1 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
5 2018-04-26 Tom de Vries <tom@codesourcery.com>
8 * config/nvptx/nvptx.c (verify_neutering_jumps)
9 (verify_neutering_labels): New function
10 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
12 2018-04-26 Tom de Vries <tom@codesourcery.com>
15 * config/nvptx/nvptx.c (needs_neutering_p): New function.
16 (nvptx_single): Use needs_neutering_p to skip over insns that do not
19 2018-04-26 Richard Biener <rguenther@suse.de>
20 Tom de Vries <tom@codesourcery.com>
23 * lto-streamer-out.c (output_function): Fixup loops if required to match
24 discovery done in the reader.
26 2018-04-26 Richard Biener <rguenther@suse.de>
28 PR tree-optimization/85116
29 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
30 have a loop exit from the single latch predecessor. Remove
31 case of header with just condition.
32 (ch_base::copy_headers): Exclude infinite loops from any
34 (pass_ch::execute): Record exits.
36 2018-04-26 Richard Biener <rguenther@suse.de>
38 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
39 prologue cost vector and pass it to vect_get_load_cost.
40 (vect_get_peeling_costs_all_drs): Likewise.
41 (vect_peeling_hash_get_lowest_cost): Likewise.
42 (vect_enhance_data_refs_alignment): Likewise.
44 2018-04-26 Richard Biener <rguenther@suse.de>
47 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
48 checking of integer<->pointer conversions.
49 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
50 sign-/zero-extending pointer types.
51 (expand_omp_for_static_chunk): Likewise.
53 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
54 Jean Lee <xiaoyur347@gmail.com>
56 * config/mips/mips.c (mips_asan_shadow_offset): New function.
57 (TARGET_ASAN_SHADOW_OFFSET): Define.
58 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
59 true for -fsanitize=address.
61 2018-04-25 Mark Wielaard <mark@klomp.org>
63 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
66 2018-04-25 Jakub Jelinek <jakub@redhat.com>
68 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
69 than "alu", remove explicit "memory" and "imm_disp" attributes.
70 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
73 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
74 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
77 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
80 * config/i386/i386.c (ix86_expand_builtin): Change memory
81 operand to XI, extend p0 to Pmode.
82 * config/i386/i386.md: Change unspec volatile and operand
83 1 mode to XI, change operand 0 mode to P.
85 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
87 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
88 GET_MODE_MASK before any checking.
89 (nds32_can_use_bset_p): Likewise.
90 (nds32_can_use_btgl_p): Likewise.
92 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
94 * config/nds32/nds32-doubleword.md: New define_split pattern for
95 illegal register number.
97 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
99 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
101 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
103 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
105 2018-04-25 Richard Biener <rguenther@suse.de>
107 * lto-streamer.h (LTO_major_version): Bump to 8.
109 2018-04-25 Jakub Jelinek <jakub@redhat.com>
111 * BASE-VER: Set to 9.0.0.
113 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
115 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
116 in __abskf2 and __powikf2.
118 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
121 * config/aarch64/constraints.md (Usg, Usj): New constraints.
122 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
123 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
124 Use the above on operand 2. Reindent.
125 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
127 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
130 * common/config/i386/i386-common.c (ix86_handle_option): Don't
132 * config/i386/i386.opt (mcet): Removed.
133 * doc/install.texi: Remove -mcet documentation.
134 * doc/invoke.texi: Likewise.
136 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
139 * doc/install.texi: Remove -mcet from bootstrap-cet.
141 2018-04-24 Jakub Jelinek <jakub@redhat.com>
144 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
145 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
149 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
150 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
151 containing a CONST_VECTOR.
153 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
155 * doc/install.texi: Update newlib dependency for nvptx.
157 2018-04-24 Jakub Jelinek <jakub@redhat.com>
160 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
161 instead of INTVAL when shifting x left.
163 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
165 PR tree-optimization/85478
166 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
167 vect_grouped_store_supported for single element vectors.
169 2018-04-24 Richard Biener <rguenther@suse.de>
172 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
173 load cost increase to the case of non-constant step.
175 2018-04-24 Jakub Jelinek <jakub@redhat.com>
178 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
179 destination if any_malformed_asm.
181 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
184 * expr.c (store_field): In the bitfield case, if the value comes from
185 a function call and is returned in registers by means of a PARALLEL,
186 do not change the mode of the temporary unless BLKmode and VOIDmode.
188 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
190 PR rtl-optimization/85423
191 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
192 dependencies to debug insns when the previous insn is non-debug.
194 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
196 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
197 enums into a single definition.
198 (fls): Fix predicates and printing.
201 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
203 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
204 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
205 and short u6 immediate.
206 (check_if_valid_sleep_operand): Remove.
207 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
209 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
211 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
212 flag_always_save_lp condition.
213 * config/nds32/nds32.opt (malways-save-lp): New option.
215 2018-04-22 Shiva Chen <shiva0217@gmail.com>
217 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
218 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
219 * config/nds32/nds32.h
220 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
221 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
223 2018-04-22 Shiva Chen <shiva0217@gmail.com>
225 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
226 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
228 2018-04-22 Shiva Chen <shiva0217@gmail.com>
229 Chung-Ju Wu <jasonwucj@gmail.com>
231 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
233 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
234 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
236 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
238 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
240 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
242 * config/nds32/nds32-protos.h (nds32_data_alignment,
243 nds32_local_alignment): Declare.
244 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
245 nds32_local_alignment): New functions.
246 (TARGET_CONSTANT_ALIGNMENT): Define.
247 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
249 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
251 * config/nds32/nds32.c
252 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
253 (TARGET_MODES_TIEABLE_P): Likewise.
