
5305 2018-02-07 Alan Hayward <alan.hayward@arm.com>
5307 * genextract.c (push_pathstr_operand): New function to support
5309 (walk_rtx): Call push_pathstr_operand.
5310 (print_path): Support [a-zA-Z].
5312 2018-02-07 Richard Biener <rguenther@suse.de>
5314 PR tree-optimization/84037
5315 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
5316 (cse_and_gimplify_to_preheader): Declare.
5317 (vect_get_place_in_interleaving_chain): Likewise.
5318 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5320 (_loop_vec_info::~_loop_vec_info): Delete it.
5321 (cse_and_gimplify_to_preheader): New function.
5322 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
5323 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
5324 (vectorizable_load): Likewise. For grouped stores always base
5325 the IV on the first element.
5326 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
5327 condition before gimplifying.
5329 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5331 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
5332 *DIV_EXPR and *MOD_EXPR.
5334 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
5337 * config/i386/i386.c (ix86_option_override_internal): Mask out
5338 the CF_SET bit when checking -fcf-protection.
5340 2018-02-07 Tom de Vries <tom@codesourcery.com>
5343 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
5346 2018-02-07 Richard Biener <rguenther@suse.de>
5348 PR tree-optimization/84204
5349 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
5352 PR tree-optimization/84205
5353 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
5354 special-case isl_ast_op_zdiv_r.
5356 PR tree-optimization/84223
5357 * graphite-scop-detection.c (gather_bbs::before_dom_children):
5358 Only add conditions from within the region.
5359 (gather_bbs::after_dom_children): Adjust.
5361 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
5364 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
5365 * config/avr/avr.md: Only post-reload split REG-REG moves if
5366 either register is GENERAL_REG_P.
5368 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5370 PR tree-optimization/84235
5371 * tree-ssa-scopedtables.c
5372 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
5373 if the subtraction is performed in floating point type where NaNs are
5374 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
5375 build 1. Formatting fix.
5377 2018-02-06 Jakub Jelinek <jakub@redhat.com>
5380 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
5381 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
5382 and skip it regardless of bb boundaries. Use CALL_P macro,
5383 don't test INSN_P (insn) together with CALL_P or JUMP_P check
5384 unnecessarily, formatting fix.
5386 2018-02-06 Michael Collison <michael.collison@arm.com>
5388 * config/arm/thumb2.md:
5389 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
5390 (*thumb_mov_notscc): Ditto.
5392 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
5395 * config/rs6000/rs6000.md (su code attribute): Use "u" for
5396 unsigned_fix, not "s".
5398 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5400 * configure.ac (gcc_fn_eh_frame_ro): New function.
5401 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
5402 correct .eh_frame permissions.
5403 * configure: Regenerate.
5405 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
5407 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
5410 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5412 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5413 Display warning message for -mno-speculate-indirect-jumps.
5415 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
5417 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
5419 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
5421 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
5423 PR tree-optimization/84225
5424 * tree-eh.c (find_trapping_overflow): Only call
5425 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
5427 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5430 * config/i386/i386.c: Reimplement the check of possible options
5431 -mibt/-mshstk conbination. Change error messages.
5432 * doc/invoke.texi: Fix a typo: remove extra '='.
5434 2018-02-06 Marek Polacek <polacek@redhat.com>
5436 PR tree-optimization/84228
5437 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
5439 2018-02-06 Tamar Christina <tamar.christina@arm.com>
5442 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
5443 emitted arch directives.
5444 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
5445 __ARM_FEATURE_COPROC before changing architectures.
5447 2018-02-06 Richard Biener <rguenther@suse.de>
5449 * config/i386/i386.c (print_reg): Fix typo.
5450 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
5452 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
5454 * configure: Regenerate.
5456 2018-02-05 Martin Sebor <msebor@redhat.com>
5458 PR tree-optimization/83369
5459 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
5462 2018-02-05 Martin Liska <mliska@suse.cz>
5464 * doc/invoke.texi: Cherry-pick upstream r323995.
5466 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
5468 * ira.c (ira_init_register_move_cost): Adjust comment.
5470 2018-02-05 Martin Liska <mliska@suse.cz>
5472 PR gcov-profile/84137
5473 * doc/gcov.texi: Fix typo in documentation.
5475 2018-02-05 Martin Liska <mliska@suse.cz>
5477 PR gcov-profile/83879
5478 * doc/gcov.texi: Document necessity of --dynamic-list-data when
5479 using dlopen functionality.
5481 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
5483 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
5484 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
5485 _mm_maskz_range_ss, _mm_mask_range_round_ss,
5486 _mm_maskz_range_round_ss): New intrinsics.
5487 (__builtin_ia32_rangesd128_round)
5488 (__builtin_ia32_rangess128_round): Remove.
5489 (__builtin_ia32_rangesd128_mask_round,
5490 __builtin_ia32_rangess128_mask_round): New builtins.
5491 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
5492 __builtin_ia32_rangess128_round): Remove.
5493 (__builtin_ia32_rangesd128_mask_round,
5494 __builtin_ia32_rangess128_mask_round): New builtins.
5495 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
5496 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
5497 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
5498 "<round_saeonly_constraint>")): Changed to ...
5499 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
5500 "<round_saeonly_scalar_constraint>")): ... this.
5501 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
5502 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
5503 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
5504 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
5505 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
5507 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
5509 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
5511 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
5512 Remove all values except native, 8540 and 8548.
5514 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
5516 * config/i386/i386.c (ix86_output_function_return): Pass
5517 INVALID_REGNUM, instead of -1, as invalid register number to
5518 indirect_thunk_name and output_indirect_thunk.
5520 2018-02-02 Julia Koval <julia.koval@intel.com>
5522 * config.gcc: Add -march=icelake.
5523 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
5524 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
5525 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
5526 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
5527 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
5528 (processor_target_table): Add icelake.
5529 (ix86_option_override_internal): Handle new PTAs.
5530 (get_builtin_code_for_version): Handle icelake.
5531 (M_INTEL_COREI7_ICELAKE): New.
5532 (fold_builtin_cpu): Handle icelake.
5533 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
5534 * doc/invoke.texi: Add -march=icelake.
5536 2018-02-02 Julia Koval <julia.koval@intel.com>
5538 * config/i386/i386.c (ix86_option_override_internal): Change flags type
5539 to wide_int_bitmask.
5540 * wide-int-bitmask.h: New.
5542 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5545 * config/i386/i386.md: Replace Pmode with word_mode in
5546 builtin_setjmp_setup and builtin_longjmp to support x32.
5548 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
5552 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
5554 (rs6000_supported_cpu_names): New static variable.
5555 (linux_cpu_translation_table): Likewise.
5556 (elf_platform) <cpu>: Define new static variable and use it.
5557 Translate kernel AT_PLATFORM name to canonical name if needed.
5558 Error if platform name is unknown.
5560 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
5563 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
5565 2018-02-01 Jeff Law <law@redhat.com>
5568 * config/i386/i386.c (release_scratch_register_on_entry): Add new
5569 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
5570 the scratch if RELEASE_VIA_POP is false.
5571 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
5572 If we have to save a temporary register, decrement SIZE appropriately.
5573 Pass new arguments to release_scratch_register_on_entry.
5574 (ix86_adjust_stack_and_probe): Likewise.
5575 (ix86_emit_probe_stack_range): Pass new arguments to
5576 release_scratch_register_on_entry.
5578 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
5580 PR rtl-optimization/84157
5581 * combine.c (change_zero_ext): Use REG_P predicate in
5582 front of HARD_REGISTER_P predicate.
5584 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
5586 * config/avr/avr.c (avr_option_override): Move disabling of
5587 -fdelete-null-pointer-checks to...
5588 * common/config/avr/avr-common.c (avr_option_optimization_table):
5591 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5593 PR tree-optimization/81635
5594 * tree-data-ref.c (split_constant_offset_1): For types that
5595 wrap on overflow, try to use range info to prove that wrapping
5598 2018-02-01 Renlin Li <renlin.li@arm.com>
5601 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
5603 (aarch64_register_move_cost): Likewise.
5604 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
5606 (REG_CLASS_NAMES): Likewise.
5607 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
5608 TAILCALL_ADDR_REGS. Remove IP registers.
5609 * config/aarch64/aarch64.md (Ucs): Update register constraint.
5611 2018-02-01 Richard Biener <rguenther@suse.de>
5613 * domwalk.h (dom_walker::dom_walker): Add additional constructor
5614 for specifying RPO order and allow NULL for that.
5615 * domwalk.c (dom_walker::dom_walker): Likewise.
5616 (dom_walker::walk): Handle NULL RPO order.
5617 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
5619 (rewrite_update_dom_walker): Likewise.
5620 (mark_def_dom_walker): Likewise.
5622 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5624 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
5625 (aarch64_maybe_expand_sve_subreg_move): Declare.
5626 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
5627 * config/aarch64/predicates.md (aarch64_any_register_operand): New
5629 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
5630 that are semantically a reverse operation.
5631 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
5632 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
5633 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
5635 (aarch64_can_change_mode_class): For big-endian, forbid changes
5636 between two SVE modes if they have different element sizes.
5638 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5640 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
5641 the TImode handling for big-endian targets.
5643 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5645 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
5646 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
5648 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
5649 Remove BSWAP handing for big-endian targets and use the form of
5650 LD1RQ appropariate for the mode.
5652 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5654 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
5655 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
5658 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5661 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
5662 check for operands that need to go through aarch64_sve_reload_be.
5664 2018-02-01 Jakub Jelinek <jakub@redhat.com>
5666 PR tree-optimization/81661
5667 PR tree-optimization/84117
5668 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
5669 * tree-eh.c: Include gimplify.h.
5670 (find_trapping_overflow, replace_trapping_overflow,
5671 rewrite_to_non_trapping_overflow): New functions.
5672 * tree-vect-loop.c: Include tree-eh.h.
5673 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
5674 * tree-data-ref.c: Include tree-eh.h.
5675 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
5677 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
5679 PR rtl-optimization/84123
5680 * combine.c (change_zero_ext): Check if hard register satisfies
5681 can_change_dest_mode before calling gen_lowpart_SUBREG.
5683 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
5686 * ira.c (ira_init_register_move_cost): Remove assert.
5688 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5690 PR rtl-optimization/84071
5691 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
5692 * doc/tm.texi: Regenerate.
5694 2018-01-31 Richard Biener <rguenther@suse.de>
5696 PR tree-optimization/84132
5697 * tree-data-ref.c (analyze_miv_subscript): Properly
5698 check whether evolution_function_is_affine_multivariate_p
5699 before calling gcd_of_steps_may_divide_p.
5701 2018-01-31 Julia Koval <julia.koval@intel.com>
5704 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
5705 * config/i386/i386.md (rdpid_rex64) New.
5706 (rdpid): Make 32bit only.
5708 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
5711 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
5712 an IDENTIFIER_NODE for FUNCTION_TYPE's.
5714 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5717 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5719 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5721 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5723 PR rtl-optimization/84071
5724 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
5725 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
5727 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5729 * config/arc/arc.c (arc_handle_aux_attribute): New function.
5730 (arc_attribute_table): Add 'aux' attribute.
5731 (arc_in_small_data_p): Consider aux like variables.
5732 (arc_is_aux_reg_p): New function.
5733 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
5734 (arc_get_aux_arg): New function.
5735 (prepare_move_operands): Handle aux-register access.
5736 (arc_handle_aux_attribute): New function.
5737 * doc/extend.texi (ARC Variable attributes): Add subsection.
5739 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5741 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
5742 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
5743 (arc_attribute_table): Add 'uncached' attribute.
5744 (arc_print_operand): Print '.di' flag for uncached memory
5746 (arc_in_small_data_p): Do not consider for small data the uncached
5748 (arc_is_uncached_mem_p): New function.
5749 * config/arc/predicates.md (compact_store_memory_operand): Check
5750 for uncached memory accesses.
5751 (nonvol_nonimm_operand): Likewise.
5752 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
5754 2018-01-31 Jakub Jelinek <jakub@redhat.com>
5757 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
5758 falign-loops=): Add Optimization flag.
5760 2018-01-30 Jeff Law <law@redhat.com>
5763 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
5764 INT_REGISTERS_SAVED. Check it prior to calling
5765 get_scratch_register_on_entry.
5766 (ix86_adjust_stack_and_probe): Similarly.
5767 (ix86_emit_probe_stack_range): Similarly.
5768 (ix86_expand_prologue): Corresponding changes.
5770 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5773 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
5774 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
5776 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
5779 * lra-constraints.c (curr_insn_transform): Process AND in the
5782 2018-01-30 Jakub Jelinek <jakub@redhat.com>
5784 PR rtl-optimization/83986
5785 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
5786 dependence against last_pending_memory_flush in addition to
5789 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
5791 PR tree-optimization/81611
5792 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
5795 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5798 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
5801 2018-01-30 Richard Biener <rguenther@suse.de>
5802 Jakub Jelinek <jakub@redhat.com>
5804 PR tree-optimization/84111
5805 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
5806 inner loops added during recursion, as they don't have up-to-date
5809 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5812 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
5813 (can_inline_edge_by_limits_p): ... here.
5814 (can_early_inline_edge_p, check_callers,
5815 update_caller_keys, update_callee_keys, recursive_inlining,
5816 add_new_edges_to_heap, speculation_useful_p,
5817 inline_small_functions,
5818 inline_small_functions, flatten_function,
5819 inline_to_all_callers_1): Update.
5821 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5823 * profile-count.c (profile_count::combine_with_ipa_count): Handle
5826 2018-01-30 Richard Biener <rguenther@suse.de>
5828 PR tree-optimization/83008
5829 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
5830 invariant and constant vector uses in stmts when they need
5833 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5836 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
5837 * configure: Regenerate.
5839 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5841 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
5843 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
5844 Use gen_rtx_REG rather than gen_lowpart.
5846 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5848 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
5849 rather than 0 when creating partial subregs.
5851 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5853 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
5856 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
5859 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
5860 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
5861 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
5862 flags. This restores the settings used before the 2017-07-24.
5863 Turning off pre increment/decrement/modify allows IVOPTS to
5864 optimize DF/SF loops where the index is an int.
5866 2018-01-29 Richard Biener <rguenther@suse.de>
5867 Kelvin Nilsen <kelvin@gcc.gnu.org>
5870 * tree-vect-stmts.c (vectorizable_call): Don't call
5871 targetm.vectorize_builtin_md_vectorized_function if callee is
5874 2018-01-22 Carl Love <cel@us.ibm.com>
5876 * doc/extend.tex: Fix typo in second arg in
5877 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
5879 2018-01-29 Richard Biener <rguenther@suse.de>
5881 PR tree-optimization/84086
5882 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
5883 (flush_ssaname_freelist): When SSA names were released reset
5884 the SCEV hash table.
5886 2018-01-29 Richard Biener <rguenther@suse.de>
5888 PR tree-optimization/84057
5889 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
5890 removed paths when removing edges.
5892 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
5894 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
5895 -mfunction-return=@var{choice}.
5897 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
5900 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
5902 (layout::print_source_line): Likewise.
5903 (test_get_line_width_without_trailing_whitespace): Add test cases.
5905 2018-01-27 Jakub Jelinek <jakub@redhat.com>
5908 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
5911 2018-01-26 Jim Wilson <jimw@sifive.com>
5913 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
5915 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
5918 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5920 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
5921 and CMP + SUB-immediate -> SUBS.
5923 2018-01-26 Martin Sebor <msebor@redhat.com>
5925 PR tree-optimization/83896
5926 * tree-ssa-strlen.c (get_string_len): Rename...
5927 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
5928 Avoid assuming length is constant.
5929 (handle_char_store): Use HOST_WIDE_INT for string length.
5931 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
5934 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
5935 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
5937 2018-01-26 Richard Biener <rguenther@suse.de>
5939 PR rtl-optimization/84003
5940 * dse.c (record_store): Only record redundant stores when
5941 the earlier store aliases at least all accesses the later one does.
5943 2018-01-26 Jakub Jelinek <jakub@redhat.com>
5945 PR rtl-optimization/83985
5946 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
5947 REG_CFA_RESTORE insns.
5948 (delete_unmarked_insns): Don't ignore separate shrink wrapping
5949 REG_CFA_RESTORE insns here.
5952 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
5953 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
5955 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5957 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
5958 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
5959 (arc_init): Likewise.
5960 (arc_override_options): Likewise.
5961 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
5963 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
5965 * config/arc/arc.h (TARGET_DBNZ): Define.
5966 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
5967 properly set the tune attribute.
5968 (dbnz): Use TARGET_DBNZ guard.
5969 * config/arc/arc.opt (mtune): Add core3 option.
5971 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5973 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
5974 recognize new pic like addresses.
5975 (arc_delegitimize_address): Clean up.
5977 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5979 * config/arc/arc-arches.def: Option mrf16 valid for all
5981 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
5982 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
5983 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
5984 * config/arc/arc-tables.opt: Regenerate.
5985 * config/arc/arc.c (arc_conditional_register_usage): Handle
5986 reduced register file case.
5987 (arc_file_start): Set must have build attributes.
5988 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
5990 * config/arc/arc.opt (mrf16): Add new option.
5991 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
5992 * config/arc/genmultilib.awk: Handle new mrf16 option.
5993 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
5994 * config/arc/t-multilib: Regenerate.
5995 * doc/invoke.texi (ARC Options): Document mrf16 option.
5997 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5999 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
6000 * config/arc/arc.c (arc_handle_secure_attribute): New function.
6001 (arc_attribute_table): Add 'secure_call' attribute.
6002 (arc_print_operand): Print secure call operand.
6003 (arc_function_ok_for_sibcall): Don't optimize tail calls when
6005 (arc_is_secure_call_p): New function. * config/arc/arc.md
6006 (call_i): Add support for sjli instruction.
6007 (call_value_i): Likewise.
6008 * config/arc/constraints.md (Csc): New constraint.
6010 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6011 John Eric Martin <John.Martin@emmicro-us.com>
6013 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
6014 * config/arc/arc.c (_arc_jli_section): New struct.
6015 (arc_jli_section): New type.
6016 (rc_jli_sections): New static variable.
6017 (arc_handle_jli_attribute): New function.
6018 (arc_attribute_table): Add jli_always and jli_fixed attribute.
6019 (arc_file_end): New function.
6020 (TARGET_ASM_FILE_END): Define.
6021 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
6022 (arc_add_jli_section): New function.
6023 (jli_call_scan): Likewise.
6024 (arc_reorg): Call jli_call_scan.
6025 (arc_output_addsi): Remove 'S' from printing asm operand.
6026 (arc_is_jli_call_p): New function.
6027 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
6029 (movhi_insn): Likewise.
6030 (movsi_insn): Likewise.
6031 (movsi_set_cc_insn): Likewise.
6032 (loadqi_update): Likewise.
6033 (load_zeroextendqisi_update): Likewise.
6034 (load_signextendqisi_update): Likewise.
6035 (loadhi_update): Likewise.
6036 (load_zeroextendhisi_update): Likewise.
6037 (load_signextendhisi_update): Likewise.
6038 (loadsi_update): Likewise.
6039 (loadsf_update): Likewise.
6040 (movsicc_insn): Likewise.
6041 (bset_insn): Likewise.
6042 (bxor_insn): Likewise.
6043 (bclr_insn): Likewise.
6044 (bmsk_insn): Likewise.
6045 (bicsi3_insn): Likewise.
6046 (cmpsi_cc_c_insn): Likewise.
6047 (movsi_ne): Likewise.
6048 (movsi_cond_exec): Likewise.
6049 (clrsbsi2): Likewise.
6060 (call_i): Remove 'S' asm letter, add jli instruction.
6061 (call_value_i): Likewise.
6062 * config/arc/arc.op (mjli-always): New option.
6063 * config/arc/constraints.md (Cji): New constraint.
6064 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
6066 (subsf3_fpx): Likewise.
6067 (mulsf3_fpx): Likewise.
6068 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
6070 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
6072 * doc/invoke.texi (ARC): Document mjli-always option.
6074 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
6076 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
6077 avoid addition with 0 and use incw and decw where possible.
6079 2018-01-26 Richard Biener <rguenther@suse.de>
6081 PR tree-optimization/81082
6082 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
6083 association if it requires casting to unsigned.
6084 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
6085 from fold_plusminus_mult_expr to catch important cases late when
6086 range info is available.
6088 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6090 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
6091 * configure.ac (hidden_linkonce): New test.
6092 * configure: Regenerate.
6093 * config.in: Regenerate.
6095 2018-01-26 Julia Koval <julia.koval@intel.com>
6097 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
6098 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
6099 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
6100 _mm_mask_bitshuffle_epi64_mask): Fix type.
6101 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
6102 USI_FTYPE_V4DI_V4DI_USI): Remove.
6103 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
6104 __builtin_ia32_vpshufbitqmb256_mask,
6105 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
6106 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
6107 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
6109 2018-01-26 Alan Modra <amodra@gmail.com>
6112 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
6113 UNSPEC_VBPERMQ. Sort other unspecs.
6115 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
6117 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
6119 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
6122 * predict.c (drop_profile): Do not push/pop cfun; update also
6124 (handle_missing_profiles): Fix logic looking for zero profiles.
6126 2018-01-25 Jakub Jelinek <jakub@redhat.com>
6129 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
6130 on functions with #pragma omp declare simd or functions with simd
6132 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
6133 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
6134 Remove trailing \n from warning_at calls.
6136 2018-01-25 Tom de Vries <tom@codesourcery.com>
6139 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6140 for neutered workers.
6142 2018-01-24 Joseph Myers <joseph@codesourcery.com>
6145 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
6146 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
6148 2018-01-24 Jeff Law <law@redhat.com>
6151 * i386.c (get_probe_interval): Move to earlier point.
6152 (ix86_compute_frame_layout): If -fstack-clash-protection and
6153 the frame is larger than the probe interval, then use pushes
6154 to save registers rather than reg->mem moves.
6155 (ix86_expand_prologue): Remove conditional for int_registers_saved
6158 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
6161 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
6162 min/max for never referenced object.
6164 2018-01-24 Jakub Jelinek <jakub@redhat.com>
6167 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
6169 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
6170 attributes from DECL_ATTRIBUTES (decl) without affecting
6171 DECL_ATTRIBUTES (current_function_decl).
6172 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
6173 functions with non-NULL DECL_ABSTRACT_ORIGIN.
6175 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
6177 PR tree-optimization/83979
6178 * fold-const.c (fold_comparison): Use constant_boolean_node
6179 instead of boolean_{true,false}_node.
6181 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
6183 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
6186 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6188 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6189 Simplify the clause that sets the length attribute.
6190 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6191 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
6192 clause that sets the length attribute.
6193 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6195 2018-01-24 Tom de Vries <tom@codesourcery.com>
6198 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
6199 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
6200 Add strict parameter.
6201 (prevent_branch_around_nothing): Insert dummy insn between branch to
6202 label and label with no ptx insn inbetween.
6203 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
6205 2018-01-24 Tom de Vries <tom@codesourcery.com>
6208 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6209 for neutered threads in warp.
6210 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
6212 2018-01-24 Richard Biener <rguenther@suse.de>
6214 PR tree-optimization/83176
6215 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
6218 2018-01-24 Richard Biener <rguenther@suse.de>
6220 PR tree-optimization/82819
6221 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
6222 code generating pluses that are no-ops in the target precision.
6224 2018-01-24 Richard Biener <rguenther@suse.de>
6227 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
6229 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6231 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
6232 to merge probabilities.
6233 * predict.c (probably_never_executed): Also mark as cold functions
6234 with global 0 profile and guessed local profile.
6235 * profile-count.c (profile_probability::combine_with_count): New
6237 * profile-count.h (profile_probability::operator*,
6238 profile_probability::operator*=, profile_probability::operator/,
6239 profile_probability::operator/=): Reduce precision to adjusted
6240 and set value to guessed on contradictory divisions.
6241 (profile_probability::combine_with_freq): Remove.