255 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
257 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
260 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
261 Chung-Ju Wu <jasonwucj@gmail.com>
263 * config/nds32/constants.md (unspec_volatile_element): Add enum values
264 for unaligned access.
265 * config/nds32/nds32-intrinsic.c: Implementation of expanding
267 * config/nds32/nds32-intrinsic.md: Likewise.
268 * config/nds32/nds32_intrinsic.h: Likewise.
269 * config/nds32/nds32.h (nds32_builtins): Likewise.
270 * config/nds32/nds32.opt (munaligned-access): New option.
271 * config/nds32/nds32.c (nds32_asm_file_start): Display
272 flag_unaligned_access status.
274 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
276 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
277 -mno-relax is present.
278 * config/riscv/linux.h (LINK_SPEC): Ditto.
280 2018-04-20 Martin Sebor <msebor@redhat.com>
283 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
285 (gimple_fold_builtin_stxcpy_chk): Same.
286 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
288 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
291 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
292 __powikf2 when long double is IEEE 128-bit.
294 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
296 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
297 step to make sure stack always aligned.
299 2018-04-20 Carl Love <cel@us.ibm.com>
302 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
305 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
306 Tom de Vries <tom@codesourcery.com>
309 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
310 Emit insns for calls too.
311 (nvptx_find_par): Always look for worker-level predecessor insn.
312 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
314 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
315 (nvptx_process_pars): Propagate frames for calls.
317 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
320 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
322 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
323 (ix86_handle_option): Don't handle OPT_mibt.
324 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
326 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
328 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
329 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
330 (ix86_target_macros): Define __CET__ with flag_cf_protection
332 * config/i386/i386.c (isa2_opts): Remove -mibt.
333 * config/i386/i386.h (TARGET_IBT): Removed.
334 (TARGET_IBT_P): Likewise.
335 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
336 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
337 * config/i386/i386.opt (mcet): Update help message.
340 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
341 -mcet as an alias for -mshstk.
343 2018-04-20 Richard Biener <rguenther@suse.de>
346 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
347 complexity by forcing a single use of the multiply operand.
349 2018-04-20 Martin Jambor <mjambor@suse.cz>
352 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
353 recursion dependency to only apply to non-clones.
355 2018-04-20 Martin Jambor <mjambor@suse.cz>
358 * ipa-cp.c (create_specialized_node): Check that clones of
359 self-recursive edges exist during IPA-CP.
361 2018-04-19 Toon Moene <toon@moene.org>
363 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
366 2018-04-19 Jakub Jelinek <jakub@redhat.com>
368 PR tree-optimization/85467
369 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
370 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
371 VECTOR_CST element to type.
373 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
376 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
377 * config/i386/i386.md (builtin_setjmp_setup): Removed.
378 (builtin_longjmp): Likewise.
379 (save_stack_nonlocal): New pattern.
380 (restore_stack_nonlocal): Likewise.
382 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
385 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
386 Replace ASM_OUTPUT_LABEL with fprintf.
388 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
391 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
392 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
393 * config/i386/i386-c.c (ix86_target_macros_internal): Also
394 define __IBT__ and __SHSTK__ for -fcf-protection.
395 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
397 (ix86_trampoline_init): Likewise.
398 (x86_output_mi_thunk): Likewise.
399 (ix86_notrack_prefixed_insn_p): Likewise.
400 (ix86_option_override_internal): Don't disallow -fcf-protection.
401 * config/i386/i386.md (rdssp<mode>): Also enable for
403 (incssp<mode>): Likewise.
404 (nop_endbr): Likewise.
405 * config/i386/i386.opt (mcet): Change help message to built-in
409 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
410 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
411 enable CET built-in functions.
413 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
415 * common/config/i386/i386-common.c
416 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
417 OPTION_MASK_ISA_MOVDIRI_UNSET,
418 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
419 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
420 * config.gcc (movdirintrin.h): New header.
421 * config/i386/cpuid.h (bit_MOVDIRI,
422 bit_MOVDIR64B): New bits.
423 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
425 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
426 (VOID, PVOID, PCVOID)): New function types.
427 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
428 __builtin_ia32_directstoreu_u64,
429 __builtin_ia32_movdir64b): New builtins.
430 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
431 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
433 (ix86_valid_target_attribute_inner_p): Ditto.
434 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
435 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
436 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
437 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
438 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
439 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
440 (movdiri<mode>, movdir64b_<mode>): New.
441 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
442 * config/i386/immintrin.h: Include movdirintrin.h.
443 * config/i386/movdirintrin.h: New file.
444 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
446 2018-04-19 Richard Biener <rguenther@suse.de>
449 * cfg.c (clear_bb_flags): When loop state says we have
450 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
452 2018-04-19 Richard Biener <rguenther@suse.de>
454 PR tree-optimization/84737
455 * tree-vect-data-refs.c (vect_copy_ref_info): New function
456 copying restrict info.
457 (vect_setup_realignment): Use it.
458 * tree-vectorizer.h (vect_copy_ref_info): Declare.
459 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
460 the first DR to all generated stores.
461 (vectorizable_load): Likewise for loads.
463 2018-04-19 Jakub Jelinek <jakub@redhat.com>
465 PR tree-optimization/85446
466 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
467 the integral and pointer types to have the same precision.
469 * doc/install.texi: Document --disable-cet being the default and
472 2018-04-18 Martin Liska <mliska@suse.cz>
474 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
477 2018-04-18 Martin Liska <mliska@suse.cz>
480 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
483 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
484 arguments if they are comparable.
486 2018-04-18 Martin Liska <mliska@suse.cz>
489 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
492 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
495 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
498 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
499 ENDBR after calling __morestack.
501 2018-04-18 David Malcolm <dmalcolm@redhat.com>
504 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
505 by using gcc_base_ver to generate a gcc_driver_version, and use
506 it when generating GCC_DRIVER_NAME.
507 * configure: Regenerate.
509 2018-04-18 Jakub Jelinek <jakub@redhat.com>
512 * config.gcc: Obsolete powerpc*-*-*spe*.
514 2018-04-17 Jakub Jelinek <jakub@redhat.com>
517 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
518 (stabstr_D): Change type of unum from unsigned int to
519 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
522 2018-04-17 Jim Wilson <jimw@sifive.com>
525 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
526 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
527 Set arg_pointer_offset after using pretend_args_size.
529 2018-04-17 Jakub Jelinek <jakub@redhat.com>
531 PR rtl-optimization/85431
532 * dse.c (record_store): Ignore zero width stores.
535 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
536 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
537 __builtin_stack_restore rather than after it.
538 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
539 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
540 argument instead of virtual_dynamic_stack_rtx.
542 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
544 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
546 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
547 Add note to error message to explain internal mapping of overloaded
548 built-in function name to non-overloaded built-in function name.
549 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
552 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
555 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
556 where the inputs overlap with the output.
558 2018-04-17 Jakub Jelinek <jakub@redhat.com>
560 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
561 (=v, v) alternative and explicit "memory" attribute.
562 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
563 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
565 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
566 "sselog1" type instead of "sselog".
567 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
568 "sselog". Remove explicit "memory" attribute.
569 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
570 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
572 (vec_extract_hi_v32hi): Merge all alternatives into one, use
573 "sselog1" type instead of "sselog". Remove explicit "memory"
575 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
576 use "sselog1" type instead of "sselog". Remove explicit "memory"
578 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
579 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
581 (vec_extract_hi_v64qi): Merge all alternatives into one, use
582 "sselog1" type instead of "sselog". Remove explicit "memory"
584 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
585 use "sselog1" type instead of "sselog". Remove explicit "memory"
589 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
592 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
595 2018-04-17 Martin Jambor <mjambor@suse.cz>
598 * ipa-cp.c (create_specialized_node): Call
599 expand_all_artificial_thunks if necessary.
601 2018-04-17 Martin Liska <mliska@suse.cz>
604 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
605 in message, remote space in between '_G' and '('.
607 2018-04-17 Jakub Jelinek <jakub@redhat.com>
610 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
611 avx512f_vmcmp<mode>3<round_saeonly_name>,
612 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
613 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
614 avx512f_rndscale<mode><round_saeonly_name>,
615 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
616 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
617 Use %<iptr>2 instead of %2 for -masm=intel.
618 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
619 avx512f_vcvttss2usi<round_saeonly_name>,
620 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
622 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
623 avx512f_vcvttsd2usi<round_saeonly_name>,
624 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
625 Use %q1 instead of %1 for -masm=intel.
626 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
627 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
628 of %3 for -masm=intel.
629 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
631 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
633 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
635 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
636 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
638 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
640 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
641 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
642 %g1 and one with %0 and %1.
643 (avx512er_vmrcp28<mode><round_saeonly_name>,
644 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
646 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
647 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
648 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
649 of %0 and %{%4%} for -masm=intel.
650 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
651 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
652 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
653 order of %0 and %{%5%}%{z%} for -masm=intel.
655 2018-04-17 Jan Hubicka <jh@suse.cz>
658 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
660 2018-04-17 Martin Liska <mliska@suse.cz>
663 * multiple_target.c (create_dispatcher_calls): Set apostrophes
664 for target_clone error message. Make default implementation
665 clone to be a local declaration.
666 (separate_attrs): Add new argument and check for an empty
668 (expand_target_clones): Handle it.
669 (ipa_target_clone): Make redirection just for target_clones
672 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
673 Tom de Vries <tom@codesourcery.com>
676 * omp-expand.c (expand_oacc_for): Add dummy false branch for
677 tiled basic blocks without omp continue statements.
679 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
682 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
683 vec_extract expression as having side effects to make sure it gets
686 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
689 * config/i386/i386.c (get_builtin_code_for_version): Check
692 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
695 * gcc/config.gcc: Support "skylake".
696 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
698 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
699 (processor_target_table): Add "skylake".
700 (ix86_option_override_internal): Add "skylake".
701 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
702 PROCESSOR_CANNONLAKE.
703 (get_builtin_code_for_version): Fix priority for
704 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
705 PROCESSOR_SKYLAKE-AVX512.
706 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
707 (processor_type): Add PROCESSOR_SKYLAKE.
709 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
710 Jason Merrill <jason@redhat.com>
713 * convert.c (convert_to_integer_1): Use direct recursion for
714 enumeral types and types with a precision less than the number
715 of bits in their mode.
717 2018-04-16 Julia Koval <julia.koval@intel.com>
720 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
721 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
723 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
726 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
727 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
728 and -mno-direct-move.
730 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
733 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
734 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
735 Ensure negative shifts result in {0}.
737 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
739 PR rtl-optimization/79916
740 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
741 regs (if any) to define how to gnerate SD moves when LRA is in
744 2018-04-13 Jakub Jelinek <jakub@redhat.com>
746 PR rtl-optimization/85393
747 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
748 * except.c (expand_dw2_landing_pad_for_region): Make static.
749 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
750 a label and unconditional jump to old_bb, rather than
751 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
754 PR rtl-optimization/85376
755 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
756 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
757 instead of a specific value.
759 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
760 Bin Cheng <bin.cheng@arm.com>
762 PR tree-optimization/82965
763 PR tree-optimization/83991
764 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
765 by_profile_only parameter.