6242 (profile_probability::combine_wiht_count): Declare.
6243 (profile_count::force_nonzero):: Set to adjusted.
6244 (profile_count::probability_in):: Set quality to adjusted.
6245 * tree-ssa-tail-merge.c (replace_block_by): Use
6248 2018-01-23 Andrew Waterman <andrew@sifive.com>
6249 Jim Wilson <jimw@sifive.com>
6251 * config/riscv/riscv.c (riscv_stack_boundary): New.
6252 (riscv_option_override): Set riscv_stack_boundary. Handle
6253 riscv_preferred_stack_boundary_arg.
6254 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
6255 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
6256 (STACK_BOUNDARY): Set to riscv_stack_boundary.
6257 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
6258 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
6259 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
6261 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
6264 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
6265 of struct ix86_frame.
6266 (ix86_expand_epilogue): Likewise. Add a local variable for
6267 the reg_save_offset field in struct ix86_frame.
6269 2018-01-23 Bin Cheng <bin.cheng@arm.com>
6271 PR tree-optimization/82604
6272 * tree-loop-distribution.c (enum partition_kind): New enum item
6273 PKIND_PARTIAL_MEMSET.
6274 (partition_builtin_p): Support above new enum item.
6275 (generate_code_for_partition): Ditto.
6276 (compute_access_range): Differentiate cases that equality can be
6277 proven at all loops, the innermost loops or no loops.
6278 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
6279 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
6280 (finalize_partitions, distribute_loop): Don't fuse partition of
6281 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
6282 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
6285 2018-01-23 Martin Liska <mliska@suse.cz>
6287 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
6288 order to ignore the predictor.
6289 (PRED_POLYMORPHIC_CALL): Likewise.
6290 (PRED_RECURSIVE_CALL): Likewise.
6292 2018-01-23 Martin Liska <mliska@suse.cz>
6294 * tree-profile.c (tree_profiling): Print function header to
6295 aware reader which function we are working on.
6296 * value-prof.c (gimple_find_values_to_profile): Do not print
6297 not interesting value histograms.
6299 2018-01-23 Martin Liska <mliska@suse.cz>
6301 * profile-count.h (enum profile_quality): Add
6302 profile_uninitialized as the first value. Do not number values
6303 as they are zero based.
6304 (profile_count::verify): Update sanity check.
6305 (profile_probability::verify): Likewise.
6307 2018-01-23 Nathan Sidwell <nathan@acm.org>
6309 * doc/invoke.texi (ffor-scope): Deprecate.
6311 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6313 PR tree-optimization/83510
6314 * domwalk.c (set_all_edges_as_executable): New function.
6315 (dom_walker::dom_walker): Convert bool param
6316 "skip_unreachable_blocks" to enum reachability. Move setup of
6317 edge flags to set_all_edges_as_executable and only do it when
6318 reachability is REACHABLE_BLOCKS.
6319 * domwalk.h (enum dom_walker::reachability): New enum.
6320 (dom_walker::dom_walker): Convert bool param
6321 "skip_unreachable_blocks" to enum reachability.
6322 (set_all_edges_as_executable): New decl.
6323 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
6324 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
6326 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
6327 but converting true to REACHABLE_BLOCKS.
6328 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
6330 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
6331 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
6332 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
6334 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
6335 if check_all_array_refs will be called.
6337 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6339 * tree.c (selftest::test_location_wrappers): Add more test
6342 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6344 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
6345 (selftest::test_bit_in_range): Likewise.
6347 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6350 * doc/sourcebuild.texi (vect_float): Say that the selector
6351 only describes the situation when -funsafe-math-optimizations is on.
6352 (vect_float_strict): Document.
6354 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6356 PR tree-optimization/83965
6357 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
6358 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
6359 instead of checking only for a reduction.
6360 (vect_recog_widen_sum_pattern): Likewise.
6362 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6364 * predict.c (probably_never_executed): Only use precise profile info.
6365 (compute_function_frequency): Skip after inlining hack since we now
6366 have quality checking.
6368 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6370 * profile-count.h (profile_probability::very_unlikely,
6371 profile_probability::unlikely, profile_probability::even): Set
6372 precision to guessed.
6374 2018-01-23 Richard Biener <rguenther@suse.de>
6376 PR tree-optimization/83963
6377 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
6378 Properly terminate dominator walk when crossing the exit edge not
6379 when visiting its source block.
6381 2018-01-23 Jakub Jelinek <jakub@redhat.com>
6384 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
6385 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
6387 2018-01-22 Jakub Jelinek <jakub@redhat.com>
6389 PR tree-optimization/83957
6390 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
6391 semicolon after for body surrounded by braces.
6393 PR tree-optimization/83081
6394 * profile-count.h (profile_probability::split): New method.
6395 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
6396 Use profile_probability::split.
6397 (do_compare_rtx_and_jump): Fix adjustment of probabilities
6398 when splitting a single conditional jump into 2.
6400 2018-01-22 David Malcolm <dmalcolm@redhat.com>
6402 PR tree-optimization/69452
6403 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
6406 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6408 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
6409 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
6410 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
6412 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6414 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
6415 * config/rl78/rl78.md (movdi): New define_expand.
6416 * config/rl78/rl78.c (rl78_split_movdi): New function.
6418 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
6421 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
6423 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
6424 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
6425 128-bit to produce an UNSPEC move to get the double word with the
6426 signbit and then a shift directly to do signbit.
6427 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
6428 implementation with a new version that just does either a direct
6429 move or a regular move. Move memory interface to separate insns.
6430 Move insns so they are next to the expander.
6431 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
6432 with signbit move. Split big and little endian case.
6433 (signbit<mode>2_dm_mem_le): Likewise.
6434 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
6435 (signbit<mode>2_dm2): Likewise.
6437 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6439 * config/rl78/rl78.md (anddi3): New define_expand.
6441 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6443 * config/rl78/rl78.md (umindi3): New define_expand.
6445 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6447 * config/rl78/rl78.md (smindi3): New define_expand.
6449 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6451 * config/rl78/rl78.md (smaxdi3): New define_expand.
6453 2018-01-22 Carl Love <cel@us.ibm.com>
6455 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
6456 LVX_V1TI): Add macro expansion.
6457 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
6458 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
6459 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
6460 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
6461 Change check to determine if the instruction is a byte reversing
6462 entry. Fix typo in comment.
6463 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
6464 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
6465 Add def_builtin calls for new builtins.
6466 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
6467 Add define_insn expansion.
6469 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6471 * config/rl78/rl78.md (umaxdi3): New define_expand.
6473 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6475 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
6476 for non-QImode registers.
6478 2018-01-22 Richard Biener <rguenther@suse.de>
6480 PR tree-optimization/83963
6481 * graphite-scop-detection.c (scop_detection::get_sese): Delay
6482 including the loop exit block.
6483 (scop_detection::merge_sese): Likewise.
6484 (scop_detection::add_scop): Do it here instead.
6486 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6488 * doc/sourcebuild.texi (arm_softfloat): Document.
6490 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
6493 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
6494 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
6495 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
6497 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6498 David Edelsohn <dje.gcc@gmail.com>
6501 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6502 Change "crset eq" to "crset 2".
6503 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6504 (*call_indirect_aix<mode>_nospec): Likewise.
6505 (*call_value_indirect_aix<mode>_nospec): Likewise.
6506 (*call_indirect_elfv2<mode>_nospec): Likewise.
6507 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
6508 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
6509 change assembly output from . to $.
6510 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6511 (indirect_jump<mode>_nospec): Change assembly output from . to $.
6512 (*tablejump<mode>_internal1_nospec): Likewise.
6514 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
6517 * config/sh/sh_optimize_sett_clrt.cc:
6518 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
6520 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
6522 PR tree-optimization/83940
6523 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
6524 offset_dt to vect_constant_def rather than vect_unknown_def_type.
6525 (vect_check_load_store_mask): Add a mask_dt_out parameter and
6526 use it to pass back the definition type.
6527 (vect_check_store_rhs): Likewise rhs_dt_out.
6528 (vect_build_gather_load_calls): Add a mask_dt argument and use
6529 it instead of a call to vect_is_simple_use.
6530 (vectorizable_store): Update calls to vect_check_load_store_mask
6531 and vect_check_store_rhs. Use the dt returned by the latter instead
6532 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
6533 instead of calls to vect_is_simple_use. Pass the scalar rather
6534 than the vector operand to vect_is_simple_use when handling
6535 second and subsequent copies of an rhs value.
6536 (vectorizable_load): Update calls to vect_check_load_store_mask
6537 and vect_build_gather_load_calls. Use the cached mask_dt and
6538 gs_info.offset_dt instead of calls to vect_is_simple_use.
6540 2018-01-20 Jakub Jelinek <jakub@redhat.com>
6543 * tree-emutls.c: Include gimplify.h.
6544 (lower_emutls_2): New function.
6545 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
6546 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
6547 it before further processing.
6550 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
6551 UINTVAL (trueop1) instead of INTVAL (op1).
6553 2018-01-19 Jakub Jelinek <jakub@redhat.com>
6557 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
6558 INCOMING_FRAME_SP_OFFSET if not defined.
6559 (scan_trace): Add ENTRY argument. If true and
6560 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
6561 emit a note to adjust the CFA offset.
6562 (create_cfi_notes): Adjust scan_trace callers.
6563 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
6564 INCOMING_FRAME_SP_OFFSET in the CIE.
6565 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
6566 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
6568 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
6569 * doc/tm.texi: Regenerated.
6571 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6573 PR rtl-optimization/83147
6574 * lra-constraints.c (remove_inheritance_pseudos): Use
6575 lra_substitute_pseudo_within_insn.
6577 2018-01-19 Tom de Vries <tom@codesourcery.com>
6578 Cesar Philippidis <cesar@codesourcery.com>
6581 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
6583 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
6586 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
6587 spaces for function labels.
6589 2018-01-19 Martin Liska <mliska@suse.cz>
6591 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
6592 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
6593 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
6594 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
6595 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
6596 (PRED_CONST_RETURN): Change from 69 to 65.
6597 (PRED_NULL_RETURN): Change from 91 to 71.
6598 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
6599 (PRED_LOOP_GUARD): Change from 66 to 73.
6601 2018-01-19 Martin Liska <mliska@suse.cz>
6603 * predict.c (predict_insn_def): Add new assert.
6604 (struct branch_predictor): Change type to signed integer.
6605 (test_prediction_value_range): Amend test to cover
6607 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
6608 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
6609 (PRED_LOOP_ITERATIONS_MAX): Likewise.
6610 (PRED_LOOP_IV_COMPARE): Likewise.
6611 * predict.h (PROB_UNINITIALIZED): Define new constant.
6613 2018-01-19 Martin Liska <mliska@suse.cz>
6615 * predict.c (dump_prediction): Add new format for
6616 analyze_brprob.py script which is enabled with -details
6618 * profile-count.h (precise_p): New function.
6620 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6622 PR tree-optimization/83922
6623 * tree-vect-loop.c (vect_verify_full_masking): Return false if
6624 there are no statements that need masking.
6625 (vect_active_double_reduction_p): New function.
6626 (vect_analyze_loop_operations): Use it when handling phis that
6627 are not in the loop header.
6629 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6631 PR tree-optimization/83914
6632 * tree-vect-loop.c (vectorizable_induction): Don't convert
6633 init_expr or apply the peeling adjustment for inductions
6634 that are nested within the vectorized loop.
6636 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6638 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
6641 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6645 * function.h (gimplify_parameters): Add gimple_seq * argument.
6646 * function.c: Include gimple.h and options.h.
6647 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
6648 for the added local temporaries if needed.
6649 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
6650 if there are any parameter cleanups, wrap whole body into a
6651 try/finally with the cleanups.
6653 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
6656 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
6657 Use GET_MODE_CLASS for scalar floating point.
6659 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6663 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
6664 Fix call of call_cgraph_insertion_hooks.
6666 2018-01-18 Martin Sebor <msebor@redhat.com>
6668 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
6670 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6673 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
6676 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
6679 * common.opt: (-ffile-prefix-map): New option.
6680 * opts.c (common_handle_option): Defer it.
6681 * opts-global.c (handle_common_deferred_options): Handle it.
6682 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
6683 * file-prefix-map.h: New file.
6684 (remap_debug_filename, add_debug_prefix_map): ...here.
6685 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
6686 * final.c (debug_prefix_map, add_debug_prefix_map
6687 remap_debug_filename): Move to...
6688 * file-prefix-map.c: New file.
6689 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
6690 generalize, get rid of alloca(), use strrchr() instead of strchr().
6691 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
6692 Implement in terms of add_prefix_map().
6693 (remap_macro_filename, remap_debug_filename): Implement in term of
6695 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
6696 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
6697 * dbxout.c: Include file-prefix-map.h.
6698 * varasm.c: Likewise.
6699 * vmsdbgout.c: Likewise.
6700 * xcoffout.c: Likewise.
6701 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
6702 * doc/cppopts.texi (-fmacro-prefix-map): Document.
6703 * doc/invoke.texi (-ffile-prefix-map): Document.
6704 (-fdebug-prefix-map): Update description.
6706 2018-01-18 Martin Liska <mliska@suse.cz>
6708 * config/i386/i386.c (indirect_thunk_name): Document that also
6710 (output_indirect_thunk): Document why both instructions
6711 (pause and lfence) are generated.
6713 2018-01-18 Richard Biener <rguenther@suse.de>
6715 PR tree-optimization/83887
6716 * graphite-scop-detection.c
6717 (scop_detection::get_nearest_dom_with_single_entry): Remove.
6718 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
6719 (scop_detection::merge_sese): Re-implement with a flood-fill
6720 algorithm that properly finds a SESE region if it exists.
6722 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6725 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
6726 pointer_diff optimizations use view_convert instead of convert.
6728 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6730 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6731 Generate different code for -mno-speculate-indirect-jumps.
6732 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6733 (*call_indirect_aix<mode>): Disable for
6734 -mno-speculate-indirect-jumps.
6735 (*call_indirect_aix<mode>_nospec): New define_insn.
6736 (*call_value_indirect_aix<mode>): Disable for
6737 -mno-speculate-indirect-jumps.
6738 (*call_value_indirect_aix<mode>_nospec): New define_insn.
6739 (*sibcall_nonlocal_sysv<mode>): Generate different code for
6740 -mno-speculate-indirect-jumps.
6741 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6743 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6745 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
6746 long double type, set the flags for noting the default long double
6747 type, even if we don't pass or return a long double type.
6749 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
6752 * ipa-inline.c (flatten_function): Do not overwrite final inlining
6755 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
6757 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
6758 support for merge[hl].
6759 (fold_mergehl_helper): New helper function.
6760 (tree-vector-builder.h): New #include for tree_vector_builder usage.
6761 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
6762 (altivec_vmrglw_direct): Add xxmrglw insn.
6764 2018-01-17 Andrew Waterman <andrew@sifive.com>
6766 * config/riscv/riscv.c (riscv_conditional_register_usage): If
6767 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
6769 2018-01-17 David Malcolm <dmalcolm@redhat.com>
6772 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
6773 call the lto_location_cache before reading the
6774 DECL_SOURCE_LOCATION of the types.
6776 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
6777 Richard Sandiford <richard.sandiford@linaro.org>
6779 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
6780 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
6781 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
6782 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
6783 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
6785 * config/aarch64/constraints.md (aarch64_movti_operand):
6787 * config/aarch64/predicates.md (Uti): Add new constraint.
6789 2018-01-17 Carl Love <cel@us.ibm.com>
6791 * config/rs6000/vsx.md (define_expand xl_len_r,
6792 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
6793 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
6795 (define_expand, define_insn): Move the shift left from the
6796 define_insn to the define_expand for lxvl and stxvl instructions.
6797 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
6798 and XL_LEN_R definitions to PURE.
6800 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
6802 * config/i386/i386.c (indirect_thunk_name): Declare regno
6803 as unsigned int. Compare regno with INVALID_REGNUM.
6804 (output_indirect_thunk): Ditto.
6805 (output_indirect_thunk_function): Ditto.
6806 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
6807 in the call to output_indirect_thunk_function.
6809 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
6812 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
6813 rather than the size of inner_type to determine the stack slot size
6814 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
6816 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
6819 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
6822 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
6824 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
6825 endian Linux systems to optionally enable multilibs for selecting
6826 the long double type if the user configured an explicit type.
6827 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
6828 have no long double multilibs if not defined.
6829 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
6830 warn if the user used -mabi={ieee,ibm}longdouble and we built
6831 multilibs for long double.
6832 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
6833 appropriate multilib option.
6834 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
6836 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
6837 for building long double multilibs.
6838 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
6840 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
6842 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
6845 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
6847 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
6850 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
6853 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
6856 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6858 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
6859 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
6861 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
6863 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
6864 different rtl trees depending on TARGET_64BIT.
6865 (rs6000_gen_lvx): Likewise.
6867 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6869 * config/visium/visium.md (nop): Tweak comment.
6870 (hazard_nop): Likewise.
6872 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6874 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
6875 -mspeculate-indirect-jumps.
6876 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
6877 for -mno-speculate-indirect-jumps.
6878 (*call_indirect_elfv2<mode>_nospec): New define_insn.
6879 (*call_value_indirect_elfv2<mode>): Disable for
6880 -mno-speculate-indirect-jumps.
6881 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
6882 (indirect_jump): Emit different RTL for
6883 -mno-speculate-indirect-jumps.
6884 (*indirect_jump<mode>): Disable for
6885 -mno-speculate-indirect-jumps.
6886 (*indirect_jump<mode>_nospec): New define_insn.
6887 (tablejump): Emit different RTL for
6888 -mno-speculate-indirect-jumps.
6889 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
6890 (tablejumpsi_nospec): New define_expand.
6891 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
6892 (tablejumpdi_nospec): New define_expand.
6893 (*tablejump<mode>_internal1): Disable for
6894 -mno-speculate-indirect-jumps.
6895 (*tablejump<mode>_internal1_nospec): New define_insn.
6896 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
6899 2018-01-16 Artyom Skrobov tyomitch@gmail.com
6901 * caller-save.c (insert_save): Drop unnecessary parameter. All
6904 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6905 Richard Biener <rguenth@suse.de>
6908 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
6909 return early, inline manually is_gimple_sizepos. Make sure if we
6910 call gimplify_expr we don't end up with a gimple constant.
6911 * tree.c (variably_modified_type_p): Don't return true for
6912 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
6913 * gimplify.h (is_gimple_sizepos): Remove.
6915 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6917 PR tree-optimization/83857
6918 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
6919 vectorizable_live_operation for pure SLP statements.
6920 (vectorizable_live_operation): Handle PHIs.
6922 2018-01-16 Richard Biener <rguenther@suse.de>
6924 PR tree-optimization/83867
6925 * tree-vect-stmts.c (vect_transform_stmt): Precompute
6926 nested_in_vect_loop_p since the scalar stmt may get invalidated.
6928 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6931 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
6932 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
6933 If off is not INTEGER_CST, issue a may not be aligned warning
6934 rather than isn't aligned. Use isn%'t rather than isn't.
6935 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
6937 <case MULT_EXPR>: Improve the case when bottom and one of the
6938 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
6939 operand, in that case check if the other operand is multiple of
6940 bottom divided by the INTEGER_CST operand.
6942 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6945 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
6946 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
6947 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
6948 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
6949 * config/pa/pa.c (pa_function_arg_advance): Likewise.
6950 (pa_function_arg, pa_arg_partial_bytes): Likewise.
6951 (pa_function_arg_size): New function.
6953 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6955 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
6956 in a separate statement.
6958 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6960 PR tree-optimization/83847
6961 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
6962 group gathers and scatters.
6964 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6966 PR rtl-optimization/86620
6967 * params.def (max-sched-ready-insns): Bump minimum value to 1.
6969 PR rtl-optimization/83213
6970 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
6971 to last if both are JUMP_INSNs.
6973 PR tree-optimization/83843
6974 * gimple-ssa-store-merging.c
6975 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
6976 store_immediate_info for bswap/nop orig_stores.
6978 2018-01-15 Andrew Waterman <andrew@sifive.com>
6980 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
6982 <UDIV>: Increase cost if !TARGET_DIV.
6984 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
6986 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
6987 (define_attr "cr_logical_3op"): New.
6988 (cceq_ior_compare): Adjust.
6989 (cceq_ior_compare_complement): Adjust.
6990 (*cceq_rev_compare): Adjust.
6991 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
6992 (is_cracked_insn): Adjust.
6993 (insn_must_be_first_in_group): Adjust.
6994 * config/rs6000/40x.md: Adjust.
6995 * config/rs6000/440.md: Adjust.
6996 * config/rs6000/476.md: Adjust.
6997 * config/rs6000/601.md: Adjust.
6998 * config/rs6000/603.md: Adjust.
6999 * config/rs6000/6xx.md: Adjust.
7000 * config/rs6000/7450.md: Adjust.
7001 * config/rs6000/7xx.md: Adjust.
7002 * config/rs6000/8540.md: Adjust.
7003 * config/rs6000/cell.md: Adjust.
7004 * config/rs6000/e300c2c3.md: Adjust.
7005 * config/rs6000/e500mc.md: Adjust.
7006 * config/rs6000/e500mc64.md: Adjust.
7007 * config/rs6000/e5500.md: Adjust.
7008 * config/rs6000/e6500.md: Adjust.
7009 * config/rs6000/mpc.md: Adjust.
7010 * config/rs6000/power4.md: Adjust.
7011 * config/rs6000/power5.md: Adjust.
7012 * config/rs6000/power6.md: Adjust.
7013 * config/rs6000/power7.md: Adjust.
7014 * config/rs6000/power8.md: Adjust.
7015 * config/rs6000/power9.md: Adjust.
7016 * config/rs6000/rs64.md: Adjust.
7017 * config/rs6000/titan.md: Adjust.
7019 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7021 * config/i386/predicates.md (indirect_branch_operand): Rewrite
7022 ix86_indirect_branch_register logic.
7024 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7026 * config/i386/constraints.md (Bs): Update
7027 ix86_indirect_branch_register check. Don't check
7028 ix86_indirect_branch_register with GOT_memory_operand.
7030 * config/i386/predicates.md (GOT_memory_operand): Don't check
7031 ix86_indirect_branch_register here.
7032 (GOT32_symbol_operand): Likewise.
7034 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7036 * config/i386/predicates.md (constant_call_address_operand):
7037 Rewrite ix86_indirect_branch_register logic.
7038 (sibcall_insn_operand): Likewise.
7040 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7042 * config/i386/constraints.md (Bs): Replace
7043 ix86_indirect_branch_thunk_register with
7044 ix86_indirect_branch_register.
7046 * config/i386/i386.md (indirect_jump): Likewise.
7047 (tablejump): Likewise.
7048 (*sibcall_memory): Likewise.
7049 (*sibcall_value_memory): Likewise.
7050 Peepholes of indirect call and jump via memory: Likewise.
7051 * config/i386/i386.opt: Likewise.
7052 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7053 (GOT_memory_operand): Likewise.
7054 (call_insn_operand): Likewise.
7055 (sibcall_insn_operand): Likewise.
7056 (GOT32_symbol_operand): Likewise.
7058 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7061 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
7062 type rather than type addr's type points to.
7063 (expand_omp_atomic_mutex): Likewise.
7064 (expand_omp_atomic): Likewise.
7066 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7069 * config/i386/i386.c (output_indirect_thunk_function): Use
7070 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
7071 for __x86_return_thunk.
7073 2018-01-15 Richard Biener <rguenther@suse.de>
7076 * expmed.c (extract_bit_field_1): Fix typo.
7078 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7081 * config/arm/iterators.md (VF): New mode iterator.
7082 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
7083 Remove integer-related logic from pattern.
7084 (neon_vabd<mode>_3): Likewise.
7086 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7089 * common.opt (fstrict-overflow): No longer an alias.
7090 (fwrapv-pointer): New option.
7091 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
7092 also for pointer types based on flag_wrapv_pointer.