766 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
767 information if the loop was predicted to iterate too many times.
768 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
770 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
773 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
776 2018-04-13 Martin Liska <mliska@suse.cz>
777 Jakub Jelinek <jakub@redhat.com>
780 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
781 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
782 * builtins.c (expand_builtin_memory_copy_args): Use
783 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
784 handle dest_addr == pc_rtx.
786 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
789 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
790 asked to not generate direct moves.
791 (fix_trunc<mode>si2_stfiwx): Similar.
792 (fix_trunc<mode>si2_internal): Similar.
794 2018-04-12 Jakub Jelinek <jakub@redhat.com>
797 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
798 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
799 lookup if dest in some wider mode is known to be const0_rtx and
800 if so, record permanent equivalence for it to be ZERO_EXTEND of
801 the narrower mode destination.
803 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
805 * lto-streamer-out.c (output_function): Revert 259346.
806 * omp-expand.c (expand_oacc_for): Likewise.
808 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
810 PR rtl-optimization/85354
811 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
812 * sel-sched.c (sel_global_init): ... here.
814 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
817 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
818 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
819 mode for PE-COFF targets.
820 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
821 (i386_pe_asm_lto_end): Likewise.
822 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
823 (TARGET_ASM_LTO_END): Likewise.
824 * config/i386/winnt.c (saved_debug_info_level): New static variable.
825 (i386_pe_asm_lto_start): New function.
826 (i386_pe_asm_lto_end): Likewise.
828 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
829 Richard Biener <rguenther@suse.de>
832 * lto-streamer-out.c (output_function): Fix CFG loop state before
834 * omp-expand.c (expand_oacc_for): Handle calls to internal
835 functions like regular functions.
837 2018-04-12 Richard Biener <rguenther@suse.de>
840 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
841 for the early LTO debug to properly generate references to it
842 during DIE emission. Do not re-use that for the skeleton for
844 (dwarf2out_early_finish): Likewise.
846 2018-04-12 Jakub Jelinek <jakub@redhat.com>
850 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
851 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
852 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
853 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
854 and output is a reg, avoid creating invalid lowpart subreg, but
855 instead split into a 512-bit move. Don't split if not AVX512VL,
856 input is xmm16+ reg and output is a mem.
857 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
858 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
859 xmm16+ reg and output is a mem.
861 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
863 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
864 also for flag_dwarf2_cfi_asm.
866 2018-04-12 Jakub Jelinek <jakub@redhat.com>
868 PR rtl-optimization/85342
869 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
870 a bool scalar var inside of the loop instead. Don't try to update
871 recog_data.operand after failed apply_change_group.
873 2018-04-12 Tom de Vries <tom@codesourcery.com>
876 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
877 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
878 array with flexible array member as array without given dimension.
879 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
880 argument for undefined param to true.
882 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
885 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
886 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
887 from PowerPC section.
888 * config/rs6000/sysv4.opt (mcall-): Improve help text.
889 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
890 help text that is too long.
891 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
892 help text that is too long.
893 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
894 help text that is too long.
896 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
898 * config/alpha/alpha.md (stack_probe_internal): Rename
899 from "probe_stack". Update all callers.
901 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
903 PR rtl-optimization/84566
904 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
905 sched_macro_fuse_insns.
907 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
910 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
911 (compute_block_dependences): ... from here.
913 2018-04-11 Jakub Jelinek <jakub@redhat.com>
915 PR tree-optimization/85331
916 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
917 from int to HOST_WIDE_INT.
919 2018-04-11 Martin Jambor <mjambor@suse.cz>
922 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
923 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
924 not the same as the source val.
925 (cgraph_edge_brings_value_p): New parameter.
926 (gather_edges_for_value): Pass destination value to
927 cgraph_edge_brings_value_p.
928 (perhaps_add_new_callers): Likewise.
929 (get_info_about_necessary_edges): Likewise and exclude values brought
930 only by self-recursive edges.
931 (create_specialized_node): Redirect only clones of self-calling edges.
932 (+self_recursive_pass_through_p): New function.
933 (find_more_scalar_values_for_callers_subset): Use it.
934 (find_aggregate_values_for_callers_subset): Likewise.
935 (known_aggs_to_agg_replacement_list): Removed.
936 (decide_whether_version_node): Re-calculate known constants for all
937 remaining context clones.
939 2018-04-11 Richard Biener <rguenther@suse.de>
942 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
943 from early DWARF output.
944 (dwarf2out_early_finish): Output line info unconditionally into
945 early DWARF and add reference to it.
947 2018-04-11 Jakub Jelinek <jakub@redhat.com>
950 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
951 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
952 other than V2DFmode using iptr mode attribute.
953 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
955 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
957 PR rtl-optimization/84659
958 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
960 2018-04-11 Jakub Jelinek <jakub@redhat.com>
963 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
965 (output_loc_list): Pass address of a dummy size variable even in the
966 locview handling loop.
967 (index_location_lists): Add comment on why skip_loc_list_entry can't
970 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
973 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
976 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
979 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
980 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
981 and -mstring-compare-inline-limit.
983 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
986 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
987 for stack clash protection in a register whenever we need it to be in
990 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
992 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
993 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
995 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
998 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
1000 (mlong-double-): Ditto.
1001 * config/rs6000/sysv4.opt (msdata=): Ditto.
1002 (mtls-size=): Ditto.
1004 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
1006 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1007 erroneous entries for
1008 "vector int vec_ldl (int, long int *)", and
1009 "vector unsigned int vec_ldl (int, unsigned long int *)".
1010 Add comments and entries for
1011 "vector bool char vec_ldl (int, bool char *)",
1012 "vector bool short vec_ldl (int, bool short *)",
1013 "vector bool int vec_ldl (int, bool int *)",
1014 "vector bool long long vec_ldl (int, bool long long *)",
1015 "vector pixel vec_ldl (int, pixel *)",
1016 "vector long long vec_ldl (int, long long *)",
1017 "vector unsigned long long vec_ldl (int, unsigned long long *)".
1018 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
1019 type tree bool_long_long_type_node and correct definition of
1020 bool_V2DI_type_node to make reference to this new type tree.
1021 (rs6000_mangle_type): Replace erroneous reference to
1022 bool_long_type_node with bool_long_long_type_node.
1023 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
1024 comments to emphasize sign distinctions for char and int types and
1025 replace RS6000_BTI_bool_long constant with
1026 RS6000_BTI_bool_long_long constant. Also add comment to restrict
1027 use of RS6000_BTI_pixel.
1028 (bool_long_type_node): Remove this macro definition.
1029 (bool_long_long_type_node): New macro definition
1031 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1033 PR rtl-optimization/85300
1034 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
1035 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
1036 simplify_unary_operation fails.
1038 2018-04-10 Martin Liska <mliska@suse.cz>
1040 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
1041 cgraph_edge and ipa_ref.
1043 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1047 * config/i386/sse.md
1048 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
1049 computation of the VEC_MERGE selector from mask.
1050 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
1051 Fix decoding of the VEC_MERGE selector into mask.
1053 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1055 PR tree-optimization/85286
1056 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
1058 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1060 * final.c (final_1): Set insn_last_address as well as
1061 insn_current_address.
1063 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1066 * explow.c (emit_stack_probe): Call validize_mem on memory location
1067 before passing it to gen_probe_stack. Create address operand and
1068 legitimize it for the probe_stack_address case.
1070 2018-04-09 Jan Hubicka <jh@suse.cz>
1073 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
1074 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
1075 * tree.c (free_lang_data_in_type): Fix handling of binfos;
1077 (free_lang_data): Rebuild type inheritance graph.
1079 2018-04-09 Martin Sebor <msebor@redhat.com>
1081 * invoke.texi (-finline-small-functions): Mention other optimization
1083 (-findirect-inlining, -fpartial-inlining): Same.
1084 (-finline-functions-called-once): Same.
1085 (-freorder-blocks-and-partition): Same.
1087 2018-04-09 Jan Hubicka <jh@suse.cz>
1090 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
1091 jumps; choose last target that matches the criteria (i.e.
1092 no partition changes for non-crossing jumps).
1093 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
1094 support for redirecting crossing jumps to non-crossing.
1096 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
1098 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
1099 also for naked functions.
1101 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
1103 * config/arc/arc.md (add_shift): New pattern.
1104 (add_shift2): Likewise.
1105 (sub_shift): Likewise.
1106 (sub_shift_cmp0_noout): Likewise.
1107 (compare_si_ashiftsi): Likewise.
1108 (xbfu_cmp0_noout): New combine pattern.
1109 (xbfu_cmp0"): Likewise.
1110 (movsi_set_cc_insn): Place the predicable variant first.
1111 (commutative_binary_cmp0_noout): Remove clobber.
1112 (commutative_binary_cmp0): New pattern.
1113 (noncommutative_binary_cmp0): Likewise.
1114 (noncommutative_binary_cmp0_noout): Likewise.
1115 (noncommutative_binary_comparison_result_used): Removed.
1116 (rsub_cmp0): New pattern.
1117 (rsub_cmp0_noout): Likewise.
1118 (extzvsi): Changed, keep only meaningful variants.
1119 (SQH, SEZ): New iterators.
1120 (SQH_postfix): New mode attribute.
1121 (SEZ_prefix): New code attribute.
1122 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
1123 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
1124 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
1126 (noncommutative_operator): Check the availability of barrel
1129 2018-04-09 Richard Biener <rguenther@suse.de>
1131 PR tree-optimization/85284
1132 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
1133 Only use the niter constraining form of simple_iv when the exit
1136 2018-04-09 Tom de Vries <tom@codesourcery.com>
1139 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1140 (define_expand "*memory_barrier"): New define_expand.
1141 (define_insn "memory_barrier"): New insn.
1143 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1145 PR rtl-optimization/80463
1146 PR rtl-optimization/83972
1147 PR rtl-optimization/83480
1149 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
1150 correct producer for the insn.
1151 (tidy_control_flow): Fixup seqnos in case of debug insns.
1153 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1155 PR rtl-optimization/83913
1157 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
1158 different sched-times when merging exprs.
1160 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1162 PR rtl-optimization/83962
1164 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
1165 tidy_fallthru_edge and tidy_control_flow.
1167 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1169 PR rtl-optimization/83530
1171 * sel-sched.c (force_next_insn): New global variable.
1172 (remove_insn_for_debug): When force_next_insn is true, also leave only
1173 next insn in the ready list.
1174 (sel_sched_region): When the region wasn't scheduled, make another pass
1175 over it with force_next_insn set to 1.
1177 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1179 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
1181 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1182 for interrupt control.
1183 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1184 functions for interrupt control.
1185 * config/nds32/nds32-intrinsic.md: Likewise.
1186 * config/nds32/nds32_intrinsic.h: Likewise.
1187 * config/nds32/nds32.h (nds32_builtins): Likewise.
1189 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1191 * config/nds32/nds32.c (nds32_init_machine_status,
1192 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
1193 strict_aligned_p field.
1194 (nds32_expand_to_rtl_hook): New function.
1195 (TARGET_EXPAND_TO_RTL_HOOK): Define.