7093 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
7094 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
7095 opts->x_flag_wrapv got set.
7096 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
7097 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
7098 POINTER_TYPE_OVERFLOW_UNDEFINED.
7099 * match.pd: Likewise in address comparison pattern.
7100 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
7102 2018-01-15 Richard Biener <rguenther@suse.de>
7105 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
7106 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
7107 Reset type names to their identifier if their TYPE_DECL doesn't
7108 have linkage (and thus is used for ODR and devirt).
7109 (save_debug_info_for_decl): Remove.
7110 (save_debug_info_for_type): Likewise.
7111 (add_tree_to_fld_list): Adjust.
7112 * tree-pretty-print.c (dump_generic_node): Make dumping of
7113 type names more robust.
7115 2018-01-15 Richard Biener <rguenther@suse.de>
7117 * BASE-VER: Bump to 8.0.1.
7119 2018-01-14 Martin Sebor <msebor@redhat.com>
7122 * builtins.c (check_access): Avoid warning when the no-warning bit
7125 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
7127 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
7128 * ira-color (allocno_hard_regs_compare): Likewise.
7130 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
7133 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
7134 Use .pushsection/.popsection.
7136 2018-01-14 Martin Sebor <msebor@redhat.com>
7139 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
7141 2018-01-14 Jakub Jelinek <jakub@redhat.com>
7143 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
7144 entry from extra_headers.
7145 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
7146 extra_headers, make the list bitwise identical to the i?86-*-* one.
7148 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7150 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
7151 -mcmodel=large with -mindirect-branch=thunk,
7152 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
7153 -mfunction-return=thunk-extern.
7154 * doc/invoke.texi: Document -mcmodel=large is incompatible with
7155 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
7156 -mfunction-return=thunk and -mfunction-return=thunk-extern.
7158 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7160 * config/i386/i386.c (print_reg): Print the name of the full
7161 integer register without '%'.
7162 (ix86_print_operand): Handle 'V'.
7163 * doc/extend.texi: Document 'V' modifier.
7165 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7167 * config/i386/constraints.md (Bs): Disallow memory operand for
7168 -mindirect-branch-register.
7170 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7171 (GOT_memory_operand): Likewise.
7172 (call_insn_operand): Likewise.
7173 (sibcall_insn_operand): Likewise.
7174 (GOT32_symbol_operand): Likewise.
7175 * config/i386/i386.md (indirect_jump): Call convert_memory_address
7176 for -mindirect-branch-register.
7177 (tablejump): Likewise.
7178 (*sibcall_memory): Likewise.
7179 (*sibcall_value_memory): Likewise.
7180 Disallow peepholes of indirect call and jump via memory for
7181 -mindirect-branch-register.
7182 (*call_pop): Replace m with Bw.
7183 (*call_value_pop): Likewise.
7184 (*sibcall_pop_memory): Replace m with Bs.
7185 * config/i386/i386.opt (mindirect-branch-register): New option.
7186 * doc/invoke.texi: Document -mindirect-branch-register option.
7188 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7190 * config/i386/i386-protos.h (ix86_output_function_return): New.
7191 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
7192 set function_return_type.
7193 (indirect_thunk_name): Add ret_p to indicate thunk for function
7195 (output_indirect_thunk_function): Pass false to
7196 indirect_thunk_name.
7197 (ix86_output_indirect_branch_via_reg): Likewise.
7198 (ix86_output_indirect_branch_via_push): Likewise.
7199 (output_indirect_thunk_function): Create alias for function
7200 return thunk if regno < 0.
7201 (ix86_output_function_return): New function.
7202 (ix86_handle_fndecl_attribute): Handle function_return.
7203 (ix86_attribute_table): Add function_return.
7204 * config/i386/i386.h (machine_function): Add
7205 function_return_type.
7206 * config/i386/i386.md (simple_return_internal): Use
7207 ix86_output_function_return.
7208 (simple_return_internal_long): Likewise.
7209 * config/i386/i386.opt (mfunction-return=): New option.
7210 (indirect_branch): Mention -mfunction-return=.
7211 * doc/extend.texi: Document function_return function attribute.
7212 * doc/invoke.texi: Document -mfunction-return= option.
7214 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7216 * config/i386/i386-opts.h (indirect_branch): New.
7217 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
7218 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
7219 with local indirect jump when converting indirect call and jump.
7220 (ix86_set_indirect_branch_type): New.
7221 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
7222 (indirectlabelno): New.
7223 (indirect_thunk_needed): Likewise.
7224 (indirect_thunk_bnd_needed): Likewise.
7225 (indirect_thunks_used): Likewise.
7226 (indirect_thunks_bnd_used): Likewise.
7227 (INDIRECT_LABEL): Likewise.
7228 (indirect_thunk_name): Likewise.
7229 (output_indirect_thunk): Likewise.
7230 (output_indirect_thunk_function): Likewise.
7231 (ix86_output_indirect_branch_via_reg): Likewise.
7232 (ix86_output_indirect_branch_via_push): Likewise.
7233 (ix86_output_indirect_branch): Likewise.
7234 (ix86_output_indirect_jmp): Likewise.
7235 (ix86_code_end): Call output_indirect_thunk_function if needed.
7236 (ix86_output_call_insn): Call ix86_output_indirect_branch if
7238 (ix86_handle_fndecl_attribute): Handle indirect_branch.
7239 (ix86_attribute_table): Add indirect_branch.
7240 * config/i386/i386.h (machine_function): Add indirect_branch_type
7241 and has_local_indirect_jump.
7242 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
7244 (tablejump): Likewise.
7245 (*indirect_jump): Use ix86_output_indirect_jmp.
7246 (*tablejump_1): Likewise.
7247 (simple_return_indirect_internal): Likewise.
7248 * config/i386/i386.opt (mindirect-branch=): New option.
7249 (indirect_branch): New.
7252 (thunk-inline): Likewise.
7253 (thunk-extern): Likewise.
7254 * doc/extend.texi: Document indirect_branch function attribute.
7255 * doc/invoke.texi: Document -mindirect-branch= option.
7257 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
7260 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
7262 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
7264 * ipa-inline.c (want_inline_small_function_p): Return false if
7265 inlining has already failed with CIF_FINAL_ERROR.
7266 (update_caller_keys): Call want_inline_small_function_p before
7268 (update_callee_keys): Likewise.
7270 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7272 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
7274 (rs6000_quadword_masked_address_p): Likewise.
7275 (quad_aligned_load_p): Likewise.
7276 (quad_aligned_store_p): Likewise.
7277 (const_load_sequence_p): Add comment to describe the outer-most loop.
7278 (mimic_memory_attributes_and_flags): New function.
7279 (rs6000_gen_stvx): Likewise.
7280 (replace_swapped_aligned_store): Likewise.
7281 (rs6000_gen_lvx): Likewise.
7282 (replace_swapped_aligned_load): Likewise.
7283 (replace_swapped_load_constant): Capitalize argument name in
7284 comment describing this function.
7285 (rs6000_analyze_swaps): Add a third pass to search for vector loads
7286 and stores that access quad-word aligned addresses and replace
7287 with stvx or lvx instructions when appropriate.
7288 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
7289 New function prototype.
7290 (rs6000_quadword_masked_address_p): Likewise.
7291 (rs6000_gen_lvx): Likewise.
7292 (rs6000_gen_stvx): Likewise.
7293 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
7294 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
7295 when memory address is aligned.
7296 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
7297 this split to select lvx instruction when memory address is aligned.
7298 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
7299 instruction when memory address is aligned.
7300 (*vsx_le_perm_load_v16qi): Likewise.
7301 (four unnamed splitters): Modify to select the stvx instruction
7302 when memory is aligned.
7304 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
7306 * predict.c (determine_unlikely_bbs): Handle correctly BBs
7307 which appears in the queue multiple times.
7309 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7310 Alan Hayward <alan.hayward@arm.com>
7311 David Sherwood <david.sherwood@arm.com>
7313 * tree-vectorizer.h (vec_lower_bound): New structure.
7314 (_loop_vec_info): Add check_nonzero and lower_bounds.
7315 (LOOP_VINFO_CHECK_NONZERO): New macro.
7316 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
7317 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
7318 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
7319 fields. Make seg_len the distance travelled, not including the
7321 (dr_direction_indicator): Declare.
7322 (dr_zero_step_indicator): Likewise.
7323 (dr_known_forward_stride_p): Likewise.
7324 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
7326 (runtime_alias_check_p): Allow runtime alias checks with
7328 (operator ==): Compare access_size and align.
7329 (prune_runtime_alias_test_list): Rework for new distinction between
7330 the access_size and seg_len.
7331 (create_intersect_range_checks_index): Likewise. Cope with polynomial
7333 (get_segment_min_max): New function.
7334 (create_intersect_range_checks): Use it.
7335 (dr_step_indicator): New function.
7336 (dr_direction_indicator): Likewise.
7337 (dr_zero_step_indicator): Likewise.
7338 (dr_known_forward_stride_p): Likewise.
7339 * tree-loop-distribution.c (data_ref_segment_size): Return
7340 DR_STEP * (niters - 1).
7341 (compute_alias_check_pairs): Update call to the dr_with_seg_len
7343 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
7344 (vect_preserves_scalar_order_p): New function, split out from...
7345 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
7346 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
7347 (vect_vfa_access_size): New function.
7348 (vect_vfa_align): Likewise.
7349 (vect_compile_time_alias): Take access_size_a and access_b arguments.
7350 (dump_lower_bound): New function.
7351 (vect_check_lower_bound): Likewise.
7352 (vect_small_gap_p): Likewise.
7353 (vectorizable_with_step_bound_p): Likewise.
7354 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
7355 depencies if the vectorization factor is 1. Convert the checks
7356 for nonzero steps into checks on the bounds of DR_STEP. Try using
7357 a bunds check for variable steps if the minimum required step is
7358 relatively small. Update calls to the dr_with_seg_len
7359 constructor and to vect_compile_time_alias.
7360 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
7362 (vect_loop_versioning): Call it.
7363 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
7365 (vect_estimate_min_profitable_iters): Account for any bounds checks.
7367 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7368 Alan Hayward <alan.hayward@arm.com>
7369 David Sherwood <david.sherwood@arm.com>
7371 * doc/sourcebuild.texi (vect_scatter_store): Document.
7372 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
7374 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
7376 * genopinit.c (main): Add supports_vec_scatter_store and
7377 supports_vec_scatter_store_cached to target_optabs.
7378 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
7379 IFN_MASK_SCATTER_STORE.
7380 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
7382 * internal-fn.h (internal_store_fn_p): Declare.
7383 (internal_fn_stored_value_index): Likewise.
7384 * internal-fn.c (scatter_store_direct): New macro.
7385 (expand_scatter_store_optab_fn): New function.
7386 (direct_scatter_store_optab_supported_p): New macro.
7387 (internal_store_fn_p): New function.
7388 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
7389 IFN_MASK_SCATTER_STORE.
7390 (internal_fn_mask_index): Likewise.
7391 (internal_fn_stored_value_index): New function.
7392 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
7394 * optabs-query.h (supports_vec_scatter_store_p): Declare.
7395 * optabs-query.c (supports_vec_scatter_store_p): New function.
7396 * tree-vectorizer.h (vect_get_store_rhs): Declare.
7397 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
7398 true for scatter stores.
7399 (vect_gather_scatter_fn_p): Handle scatter stores too.
7400 (vect_check_gather_scatter): Consider using scatter stores if
7401 supports_vec_scatter_store_p.
7402 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
7404 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7405 internal_fn_stored_value_index.
7406 (check_load_store_masking): Handle scatter stores too.
7407 (vect_get_store_rhs): Make public.
7408 (vectorizable_call): Use internal_store_fn_p.
7409 (vectorizable_store): Handle scatter store internal functions.
7410 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
7411 when deciding whether the end of the group has been reached.
7412 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
7413 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
7414 (mask_scatter_store<mode>): New insns.
7416 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7417 Alan Hayward <alan.hayward@arm.com>
7418 David Sherwood <david.sherwood@arm.com>
7420 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
7421 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
7422 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
7424 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
7425 Use vect_truncate_gather_scatter_offset if we can't treat the
7426 operation as a normal gather load or scatter store.
7427 (get_group_load_store_type): Take the gather_scatter_info
7428 as argument. Try using a gather load or scatter store for
7429 single-element groups.
7430 (get_load_store_type): Update calls to get_group_load_store_type
7431 and vect_use_strided_gather_scatters_p.
7433 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7434 Alan Hayward <alan.hayward@arm.com>
7435 David Sherwood <david.sherwood@arm.com>
7437 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
7438 optional tree argument.
7439 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
7441 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
7442 but continue to use the current value as a fallback.
7443 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
7444 to compare the updates.
7445 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
7446 (get_load_store_type): Use it when handling a strided access.
7447 (vect_get_strided_load_store_ops): New function.
7448 (vect_get_data_ptr_increment): Likewise.
7449 (vectorizable_load): Handle strided gather loads. Always pass
7450 a step to vect_create_data_ref_ptr and bump_vector_ptr.
7452 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7453 Alan Hayward <alan.hayward@arm.com>
7454 David Sherwood <david.sherwood@arm.com>
7456 * doc/md.texi (gather_load@var{m}): Document.
7457 (mask_gather_load@var{m}): Likewise.
7458 * genopinit.c (main): Add supports_vec_gather_load and
7459 supports_vec_gather_load_cached to target_optabs.
7460 * optabs-tree.c (init_tree_optimization_optabs): Use
7461 ggc_cleared_alloc to allocate target_optabs.
7462 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
7463 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
7465 * internal-fn.h (internal_load_fn_p): Declare.
7466 (internal_gather_scatter_fn_p): Likewise.
7467 (internal_fn_mask_index): Likewise.
7468 (internal_gather_scatter_fn_supported_p): Likewise.
7469 * internal-fn.c (gather_load_direct): New macro.
7470 (expand_gather_load_optab_fn): New function.
7471 (direct_gather_load_optab_supported_p): New macro.
7472 (direct_internal_fn_optab): New function.
7473 (internal_load_fn_p): Likewise.
7474 (internal_gather_scatter_fn_p): Likewise.
7475 (internal_fn_mask_index): Likewise.
7476 (internal_gather_scatter_fn_supported_p): Likewise.
7477 * optabs-query.c (supports_at_least_one_mode_p): New function.
7478 (supports_vec_gather_load_p): Likewise.
7479 * optabs-query.h (supports_vec_gather_load_p): Declare.
7480 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
7481 and memory_type field.
7482 (NUM_PATTERNS): Bump to 15.
7483 * tree-vect-data-refs.c: Include internal-fn.h.
7484 (vect_gather_scatter_fn_p): New function.
7485 (vect_describe_gather_scatter_call): Likewise.
7486 (vect_check_gather_scatter): Try using internal functions for
7487 gather loads. Recognize existing calls to a gather load function.
7488 (vect_analyze_data_refs): Consider using gather loads if
7489 supports_vec_gather_load_p.
7490 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
7491 (vect_get_gather_scatter_offset_type): Likewise.
7492 (vect_convert_mask_for_vectype): Likewise.
7493 (vect_add_conversion_to_patterm): Likewise.
7494 (vect_try_gather_scatter_pattern): Likewise.
7495 (vect_recog_gather_scatter_pattern): New pattern recognizer.
7496 (vect_vect_recog_func_ptrs): Add it.
7497 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7498 internal_fn_mask_index and internal_gather_scatter_fn_p.
7499 (check_load_store_masking): Take the gather_scatter_info as an
7500 argument and handle gather loads.
7501 (vect_get_gather_scatter_ops): New function.
7502 (vectorizable_call): Check internal_load_fn_p.
7503 (vectorizable_load): Likewise. Handle gather load internal
7505 (vectorizable_store): Update call to check_load_store_masking.
7506 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
7507 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
7508 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
7509 (aarch64_gather_scale_operand_d): New predicates.
7510 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
7511 (mask_gather_load<mode>): New insns.
7513 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7514 Alan Hayward <alan.hayward@arm.com>
7515 David Sherwood <david.sherwood@arm.com>
7517 * optabs.def (fold_left_plus_optab): New optab.
7518 * doc/md.texi (fold_left_plus_@var{m}): Document.
7519 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
7520 * internal-fn.c (fold_left_direct): Define.
7521 (expand_fold_left_optab_fn): Likewise.
7522 (direct_fold_left_optab_supported_p): Likewise.
7523 * fold-const-call.c (fold_const_fold_left): New function.
7524 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
7525 * tree-parloops.c (valid_reduction_p): New function.
7526 (gather_scalar_reductions): Use it.
7527 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
7528 (vect_finish_replace_stmt): Declare.
7529 * tree-vect-loop.c (fold_left_reduction_fn): New function.
7530 (needs_fold_left_reduction_p): New function, split out from...
7531 (vect_is_simple_reduction): ...here. Accept reductions that
7532 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
7533 (vect_force_simple_reduction): Also store the reduction type in
7534 the assignment's STMT_VINFO_REDUC_TYPE.
7535 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
7536 (merge_with_identity): New function.
7537 (vect_expand_fold_left): Likewise.
7538 (vectorize_fold_left_reduction): Likewise.
7539 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
7540 scalar phi in place for it. Check for target support and reject
7541 cases that would reassociate the operation. Defer the transform
7542 phase to vectorize_fold_left_reduction.
7543 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
7544 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
7545 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
7547 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7549 * tree-if-conv.c (predicate_mem_writes): Remove redundant
7550 call to ifc_temp_var.
7552 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7553 Alan Hayward <alan.hayward@arm.com>
7554 David Sherwood <david.sherwood@arm.com>
7556 * target.def (legitimize_address_displacement): Take the original
7557 offset as a poly_int.
7558 * targhooks.h (default_legitimize_address_displacement): Update
7560 * targhooks.c (default_legitimize_address_displacement): Likewise.
7561 * doc/tm.texi: Regenerate.
7562 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
7563 as an argument, moving assert of ad->disp == ad->disp_term to...
7564 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
7565 Try calling targetm.legitimize_address_displacement before expanding
7566 the address rather than afterwards, and adjust for the new interface.
7567 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
7568 Match the new hook interface. Handle SVE addresses.
7569 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
7572 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7574 * Makefile.in (OBJS): Add early-remat.o.
7575 * target.def (select_early_remat_modes): New hook.
7576 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
7577 * doc/tm.texi: Regenerate.
7578 * targhooks.h (default_select_early_remat_modes): Declare.
7579 * targhooks.c (default_select_early_remat_modes): New function.
7580 * timevar.def (TV_EARLY_REMAT): New timevar.
7581 * passes.def (pass_early_remat): New pass.
7582 * tree-pass.h (make_pass_early_remat): Declare.
7583 * early-remat.c: New file.
7584 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
7586 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
7588 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7589 Alan Hayward <alan.hayward@arm.com>
7590 David Sherwood <david.sherwood@arm.com>
7592 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
7593 vfm1 with a bound_epilog parameter.
7594 (vect_do_peeling): Update calls accordingly, and move the prologue
7595 call earlier in the function. Treat the base bound_epilog as 0 for
7596 fully-masked loops and retain vf - 1 for other loops. Add 1 to
7597 this base when peeling for gaps.
7598 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
7599 with fully-masked loops.
7600 (vect_estimate_min_profitable_iters): Handle the single peeled
7601 iteration in that case.
7603 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7604 Alan Hayward <alan.hayward@arm.com>
7605 David Sherwood <david.sherwood@arm.com>
7607 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
7608 single-element interleaving even if the size is not a power of 2.
7609 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
7610 accesses for single-element interleaving if the group size is
7613 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7614 Alan Hayward <alan.hayward@arm.com>
7615 David Sherwood <david.sherwood@arm.com>
7617 * doc/md.texi (fold_extract_last_@var{m}): Document.
7618 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
7619 * optabs.def (fold_extract_last_optab): New optab.
7620 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
7621 * internal-fn.c (fold_extract_direct): New macro.
7622 (expand_fold_extract_optab_fn): Likewise.
7623 (direct_fold_extract_optab_supported_p): Likewise.
7624 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
7625 * tree-vect-loop.c (vect_model_reduction_cost): Handle
7626 EXTRACT_LAST_REDUCTION.
7627 (get_initial_def_for_reduction): Do not create an initial vector
7628 for EXTRACT_LAST_REDUCTION reductions.
7629 (vectorizable_reduction): Leave the scalar phi in place for
7630 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
7631 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
7632 epilogue code for EXTRACT_LAST_REDUCTION and defer the
7633 transform phase to vectorizable_condition.
7634 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
7636 (vect_finish_stmt_generation): ...here.
7637 (vect_finish_replace_stmt): New function.
7638 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
7639 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
7641 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
7643 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7644 Alan Hayward <alan.hayward@arm.com>
7645 David Sherwood <david.sherwood@arm.com>
7647 * doc/md.texi (extract_last_@var{m}): Document.
7648 * optabs.def (extract_last_optab): New optab.
7649 * internal-fn.def (EXTRACT_LAST): New internal function.
7650 * internal-fn.c (cond_unary_direct): New macro.
7651 (expand_cond_unary_optab_fn): Likewise.
7652 (direct_cond_unary_optab_supported_p): Likewise.
7653 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
7654 loops using EXTRACT_LAST.
7655 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
7656 (extract_last_<mode>): ...this optab.
7657 (vec_extract<mode><Vel>): Update accordingly.
7659 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7660 Alan Hayward <alan.hayward@arm.com>
7661 David Sherwood <david.sherwood@arm.com>
7663 * target.def (empty_mask_is_expensive): New hook.
7664 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
7665 * doc/tm.texi: Regenerate.
7666 * targhooks.h (default_empty_mask_is_expensive): Declare.
7667 * targhooks.c (default_empty_mask_is_expensive): New function.
7668 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
7669 if the target says that empty masks are expensive.
7670 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
7672 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
7674 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7675 Alan Hayward <alan.hayward@arm.com>
7676 David Sherwood <david.sherwood@arm.com>
7678 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
7679 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
7680 (vect_use_loop_mask_for_alignment_p): New function.
7681 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
7682 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
7683 niters_skip argument. Make sure that the first niters_skip elements
7684 of the first iteration are inactive.
7685 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
7686 Update call to vect_set_loop_masks_directly.
7687 (get_misalign_in_elems): New function, split out from...
7688 (vect_gen_prolog_loop_niters): ...here.
7689 (vect_update_init_of_dr): Take a code argument that specifies whether
7690 the adjustment should be added or subtracted.
7691 (vect_update_init_of_drs): Likewise.
7692 (vect_prepare_for_masked_peels): New function.
7693 (vect_do_peeling): Skip prologue peeling if we're using a mask
7694 instead. Update call to vect_update_inits_of_drs.
7695 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7697 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
7698 alignment. Do not include the number of peeled iterations in
7699 the minimum threshold in that case.
7700 (vectorizable_induction): Adjust the start value down by
7701 LOOP_VINFO_MASK_SKIP_NITERS iterations.
7702 (vect_transform_loop): Call vect_prepare_for_masked_peels.
7703 Take the number of skipped iterations into account when calculating
7705 * tree-vect-stmts.c (vect_gen_while_not): New function.
7707 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7708 Alan Hayward <alan.hayward@arm.com>
7709 David Sherwood <david.sherwood@arm.com>
7711 * doc/sourcebuild.texi (vect_fully_masked): Document.
7712 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
7714 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
7716 (vect_analyze_loop_2): ...here. Don't check the vectorization
7717 factor against the number of loop iterations if the loop is
7720 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7721 Alan Hayward <alan.hayward@arm.com>
7722 David Sherwood <david.sherwood@arm.com>
7724 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
7725 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
7726 (dump_groups): Update accordingly.
7727 (iv_use::mem_type): New member variable.
7728 (address_p): New function.
7729 (record_use): Add a mem_type argument and initialize the new
7731 (record_group_use): Add a mem_type argument. Use address_p.
7732 Remove obsolete null checks of base_object. Update call to record_use.
7733 (find_interesting_uses_op): Update call to record_group_use.
7734 (find_interesting_uses_cond): Likewise.
7735 (find_interesting_uses_address): Likewise.
7736 (get_mem_type_for_internal_fn): New function.
7737 (find_address_like_use): Likewise.
7738 (find_interesting_uses_stmt): Try find_address_like_use before
7739 calling find_interesting_uses_op.
7740 (addr_offset_valid_p): Use the iv mem_type field as the type
7741 of the addressed memory.
7742 (add_autoinc_candidates): Likewise.
7743 (get_address_cost): Likewise.
7744 (split_small_address_groups_p): Use address_p.
7745 (split_address_groups): Likewise.
7746 (add_iv_candidate_for_use): Likewise.
7747 (autoinc_possible_for_pair): Likewise.
7748 (rewrite_groups): Likewise.
7749 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
7750 (determine_group_iv_cost): Update after split of USE_ADDRESS.
7751 (get_alias_ptr_type_for_ptr_address): New function.
7752 (rewrite_use_address): Rewrite address uses in calls that were
7753 identified by find_address_like_use.
7755 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7756 Alan Hayward <alan.hayward@arm.com>
7757 David Sherwood <david.sherwood@arm.com>
7759 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
7761 * gimple-expr.h (is_gimple_addressable: Likewise.
7762 * gimple-expr.c (is_gimple_address): Likewise.
7763 * internal-fn.c (expand_call_mem_ref): New function.
7764 (expand_mask_load_optab_fn): Use it.
7765 (expand_mask_store_optab_fn): Likewise.
7767 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7768 Alan Hayward <alan.hayward@arm.com>
7769 David Sherwood <david.sherwood@arm.com>
7771 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
7772 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
7773 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
7774 (cond_umax@var{mode}): Document.
7775 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
7776 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
7777 (cond_umin_optab, cond_umax_optab): New optabs.
7778 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
7779 (COND_IOR, COND_XOR): New internal functions.
7780 * internal-fn.h (get_conditional_internal_fn): Declare.
7781 * internal-fn.c (cond_binary_direct): New macro.
7782 (expand_cond_binary_optab_fn): Likewise.
7783 (direct_cond_binary_optab_supported_p): Likewise.
7784 (get_conditional_internal_fn): New function.
7785 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
7786 Cope with reduction statements that are vectorized as calls rather
7788 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
7789 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
7790 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
7791 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
7792 (UNSPEC_COND_EOR): New unspecs.
7793 (optab): Add mappings for them.
7794 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
7795 (sve_int_op, sve_fp_op): New int attributes.
7797 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7798 Alan Hayward <alan.hayward@arm.com>
7799 David Sherwood <david.sherwood@arm.com>
7801 * optabs.def (while_ult_optab): New optab.
7802 * doc/md.texi (while_ult@var{m}@var{n}): Document.
7803 * internal-fn.def (WHILE_ULT): New internal function.
7804 * internal-fn.h (direct_internal_fn_supported_p): New override
7805 that takes two types as argument.
7806 * internal-fn.c (while_direct): New macro.
7807 (expand_while_optab_fn): New function.
7808 (convert_optab_supported_p): Likewise.
7809 (direct_while_optab_supported_p): New macro.
7810 * wide-int.h (wi::udiv_ceil): New function.
7811 * tree-vectorizer.h (rgroup_masks): New structure.
7812 (vec_loop_masks): New typedef.
7813 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
7815 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
7816 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
7817 (vect_max_vf): New function.
7818 (slpeel_make_loop_iterate_ntimes): Delete.
7819 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
7820 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
7821 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
7822 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
7823 internal-fn.h, stor-layout.h and optabs-query.h.
7824 (vect_set_loop_mask): New function.
7825 (add_preheader_seq): Likewise.
7826 (add_header_seq): Likewise.
7827 (interleave_supported_p): Likewise.
7828 (vect_maybe_permute_loop_masks): Likewise.
7829 (vect_set_loop_masks_directly): Likewise.
7830 (vect_set_loop_condition_masked): Likewise.
7831 (vect_set_loop_condition_unmasked): New function, split out from
7832 slpeel_make_loop_iterate_ntimes.
7833 (slpeel_make_loop_iterate_ntimes): Rename to..
7834 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
7835 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
7836 (vect_do_peeling): Update call accordingly.
7837 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
7839 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7840 mask_compare_type, can_fully_mask_p and fully_masked_p.
7841 (release_vec_loop_masks): New function.
7842 (_loop_vec_info): Use it to free the loop masks.
7843 (can_produce_all_loop_masks_p): New function.
7844 (vect_get_max_nscalars_per_iter): Likewise.
7845 (vect_verify_full_masking): Likewise.
7846 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
7847 retries, and free the mask rgroups before retrying. Check loop-wide
7848 reasons for disallowing fully-masked loops. Make the final decision
7849 about whether use a fully-masked loop or not.
7850 (vect_estimate_min_profitable_iters): Do not assume that peeling
7851 for the number of iterations will be needed for fully-masked loops.
7852 (vectorizable_reduction): Disable fully-masked loops.
7853 (vectorizable_live_operation): Likewise.
7854 (vect_halve_mask_nunits): New function.
7855 (vect_double_mask_nunits): Likewise.
7856 (vect_record_loop_mask): Likewise.
7857 (vect_get_loop_mask): Likewise.
7858 (vect_transform_loop): Handle the case in which the final loop
7859 iteration might handle a partial vector. Call vect_set_loop_condition
7860 instead of slpeel_make_loop_iterate_ntimes.
7861 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
7862 (check_load_store_masking): New function.
7863 (prepare_load_store_mask): Likewise.
7864 (vectorizable_store): Handle fully-masked loops.
7865 (vectorizable_load): Likewise.
7866 (supportable_widening_operation): Use vect_halve_mask_nunits for
7868 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
7869 (vect_gen_while): New function.
7870 * config/aarch64/aarch64.md (umax<mode>3): New expander.
7871 (aarch64_uqdec<mode>): New insn.
7873 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7874 Alan Hayward <alan.hayward@arm.com>
7875 David Sherwood <david.sherwood@arm.com>
7877 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
7878 (reduc_xor_scal_optab): New optabs.
7879 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
7880 (reduc_xor_scal_@var{m}): Document.
7881 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
7882 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
7884 * fold-const-call.c (fold_const_call): Handle them.
7885 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
7886 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
7887 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
7888 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
7889 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
7890 (UNSPEC_XORV): New unspecs.
7891 (optab): Add entries for them.
7892 (BITWISEV): New int iterator.
7893 (bit_reduc_op): New int attributes.
7895 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7896 Alan Hayward <alan.hayward@arm.com>
7897 David Sherwood <david.sherwood@arm.com>
7899 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
7900 * internal-fn.def (VEC_SHL_INSERT): New internal function.
7901 * optabs.def (vec_shl_insert_optab): New optab.
7902 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
7903 (duplicate_and_interleave): Likewise.
7904 * tree-vect-loop.c: Include internal-fn.h.
7905 (neutral_op_for_slp_reduction): New function, split out from
7906 get_initial_defs_for_reduction.
7907 (get_initial_def_for_reduction): Handle option 2 for variable-length
7908 vectors by loading the neutral value into a vector and then shifting
7909 the initial value into element 0.
7910 (get_initial_defs_for_reduction): Replace the code argument with
7911 the neutral value calculated by neutral_op_for_slp_reduction.
7912 Use gimple_build_vector for constant-length vectors.
7913 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
7914 but the first group_size elements have a neutral value.
7915 Use duplicate_and_interleave otherwise.
7916 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
7917 Update call to get_initial_defs_for_reduction. Handle SLP
7918 reductions for variable-length vectors by creating one vector
7919 result for each scalar result, with the elements associated
7920 with other scalar results stubbed out with the neutral value.
7921 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
7922 Require IFN_VEC_SHL_INSERT for double reductions on
7923 variable-length vectors, or SLP reductions that have
7924 a neutral value. Require can_duplicate_and_interleave_p
7925 support for variable-length unchained SLP reductions if there
7926 is no neutral value, such as for MIN/MAX reductions. Also require
7927 the number of vector elements to be a multiple of the number of
7928 SLP statements when doing variable-length unchained SLP reductions.
7929 Update call to vect_create_epilog_for_reduction.
7930 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
7931 and remove initial values.
7932 (duplicate_and_interleave): Make public.
7933 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
7934 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
7936 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7937 Alan Hayward <alan.hayward@arm.com>
7938 David Sherwood <david.sherwood@arm.com>
7940 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
7941 (can_duplicate_and_interleave_p): New function.
7942 (vect_get_and_check_slp_defs): Take the vector of statements
7943 rather than just the current one. Remove excess parentheses.
7944 Restriction rejectinon of vect_constant_def and vect_external_def
7945 for variable-length vectors to boolean types, or types for which
7946 can_duplicate_and_interleave_p is false.
7947 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
7948 (duplicate_and_interleave): New function.
7949 (vect_get_constant_vectors): Use gimple_build_vector for
7950 constant-length vectors and suitable variable-length constant
7951 vectors. Use duplicate_and_interleave for other variable-length
7952 vectors. Don't defer the update when inserting new statements.
7954 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7955 Alan Hayward <alan.hayward@arm.com>
7956 David Sherwood <david.sherwood@arm.com>
7958 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
7959 min_profitable_iters doesn't go negative.
7961 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7962 Alan Hayward <alan.hayward@arm.com>
7963 David Sherwood <david.sherwood@arm.com>
7965 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
7966 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
7967 * optabs.def (vec_mask_load_lanes_optab): New optab.
7968 (vec_mask_store_lanes_optab): Likewise.
7969 * internal-fn.def (MASK_LOAD_LANES): New internal function.
7970 (MASK_STORE_LANES): Likewise.
7971 * internal-fn.c (mask_load_lanes_direct): New macro.
7972 (mask_store_lanes_direct): Likewise.
7973 (expand_mask_load_optab_fn): Handle masked operations.
7974 (expand_mask_load_lanes_optab_fn): New macro.
7975 (expand_mask_store_optab_fn): Handle masked operations.
7976 (expand_mask_store_lanes_optab_fn): New macro.
7977 (direct_mask_load_lanes_optab_supported_p): Likewise.
7978 (direct_mask_store_lanes_optab_supported_p): Likewise.
7979 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
7981 (vect_load_lanes_supported): Likewise.
7982 * tree-vect-data-refs.c (strip_conversion): New function.
7983 (can_group_stmts_p): Likewise.
7984 (vect_analyze_data_ref_accesses): Use it instead of checking
7985 for a pair of assignments.
7986 (vect_store_lanes_supported): Take a masked_p parameter.
7987 (vect_load_lanes_supported): Likewise.
7988 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
7989 vect_store_lanes_supported and vect_load_lanes_supported.
7990 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7991 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
7992 parameter. Don't allow gaps for masked accesses.
7993 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
7994 and vect_load_lanes_supported.
7995 (get_load_store_type): Take a masked_p parameter and update
7996 call to get_group_load_store_type.
7997 (vectorizable_store): Update call to get_load_store_type.
7998 Handle IFN_MASK_STORE_LANES.
7999 (vectorizable_load): Update call to get_load_store_type.
8000 Handle IFN_MASK_LOAD_LANES.
8002 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8003 Alan Hayward <alan.hayward@arm.com>
8004 David Sherwood <david.sherwood@arm.com>
8006 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
8008 * config/aarch64/aarch64-protos.h
8009 (aarch64_sve_struct_memory_operand_p): Declare.
8010 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
8011 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
8012 (VPRED, vpred): Handle SVE structure modes.
8013 * config/aarch64/constraints.md (Utx): New constraint.
8014 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
8015 (aarch64_sve_struct_nonimmediate_operand): New predicates.
8016 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
8017 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
8018 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
8019 structure modes. Split into pieces after RA.
8020 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
8021 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
8023 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
8024 SVE structure modes.
8025 (aarch64_classify_address): Likewise.
8026 (sizetochar): Move earlier in file.
8027 (aarch64_print_operand): Handle SVE register lists.
8028 (aarch64_array_mode): New function.
8029 (aarch64_sve_struct_memory_operand_p): Likewise.
8030 (TARGET_ARRAY_MODE): Redefine.
8032 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8033 Alan Hayward <alan.hayward@arm.com>
8034 David Sherwood <david.sherwood@arm.com>
8036 * target.def (array_mode): New target hook.
8037 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
8038 * doc/tm.texi: Regenerate.
8039 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
8040 * hooks.c (hook_optmode_mode_uhwi_none): New function.
8041 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
8043 * stor-layout.c (mode_for_array): Likewise. Support polynomial
8046 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8047 Alan Hayward <alan.hayward@arm.com>
8048 David Sherwood <david.sherwood@arm.com>
8050 * fold-const.c (fold_binary_loc): Check the argument types
8051 rather than the result type when testing for a vector operation.
8053 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8055 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
8056 * doc/tm.texi: Regenerate.
8058 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8059 Alan Hayward <alan.hayward@arm.com>
8060 David Sherwood <david.sherwood@arm.com>
8062 * doc/invoke.texi (-msve-vector-bits=): Document new option.
8063 (sve): Document new AArch64 extension.
8064 * doc/md.texi (w): Extend the description of the AArch64
8065 constraint to include SVE vectors.
8066 (Upl, Upa): Document new AArch64 predicate constraints.
8067 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
8069 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
8070 (msve-vector-bits=): New option.
8071 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
8072 SVE when these are disabled.
8073 (sve): New extension.
8074 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
8075 modes. Adjust their number of units based on aarch64_sve_vg.
8076 (MAX_BITSIZE_MODE_ANY_MODE): Define.
8077 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
8078 aarch64_addr_query_type.
8079 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
8080 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
8081 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
8082 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
8083 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
8084 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
8085 (aarch64_simd_imm_zero_p): Delete.
8086 (aarch64_check_zero_based_sve_index_immediate): Declare.
8087 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8088 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8089 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8090 (aarch64_sve_float_mul_immediate_p): Likewise.
8091 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8093 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
8094 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
8095 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
8096 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
8097 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
8098 (aarch64_regmode_natural_size): Likewise.
8099 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
8100 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
8102 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
8103 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
8104 for VG and the SVE predicate registers.
8105 (V_ALIASES): Add a "z"-prefixed alias.
8106 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
8107 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
8108 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
8109 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
8110 (REG_CLASS_NAMES): Add entries for them.
8111 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
8112 and the predicate registers.
8113 (aarch64_sve_vg): Declare.
8114 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
8115 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
8116 (REGMODE_NATURAL_SIZE): Define.
8117 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
8119 * config/aarch64/aarch64.c: Include cfgrtl.h.
8120 (simd_immediate_info): Add a constructor for series vectors,
8121 and an associated step field.
8122 (aarch64_sve_vg): New variable.
8123 (aarch64_dbx_register_number): Handle VG and the predicate registers.
8124 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
8125 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
8126 (VEC_ANY_DATA, VEC_STRUCT): New constants.
8127 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
8128 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
8129 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
8130 (aarch64_get_mask_mode): New functions.
8131 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
8132 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
8133 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
8134 predicate modes and predicate registers. Explicitly restrict
8135 GPRs to modes of 16 bytes or smaller. Only allow FP registers
8136 to store a vector mode if it is recognized by
8137 aarch64_classify_vector_mode.
8138 (aarch64_regmode_natural_size): New function.
8139 (aarch64_hard_regno_caller_save_mode): Return the original mode
8141 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
8142 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
8143 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
8144 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
8146 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
8147 does not overlap dest if the function is frame-related. Handle
8149 (aarch64_split_add_offset): New function.
8150 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
8151 them aarch64_add_offset.
8152 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
8153 and update call to aarch64_sub_sp.
8154 (aarch64_add_cfa_expression): New function.
8155 (aarch64_expand_prologue): Pass extra temporary registers to the
8156 functions above. Handle the case in which we need to emit new
8157 DW_CFA_expressions for registers that were originally saved
8158 relative to the stack pointer, but now have to be expressed
8159 relative to the frame pointer.
8160 (aarch64_output_mi_thunk): Pass extra temporary registers to the
8162 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
8163 IP0 and IP1 values for SVE frames.
8164 (aarch64_expand_vec_series): New function.
8165 (aarch64_expand_sve_widened_duplicate): Likewise.
8166 (aarch64_expand_sve_const_vector): Likewise.
8167 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
8168 Handle SVE constants. Use emit_move_insn to move a force_const_mem
8169 into the register, rather than emitting a SET directly.
8170 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
8171 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
8172 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
8173 (offset_9bit_signed_scaled_p): New functions.
8174 (aarch64_replicate_bitmask_imm): New function.
8175 (aarch64_bitmask_imm): Use it.
8176 (aarch64_cannot_force_const_mem): Reject expressions involving
8177 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
8178 (aarch64_classify_index): Handle SVE indices, by requiring
8179 a plain register index with a scale that matches the element size.
8180 (aarch64_classify_address): Handle SVE addresses. Assert that
8181 the mode of the address is VOIDmode or an integer mode.
8182 Update call to aarch64_classify_symbol.
8183 (aarch64_classify_symbolic_expression): Update call to
8184 aarch64_classify_symbol.
8185 (aarch64_const_vec_all_in_range_p): New function.
8186 (aarch64_print_vector_float_operand): Likewise.
8187 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
8188 "vN" for FP registers with SVE modes. Handle (const ...) vectors
8189 and the FP immediates 1.0 and 0.5.
8190 (aarch64_print_address_internal): Handle SVE addresses.
8191 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
8192 (aarch64_regno_regclass): Handle predicate registers.
8193 (aarch64_secondary_reload): Handle big-endian reloads of SVE
8195 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
8196 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
8197 (aarch64_convert_sve_vector_bits): New function.
8198 (aarch64_override_options): Use it to handle -msve-vector-bits=.
8199 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8201 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
8202 Handle SVE vector and predicate modes. Accept VL-based constants
8203 that need only one temporary register, and VL offsets that require
8204 no temporary registers.
8205 (aarch64_conditional_register_usage): Mark the predicate registers
8206 as fixed if SVE isn't available.
8207 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
8208 Return true for SVE vector and predicate modes.
8209 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
8210 rather than an unsigned int. Handle SVE modes.
8211 (aarch64_preferred_simd_mode): Update call accordingly. Handle
8213 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
8215 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8216 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8217 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8218 (aarch64_sve_float_mul_immediate_p): New functions.
8219 (aarch64_sve_valid_immediate): New function.
8220 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
8221 Explicitly reject structure modes. Check for INDEX constants.
8222 Handle PTRUE and PFALSE constants.
8223 (aarch64_check_zero_based_sve_index_immediate): New function.
8224 (aarch64_simd_imm_zero_p): Delete.
8225 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
8226 vector modes. Accept constants in the range of CNT[BHWD].
8227 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
8228 ask for an Advanced SIMD mode.
8229 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
8230 (aarch64_simd_vector_alignment): Handle SVE predicates.
8231 (aarch64_vectorize_preferred_vector_alignment): New function.
8232 (aarch64_simd_vector_alignment_reachable): Use it instead of
8234 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
8235 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
8237 (MAX_VECT_LEN): Delete.
8238 (expand_vec_perm_d): Add a vec_flags field.
8239 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
8240 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
8241 (aarch64_evpc_ext): Don't apply a big-endian lane correction
8243 (aarch64_evpc_rev): Rename to...
8244 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
8245 (aarch64_evpc_rev_global): New function.
8246 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
8247 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
8249 (aarch64_evpc_sve_tbl): New function.
8250 (aarch64_expand_vec_perm_const_1): Update after rename of
8251 aarch64_evpc_rev. Handle SVE permutes too, trying
8252 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
8253 than aarch64_evpc_tbl.
8254 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
8255 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
8256 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
8257 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
8258 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
8259 (aarch64_expand_sve_vcond): New functions.
8260 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
8261 of aarch64_vector_mode_p.
8262 (aarch64_dwarf_poly_indeterminate_value): New function.
8263 (aarch64_compute_pressure_classes): Likewise.
8264 (aarch64_can_change_mode_class): Likewise.
8265 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
8266 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
8267 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
8268 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
8269 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
8270 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
8271 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
8272 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
8274 (Dn, Dl, Dr): Accept const as well as const_vector.
8275 (Dz): Likewise. Compare against CONST0_RTX.
8276 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
8277 of "vector" where appropriate.
8278 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
8279 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
8280 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
8281 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
8282 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
8283 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
8284 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
8285 (v_int_equiv): Extend to SVE modes.
8286 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
8288 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
8289 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
8290 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
8291 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
8292 (SVE_COND_FP_CMP): New int iterators.
8293 (perm_hilo): Handle the new unpack unspecs.
8294 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
8296 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
8297 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
8298 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
8299 (aarch64_equality_operator, aarch64_constant_vector_operand)
8300 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
8301 (aarch64_sve_nonimmediate_operand): Likewise.
8302 (aarch64_sve_general_operand): Likewise.
8303 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
8304 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
8305 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
8306 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
8307 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
8308 (aarch64_sve_float_arith_immediate): Likewise.
8309 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
8310 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
8311 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
8312 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
8313 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
8314 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
8315 (aarch64_sve_float_arith_operand): Likewise.
8316 (aarch64_sve_float_arith_with_sub_operand): Likewise.
8317 (aarch64_sve_float_mul_operand): Likewise.
8318 (aarch64_sve_vec_perm_operand): Likewise.
8319 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
8320 (aarch64_mov_operand): Accept const_poly_int and const_vector.
8321 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
8322 as well as const_vector.
8323 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
8324 in file. Use CONST0_RTX and CONSTM1_RTX.
8325 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
8326 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
8327 Use aarch64_simd_imm_zero.
8328 * config/aarch64/aarch64-sve.md: New file.
8329 * config/aarch64/aarch64.md: Include it.
8330 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
8331 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
8332 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
8333 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
8334 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
8335 (sve): New attribute.
8336 (enabled): Disable instructions with the sve attribute unless
8338 (movqi, movhi): Pass CONST_POLY_INT operaneds through
8339 aarch64_expand_mov_immediate.
8340 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
8341 CNT[BHSD] immediates.
8342 (movti): Split CONST_POLY_INT moves into two halves.
8343 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
8344 Split additions that need a temporary here if the destination
8345 is the stack pointer.
8346 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
8347 (*add<mode>3_poly_1): New instruction.
8348 (set_clobber_cc): New expander.
8350 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8352 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
8353 parameter and use it instead of GET_MODE_SIZE (innermode). Use
8354 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
8355 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
8356 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
8357 Change innermode from fixed_mode_size to machine_mode.
8358 (simplify_subreg): Update call accordingly. Handle a constant-sized
8359 subreg of a variable-length CONST_VECTOR.
8361 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8362 Alan Hayward <alan.hayward@arm.com>
8363 David Sherwood <david.sherwood@arm.com>
8365 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
8366 (add_offset_to_base): New function, split out from...
8367 (create_mem_ref): ...here. When handling a scale other than 1,
8368 check first whether the address is valid without the offset.
8369 Add it into the base if so, leaving the index and scale as-is.
8371 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8374 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
8375 fold_for_warn before checking if arg2 is INTEGER_CST.
8377 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
8379 * config/rs6000/predicates.md (load_multiple_operation): Delete.
8380 (store_multiple_operation): Delete.
8381 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
8382 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
8383 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
8384 guarded by TARGET_STRING.
8385 (rs6000_output_load_multiple): Delete.
8386 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
8387 OPTION_MASK_STRING / TARGET_STRING handling.
8388 (print_operand) <'N', 'O'>: Add comment that these are unused now.
8389 (const rs6000_opt_masks) <"string">: Change mask to 0.
8390 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
8391 (MASK_STRING): Delete.
8392 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
8394 (load_multiple): Delete.
8401 (store_multiple): Delete.
8408 (movmemsi_8reg): Delete.
8409 (corresponding unnamed define_insn): Delete.
8410 (movmemsi_6reg): Delete.
8411 (corresponding unnamed define_insn): Delete.