1196 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
1198 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1199 Chung-Ju Wu <jasonwucj@gmail.com>
1201 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
1202 * config/nds32/nds32-n7.md: New file.
1203 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
1204 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
1206 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
1207 * config/nds32/nds32.md (pipeline_model): Add n7.
1208 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
1209 * config/nds32/pipelines.md: Include n7 settings.
1211 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1212 Chung-Ju Wu <jasonwucj@gmail.com>
1214 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
1215 * config/nds32/nds32-e8.md: New file.
1216 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
1217 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
1219 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
1220 * config/nds32/nds32.md (pipeline_model): Add e8.
1221 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
1222 * config/nds32/pipelines.md: Include e8 settings.
1224 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1225 Chung-Ju Wu <jasonwucj@gmail.com>
1227 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
1228 * config/nds32/nds32-n8.md: New file.
1229 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
1230 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
1232 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
1233 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
1234 * config/nds32/nds32.md (pipeline_model): Add n8.
1235 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
1236 * config/nds32/pipelines.md: Include n8 settings.
1238 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1239 Chung-Ju Wu <jasonwucj@gmail.com>
1241 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
1242 * config/nds32/nds32-n9-2r1w.md: New file.
1243 * config/nds32/nds32-n9-3r2w.md: New file.
1244 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
1245 nds32_register_ports): New or modify for cpu n9.
1246 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
1248 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
1249 * config/nds32/nds32-utils.c: New file.
1250 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
1251 TARGET_MUL_SLOW): Define.
1252 * config/nds32/nds32.md (pipeline_model): New attribute.
1253 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
1254 New options that support cpu n9.
1255 * config/nds32/pipelines.md: Include n9 settings.
1256 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
1258 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1260 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
1261 information if necessary.
1262 (output_cond_branch_compare_zero): Likewise.
1263 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
1264 (nds32_target_alignment): Refine for alignment.
1265 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
1266 (FUNCTION_BOUNDARY): Modify.
1267 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
1269 * config/nds32/nds32.opt (malways-align, malign-functions): New.
1271 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1273 * config/nds32/constants.md (unspec_volatile_element): Add values for
1274 TLB operation and data prefetch.
1275 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1276 functions for TLB operation and data prefetch.
1277 * config/nds32/nds32-intrinsic.md: Likewise.
1278 * config/nds32/nds32_intrinsic.h: Likewise.
1279 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
1280 (nds32_print_operand): Likewise.
1281 * config/nds32/nds32.h (nds32_builtins): Likewise.
1283 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
1284 Andrew Pinski <pinsika@gcc.gnu.org>
1287 * match.pd: Use constant_boolean_node of correct type instead of
1288 boolean_true_node or boolean_false_node for simplifying
1289 pointer comparisons to zero.
1291 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1293 PR tree-optimization/80021
1294 * tree.c (verify_type_variant): Make error call in verify_variant_match
1295 translatable and remove final full stop.
1297 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1299 * config/nds32/constants.md (unspec_volatile_element): Add
1300 UNSPEC_VOLATILE_EH_RETURN.
1301 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
1302 nds32_output_stack_pop): Support dwarf exception handling process.
1303 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
1304 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
1305 exception handling process.
1306 (nds32_compute_stack_frame): Likewise.
1307 (nds32_return_addr_rtx): Likewise.
1308 (nds32_initial_elimination_offset): Likewise.
1309 (nds32_expand_prologue): Likewise.
1310 (nds32_expand_epilogue): Likewise.
1311 (nds32_dynamic_chain_address): New function.
1312 * config/nds32/nds32.h (machine_function): Add fields for dwarf
1314 (DYNAMIC_CHAIN_ADDRESS): Define.
1315 (EH_RETURN_DATA_REGNO): Define.
1316 (EH_RETURN_STACKADJ_RTX): Define.
1317 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
1318 patterns for dwarf exception handling.
1320 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1322 * config/nds32/nds32.h: Clean up obsolete macros.
1324 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1326 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1327 Add enum values for particular instructions.
1328 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1329 particular intrinsic functions.
1330 * config/nds32/nds32-intrinsic.md: Likewise.
1331 * config/nds32/nds32_intrinsic.h: Likewise.
1332 * config/nds32/nds32.h (nds32_builtins): Likewise.
1333 * config/nds32/nds32.md (type): Add pbsad and pbsada.
1334 (btst, ave): New patterns for particular instructions.
1336 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1338 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1339 Add enum values for atomic load/store and memory sync.
1340 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
1342 * config/nds32/nds32-intrinsic.md: Likewise.
1343 * config/nds32/nds32_intrinsic.h: Likewise.
1344 * config/nds32/nds32.h (nds32_builtins): Likewise.
1346 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1348 PR tree-optimization/85257
1349 * fold-const.c (native_encode_vector): If not all elts could fit
1350 and off is -1, return 0 rather than offset.
1351 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
1352 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
1353 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
1354 adjust buffer in native_interpret_expr call.
1356 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1358 * config/nds32/constants.md (unspec_volatile_element): Add cache
1359 control enum values.
1360 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
1361 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
1362 * config/nds32/nds32.c (nds32_cctl_names): New.
1363 (nds32_print_operand): Handle cache control register names.
1364 * config/nds32/nds32.h (nds32_builtins): New enum values.
1365 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
1367 * config/nds32/nds32.md (type): Add mmu.
1368 * config/nds32/pipelines.md (simple_insn): Add mmu.
1370 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1372 * config/nds32/nds32.md (type): Remove call.
1373 * config/nds32/pipelines.md (simple_insn): Likewise.