8412 (movmemsi_4reg): Delete.
8413 (corresponding unnamed define_insn): Delete.
8414 (movmemsi_2reg): Delete.
8415 (corresponding unnamed define_insn): Delete.
8416 (movmemsi_1reg): Delete.
8417 (corresponding unnamed define_insn): Delete.
8418 * config/rs6000/rs6000.opt (mno-string): New.
8419 (mstring): Replace by deprecation warning stub.
8420 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
8422 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8424 * regrename.c (regrename_do_replace): If replacing the same
8425 reg multiple times, try to reuse last created gen_raw_REG.
8428 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
8429 main to workaround a bug in GDB.
8431 2018-01-12 Tom de Vries <tom@codesourcery.com>
8434 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
8436 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
8438 PR rtl-optimization/80481
8439 * ira-color.c (get_cap_member): New function.
8440 (allocnos_conflict_by_live_ranges_p): Use it.
8441 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
8442 (setup_slot_coalesced_allocno_live_ranges): Ditto.
8444 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
8447 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
8448 (*saddl_se_1): Ditto.
8450 (*ssubl_se_1): Ditto.
8452 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8454 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
8455 rather than wi::to_widest for DR_INITs.
8456 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
8457 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
8458 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
8460 (vect_analyze_group_access_1): Note that here.
8462 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8464 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
8465 polynomial type sizes.
8467 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8469 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
8470 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
8471 (gimple_add_tmp_var): Likewise.
8473 2018-01-12 Martin Liska <mliska@suse.cz>
8475 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
8476 (gimple_alloc_sizes): Likewise.
8477 (dump_gimple_statistics): Use PRIu64 in printf format.
8478 * gimple.h: Change uint64_t to int.
8480 2018-01-12 Martin Liska <mliska@suse.cz>
8482 * tree-core.h: Use uint64_t instead of int.
8483 * tree.c (tree_node_counts): Likewise.
8484 (tree_node_sizes): Likewise.
8485 (dump_tree_statistics): Use PRIu64 in printf format.
8487 2018-01-12 Martin Liska <mliska@suse.cz>
8489 * Makefile.in: As qsort_chk is implemented in vec.c, add
8490 vec.o to linkage of gencfn-macros.
8491 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
8492 passing the info to record_node_allocation_statistics.
8493 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
8495 * ggc-common.c (struct ggc_usage): Add operator== and use
8496 it in operator< and compare function.
8497 * mem-stats.h (struct mem_usage): Likewise.
8498 * vec.c (struct vec_usage): Remove operator< and compare
8499 function. Can be simply inherited.
8501 2018-01-12 Martin Jambor <mjambor@suse.cz>
8504 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
8505 * tree-ssa-math-opts.c: Include domwalk.h.
8506 (convert_mult_to_fma_1): New function.
8507 (fma_transformation_info): New type.
8508 (fma_deferring_state): Likewise.
8509 (cancel_fma_deferring): New function.
8510 (result_of_phi): Likewise.
8511 (last_fma_candidate_feeds_initial_phi): Likewise.
8512 (convert_mult_to_fma): Added deferring logic, split actual
8513 transformation to convert_mult_to_fma_1.
8514 (math_opts_dom_walker): New type.
8515 (math_opts_dom_walker::after_dom_children): New method, body moved
8516 here from pass_optimize_widening_mul::execute, added deferring logic
8518 (pass_optimize_widening_mul::execute): Moved most of code to
8519 math_opts_dom_walker::after_dom_children.
8520 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
8521 * config/i386/i386.c (ix86_option_override_internal): Added
8522 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
8524 2018-01-12 Richard Biener <rguenther@suse.de>
8527 * dwarf2out.c (gen_variable_die): Do not reset old_die for
8528 inline instance vars.
8530 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
8533 * config/rx/rx.c (rx_is_restricted_memory_address):
8536 2018-01-12 Richard Biener <rguenther@suse.de>
8538 PR tree-optimization/80846
8539 * target.def (split_reduction): New target hook.
8540 * targhooks.c (default_split_reduction): New function.
8541 * targhooks.h (default_split_reduction): Declare.
8542 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
8543 target requests first reduce vectors by combining low and high
8545 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
8546 (get_vectype_for_scalar_type_and_size): Export.
8547 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
8548 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
8549 * doc/tm.texi: Regenerate.
8550 * config/i386/i386.c (ix86_split_reduction): Implement
8551 TARGET_VECTORIZE_SPLIT_REDUCTION.
8553 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8556 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
8557 in PIC mode except for TARGET_VXWORKS_RTP.
8558 * config/sparc/sparc.c: Include cfgrtl.h.
8559 (TARGET_INIT_PIC_REG): Define.
8560 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
8561 (sparc_pic_register_p): New predicate.
8562 (sparc_legitimate_address_p): Use it.
8563 (sparc_legitimize_pic_address): Likewise.
8564 (sparc_delegitimize_address): Likewise.
8565 (sparc_mode_dependent_address_p): Likewise.
8566 (gen_load_pcrel_sym): Remove 4th parameter.
8567 (load_got_register): Adjust call to above. Remove obsolete stuff.
8568 (sparc_expand_prologue): Do not call load_got_register here.
8569 (sparc_flat_expand_prologue): Likewise.
8570 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
8571 (sparc_use_pseudo_pic_reg): New function.
8572 (sparc_init_pic_reg): Likewise.
8573 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8574 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
8576 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
8578 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
8579 Add item for branch_cost.
8581 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8583 PR rtl-optimization/83565
8584 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
8585 not extend the result to a larger mode for rotate operations.
8586 (num_sign_bit_copies1): Likewise.
8588 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8591 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
8593 Use values-Xc.o for -pedantic.
8594 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
8596 2018-01-12 Martin Liska <mliska@suse.cz>
8599 * ipa-devirt.c (final_warning_record::grow_type_warnings):
8601 (possible_polymorphic_call_targets): Use it.
8602 (ipa_devirt): Likewise.
8604 2018-01-12 Martin Liska <mliska@suse.cz>
8606 * profile-count.h (enum profile_quality): Use 0 as invalid
8607 enum value of profile_quality.
8609 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
8611 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
8612 -mext-string options.
8614 2018-01-12 Richard Biener <rguenther@suse.de>
8616 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
8617 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
8618 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
8620 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
8622 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
8624 * configure.ac (--with-long-double-format): Add support for the
8625 configuration option to change the default long double format on
8627 * config.gcc (powerpc*-linux*-*): Likewise.
8628 * configure: Regenerate.
8629 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
8630 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
8631 used without modification.
8633 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8635 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
8636 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
8637 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
8638 MISC_BUILTIN_SPEC_BARRIER.
8639 (rs6000_init_builtins): Likewise.
8640 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
8642 (speculation_barrier): New define_insn.
8643 * doc/extend.texi: Document __builtin_speculation_barrier.
8645 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8648 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
8649 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
8650 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
8652 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
8653 integral modes instead of "ss" and "sd".
8654 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
8655 vectors with 32-bit and 64-bit elements.
8656 (vecdupssescalarmodesuffix): New mode attribute.
8657 (vec_dup<mode>): Use it.
8659 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
8662 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
8663 frame if argument is passed on stack.
8665 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8668 * ree.c (combine_reaching_defs): Optimize also
8669 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
8670 reg2=any_extend(exp); reg1=reg2;, formatting fix.
8672 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8675 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
8677 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8680 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
8681 after they are computed.
8683 2018-01-11 Bin Cheng <bin.cheng@arm.com>
8685 PR tree-optimization/83695
8686 * gimple-loop-linterchange.cc
8687 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
8688 reset cached scev information after interchange.
8689 (pass_linterchange::execute): Remove call to scev_reset_htab.
8691 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8693 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
8694 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
8695 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
8696 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
8697 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
8698 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
8699 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
8700 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
8701 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
8702 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
8703 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
8704 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
8705 (V_lane_reg): Likewise.
8706 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
8708 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
8709 (vfmal_lane_low<mode>_intrinsic,
8710 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
8711 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
8712 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
8713 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
8714 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
8715 vfmsl_lane_high<mode>_intrinsic): New define_insns.
8717 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8719 * config/arm/arm-cpus.in (fp16fml): New feature.
8720 (ALL_SIMD): Add fp16fml.
8721 (armv8.2-a): Add fp16fml as an option.
8722 (armv8.3-a): Likewise.
8723 (armv8.4-a): Add fp16fml as part of fp16.
8724 * config/arm/arm.h (TARGET_FP16FML): Define.
8725 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
8727 * config/arm/arm-modes.def (V2HF): Define.
8728 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
8729 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
8730 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
8731 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
8732 vfmsl_low, vfmsl_high): New set of builtins.
8733 * config/arm/iterators.md (PLUSMINUS): New code iterator.
8734 (vfml_op): New code attribute.
8735 (VFMLHALVES): New int iterator.
8736 (VFML, VFMLSEL): New mode attributes.
8737 (V_reg): Define mapping for V2HF.
8738 (V_hi, V_lo): New mode attributes.
8739 (VF_constraint): Likewise.
8740 (vfml_half, vfml_half_selector): New int attributes.
8741 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
8743 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
8744 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
8746 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
8747 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
8748 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
8749 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
8751 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
8752 Document new effective target and option set.
8754 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8756 * config/arm/arm-cpus.in (armv8_4): New feature.
8757 (ARMv8_4a): New fgroup.
8758 (armv8.4-a): New arch.
8759 * config/arm/arm-tables.opt: Regenerate.
8760 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
8761 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
8762 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
8763 Add matching rules for -march=armv8.4-a and extensions.
8764 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
8766 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
8769 * config/rx/rx.md (BW): New mode attribute.
8770 (sync_lock_test_and_setsi): Add mode suffix to insn output.
8772 2018-01-11 Richard Biener <rguenther@suse.de>
8774 PR tree-optimization/83435
8775 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
8776 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
8777 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
8779 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8780 Alan Hayward <alan.hayward@arm.com>
8781 David Sherwood <david.sherwood@arm.com>
8783 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
8785 (aarch64_classify_address): Initialize it. Track polynomial offsets.
8786 (aarch64_print_address_internal): Use it to check for a zero offset.
8788 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8789 Alan Hayward <alan.hayward@arm.com>
8790 David Sherwood <david.sherwood@arm.com>
8792 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
8793 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
8794 Return a poly_int64 rather than a HOST_WIDE_INT.
8795 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
8796 rather than a HOST_WIDE_INT.
8797 * config/aarch64/aarch64.h (aarch64_frame): Protect with
8798 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
8799 hard_fp_offset, frame_size, initial_adjust, callee_offset and
8800 final_offset from HOST_WIDE_INT to poly_int64.
8801 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
8802 to_constant when getting the number of units in an Advanced SIMD
8804 (aarch64_builtin_vectorized_function): Check for a constant number
8806 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
8808 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
8809 attribute instead of GET_MODE_NUNITS.
8810 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
8811 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
8812 GET_MODE_SIZE for fixed-size registers.
8813 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
8814 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
8815 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
8816 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
8817 (aarch64_print_operand, aarch64_print_address_internal)
8818 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
8819 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
8820 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
8821 Handle polynomial GET_MODE_SIZE.
8822 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
8823 wider than SImode without modification.
8824 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
8825 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
8826 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
8827 passing and returning SVE modes.
8828 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
8829 rather than GEN_INT.
8830 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
8831 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
8832 (aarch64_allocate_and_probe_stack_space): Likewise.
8833 (aarch64_layout_frame): Cope with polynomial offsets.
8834 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
8835 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
8837 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
8838 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
8839 poly_int64 rather than a HOST_WIDE_INT.
8840 (aarch64_get_separate_components, aarch64_process_components)
8841 (aarch64_expand_prologue, aarch64_expand_epilogue)
8842 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
8843 (aarch64_anchor_offset): New function, split out from...
8844 (aarch64_legitimize_address): ...here.
8845 (aarch64_builtin_vectorization_cost): Handle polynomial
8846 TYPE_VECTOR_SUBPARTS.
8847 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
8849 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
8850 number of elements from the PARALLEL rather than the mode.
8851 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
8852 rather than GET_MODE_BITSIZE.
8853 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
8854 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
8855 (aarch64_expand_vec_perm_const_1): Handle polynomial
8856 d->perm.length () and d->perm elements.
8857 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
8858 Apply to_constant to d->perm elements.
8859 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
8860 polynomial CONST_VECTOR_NUNITS.
8861 (aarch64_move_pointer): Take amount as a poly_int64 rather
8863 (aarch64_progress_pointer): Avoid temporary variable.
8864 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
8865 the mode attribute instead of GET_MODE.
8867 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8868 Alan Hayward <alan.hayward@arm.com>
8869 David Sherwood <david.sherwood@arm.com>
8871 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
8872 x exists before using it.
8873 (aarch64_add_constant_internal): Rename to...
8874 (aarch64_add_offset_1): ...this. Replace regnum with separate
8875 src and dest rtxes. Handle the case in which they're different,
8876 including when the offset is zero. Replace scratchreg with an rtx.
8877 Use 2 additions if there is no spare register into which we can
8878 move a 16-bit constant.
8879 (aarch64_add_constant): Delete.
8880 (aarch64_add_offset): Replace reg with separate src and dest
8881 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
8882 Use aarch64_add_offset_1.
8883 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
8884 an rtx rather than an int. Take the delta as a poly_int64
8885 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
8886 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
8887 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
8888 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
8889 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
8891 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
8892 aarch64_add_constant.
8894 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8896 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
8897 Use scalar_float_mode.
8899 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8901 * config/aarch64/aarch64-simd.md
8902 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
8903 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
8904 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
8905 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
8906 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
8907 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
8908 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
8909 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
8910 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
8911 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
8913 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8916 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
8917 targ_options->x_arm_arch_string is non NULL.
8919 2018-01-11 Tamar Christina <tamar.christina@arm.com>
8921 * config/aarch64/aarch64.h
8922 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
8924 2018-01-11 Sudakshina Das <sudi.das@arm.com>
8927 * expmed.c (emit_store_flag_force): Swap if const op0
8928 and change VOIDmode to mode of op0.
8930 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8932 PR rtl-optimization/83761
8933 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
8934 than bytes to mode_for_size.
8936 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8939 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
8940 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
8943 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8946 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
8947 when in layout mode.
8948 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
8949 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
8952 2018-01-10 Michael Collison <michael.collison@arm.com>
8954 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
8955 * config/aarch64/aarch64-option-extension.def: Add
8956 AARCH64_OPT_EXTENSION of 'fp16fml'.
8957 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8958 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
8959 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
8960 * config/aarch64/constraints.md (Ui7): New constraint.
8961 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
8962 (VFMLA_SEL_W): Ditto.
8965 (VFMLA16_LOW): New int iterator.
8966 (VFMLA16_HIGH): Ditto.
8967 (UNSPEC_FMLAL): New unspec.
8968 (UNSPEC_FMLSL): Ditto.
8969 (UNSPEC_FMLAL2): Ditto.
8970 (UNSPEC_FMLSL2): Ditto.
8971 (f16mac): New code attribute.
8972 * config/aarch64/aarch64-simd-builtins.def
8973 (aarch64_fmlal_lowv2sf): Ditto.
8974 (aarch64_fmlsl_lowv2sf): Ditto.
8975 (aarch64_fmlalq_lowv4sf): Ditto.
8976 (aarch64_fmlslq_lowv4sf): Ditto.
8977 (aarch64_fmlal_highv2sf): Ditto.
8978 (aarch64_fmlsl_highv2sf): Ditto.
8979 (aarch64_fmlalq_highv4sf): Ditto.
8980 (aarch64_fmlslq_highv4sf): Ditto.
8981 (aarch64_fmlal_lane_lowv2sf): Ditto.
8982 (aarch64_fmlsl_lane_lowv2sf): Ditto.
8983 (aarch64_fmlal_laneq_lowv2sf): Ditto.
8984 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
8985 (aarch64_fmlalq_lane_lowv4sf): Ditto.
8986 (aarch64_fmlsl_lane_lowv4sf): Ditto.
8987 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
8988 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
8989 (aarch64_fmlal_lane_highv2sf): Ditto.
8990 (aarch64_fmlsl_lane_highv2sf): Ditto.
8991 (aarch64_fmlal_laneq_highv2sf): Ditto.
8992 (aarch64_fmlsl_laneq_highv2sf): Ditto.
8993 (aarch64_fmlalq_lane_highv4sf): Ditto.
8994 (aarch64_fmlsl_lane_highv4sf): Ditto.
8995 (aarch64_fmlalq_laneq_highv4sf): Ditto.
8996 (aarch64_fmlsl_laneq_highv4sf): Ditto.
8997 * config/aarch64/aarch64-simd.md:
8998 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
8999 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9000 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
9001 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9002 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
9003 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
9004 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
9005 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
9006 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
9007 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
9008 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
9009 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
9010 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
9011 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
9012 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
9013 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
9014 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
9015 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
9016 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
9017 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
9018 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
9019 (vfmlsl_low_u32): Ditto.
9020 (vfmlalq_low_u32): Ditto.
9021 (vfmlslq_low_u32): Ditto.
9022 (vfmlal_high_u32): Ditto.
9023 (vfmlsl_high_u32): Ditto.
9024 (vfmlalq_high_u32): Ditto.
9025 (vfmlslq_high_u32): Ditto.
9026 (vfmlal_lane_low_u32): Ditto.
9027 (vfmlsl_lane_low_u32): Ditto.
9028 (vfmlal_laneq_low_u32): Ditto.
9029 (vfmlsl_laneq_low_u32): Ditto.
9030 (vfmlalq_lane_low_u32): Ditto.
9031 (vfmlslq_lane_low_u32): Ditto.
9032 (vfmlalq_laneq_low_u32): Ditto.
9033 (vfmlslq_laneq_low_u32): Ditto.
9034 (vfmlal_lane_high_u32): Ditto.
9035 (vfmlsl_lane_high_u32): Ditto.
9036 (vfmlal_laneq_high_u32): Ditto.
9037 (vfmlsl_laneq_high_u32): Ditto.
9038 (vfmlalq_lane_high_u32): Ditto.
9039 (vfmlslq_lane_high_u32): Ditto.
9040 (vfmlalq_laneq_high_u32): Ditto.
9041 (vfmlslq_laneq_high_u32): Ditto.
9042 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
9043 (AARCH64_FL_FOR_ARCH8_4): New.
9044 (AARCH64_ISA_F16FML): New ISA flag.
9045 (TARGET_F16FML): New feature flag for fp16fml.
9046 (doc/invoke.texi): Document new fp16fml option.
9048 2018-01-10 Michael Collison <michael.collison@arm.com>
9050 * config/aarch64/aarch64-builtins.c:
9051 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
9052 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9053 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
9054 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
9055 (AARCH64_ISA_SHA3): New ISA flag.
9056 (TARGET_SHA3): New feature flag for sha3.
9057 * config/aarch64/iterators.md (sha512_op): New int attribute.
9058 (CRYPTO_SHA512): New int iterator.
9059 (UNSPEC_SHA512H): New unspec.
9060 (UNSPEC_SHA512H2): Ditto.
9061 (UNSPEC_SHA512SU0): Ditto.
9062 (UNSPEC_SHA512SU1): Ditto.
9063 * config/aarch64/aarch64-simd-builtins.def
9064 (aarch64_crypto_sha512hqv2di): New builtin.
9065 (aarch64_crypto_sha512h2qv2di): Ditto.
9066 (aarch64_crypto_sha512su0qv2di): Ditto.
9067 (aarch64_crypto_sha512su1qv2di): Ditto.
9068 (aarch64_eor3qv8hi): Ditto.
9069 (aarch64_rax1qv2di): Ditto.
9070 (aarch64_xarqv2di): Ditto.
9071 (aarch64_bcaxqv8hi): Ditto.
9072 * config/aarch64/aarch64-simd.md:
9073 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
9074 (aarch64_crypto_sha512su0qv2di): Ditto.
9075 (aarch64_crypto_sha512su1qv2di): Ditto.
9076 (aarch64_eor3qv8hi): Ditto.
9077 (aarch64_rax1qv2di): Ditto.
9078 (aarch64_xarqv2di): Ditto.
9079 (aarch64_bcaxqv8hi): Ditto.
9080 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
9081 (vsha512h2q_u64): Ditto.
9082 (vsha512su0q_u64): Ditto.
9083 (vsha512su1q_u64): Ditto.
9084 (veor3q_u16): Ditto.
9085 (vrax1q_u64): Ditto.
9087 (vbcaxq_u16): Ditto.
9088 * config/arm/types.md (crypto_sha512): New type attribute.
9089 (crypto_sha3): Ditto.
9090 (doc/invoke.texi): Document new sha3 option.
9092 2018-01-10 Michael Collison <michael.collison@arm.com>
9094 * config/aarch64/aarch64-builtins.c:
9095 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
9096 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9097 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
9098 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
9099 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
9100 (AARCH64_ISA_SM4): New ISA flag.
9101 (TARGET_SM4): New feature flag for sm4.
9102 * config/aarch64/aarch64-simd-builtins.def
9103 (aarch64_sm3ss1qv4si): Ditto.
9104 (aarch64_sm3tt1aq4si): Ditto.
9105 (aarch64_sm3tt1bq4si): Ditto.
9106 (aarch64_sm3tt2aq4si): Ditto.
9107 (aarch64_sm3tt2bq4si): Ditto.
9108 (aarch64_sm3partw1qv4si): Ditto.
9109 (aarch64_sm3partw2qv4si): Ditto.
9110 (aarch64_sm4eqv4si): Ditto.
9111 (aarch64_sm4ekeyqv4si): Ditto.
9112 * config/aarch64/aarch64-simd.md:
9113 (aarch64_sm3ss1qv4si): Ditto.
9114 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
9115 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
9116 (aarch64_sm4eqv4si): Ditto.
9117 (aarch64_sm4ekeyqv4si): Ditto.
9118 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
9119 (sm3part_op): Ditto.
9120 (CRYPTO_SM3TT): Ditto.
9121 (CRYPTO_SM3PART): Ditto.
9122 (UNSPEC_SM3SS1): New unspec.
9123 (UNSPEC_SM3TT1A): Ditto.
9124 (UNSPEC_SM3TT1B): Ditto.
9125 (UNSPEC_SM3TT2A): Ditto.
9126 (UNSPEC_SM3TT2B): Ditto.
9127 (UNSPEC_SM3PARTW1): Ditto.
9128 (UNSPEC_SM3PARTW2): Ditto.
9129 (UNSPEC_SM4E): Ditto.
9130 (UNSPEC_SM4EKEY): Ditto.
9131 * config/aarch64/constraints.md (Ui2): New constraint.
9132 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
9133 * config/arm/types.md (crypto_sm3): New type attribute.
9134 (crypto_sm4): Ditto.
9135 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
9136 (vsm3tt1aq_u32): Ditto.
9137 (vsm3tt1bq_u32): Ditto.
9138 (vsm3tt2aq_u32): Ditto.
9139 (vsm3tt2bq_u32): Ditto.
9140 (vsm3partw1q_u32): Ditto.
9141 (vsm3partw2q_u32): Ditto.
9142 (vsm4eq_u32): Ditto.
9143 (vsm4ekeyq_u32): Ditto.
9144 (doc/invoke.texi): Document new sm4 option.
9146 2018-01-10 Michael Collison <michael.collison@arm.com>
9148 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
9149 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
9150 (AARCH64_FL_FOR_ARCH8_4): New.
9151 (AARCH64_FL_V8_4): New flag.
9152 (doc/invoke.texi): Document new armv8.4-a option.
9154 2018-01-10 Michael Collison <michael.collison@arm.com>
9156 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9157 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
9158 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
9159 * config/aarch64/aarch64-option-extension.def: Add
9160 AARCH64_OPT_EXTENSION of 'sha2'.
9161 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
9162 (crypto): Disable sha2 and aes if crypto disabled.
9163 (crypto): Enable aes and sha2 if enabled.
9164 (simd): Disable sha2 and aes if simd disabled.
9165 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
9167 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
9168 (TARGET_SHA2): New feature flag for sha2.
9169 (TARGET_AES): New feature flag for aes.
9170 * config/aarch64/aarch64-simd.md:
9171 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
9172 conditional on TARGET_AES.
9173 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
9174 (aarch64_crypto_sha1hsi): Make pattern conditional
9176 (aarch64_crypto_sha1hv4si): Ditto.
9177 (aarch64_be_crypto_sha1hv4si): Ditto.
9178 (aarch64_crypto_sha1su1v4si): Ditto.
9179 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
9180 (aarch64_crypto_sha1su0v4si): Ditto.
9181 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
9182 (aarch64_crypto_sha256su0v4si): Ditto.
9183 (aarch64_crypto_sha256su1v4si): Ditto.
9184 (doc/invoke.texi): Document new aes and sha2 options.
9186 2018-01-10 Martin Sebor <msebor@redhat.com>
9188 PR tree-optimization/83781
9189 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
9192 2018-01-11 Martin Sebor <msebor@gmail.com>
9193 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9195 PR tree-optimization/83501
9196 PR tree-optimization/81703
9198 * tree-ssa-strlen.c (get_string_cst): Rename...
9199 (get_string_len): ...to this. Handle global constants.
9200 (handle_char_store): Adjust.
9202 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
9203 Jim Wilson <jimw@sifive.com>
9205 * config/riscv/riscv-protos.h (riscv_output_return): New.
9206 * config/riscv/riscv.c (struct machine_function): New naked_p field.
9207 (riscv_attribute_table, riscv_output_return),
9208 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
9209 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
9210 (riscv_compute_frame_info): Only compute frame->mask if not a naked
9212 (riscv_expand_prologue): Add early return for naked function.
9213 (riscv_expand_epilogue): Likewise.
9214 (riscv_function_ok_for_sibcall): Return false for naked function.
9215 (riscv_set_current_function): New.
9216 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
9217 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
9218 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
9219 * doc/extend.texi (RISC-V Function Attributes): New.
9221 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
9223 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
9224 check for 128-bit long double before checking TCmode.
9225 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
9226 128-bit long doubles before checking TFmode or TCmode.
9227 (FLOAT128_IBM_P): Likewise.
9229 2018-01-10 Martin Sebor <msebor@redhat.com>
9231 PR tree-optimization/83671
9232 * builtins.c (c_strlen): Unconditionally return zero for the empty
9234 Use -Warray-bounds for warnings.
9235 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
9236 for non-constant array indices with COMPONENT_REF, arrays of
9237 arrays, and pointers to arrays.
9238 (gimple_fold_builtin_strlen): Determine and set length range for
9239 non-constant character arrays.
9241 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
9244 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
9247 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
9249 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
9251 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9254 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
9255 VECTOR_MEM_ALTIVEC_OR_VSX_P.
9256 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
9257 indexed_or_indirect_operand predicate.
9258 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
9259 (*vsx_le_perm_load_v8hi): Likewise.
9260 (*vsx_le_perm_load_v16qi): Likewise.
9261 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
9262 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
9263 (*vsx_le_perm_store_v8hi): Likewise.
9264 (*vsx_le_perm_store_v16qi): Likewise.
9265 (eight unnamed splitters): Likewise.
9267 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9269 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
9270 * config/rs6000/emmintrin.h: Likewise.
9271 * config/rs6000/mmintrin.h: Likewise.
9272 * config/rs6000/xmmintrin.h: Likewise.
9274 2018-01-10 David Malcolm <dmalcolm@redhat.com>
9277 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
9279 * tree.c (tree_nop_conversion): Return true for location wrapper
9281 (maybe_wrap_with_location): New function.
9282 (selftest::check_strip_nops): New function.
9283 (selftest::test_location_wrappers): New function.
9284 (selftest::tree_c_tests): Call it.
9285 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
9286 (maybe_wrap_with_location): New decl.
9287 (EXPR_LOCATION_WRAPPER_P): New macro.
9288 (location_wrapper_p): New inline function.
9289 (tree_strip_any_location_wrapper): New inline function.
9291 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
9294 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
9295 stack_realign_offset for the largest alignment of stack slot
9297 (ix86_find_max_used_stack_alignment): New function.
9298 (ix86_finalize_stack_frame_flags): Use it. Set
9299 max_used_stack_alignment if we don't realign stack.
9300 * config/i386/i386.h (machine_function): Add
9301 max_used_stack_alignment.
9303 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
9305 * config/arm/arm.opt (-mbranch-cost): New option.
9306 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
9309 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
9312 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
9313 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
9315 2018-01-10 Richard Biener <rguenther@suse.de>
9318 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
9319 early out so it also covers the case where we have a non-NULL
9322 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9324 PR tree-optimization/83753
9325 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
9326 for non-strided grouped accesses if the number of elements is 1.
9328 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9331 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
9332 * i386.h (TARGET_USE_GATHER): Define.
9333 * x86-tune.def (X86_TUNE_USE_GATHER): New.
9335 2018-01-10 Martin Liska <mliska@suse.cz>
9338 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
9339 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
9341 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
9342 CLEANUP_NO_PARTITIONING is not set.
9344 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9346 * doc/rtl.texi: Remove documentation of (const ...) wrappers
9347 for vectors, as a partial revert of r254296.
9348 * rtl.h (const_vec_p): Delete.
9349 (const_vec_duplicate_p): Don't test for vector CONSTs.
9350 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
9351 * expmed.c (make_tree): Likewise.
9354 * common.md (E, F): Use CONSTANT_P instead of checking for
9356 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9357 checking for CONST_VECTOR.
9359 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9362 * predict.c (force_edge_cold): Handle in more sane way edges
9365 2018-01-09 Carl Love <cel@us.ibm.com>
9367 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
9369 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
9370 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
9371 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
9372 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
9373 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
9374 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
9375 * config/rs6000/rs6000-protos.h: Add extern defition for
9376 rs6000_generate_float2_double_code.
9377 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
9379 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
9380 (float2_v2df): Add define_expand.
9382 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
9385 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
9386 op_mode in the force_to_mode call.
9388 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9390 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
9391 instead of checking each element individually.
9392 (aarch64_evpc_uzp): Likewise.
9393 (aarch64_evpc_zip): Likewise.
9394 (aarch64_evpc_ext): Likewise.
9395 (aarch64_evpc_rev): Likewise.
9396 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
9397 instead of checking each element individually. Return true without
9399 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
9400 whether all selected elements come from the same input, instead of
9401 checking each element individually. Remove calls to gen_rtx_REG,
9402 start_sequence and end_sequence and instead assert that no rtl is
9405 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9407 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
9408 order of HIGH and CONST checks.
9410 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9412 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
9413 if the destination isn't an SSA_NAME.
9415 2018-01-09 Richard Biener <rguenther@suse.de>
9417 PR tree-optimization/83668
9418 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
9420 (canonicalize_loop_form): ... here, renamed from ...
9421 (canonicalize_loop_closed_ssa_form): ... this and amended to
9422 swap successor edges for loop exit blocks to make us use
9423 the RPO order we need for initial schedule generation.
9425 2018-01-09 Joseph Myers <joseph@codesourcery.com>
9427 PR tree-optimization/64811
9428 * match.pd: When optimizing comparisons with Inf, avoid
9429 introducing or losing exceptions from comparisons with NaN.
9431 2018-01-09 Martin Liska <mliska@suse.cz>
9434 * asan.c (shadow_mem_size): Add gcc_assert.
9436 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
9438 Don't save registers in main().
9441 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
9442 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
9443 * config/avr/avr.c (avr_set_current_function): Don't error if
9444 naked, OS_task or OS_main are specified at the same time.
9445 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
9447 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
9449 * common/config/avr/avr-common.c (avr_option_optimization_table):
9450 Switch on -mmain-is-OS_task for optimizing compilations.
9452 2018-01-09 Richard Biener <rguenther@suse.de>
9454 PR tree-optimization/83572
9455 * graphite.c: Include cfganal.h.
9456 (graphite_transform_loops): Connect infinite loops to exit
9457 and remove fake edges at the end.
9459 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9461 * ipa-inline.c (edge_badness): Revert accidental checkin.
9463 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9466 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
9467 symbols; not inline clones.
9469 2018-01-09 Jakub Jelinek <jakub@redhat.com>
9472 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
9473 hard registers. Formatting fixes.
9475 PR preprocessor/83722
9476 * gcc.c (try_generate_repro): Pass
9477 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
9478 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
9481 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
9482 Kito Cheng <kito.cheng@gmail.com>
9484 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
9485 (riscv_leaf_function_p): Delete.
9486 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
9488 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9490 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
9492 (do_ifelse): New function.
9493 (do_isel): New function.
9494 (do_sub3): New function.
9495 (do_add3): New function.
9496 (do_load_mask_compare): New function.
9497 (do_overlap_load_compare): New function.
9498 (expand_compare_loop): New function.
9499 (expand_block_compare): Call expand_compare_loop() when appropriate.
9500 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
9502 (-mblock-compare-inline-loop-limit): New option.
9504 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9507 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
9508 Reverse order of second and third operands in first alternative.
9509 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
9510 of first and second elements in UNSPEC_VPERMR vector.
9511 (altivec_expand_vec_perm_le): Likewise.
9513 2018-01-08 Jeff Law <law@redhat.com>
9515 PR rtl-optimizatin/81308
9516 * tree-switch-conversion.c (cfg_altered): New file scoped static.
9517 (process_switch): If group_case_labels makes a change, then set
9519 (pass_convert_switch::execute): If a switch is converted, then
9520 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
9522 PR rtl-optimization/81308
9523 * recog.c (split_all_insns): Conditionally cleanup the CFG after
9526 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
9528 PR target/83663 - Revert r255946
9529 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
9530 generation for cases where splatting a value is not useful.
9531 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
9532 across a vec_duplicate and a paradoxical subreg forming a vector
9533 mode to a vec_concat.
9535 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9537 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
9538 -march=armv8.3-a variants.
9539 * config/arm/t-multilib: Likewise.
9540 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
9542 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9544 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
9546 (cceq_ior_compare_complement): Give it a name so I can use it, and
9547 change boolean_or_operator predicate to boolean_operator so it can
9548 be used to generate a crand.
9549 (eqne): New code iterator.
9550 (bd/bd_neg): New code_attrs.
9551 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
9552 a single define_insn.
9553 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
9554 decrement (bdnzt/bdnzf/bdzt/bdzf).
9555 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
9556 with the new names of the branch decrement patterns, and added the
9557 names of the branch decrement conditional patterns.
9559 2018-01-08 Richard Biener <rguenther@suse.de>
9561 PR tree-optimization/83563
9562 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
9565 2018-01-08 Richard Biener <rguenther@suse.de>
9568 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
9570 2018-01-08 Richard Biener <rguenther@suse.de>
9572 PR tree-optimization/83685
9573 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
9574 references to abnormals.
9576 2018-01-08 Richard Biener <rguenther@suse.de>
9579 * dwarf2out.c (output_indirect_strings): Handle empty
9580 skeleton_debug_str_hash.
9581 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
9583 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9585 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
9586 (emit_store_direct): Likewise.
9587 (arc_trampoline_adjust_address): Likewise.
9588 (arc_asm_trampoline_template): New function.
9589 (arc_initialize_trampoline): Use asm_trampoline_template.
9590 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
9591 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
9592 * config/arc/arc.md (flush_icache): Delete pattern.
9594 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9596 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
9597 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
9600 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9603 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
9604 by not USED_FOR_TARGET.
9605 (make_pass_resolve_sw_modes): Likewise.
9607 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9609 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
9612 2018-01-08 Richard Biener <rguenther@suse.de>
9615 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
9617 2018-01-08 Richard Biener <rguenther@suse.de>
9620 * match.pd ((t * 2) / 2) -> t): Add missing :c.
9622 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
9625 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
9626 basic blocks with a small number of successors.
9627 (convert_control_dep_chain_into_preds): Improve handling of
9629 (dump_predicates): Split apart into...
9630 (dump_pred_chain): ...here...
9631 (dump_pred_info): ...and here.
9632 (can_one_predicate_be_invalidated_p): Add debugging printfs.
9633 (can_chain_union_be_invalidated_p): Improve check for invalidation
9635 (uninit_uses_cannot_happen): Avoid unnecessary if
9636 convert_control_dep_chain_into_preds yielded nothing.
9638 2018-01-06 Martin Sebor <msebor@redhat.com>
9640 PR tree-optimization/83640
9641 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
9642 subtracting negative offset from size.
9643 (builtin_access::overlap): Adjust offset bounds of the access to fall
9644 within the size of the object if possible.
9646 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
9648 PR rtl-optimization/83699
9649 * expmed.c (extract_bit_field_1): Restrict the vector usage of
9650 extract_bit_field_as_subreg to cases in which the extracted
9651 value is also a vector.
9653 * lra-constraints.c (process_alt_operands): Test for the equivalence
9654 substitutions when detecting a possible reload cycle.
9656 2018-01-06 Jakub Jelinek <jakub@redhat.com>
9659 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
9660 by default if flag_selective_schedling{,2}. Formatting fixes.
9662 PR rtl-optimization/83682
9663 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
9664 if it has non-VECTOR_MODE element mode.
9665 (vec_duplicate_p): Likewise.
9668 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
9669 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
9671 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9674 * config/i386/i386-builtin.def
9675 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
9676 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
9677 Require also OPTION_MASK_ISA_AVX512F in addition to
9678 OPTION_MASK_ISA_GFNI.
9679 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
9680 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
9681 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
9682 to OPTION_MASK_ISA_GFNI.
9683 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
9684 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
9685 OPTION_MASK_ISA_AVX512BW.
9686 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
9687 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
9688 addition to OPTION_MASK_ISA_GFNI.
9689 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
9690 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
9691 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
9692 to OPTION_MASK_ISA_GFNI.
9693 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
9694 a requirement for all ISAs rather than any of them with a few
9696 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
9698 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
9699 bitmasks to be enabled with 3 exceptions, instead of requiring any
9700 enabled ISA with lots of exceptions.
9701 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
9702 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
9703 Change avx512bw in isa attribute to avx512f.
9704 * config/i386/sgxintrin.h: Add license boilerplate.
9705 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
9706 to __AVX512F__ and __AVX512VL to __AVX512VL__.
9707 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
9708 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
9710 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
9711 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
9712 temporarily sse2 rather than sse if not enabled already.
9715 * config/i386/sse.md (VI248_VLBW): Rename to ...
9716 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
9717 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
9718 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
9719 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
9720 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
9721 mode iterator instead of VI248_VLBW.
9723 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
9725 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
9726 (record_modified): Skip clobbers; add debug output.
9727 (param_change_prob): Use sreal frequencies.
9729 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9731 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
9732 punt for user-aligned variables.
9734 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9736 * tree-chrec.c (chrec_contains_symbols): Return true for
9739 2018-01-05 Sudakshina Das <sudi.das@arm.com>
9742 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
9743 of (x|y) == x for BICS pattern.
9745 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9747 PR tree-optimization/83605
9748 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
9749 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
9752 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
9754 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
9755 * config/epiphany/rtems.h: New file.
9757 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9758 Uros Bizjak <ubizjak@gmail.com>
9761 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
9762 QIreg_operand instead of register_operand predicate.
9763 * config/i386/i386.c (ix86_rop_should_change_byte_p,
9764 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
9765 comments instead of -fmitigate[-_]rop.
9767 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9770 * cgraphunit.c (symbol_table::compile): Switch to text_section
9771 before calling assembly_start debug hook.
9772 * run-rtl-passes.c (run_rtl_passes): Likewise.
9775 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9777 * tree-vrp.c (extract_range_from_binary_expr_1): Check
9778 range_int_cst_p rather than !symbolic_range_p before calling
9779 extract_range_from_multiplicative_op_1.
9781 2018-01-04 Jeff Law <law@redhat.com>
9783 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
9784 redundant test in assertion.
9786 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9788 * doc/rtl.texi: Document machine_mode wrapper classes.
9790 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9792 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
9795 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9797 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
9798 the VEC_PERM_EXPR fold to fail.
9800 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9803 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
9804 to switched_sections.
9806 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9809 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
9812 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
9815 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
9816 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
9818 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9821 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
9822 is BLKmode and bitpos not zero or mode change is needed.
9824 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9827 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
9830 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
9833 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
9834 instead of MULT rtx. Update all corresponding splitters.
9836 (*ssub<modesuffix>): Ditto.
9838 (*cmp_sadd_di): Update split patterns.
9839 (*cmp_sadd_si): Ditto.
9840 (*cmp_sadd_sidi): Ditto.
9841 (*cmp_ssub_di): Ditto.
9842 (*cmp_ssub_si): Ditto.
9843 (*cmp_ssub_sidi): Ditto.
9844 * config/alpha/predicates.md (const23_operand): New predicate.
9845 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
9846 Look for ASHIFT, not MULT inner operand.
9847 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
9849 2018-01-04 Martin Liska <mliska@suse.cz>
9851 PR gcov-profile/83669
9852 * gcov.c (output_intermediate_file): Add version to intermediate
9854 * doc/gcov.texi: Document new field 'version' in intermediate
9855 file format. Fix location of '-k' option of gcov command.
9857 2018-01-04 Martin Liska <mliska@suse.cz>
9860 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
9862 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9864 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
9866 2018-01-03 Martin Sebor <msebor@redhat.com>
9868 PR tree-optimization/83655
9869 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
9870 checking calls with invalid arguments.
9872 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9874 * tree-vect-stmts.c (vect_get_store_rhs): New function.
9875 (vectorizable_mask_load_store): Delete.
9876 (vectorizable_call): Return false for masked loads and stores.
9877 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
9878 instead of gimple_assign_rhs1.
9879 (vectorizable_load): Handle IFN_MASK_LOAD.
9880 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
9882 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9884 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
9886 (vectorizable_mask_load_store): ...here.
9887 (vectorizable_load): ...and here.
9889 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9891 * tree-vect-stmts.c (vect_build_all_ones_mask)
9892 (vect_build_zero_merge_argument): New functions, split out from...
9893 (vectorizable_load): ...here.
9895 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9897 * tree-vect-stmts.c (vect_check_store_rhs): New function,
9899 (vectorizable_mask_load_store): ...here.
9900 (vectorizable_store): ...and here.
9902 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9904 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
9906 (vectorizable_mask_load_store): ...here.
9908 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9910 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
9911 (vect_model_store_cost): Take a vec_load_store_type instead of a
9913 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
9914 (vect_model_store_cost): Take a vec_load_store_type instead of a
9916 (vectorizable_mask_load_store): Update accordingly.
9917 (vectorizable_store): Likewise.
9918 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
9920 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9922 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
9923 IFN_MASK_LOAD calls here rather than...
9924 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
9926 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9927 Alan Hayward <alan.hayward@arm.com>
9928 David Sherwood <david.sherwood@arm.com>
9930 * expmed.c (extract_bit_field_1): For vector extracts,
9931 fall back to extract_bit_field_as_subreg if vec_extract
9934 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9935 Alan Hayward <alan.hayward@arm.com>
9936 David Sherwood <david.sherwood@arm.com>
9938 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
9939 they are variable or constant sized.
9940 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
9941 slots for constant-sized data.
9943 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9944 Alan Hayward <alan.hayward@arm.com>
9945 David Sherwood <david.sherwood@arm.com>
9947 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
9948 handling COND_EXPRs with boolean comparisons, try to find a better
9949 basis for the mask type than the boolean itself.
9951 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9953 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
9954 is calculated and how it can be overridden.
9955 * genmodes.c (max_bitsize_mode_any_mode): New variable.
9956 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
9958 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
9961 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9962 Alan Hayward <alan.hayward@arm.com>
9963 David Sherwood <david.sherwood@arm.com>
9965 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
9966 Remove the mode argument.
9967 (aarch64_simd_valid_immediate): Remove the mode and inverse
9969 * config/aarch64/iterators.md (bitsize): New iterator.
9970 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
9971 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
9972 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
9973 aarch64_simd_valid_immediate.
9974 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
9975 (aarch64_reg_or_bic_imm): Likewise.
9976 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
9977 with an insn_type enum and msl with a modifier_type enum.
9978 Replace element_width with a scalar_mode. Change the shift
9979 to unsigned int. Add constructors for scalar_float_mode and
9980 scalar_int_mode elements.
9981 (aarch64_vect_float_const_representable_p): Delete.
9982 (aarch64_can_const_movi_rtx_p)
9983 (aarch64_simd_scalar_immediate_valid_for_move)
9984 (aarch64_simd_make_constant): Update call to
9985 aarch64_simd_valid_immediate.
9986 (aarch64_advsimd_valid_immediate_hs): New function.
9987 (aarch64_advsimd_valid_immediate): Likewise.
9988 (aarch64_simd_valid_immediate): Remove mode and inverse
9989 arguments. Rewrite to use the above. Use const_vec_duplicate_p
9990 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
9991 and aarch64_float_const_representable_p on the result.
9992 (aarch64_output_simd_mov_immediate): Remove mode argument.
9993 Update call to aarch64_simd_valid_immediate and use of
9994 simd_immediate_info.
9995 (aarch64_output_scalar_simd_mov_immediate): Update call
9998 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9999 Alan Hayward <alan.hayward@arm.com>
10000 David Sherwood <david.sherwood@arm.com>
10002 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
10003 (mode_nunits): Likewise CONST_MODE_NUNITS.
10004 * machmode.def (ADJUST_NUNITS): Document.
10005 * genmodes.c (mode_data::need_nunits_adj): New field.
10006 (blank_mode): Update accordingly.
10007 (adj_nunits): New variable.
10008 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
10010 (emit_mode_size_inline): Set need_bytesize_adj for all modes
10011 listed in adj_nunits.
10012 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
10013 listed in adj_nunits. Don't emit case statements for such modes.
10014 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
10015 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
10016 nothing if adj_nunits is nonnull.
10017 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
10018 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
10019 (emit_mode_fbit): Update use of print_maybe_const_decl.
10020 (emit_move_size): Likewise. Treat the array as non-const
10022 (emit_mode_adjustments): Handle adj_nunits.
10024 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10026 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
10027 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
10028 (VECTOR_MODES): Use it.
10029 (make_vector_modes): Take the prefix as an argument.
10031 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10032 Alan Hayward <alan.hayward@arm.com>
10033 David Sherwood <david.sherwood@arm.com>
10035 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
10036 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
10037 for MODE_VECTOR_BOOL.
10038 * machmode.def (VECTOR_BOOL_MODE): Document.
10039 * genmodes.c (VECTOR_BOOL_MODE): New macro.
10040 (make_vector_bool_mode): New function.
10041 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
10043 * lto-streamer-in.c (lto_input_mode_table): Likewise.
10044 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
10046 * stor-layout.c (int_mode_for_mode): Likewise.
10047 * tree.c (build_vector_type_for_mode): Likewise.
10048 * varasm.c (output_constant_pool_2): Likewise.
10049 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
10050 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
10051 for MODE_VECTOR_BOOL.
10052 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
10053 of mode class checks.
10054 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
10055 instead of a list of mode class checks.
10056 (expand_vector_scalar_condition): Likewise.
10057 (type_for_widest_vector_mode): Handle BImode as an inner mode.
10059 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10060 Alan Hayward <alan.hayward@arm.com>
10061 David Sherwood <david.sherwood@arm.com>
10063 * machmode.h (mode_size): Change from unsigned short to
10065 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
10066 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10067 or if measurement_type is not polynomial.
10068 (fixed_size_mode::includes_p): Check for constant-sized modes.
10069 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
10070 return a poly_uint16 rather than an unsigned short.
10071 (emit_mode_size): Change the type of mode_size from unsigned short
10072 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
10073 (emit_mode_adjustments): Cope with polynomial vector sizes.
10074 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10076 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10078 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
10079 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
10080 * caller-save.c (setup_save_areas): Likewise.
10081 (replace_reg_with_saved_mem): Likewise.