1375 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1377 * config/nds32/constants.md (unspec_volatile_element): Add
1378 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
1379 UNSPEC_VOLATILE_FMFCFG.
1380 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
1381 description for fmfcfg and fmfcsr.
1382 (bdesc_1arg): Add fmtcsr.
1383 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
1384 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
1385 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
1386 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
1387 unspec_fmfcfg): New patterns.
1388 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
1389 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
1390 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
1391 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
1392 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
1393 __nds32__fmfcfg): Define.
1395 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1397 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
1398 intrinsic register names.
1399 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
1400 intrinsic register enum values and macros.
1402 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1404 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
1405 for load/store addressing form.
1406 (nds32_print_operand_address): Likewise.
1408 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
1411 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
1412 based on LABEL_REF. Remove useless assertion.
1413 (pic_address_needs_scratch): Fix formatting.
1414 (sparc_legitimize_pic_address): Minor tweaks.
1415 (sparc_delegitimize_address): Adjust assertion accordingly.
1416 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
1417 into symbolic_operand.
1418 (movsi_high_pic_label_ref): Likewise.
1419 (movsi_lo_sum_pic_label_ref): Likewise.
1420 (movdi_pic_label_ref): Likewise.
1421 (movdi_high_pic_label_ref): Likewise.
1422 (movdi_lo_sum_pic_label_ref): Likewise.
1424 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
1426 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
1427 custom LIB_SPEC setup.
1429 2018-04-06 Ruslan Bukin <br@bsdpad.com>
1430 Kito Cheng <kito.cheng@gmail.com>
1432 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
1433 * config/riscv/freebsd.h: New.
1435 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1437 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
1438 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
1441 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1442 Kito Cheng <kito.cheng@gmail.com>
1444 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
1445 nds32_output_call, nds32_symbol_binds_local_p): New functions.
1446 * config/nds32/nds32-protos.h (nds32_output_call,
1447 nds32_output_return): Declare.
1448 * config/nds32/nds32.md: Refine all the call and return patterns.
1450 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1453 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
1454 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
1456 PR rtl-optimization/84872
1457 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
1458 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
1461 2018-04-06 Tamar Christina <tamar.christina@arm.com>
1463 * expr.c (copy_blkmode_to_reg): Revert 254862.
1464 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
1466 2018-04-06 Richard Biener <rguenther@suse.de>
1469 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
1470 after seeing a component reference with an adjacent field. Treat
1471 refs to arrays at struct end of external decls similar to
1472 refs to unconstrained commons.
1474 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1477 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
1478 look through SAVE_EXPRs with non-side-effects argument. Adjust
1480 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
1483 2018-04-06 Richard Biener <rguenther@suse.de>
1486 * alias.c (find_base_term): New wrapper around find_base_term
1487 unwinding CSELIB_VAL_PTR changes.
1488 (find_base_term): Do not restore CSELIB_VAL_PTR during the
1491 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1493 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
1495 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
1496 constant definitions.
1497 ("nop"): lr 0,0 -> nopr r0
1498 ("nop_lr0", "nop_lr1"): New insn definitions.
1500 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1502 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
1503 NDS32_V3PUSH_AVAILABLE_P macro.
1505 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
1506 Chung-Ju Wu <jasonwucj@gmail.com>
1508 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
1509 (nds32*-*-*): Add float and fpu_config into supported_defaults.
1510 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
1511 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
1512 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
1513 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
1514 * config/nds32/constraints.md: New constraints and checking for hard
1515 float configuration.
1516 * config/nds32/iterators.md: New mode iterator and attribute for hard
1517 float configuration.
1518 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
1520 * config/nds32/nds32-fpu.md: New file.
1521 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
1522 deal with hard float code generation.
1523 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
1525 (abi_type, float_reg_number): New enum type.
1526 * config/nds32/nds32-predicates.c: New predicates for hard float.
1527 * config/nds32/nds32-protos.h: Declare functions for hard float.
1528 * config/nds32/nds32.c: Implementation for hard float configuration.
1529 * config/nds32/nds32.h: Definitions for hard float configuration.
1530 * config/nds32/nds32.md: Include hard float machine description and
1531 modify patterns for hard float configuration.
1532 * config/nds32/nds32.opt: New options for hard float configuration.
1533 * config/nds32/predicates.md: New predicates for hard float
1536 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1538 * common/config/nds32/nds32-common.c
1539 (nds32_option_optimization_table): Enable -mreleax-hint by default.
1541 2018-04-05 Jakub Jelinek <jakub@redhat.com>
1544 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
1545 CONSTRUCTOR_ELT (ctor, ...)->value.
1547 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
1550 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
1552 2018-04-05 Tom de Vries <tom@codesourcery.com>
1555 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
1558 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1559 Kito Cheng <kito.cheng@gmail.com>
1561 * config/nds32/constraints.md (U33): Fine-tune checking condition.
1562 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
1563 * config/nds32/nds32.h (nds32_16bit_address_type): Add
1564 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
1566 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1567 Kito Cheng <kito.cheng@gmail.com>
1569 * config/nds32/constraints.md (Ufe): New memory constraint.
1570 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
1571 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
1572 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
1574 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
1575 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
1577 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1579 * config/nds32/nds32.md: Use optimize_size in the condition for
1580 alu-shift instructions.
1582 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1584 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
1586 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1588 * config/nds32/nds32.md (negsi2): Refine pattern.
1590 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1591 Chung-Ju Wu <jasonwucj@gmail.com>
1593 * config/nds32/iterators.md (shift_rotate): New code iterator.
1594 (shift): New code attribute.
1595 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
1596 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
1597 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
1598 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
1599 bit-wise operations.
1600 (andsi3, *andsi3): Ditto.