10082 * calls.c (emit_library_call_value_1): Likewise.
10083 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
10084 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
10085 (gen_lowpart_for_combine): Likewise.
10086 * convert.c (convert_to_integer_1): Likewise.
10087 * cse.c (equiv_constant, cse_insn): Likewise.
10088 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
10089 (cselib_subst_to_values): Likewise.
10090 * dce.c (word_dce_process_block): Likewise.
10091 * df-problems.c (df_word_lr_mark_ref): Likewise.
10092 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
10093 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
10094 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
10095 (rtl_for_decl_location): Likewise.
10096 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
10097 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
10098 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
10099 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
10100 (expand_expr_real_1): Likewise.
10101 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
10102 (pad_below): Likewise.
10103 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10104 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
10105 * ira.c (get_subreg_tracking_sizes): Likewise.
10106 * ira-build.c (ira_create_allocno_objects): Likewise.
10107 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
10108 (ira_sort_regnos_for_alter_reg): Likewise.
10109 * ira-costs.c (record_operand_costs): Likewise.
10110 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
10111 (resolve_simple_move): Likewise.
10112 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
10113 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
10114 (lra_constraints): Likewise.
10115 (CONST_POOL_OK_P): Reject variable-sized modes.
10116 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
10117 (add_pseudo_to_slot, lra_spill): Likewise.
10118 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10119 * optabs-query.c (get_best_extraction_insn): Likewise.
10120 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10121 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
10122 (expand_mult_highpart, valid_multiword_target_p): Likewise.
10123 * recog.c (offsettable_address_addr_space_p): Likewise.
10124 * regcprop.c (maybe_mode_change): Likewise.
10125 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
10126 * regrename.c (build_def_use): Likewise.
10127 * regstat.c (dump_reg_info): Likewise.
10128 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
10129 (find_reloads, find_reloads_subreg_address): Likewise.
10130 * reload1.c (eliminate_regs_1): Likewise.
10131 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
10132 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
10133 (simplify_binary_operation_1, simplify_subreg): Likewise.
10134 * targhooks.c (default_function_arg_padding): Likewise.
10135 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
10136 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
10137 (verify_gimple_assign_ternary): Likewise.
10138 * tree-inline.c (estimate_move_cost): Likewise.
10139 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10140 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
10141 (get_address_cost_ainc): Likewise.
10142 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
10143 (vect_supportable_dr_alignment): Likewise.
10144 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10145 (vectorizable_reduction): Likewise.
10146 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
10147 (vectorizable_operation, vectorizable_load): Likewise.
10148 * tree.c (build_same_sized_truth_vector_type): Likewise.
10149 * valtrack.c (cleanup_auto_inc_dec): Likewise.
10150 * var-tracking.c (emit_note_insn_var_location): Likewise.
10151 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
10152 (ADDR_VEC_ALIGN): Likewise.
10154 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10155 Alan Hayward <alan.hayward@arm.com>
10156 David Sherwood <david.sherwood@arm.com>
10158 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
10160 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10161 or if measurement_type is polynomial.
10162 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
10163 * combine.c (make_extraction): Likewise.
10164 * dse.c (find_shift_sequence): Likewise.
10165 * dwarf2out.c (mem_loc_descriptor): Likewise.
10166 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
10167 (extract_bit_field, extract_low_bits): Likewise.
10168 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
10169 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
10170 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
10171 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
10172 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10173 * reload.c (find_reloads): Likewise.
10174 * reload1.c (alter_reg): Likewise.
10175 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
10176 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
10177 * tree-if-conv.c (predicate_mem_writes): Likewise.
10178 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
10179 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
10180 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
10181 * valtrack.c (dead_debug_insert_temp): Likewise.
10182 * varasm.c (mergeable_constant_section): Likewise.
10183 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
10185 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10186 Alan Hayward <alan.hayward@arm.com>
10187 David Sherwood <david.sherwood@arm.com>
10189 * expr.c (expand_assignment): Cope with polynomial mode sizes
10190 when assigning to a CONCAT.
10192 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10193 Alan Hayward <alan.hayward@arm.com>
10194 David Sherwood <david.sherwood@arm.com>
10196 * machmode.h (mode_precision): Change from unsigned short to
10198 (mode_to_precision): Return a poly_uint16 rather than an unsigned
10200 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
10201 or if measurement_type is not polynomial.
10202 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
10203 in which the mode is already known to be a scalar_int_mode.
10204 * genmodes.c (emit_mode_precision): Change the type of mode_precision
10205 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
10207 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10208 for GET_MODE_PRECISION.
10209 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10210 for GET_MODE_PRECISION.
10211 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
10213 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
10214 (expand_field_assignment, make_extraction): Likewise.
10215 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
10216 (get_last_value): Likewise.
10217 * convert.c (convert_to_integer_1): Likewise.
10218 * cse.c (cse_insn): Likewise.
10219 * expr.c (expand_expr_real_1): Likewise.
10220 * lra-constraints.c (simplify_operand_subreg): Likewise.
10221 * optabs-query.c (can_atomic_load_p): Likewise.
10222 * optabs.c (expand_atomic_load): Likewise.
10223 (expand_atomic_store): Likewise.
10224 * ree.c (combine_reaching_defs): Likewise.
10225 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
10226 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
10227 * tree.h (type_has_mode_precision_p): Likewise.
10228 * ubsan.c (instrument_si_overflow): Likewise.
10230 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10231 Alan Hayward <alan.hayward@arm.com>
10232 David Sherwood <david.sherwood@arm.com>
10234 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
10235 polynomial numbers of units.
10236 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
10237 (valid_vector_subparts_p): New function.
10238 (build_vector_type): Remove temporary shim and take the number
10239 of units as a poly_uint64 rather than an int.
10240 (build_opaque_vector_type): Take the number of units as a
10241 poly_uint64 rather than an int.
10242 * tree.c (build_vector_from_ctor): Handle polynomial
10243 TYPE_VECTOR_SUBPARTS.
10244 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
10245 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
10246 (build_vector_from_val): If the number of units is variable,
10247 use build_vec_duplicate_cst for constant operands and
10248 VEC_DUPLICATE_EXPR otherwise.
10249 (make_vector_type): Remove temporary is_constant ().
10250 (build_vector_type, build_opaque_vector_type): Take the number of
10251 units as a poly_uint64 rather than an int.
10252 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
10254 * cfgexpand.c (expand_debug_expr): Likewise.
10255 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
10256 (store_constructor, expand_expr_real_1): Likewise.
10257 (const_scalar_mask_from_tree): Likewise.
10258 * fold-const-call.c (fold_const_reduction): Likewise.
10259 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
10260 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
10261 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
10262 (fold_relational_const): Likewise.
10263 (native_interpret_vector): Likewise. Change the size from an
10264 int to an unsigned int.
10265 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
10266 TYPE_VECTOR_SUBPARTS.
10267 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
10268 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
10269 duplicating a non-constant operand into a variable-length vector.
10270 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
10271 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
10272 * ipa-icf.c (sem_variable::equals): Likewise.
10273 * match.pd: Likewise.
10274 * omp-simd-clone.c (simd_clone_subparts): Likewise.
10275 * print-tree.c (print_node): Likewise.
10276 * stor-layout.c (layout_type): Likewise.
10277 * targhooks.c (default_builtin_vectorization_cost): Likewise.
10278 * tree-cfg.c (verify_gimple_comparison): Likewise.
10279 (verify_gimple_assign_binary): Likewise.
10280 (verify_gimple_assign_ternary): Likewise.
10281 (verify_gimple_assign_single): Likewise.
10282 * tree-pretty-print.c (dump_generic_node): Likewise.
10283 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10284 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
10285 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
10286 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
10287 (vect_shift_permute_load_chain): Likewise.
10288 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
10289 (expand_vector_condition, optimize_vector_constructor): Likewise.
10290 (lower_vec_perm, get_compute_type): Likewise.
10291 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10292 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
10293 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
10294 (vect_recog_mask_conversion_pattern): Likewise.
10295 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
10296 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
10297 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10298 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
10299 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
10300 (vectorizable_shift, vectorizable_operation, vectorizable_store)
10301 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
10302 (supportable_widening_operation): Likewise.
10303 (supportable_narrowing_operation): Likewise.
10304 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
10306 * varasm.c (output_constant): Likewise.
10308 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10309 Alan Hayward <alan.hayward@arm.com>
10310 David Sherwood <david.sherwood@arm.com>
10312 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
10313 so that both the length == 3 and length != 3 cases set up their
10314 own permute vectors. Add comments explaining why we know the
10315 number of elements is constant.
10316 (vect_permute_load_chain): Likewise.
10318 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10319 Alan Hayward <alan.hayward@arm.com>
10320 David Sherwood <david.sherwood@arm.com>
10322 * machmode.h (mode_nunits): Change from unsigned char to
10324 (ONLY_FIXED_SIZE_MODES): New macro.
10325 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
10326 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
10327 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
10329 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
10330 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
10331 or if measurement_type is not polynomial.
10332 * genmodes.c (ZERO_COEFFS): New macro.
10333 (emit_mode_nunits_inline): Make mode_nunits_inline return a
10335 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
10336 Use ZERO_COEFFS when emitting initializers.
10337 * data-streamer.h (bp_pack_poly_value): New function.
10338 (bp_unpack_poly_value): Likewise.
10339 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10340 for GET_MODE_NUNITS.
10341 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10342 for GET_MODE_NUNITS.
10343 * tree.c (make_vector_type): Remove temporary shim and make
10344 the real function take the number of units as a poly_uint64
10345 rather than an int.
10346 (build_vector_type_for_mode): Handle polynomial nunits.
10347 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
10348 * emit-rtl.c (const_vec_series_p_1): Likewise.
10349 (gen_rtx_CONST_VECTOR): Likewise.
10350 * fold-const.c (test_vec_duplicate_folding): Likewise.
10351 * genrecog.c (validate_pattern): Likewise.
10352 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
10353 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10354 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
10355 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
10356 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
10357 * rtlanal.c (subreg_get_info): Likewise.
10358 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10359 (vect_grouped_load_supported): Likewise.
10360 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
10361 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
10362 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
10363 (simplify_const_unary_operation, simplify_binary_operation_1)
10364 (simplify_const_binary_operation, simplify_ternary_operation)
10365 (test_vector_ops_duplicate, test_vector_ops): Likewise.
10366 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
10367 instead of CONST_VECTOR_NUNITS.
10368 * varasm.c (output_constant_pool_2): Likewise.
10369 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
10370 explicit-encoded elements in the XVEC for variable-length vectors.
10372 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10374 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
10376 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10377 Alan Hayward <alan.hayward@arm.com>
10378 David Sherwood <david.sherwood@arm.com>
10380 * coretypes.h (fixed_size_mode): Declare.
10381 (fixed_size_mode_pod): New typedef.
10382 * builtins.h (target_builtins::x_apply_args_mode)
10383 (target_builtins::x_apply_result_mode): Change type to
10384 fixed_size_mode_pod.
10385 * builtins.c (apply_args_size, apply_result_size, result_vector)
10386 (expand_builtin_apply_args_1, expand_builtin_apply)
10387 (expand_builtin_return): Update accordingly.
10389 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10391 * cse.c (hash_rtx_cb): Hash only the encoded elements.
10392 * cselib.c (cselib_hash_rtx): Likewise.
10393 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
10394 CONST_VECTOR encoding.
10396 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10397 Jeff Law <law@redhat.com>
10400 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
10401 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
10402 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
10403 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
10406 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
10407 explicitly probe *sp in a noreturn function if there were any callee
10408 register saves or frame pointer is needed.
10410 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10413 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
10414 BLKmode for ternary, binary or unary expressions.
10417 * var-tracking.c (delete_vta_debug_insn): New inline function.
10418 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
10419 insns from get_insns () to NULL instead of each bb separately.
10420 Use delete_vta_debug_insn. No longer static.
10421 (vt_debug_insns_local, variable_tracking_main_1): Adjust
10422 delete_vta_debug_insns callers.
10423 * rtl.h (delete_vta_debug_insns): Declare.
10424 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
10425 instead of variable_tracking_main.
10427 2018-01-03 Martin Sebor <msebor@redhat.com>
10429 PR tree-optimization/83603
10430 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
10431 arguments past the endof the argument list in functions declared
10432 without a prototype.
10433 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
10434 Avoid checking when arguments are null.
10436 2018-01-03 Martin Sebor <msebor@redhat.com>
10439 * doc/extend.texi (attribute const): Fix a typo.
10440 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
10441 issuing -Wsuggest-attribute for void functions.
10443 2018-01-03 Martin Sebor <msebor@redhat.com>
10445 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
10446 offset_int::from instead of wide_int::to_shwi.
10447 (maybe_diag_overlap): Remove assertion.
10448 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
10449 * gimple-ssa-sprintf.c (format_directive): Same.
10450 (parse_directive): Same.
10451 (sprintf_dom_walker::compute_format_length): Same.
10452 (try_substitute_return_value): Same.
10454 2018-01-03 Jeff Law <law@redhat.com>
10456 PR middle-end/83654
10457 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
10458 non-constant residual for zero at runtime and avoid probing in
10459 that case. Reorganize code for trailing problem to mirror handling
10462 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10464 PR tree-optimization/83501
10465 * tree-ssa-strlen.c (get_string_cst): New.
10466 (handle_char_store): Call get_string_cst.
10468 2018-01-03 Martin Liska <mliska@suse.cz>
10470 PR tree-optimization/83593
10471 * tree-ssa-strlen.c: Include tree-cfg.h.
10472 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
10473 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
10474 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
10476 (strlen_dom_walker::before_dom_children): Call
10477 gimple_purge_dead_eh_edges. Dump tranformation with details
10479 (strlen_dom_walker::before_dom_children): Update call by adding
10480 new argument cleanup_eh.
10481 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
10483 2018-01-03 Martin Liska <mliska@suse.cz>
10486 * cif-code.def (VARIADIC_THUNK): New enum value.
10487 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
10490 2018-01-03 Jan Beulich <jbeulich@suse.com>
10492 * sse.md (mov<mode>_internal): Tighten condition for when to use
10493 vmovdqu<ssescalarsize> for TI and OI modes.
10495 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10497 Update copyright years.
10499 2018-01-03 Martin Liska <mliska@suse.cz>
10502 * ipa-visibility.c (function_and_variable_visibility): Skip
10503 functions with noipa attribure.
10505 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10507 * gcc.c (process_command): Update copyright notice dates.
10508 * gcov-dump.c (print_version): Ditto.
10509 * gcov.c (print_version): Ditto.
10510 * gcov-tool.c (print_version): Ditto.
10511 * gengtype.c (create_file): Ditto.
10512 * doc/cpp.texi: Bump @copying's copyright year.
10513 * doc/cppinternals.texi: Ditto.
10514 * doc/gcc.texi: Ditto.
10515 * doc/gccint.texi: Ditto.
10516 * doc/gcov.texi: Ditto.
10517 * doc/install.texi: Ditto.
10518 * doc/invoke.texi: Ditto.
10520 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10522 * vector-builder.h (vector_builder::m_full_nelts): Change from
10523 unsigned int to poly_uint64.
10524 (vector_builder::full_nelts): Update prototype accordingly.
10525 (vector_builder::new_vector): Likewise.
10526 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
10527 (vector_builder::operator ==): Likewise.
10528 (vector_builder::finalize): Likewise.
10529 * int-vector-builder.h (int_vector_builder::int_vector_builder):
10530 Take the number of elements as a poly_uint64 rather than an
10532 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
10533 from unsigned int to poly_uint64.
10534 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
10535 (vec_perm_indices::new_vector): Likewise.
10536 (vec_perm_indices::length): Likewise.
10537 (vec_perm_indices::nelts_per_input): Likewise.
10538 (vec_perm_indices::input_nelts): Likewise.
10539 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
10540 number of elements per input as a poly_uint64 rather than an
10541 unsigned int. Use the original encoding for variable-length
10542 vectors, rather than clamping each individual element.
10543 For the second and subsequent elements in each pattern,
10544 clamp the step and base before clamping their sum.
10545 (vec_perm_indices::series_p): Handle polynomial element counts.
10546 (vec_perm_indices::all_in_range_p): Likewise.
10547 (vec_perm_indices_to_tree): Likewise.
10548 (vec_perm_indices_to_rtx): Likewise.
10549 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
10550 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
10551 (tree_vector_builder::new_binary_operation): Handle polynomial
10552 element counts. Return false if we need to know the number
10553 of elements at compile time.
10554 * fold-const.c (fold_vec_perm): Punt if the number of elements
10555 isn't known at compile time.
10557 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10559 * vec-perm-indices.h (vec_perm_builder): Change element type
10560 from HOST_WIDE_INT to poly_int64.
10561 (vec_perm_indices::element_type): Update accordingly.
10562 (vec_perm_indices::clamp): Handle polynomial element_types.
10563 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10564 (vec_perm_indices::all_in_range_p): Likewise.
10565 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
10567 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
10568 polynomial vec_perm_indices element types.
10569 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
10570 * fold-const.c (fold_vec_perm): Likewise.
10571 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
10572 * tree-vect-generic.c (lower_vec_perm): Likewise.
10573 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10574 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
10575 element type to HOST_WIDE_INT.
10577 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10578 Alan Hayward <alan.hayward@arm.com>
10579 David Sherwood <david.sherwood@arm.com>
10581 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
10582 rather than an int. Use plus_constant.
10583 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
10584 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
10586 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10587 Alan Hayward <alan.hayward@arm.com>
10588 David Sherwood <david.sherwood@arm.com>
10590 * calls.c (emit_call_1, expand_call): Change struct_value_size from
10591 a HOST_WIDE_INT to a poly_int64.
10593 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10594 Alan Hayward <alan.hayward@arm.com>
10595 David Sherwood <david.sherwood@arm.com>
10597 * calls.c (load_register_parameters): Cope with polynomial
10598 mode sizes. Require a constant size for BLKmode parameters
10599 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
10600 forces a parameter to be padded at the lsb end in order to
10601 fill a complete number of words, require the parameter size
10602 to be ordered wrt UNITS_PER_WORD.
10604 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10605 Alan Hayward <alan.hayward@arm.com>
10606 David Sherwood <david.sherwood@arm.com>
10608 * reload1.c (spill_stack_slot_width): Change element type
10609 from unsigned int to poly_uint64_pod.
10610 (alter_reg): Treat mode sizes as polynomial.
10612 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10613 Alan Hayward <alan.hayward@arm.com>
10614 David Sherwood <david.sherwood@arm.com>
10616 * reload.c (complex_word_subreg_p): New function.
10617 (reload_inner_reg_of_subreg, push_reload): Use it.
10619 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10620 Alan Hayward <alan.hayward@arm.com>
10621 David Sherwood <david.sherwood@arm.com>
10623 * lra-constraints.c (process_alt_operands): Reject matched
10624 operands whose sizes aren't ordered.
10625 (match_reload): Refer to this check here.
10627 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10628 Alan Hayward <alan.hayward@arm.com>
10629 David Sherwood <david.sherwood@arm.com>
10631 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
10632 that the mode size is in the set {1, 2, 4, 8, 16}.
10634 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10635 Alan Hayward <alan.hayward@arm.com>
10636 David Sherwood <david.sherwood@arm.com>
10638 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
10639 Use plus_constant instead of gen_rtx_PLUS.
10641 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10642 Alan Hayward <alan.hayward@arm.com>
10643 David Sherwood <david.sherwood@arm.com>
10645 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
10646 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
10647 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
10648 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
10649 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
10650 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
10651 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
10652 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
10653 * config/i386/i386.c (ix86_push_rounding): ...this new function.
10654 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
10656 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
10657 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
10658 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
10659 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
10660 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
10661 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
10662 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
10663 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
10664 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
10665 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
10667 * expr.c (emit_move_resolve_push): Treat the input and result
10668 of PUSH_ROUNDING as a poly_int64.
10669 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
10670 (emit_push_insn): Likewise.
10671 * lra-eliminations.c (mark_not_eliminable): Likewise.
10672 * recog.c (push_operand): Likewise.
10673 * reload1.c (elimination_effects): Likewise.
10674 * rtlanal.c (nonzero_bits1): Likewise.
10675 * calls.c (store_one_arg): Likewise. Require the padding to be
10676 known at compile time.
10678 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10679 Alan Hayward <alan.hayward@arm.com>
10680 David Sherwood <david.sherwood@arm.com>
10682 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
10683 Use plus_constant instead of gen_rtx_PLUS.
10685 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10686 Alan Hayward <alan.hayward@arm.com>
10687 David Sherwood <david.sherwood@arm.com>
10689 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
10690 rather than an int.
10692 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10693 Alan Hayward <alan.hayward@arm.com>
10694 David Sherwood <david.sherwood@arm.com>
10696 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
10697 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
10698 via stack temporaries. Treat the mode size as polynomial too.
10700 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10701 Alan Hayward <alan.hayward@arm.com>
10702 David Sherwood <david.sherwood@arm.com>
10704 * expr.c (expand_expr_real_2): When handling conversions involving
10705 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
10706 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
10707 as a poly_uint64 too.
10709 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10710 Alan Hayward <alan.hayward@arm.com>
10711 David Sherwood <david.sherwood@arm.com>
10713 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
10715 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10716 Alan Hayward <alan.hayward@arm.com>
10717 David Sherwood <david.sherwood@arm.com>
10719 * combine.c (can_change_dest_mode): Handle polynomial
10720 REGMODE_NATURAL_SIZE.
10721 * expmed.c (store_bit_field_1): Likewise.
10722 * expr.c (store_constructor): Likewise.
10723 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
10724 and polynomial REGMODE_NATURAL_SIZE.
10725 (gen_lowpart_common): Likewise.
10726 * reginfo.c (record_subregs_of_mode): Likewise.
10727 * rtlanal.c (read_modify_subreg_p): Likewise.
10729 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10730 Alan Hayward <alan.hayward@arm.com>
10731 David Sherwood <david.sherwood@arm.com>
10733 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
10734 numbers of elements.
10736 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10737 Alan Hayward <alan.hayward@arm.com>
10738 David Sherwood <david.sherwood@arm.com>
10740 * match.pd: Cope with polynomial numbers of vector elements.
10742 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10743 Alan Hayward <alan.hayward@arm.com>
10744 David Sherwood <david.sherwood@arm.com>
10746 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
10747 in a POINTER_PLUS_EXPR.
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 * omp-simd-clone.c (simd_clone_subparts): New function.
10754 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
10755 (ipa_simd_modify_function_body): Likewise.
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 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
10762 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
10763 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
10764 (expand_vector_condition, vector_element): Likewise.
10765 (subparts_gt): New function.
10766 (get_compute_type): Use subparts_gt.
10767 (count_type_subparts): Delete.
10768 (expand_vector_operations_1): Use subparts_gt instead of
10769 count_type_subparts.
10771 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10772 Alan Hayward <alan.hayward@arm.com>
10773 David Sherwood <david.sherwood@arm.com>
10775 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
10776 (vect_compile_time_alias): ...this new function. Do the calculation
10777 on poly_ints rather than trees.
10778 (vect_prune_runtime_alias_test_list): Update call accordingly.
10780 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10781 Alan Hayward <alan.hayward@arm.com>
10782 David Sherwood <david.sherwood@arm.com>
10784 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
10786 (vect_schedule_slp_instance): Likewise.
10788 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10789 Alan Hayward <alan.hayward@arm.com>
10790 David Sherwood <david.sherwood@arm.com>
10792 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
10793 constant and extern definitions for variable-length vectors.
10794 (vect_get_constant_vectors): Note that the number of units
10795 is known to be constant.
10797 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10798 Alan Hayward <alan.hayward@arm.com>
10799 David Sherwood <david.sherwood@arm.com>
10801 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
10802 of units as polynomial. Choose between WIDE and NARROW based
10805 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10806 Alan Hayward <alan.hayward@arm.com>
10807 David Sherwood <david.sherwood@arm.com>
10809 * tree-vect-stmts.c (simd_clone_subparts): New function.
10810 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
10812 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10813 Alan Hayward <alan.hayward@arm.com>
10814 David Sherwood <david.sherwood@arm.com>
10816 * tree-vect-stmts.c (vectorizable_call): Treat the number of
10817 vectors as polynomial. Use build_index_vector for
10818 IFN_GOMP_SIMD_LANE.
10820 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10821 Alan Hayward <alan.hayward@arm.com>
10822 David Sherwood <david.sherwood@arm.com>
10824 * tree-vect-stmts.c (get_load_store_type): Treat the number of
10825 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
10826 for variable-length vectors.
10827 (vectorizable_mask_load_store): Treat the number of units as
10828 polynomial, asserting that it is constant if the condition has
10829 already been enforced.
10830 (vectorizable_store, vectorizable_load): Likewise.
10832 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10833 Alan Hayward <alan.hayward@arm.com>
10834 David Sherwood <david.sherwood@arm.com>
10836 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
10837 of units as polynomial. Punt if we can't tell at compile time
10838 which vector contains the final result.
10840 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10841 Alan Hayward <alan.hayward@arm.com>
10842 David Sherwood <david.sherwood@arm.com>
10844 * tree-vect-loop.c (vectorizable_induction): Treat the number
10845 of units as polynomial. Punt on SLP inductions. Use an integer
10846 VEC_SERIES_EXPR for variable-length integer reductions. Use a
10847 cast of such a series for variable-length floating-point
10850 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10851 Alan Hayward <alan.hayward@arm.com>
10852 David Sherwood <david.sherwood@arm.com>
10854 * tree.h (build_index_vector): Declare.
10855 * tree.c (build_index_vector): New function.
10856 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
10857 of units as polynomial, forcibly converting it to a constant if
10858 vectorizable_reduction has already enforced the condition.
10859 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
10860 to create a {1,2,3,...} vector.
10861 (vectorizable_reduction): Treat the number of units as polynomial.
10862 Choose vectype_in based on the largest scalar element size rather
10863 than the smallest number of units. Enforce the restrictions
10866 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10867 Alan Hayward <alan.hayward@arm.com>
10868 David Sherwood <david.sherwood@arm.com>
10870 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
10871 number of units as polynomial.
10873 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10874 Alan Hayward <alan.hayward@arm.com>
10875 David Sherwood <david.sherwood@arm.com>
10877 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
10878 * target.def (autovectorize_vector_sizes): Return the vector sizes
10879 by pointer, using vector_sizes rather than a bitmask.
10880 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
10881 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
10882 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
10884 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
10885 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
10886 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
10887 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
10888 * omp-general.c (omp_max_vf): Likewise.
10889 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10890 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
10891 * tree-vect-loop.c (vect_analyze_loop): Likewise.
10892 * tree-vect-slp.c (vect_slp_bb): Likewise.
10893 * doc/tm.texi: Regenerate.
10894 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
10896 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
10897 the vector size as a poly_uint64 rather than an unsigned int.
10898 (current_vector_size): Change from an unsigned int to a poly_uint64.
10899 (get_vectype_for_scalar_type): Update accordingly.
10900 * tree.h (build_truth_vector_type): Take the size and number of
10901 units as a poly_uint64 rather than an unsigned int.
10902 (build_vector_type): Add a temporary overload that takes
10903 the number of units as a poly_uint64 rather than an unsigned int.
10904 * tree.c (make_vector_type): Likewise.
10905 (build_truth_vector_type): Take the number of units as a poly_uint64
10906 rather than an unsigned int.
10908 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10909 Alan Hayward <alan.hayward@arm.com>
10910 David Sherwood <david.sherwood@arm.com>
10912 * target.def (get_mask_mode): Take the number of units and length
10913 as poly_uint64s rather than unsigned ints.
10914 * targhooks.h (default_get_mask_mode): Update accordingly.
10915 * targhooks.c (default_get_mask_mode): Likewise.
10916 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
10917 * doc/tm.texi: Regenerate.
10919 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10920 Alan Hayward <alan.hayward@arm.com>
10921 David Sherwood <david.sherwood@arm.com>
10923 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
10924 * omp-general.c (omp_max_vf): Likewise.
10925 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
10926 (expand_omp_simd): Handle polynomial safelen.
10927 * omp-low.c (omplow_simd_context): Add a default constructor.
10928 (omplow_simd_context::max_vf): Change from int to poly_uint64.
10929 (lower_rec_simd_input_clauses): Update accordingly.
10930 (lower_rec_input_clauses): Likewise.
10932 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10933 Alan Hayward <alan.hayward@arm.com>
10934 David Sherwood <david.sherwood@arm.com>
10936 * tree-vectorizer.h (vect_nunits_for_cost): New function.
10937 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
10938 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
10939 (vect_analyze_slp_cost): Likewise.
10940 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
10941 (vect_model_load_cost): Likewise.
10943 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10944 Alan Hayward <alan.hayward@arm.com>
10945 David Sherwood <david.sherwood@arm.com>
10947 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
10948 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
10949 from an unsigned int * to a poly_uint64_pod *.
10950 (calculate_unrolling_factor): New function.
10951 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
10953 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10954 Alan Hayward <alan.hayward@arm.com>
10955 David Sherwood <david.sherwood@arm.com>
10957 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
10958 from an unsigned int to a poly_uint64.
10959 (_loop_vec_info::slp_unrolling_factor): Likewise.
10960 (_loop_vec_info::vectorization_factor): Change from an int
10962 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
10963 (vect_get_num_vectors): New function.
10964 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
10965 (vect_get_num_copies): Use vect_get_num_vectors.
10966 (vect_analyze_data_ref_dependences): Change max_vf from an int *
10967 to an unsigned int *.
10968 (vect_analyze_data_refs): Change min_vf from an int * to a
10970 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10971 than an unsigned HOST_WIDE_INT.
10972 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
10973 (vect_analyze_data_ref_dependence): Change max_vf from an int *
10974 to an unsigned int *.
10975 (vect_analyze_data_ref_dependences): Likewise.
10976 (vect_compute_data_ref_alignment): Handle polynomial vf.
10977 (vect_enhance_data_refs_alignment): Likewise.
10978 (vect_prune_runtime_alias_test_list): Likewise.
10979 (vect_shift_permute_load_chain): Likewise.
10980 (vect_supportable_dr_alignment): Likewise.
10981 (dependence_distance_ge_vf): Take the vectorization factor as a
10982 poly_uint64 rather than an unsigned HOST_WIDE_INT.
10983 (vect_analyze_data_refs): Change min_vf from an int * to a
10985 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
10986 vfm1 as a poly_uint64 rather than an int. Make the same change
10987 for the returned bound_scalar.
10988 (vect_gen_vector_loop_niters): Handle polynomial vf.
10989 (vect_do_peeling): Likewise. Update call to
10990 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
10991 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
10993 * tree-vect-loop.c (vect_determine_vectorization_factor)
10994 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
10995 (vect_get_known_peeling_cost): Likewise.
10996 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
10997 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
10998 (vect_transform_loop): Likewise. Use the lowest possible VF when
10999 updating the upper bounds of the loop.
11000 (vect_min_worthwhile_factor): Make static. Return an unsigned int
11001 rather than an int.
11002 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
11003 polynomial unroll factors.
11004 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
11005 (vect_make_slp_decision): Likewise.
11006 (vect_supported_load_permutation_p): Likewise, and polynomial
11008 (vect_analyze_slp_cost): Handle polynomial vf.
11009 (vect_slp_analyze_node_operations): Likewise.
11010 (vect_slp_analyze_bb_1): Likewise.
11011 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11012 than an unsigned HOST_WIDE_INT.
11013 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
11014 (vectorizable_load): Handle polynomial vf.
11015 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
11017 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
11019 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11020 Alan Hayward <alan.hayward@arm.com>
11021 David Sherwood <david.sherwood@arm.com>
11023 * match.pd: Handle bit operations involving three constants
11024 and try to fold one pair.
11026 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11028 * tree-vect-loop-manip.c: Include gimple-fold.h.
11029 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
11030 niters_maybe_zero parameters. Handle other cases besides a step of 1.
11031 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
11032 Add a path that uses a step of VF instead of 1, but disable it
11034 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
11035 and niters_no_overflow parameters. Update calls to
11036 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
11037 Create a new SSA name if the latter choses to use a ste other
11038 than zero, and return it via niters_vector_mult_vf_var.
11039 * tree-vect-loop.c (vect_transform_loop): Update calls to
11040 vect_do_peeling, vect_gen_vector_loop_niters and
11041 slpeel_make_loop_iterate_ntimes.
11042 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
11043 (vect_gen_vector_loop_niters): Update declarations after above changes.
11045 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
11047 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
11048 128-bit round to integer instructions.
11049 (ceil<mode>2): Likewise.
11050 (btrunc<mode>2): Likewise.
11051 (round<mode>2): Likewise.
11053 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11055 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
11056 unaligned VSX load/store on P8/P9.
11057 (expand_block_clear): Allow the use of unaligned VSX
11058 load/store on P8/P9.
11060 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11062 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
11064 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
11065 swap associated with both a load and a store.
11067 2018-01-02 Andrew Waterman <andrew@sifive.com>
11069 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
11070 * config/riscv/riscv.md (clear_cache): Use it.
11072 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
11074 * web.c: Remove out-of-date comment.
11076 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11078 * expr.c (fixup_args_size_notes): Check that any existing
11079 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
11080 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
11081 (emit_single_push_insn): ...here.
11083 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11085 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
11086 (const_vector_encoded_nelts): New function.
11087 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
11088 (const_vector_int_elt, const_vector_elt): Declare.
11089 * emit-rtl.c (const_vector_int_elt_1): New function.
11090 (const_vector_elt): Likewise.
11091 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
11092 of CONST_VECTOR_ELT.
11094 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11096 * expr.c: Include rtx-vector-builder.h.
11097 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
11098 directly on the tree encoding.
11099 (const_vector_from_tree): Likewise.
11100 * optabs.c: Include rtx-vector-builder.h.
11101 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
11102 sequence of "u" values.
11103 * vec-perm-indices.c: Include rtx-vector-builder.h.
11104 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
11105 directly on the vec_perm_indices encoding.
11107 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11109 * doc/rtl.texi (const_vector): Describe new encoding scheme.
11110 * Makefile.in (OBJS): Add rtx-vector-builder.o.
11111 * rtx-vector-builder.h: New file.
11112 * rtx-vector-builder.c: Likewise.
11113 * rtl.h (rtx_def::u2): Add a const_vector field.
11114 (CONST_VECTOR_NPATTERNS): New macro.
11115 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
11116 (CONST_VECTOR_DUPLICATE_P): Likewise.
11117 (CONST_VECTOR_STEPPED_P): Likewise.
11118 (CONST_VECTOR_ENCODED_ELT): Likewise.
11119 (const_vec_duplicate_p): Check for a duplicated vector encoding.
11120 (unwrap_const_vec_duplicate): Likewise.
11121 (const_vec_series_p): Check for a non-duplicated vector encoding.
11122 Say that the function only returns true for integer vectors.
11123 * emit-rtl.c: Include rtx-vector-builder.h.
11124 (gen_const_vec_duplicate_1): Delete.
11125 (gen_const_vector): Call gen_const_vec_duplicate instead of
11126 gen_const_vec_duplicate_1.
11127 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
11128 (gen_const_vec_duplicate): Use rtx_vector_builder.
11129 (gen_const_vec_series): Likewise.
11130 (gen_rtx_CONST_VECTOR): Likewise.
11131 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
11132 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11133 Build a new vector rather than modifying a CONST_VECTOR in-place.
11134 (handle_special_swappables): Update call accordingly.
11135 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
11136 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11137 Build a new vector rather than modifying a CONST_VECTOR in-place.
11138 (handle_special_swappables): Update call accordingly.
11140 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11142 * simplify-rtx.c (simplify_const_binary_operation): Use
11143 CONST_VECTOR_ELT instead of XVECEXP.
11145 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11147 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
11148 the selector elements to be different from the data elements
11149 if the selector is a VECTOR_CST.
11150 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
11151 ssizetype for the selector.
11153 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11155 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
11156 before testing each element individually.
11157 * tree-vect-generic.c (lower_vec_perm): Likewise.
11159 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11161 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
11162 * selftest-run-tests.c (selftest::run_tests): Call it.
11163 * vector-builder.h (vector_builder::operator ==): New function.
11164 (vector_builder::operator !=): Likewise.
11165 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
11166 (vec_perm_indices::all_from_input_p): New function.
11167 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11168 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
11169 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
11170 instead of reading the VECTOR_CST directly. Detect whether both
11171 vector inputs are the same before constructing the vec_perm_indices,
11172 and update the number of inputs argument accordingly. Use the
11173 utility functions added above. Only construct sel2 if we need to.
11175 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11177 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
11178 the broadcast of the low byte.
11179 (expand_mult_highpart): Use an explicit encoding for the permutes.
11180 * optabs-query.c (can_mult_highpart_p): Likewise.
11181 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
11182 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11183 (vectorizable_bswap): Likewise.
11184 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
11185 explicit encoding for the power-of-2 permutes.
11186 (vect_permute_store_chain): Likewise.
11187 (vect_grouped_load_supported): Likewise.
11188 (vect_permute_load_chain): Likewise.
11190 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11192 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
11193 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
11194 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
11195 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11196 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
11197 (vect_gen_perm_mask_any): Likewise.
11199 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11201 * int-vector-builder.h: New file.
11202 * vec-perm-indices.h: Include int-vector-builder.h.
11203 (vec_perm_indices): Redefine as an int_vector_builder.
11204 (auto_vec_perm_indices): Delete.
11205 (vec_perm_builder): Redefine as a stand-alone class.
11206 (vec_perm_indices::vec_perm_indices): New function.
11207 (vec_perm_indices::clamp): Likewise.
11208 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
11209 (vec_perm_indices::new_vector): New function.
11210 (vec_perm_indices::new_expanded_vector): Update for new
11211 vec_perm_indices class.
11212 (vec_perm_indices::rotate_inputs): New function.
11213 (vec_perm_indices::all_in_range_p): Operate directly on the
11214 encoded form, without computing elided elements.
11215 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
11216 encoding. Update for new vec_perm_indices class.
11217 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
11218 the given vec_perm_builder.
11219 (expand_vec_perm_var): Update vec_perm_builder constructor.
11220 (expand_mult_highpart): Use vec_perm_builder instead of
11221 auto_vec_perm_indices.
11222 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
11223 vec_perm_indices instead of auto_vec_perm_indices. Use a single
11224 or double series encoding as appropriate.
11225 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
11226 vec_perm_indices instead of auto_vec_perm_indices.
11227 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11228 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11229 (vect_permute_store_chain): Likewise.
11230 (vect_grouped_load_supported): Likewise.
11231 (vect_permute_load_chain): Likewise.
11232 (vect_shift_permute_load_chain): Likewise.
11233 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11234 (vect_transform_slp_perm_load): Likewise.
11235 (vect_schedule_slp_instance): Likewise.
11236 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11237 (vectorizable_mask_load_store): Likewise.
11238 (vectorizable_bswap): Likewise.
11239 (vectorizable_store): Likewise.
11240 (vectorizable_load): Likewise.
11241 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
11242 vec_perm_indices instead of auto_vec_perm_indices. Use
11243 tree_to_vec_perm_builder to read the vector from a tree.
11244 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
11245 vec_perm_builder instead of a vec_perm_indices.
11246 (have_whole_vector_shift): Use vec_perm_builder and
11247 vec_perm_indices instead of auto_vec_perm_indices. Leave the
11248 truncation to calc_vec_perm_mask_for_shift.
11249 (vect_create_epilog_for_reduction): Likewise.
11250 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
11251 from auto_vec_perm_indices to vec_perm_indices.
11252 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11253 instead of changing individual elements.
11254 (aarch64_vectorize_vec_perm_const): Use new_vector to install
11255 the vector in d.perm.
11256 * config/arm/arm.c (expand_vec_perm_d::perm): Change
11257 from auto_vec_perm_indices to vec_perm_indices.
11258 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11259 instead of changing individual elements.
11260 (arm_vectorize_vec_perm_const): Use new_vector to install
11261 the vector in d.perm.
11262 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
11263 Update vec_perm_builder constructor.
11264 (rs6000_expand_interleave): Likewise.
11265 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
11266 (rs6000_expand_interleave): Likewise.
11268 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11270 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
11271 to qimode could truncate the indices.
11272 * optabs.c (expand_vec_perm_var): Likewise.
11274 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11276 * Makefile.in (OBJS): Add vec-perm-indices.o.
11277 * vec-perm-indices.h: New file.
11278 * vec-perm-indices.c: Likewise.
11279 * target.h (vec_perm_indices): Replace with a forward class
11281 (auto_vec_perm_indices): Move to vec-perm-indices.h.
11282 * optabs.h: Include vec-perm-indices.h.
11283 (expand_vec_perm): Delete.
11284 (selector_fits_mode_p, expand_vec_perm_var): Declare.
11285 (expand_vec_perm_const): Declare.
11286 * target.def (vec_perm_const_ok): Replace with...
11287 (vec_perm_const): ...this new hook.
11288 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
11289 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
11290 * doc/tm.texi: Regenerate.
11291 * optabs.def (vec_perm_const): Delete.
11292 * doc/md.texi (vec_perm_const): Likewise.
11293 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
11294 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
11295 expand_vec_perm for constant permutation vectors. Assert that
11296 the mode of variable permutation vectors is the integer equivalent
11297 of the mode that is being permuted.
11298 * optabs-query.h (selector_fits_mode_p): Declare.
11299 * optabs-query.c: Include vec-perm-indices.h.
11300 (selector_fits_mode_p): New function.
11301 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
11302 is defined, instead of checking whether the vec_perm_const_optab
11303 exists. Use targetm.vectorize.vec_perm_const instead of
11304 targetm.vectorize.vec_perm_const_ok. Check whether the indices
11305 fit in the vector mode before using a variable permute.
11306 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
11307 vec_perm_indices instead of an rtx.
11308 (expand_vec_perm): Replace with...
11309 (expand_vec_perm_const): ...this new function. Take the selector
11310 as a vec_perm_indices rather than an rtx. Also take the mode of
11311 the selector. Update call to shift_amt_for_vec_perm_mask.
11312 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
11313 Use vec_perm_indices::new_expanded_vector to expand the original
11314 selector into bytes. Check whether the indices fit in the vector
11315 mode before using a variable permute.
11316 (expand_vec_perm_var): Make global.
11317 (expand_mult_highpart): Use expand_vec_perm_const.
11318 * fold-const.c: Includes vec-perm-indices.h.
11319 * tree-ssa-forwprop.c: Likewise.
11320 * tree-vect-data-refs.c: Likewise.
11321 * tree-vect-generic.c: Likewise.
11322 * tree-vect-loop.c: Likewise.
11323 * tree-vect-slp.c: Likewise.
11324 * tree-vect-stmts.c: Likewise.
11325 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
11327 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
11328 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
11329 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
11330 (aarch64_vectorize_vec_perm_const): ...this new function.
11331 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11332 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11333 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
11334 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
11335 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11336 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11337 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
11339 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
11340 check for NEON modes.
11341 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
11342 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
11343 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
11344 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
11346 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
11347 the old VEC_PERM_CONST conditions.
11348 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
11349 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
11350 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
11351 (ia64_vectorize_vec_perm_const_ok): Merge into...
11352 (ia64_vectorize_vec_perm_const): ...this new function.
11353 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
11354 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
11355 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
11356 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
11357 * config/mips/mips.c (mips_expand_vec_perm_const)
11358 (mips_vectorize_vec_perm_const_ok): Merge into...
11359 (mips_vectorize_vec_perm_const): ...this new function.
11360 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
11361 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
11362 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
11363 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
11364 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
11365 (rs6000_expand_vec_perm_const): Delete.
11366 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
11368 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11369 (altivec_expand_vec_perm_const_le): Take each operand individually.
11370 Operate on constant selectors rather than rtxes.
11371 (altivec_expand_vec_perm_const): Likewise. Update call to
11372 altivec_expand_vec_perm_const_le.
11373 (rs6000_expand_vec_perm_const): Delete.
11374 (rs6000_vectorize_vec_perm_const_ok): Delete.
11375 (rs6000_vectorize_vec_perm_const): New function.
11376 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11377 an element count and rtx array.
11378 (rs6000_expand_extract_even): Update call accordingly.
11379 (rs6000_expand_interleave): Likewise.
11380 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
11381 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
11382 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
11383 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
11384 (rs6000_expand_vec_perm_const): Delete.
11385 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11386 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11387 (altivec_expand_vec_perm_const_le): Take each operand individually.
11388 Operate on constant selectors rather than rtxes.
11389 (altivec_expand_vec_perm_const): Likewise. Update call to
11390 altivec_expand_vec_perm_const_le.
11391 (rs6000_expand_vec_perm_const): Delete.
11392 (rs6000_vectorize_vec_perm_const_ok): Delete.
11393 (rs6000_vectorize_vec_perm_const): New function. Remove stray
11394 reference to the SPE evmerge intructions.
11395 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11396 an element count and rtx array.
11397 (rs6000_expand_extract_even): Update call accordingly.
11398 (rs6000_expand_interleave): Likewise.
11399 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
11400 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
11402 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11404 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11406 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
11407 vector mode and that that mode matches the mode of the data
11409 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
11410 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
11411 directly using expand_vec_perm_1 when forcing selectors into
11413 (expand_vec_perm_var): New function, split out from expand_vec_perm.
11415 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11417 * optabs-query.h (can_vec_perm_p): Delete.
11418 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
11419 * optabs-query.c (can_vec_perm_p): Split into...
11420 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
11421 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
11422 particular selector is valid.
11423 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11424 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11425 (vect_grouped_load_supported): Likewise.
11426 (vect_shift_permute_load_chain): Likewise.
11427 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11428 (vect_transform_slp_perm_load): Likewise.
11429 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11430 (vectorizable_bswap): Likewise.
11431 (vect_gen_perm_mask_checked): Likewise.
11432 * fold-const.c (fold_ternary_loc): Likewise. Don't take
11433 implementations of variable permutation vectors into account
11434 when deciding which selector to use.
11435 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
11436 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
11437 with a false third argument.
11438 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
11439 to test whether the constant selector is valid and can_vec_perm_var_p
11440 to test whether a variable selector is valid.
11442 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11444 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
11445 * optabs-query.c (can_vec_perm_p): Likewise.
11446 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
11447 instead of vec_perm_indices.
11448 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
11449 (vect_gen_perm_mask_checked): Likewise,
11450 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
11451 (vect_gen_perm_mask_checked): Likewise,
11453 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11455 * optabs-query.h (qimode_for_vec_perm): Declare.
11456 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
11457 (qimode_for_vec_perm): ...this new function.
11458 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
11460 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11462 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
11463 does not have a conditional at the top.
11465 2018-01-02 Richard Biener <rguenther@suse.de>
11467 * ipa-inline.c (big_speedup_p): Fix expression.
11469 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11472 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
11475 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11479 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
11480 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
11481 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
11482 cond_taken_branch_cost 3->4.
11484 2018-01-01 Jakub Jelinek <jakub@redhat.com>
11486 PR tree-optimization/83581
11487 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
11488 TODO_cleanup_cfg if any changes have been made.
11490 PR middle-end/83608
11491 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
11492 convert_modes if target mode has the right side, but different mode
11495 PR middle-end/83609
11496 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
11497 last argument when extracting from CONCAT. If either from_real or
11498 from_imag is NULL, use expansion through memory. If result is not
11499 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
11500 the parts directly to inner mode, if even that fails, use expansion
11503 PR middle-end/83623
11504 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
11505 check for bswap in mode rather than HImode and use that in expand_unop
11508 Copyright (C) 2018 Free Software Foundation, Inc.
11510 Copying and distribution of this file, with or without modification,
11511 are permitted in any medium without royalty provided the copyright
11512 notice and this notice are preserved.