1601 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
1602 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
1603 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
1604 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
1605 nds32_ior_operand, nds32_xor_operand): New predicates.
1607 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1609 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
1610 (addsi3, subsi3): ... this.
1612 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1614 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
1616 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1618 * config/nds32/nds32.md: Adjust indention.
1620 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1622 * config/nds32/nds32.md (feature): New attribute.
1624 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1626 * config/nds32/nds32.md (subtype): New attribute.
1628 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
1631 * config/arm/arm-builtins.c (arm_expand_builtin): Change
1632 expansion to perform a bitwise AND of the argument followed by a
1633 boolean negation of the result.
1635 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
1637 PR rtl-optimization/84878
1638 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
1639 the basic block. Assert the use reference is not artificial and that
1640 it has an associated insn.
1642 2018-04-04 Michael Matz <matz@suse.de>
1644 * builtins.c (compute_objsize): Pass correct operand
1645 to array_at_struct_end_p.
1647 2018-04-04 Richard Biener <rguenther@suse.de>
1650 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
1651 from contexts for DINFO_LEVEL_TERSE and below.
1653 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1655 * config/nds32/nds32-doubleword.md (move_<mode>): Require
1656 resiter_operand condition.
1657 * config/nds32/nds32.md (*move<mode>): Ditto.
1659 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1660 Monk Chiang <sh.chiang04@gmail.com>
1662 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
1664 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1666 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
1668 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1669 Kito Cheng <kito.cheng@gmail.com>
1671 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
1672 nds32_cond_code_str, output_cond_branch,
1673 output_cond_branch_compare_zero, nds32_expand_cbranch,
1674 nds32_expand_cstore, nds32_expand_movcc,
1675 nds32_output_cbranchsi4_equality_zero,
1676 nds32_output_cbranchsi4_equality_reg,
1677 nds32_output_cbranchsi4_equality_reg_or_const_int,
1678 nds32_output_cbranchsi4_greater_less_zero: New functions.
1679 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
1680 nds32_expand_cstore, nds32_expand_movcc,
1681 nds32_output_cbranchsi4_equality_zero,
1682 nds32_output_cbranchsi4_equality_reg,
1683 nds32_output_cbranchsi4_equality_reg_or_const_int,
1684 nds32_output_cbranchsi4_greater_less_zero): Declare.
1685 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
1686 nds32_rimm11s_operand): New predicates.
1687 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
1688 * config/nds32/nds32.md: Rewrite all the branch and conditional move
1691 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1693 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
1694 * config/nds32/nds32.md: Ditto.
1695 * config/nds32/pipelines.md: Ditto.
1697 2018-04-04 Richard Biener <rguenther@suse.de>
1699 PR tree-optimization/85168
1700 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
1701 propagating abnormals.
1703 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1705 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
1707 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1708 Kito Cheng <kito.cheng@gmail.com>
1710 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
1711 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
1712 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
1713 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
1714 * config/nds32/nds32.md (sibcall_internal): New.
1715 (sibcall_register): Remove.
1716 (sibcall_immediate): Remove.
1717 (sibcall_value_internal): New.
1718 (sibcall_value_register): Remove.
1719 (sibcall_value_immediate): Remove.
1720 * config/nds32/predicates.md (nds32_general_register_operand): New.
1721 (nds32_call_address_operand): New.
1723 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1725 PR rtl-optimization/85167
1726 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
1727 bb_defs if *split_p, instead preinitialize it to NULL.
1729 PR tree-optimization/85156
1730 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
1731 evaluating the argument multiple times.
1733 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
1735 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
1737 (_mm_cvtpd_ps): Likewise.
1738 (_mm_cvttpd_epi32): Likewise.
1739 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
1740 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
1741 vector, pixel, and bool following altivec.h include.
1743 2018-04-03 Martin Sebor <msebor@redhat.com>
1745 * doc/extend.texi (Common Function Attributes): Clarify.
1746 (const attribute): Likewise.
1747 (pure attribute): Likewise.
1749 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1752 * config/i386/i386.c (ix86_expand_vector_set): Use
1753 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
1755 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
1757 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
1758 instructions when changing rounding bits to preserve precision bits
1759 in the x87 control word.
1761 2018-04-03 Martin Liska <mliska@suse.cz>
1763 PR tree-optimization/82491
1764 * rtl.h (strip_offset_and_add): Replace += suboffset with
1765 poly_uint64 () + suboffset.
1767 2018-03-29 Martin Liska <mliska@suse.cz>
1768 Martin Jambor <mjambor@suse.cz>
1771 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
1772 param_type is not an integral or pointer type.
1774 2018-04-03 Richard Biener <rguenther@suse.de>
1776 * sese.h (recompute_all_dominators): Remove.
1778 2018-04-02 Martin Sebor <msebor@redhat.com>
1780 * doc/invoke.texi (-Wrestrict): Fix typos.
1782 2018-04-02 Jim Wilson <jimw@sifive.com>
1784 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
1785 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
1786 (<optab>di3, <optab>si3_extend): Likewise.
1787 (<optab>si3_mask, <optab>si3_mask_1): New.
1788 (<optab>di3_mask, <optab>di3_mask_1): New.
1789 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
1790 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
1791 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
1793 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
1795 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
1798 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
1800 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
1801 (nds32_canonicalize_comparison): New function.
1803 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1804 Kito Cheng <kito.cheng@gmail.com>
1805 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1807 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
1808 * config/nds32/constants.md (unspec_volatile_element): Add
1809 UNSPEC_VOLATILE_RELAX_GROUP.
1810 * config/nds32/nds32-relax-opt.c: New file.
1811 * config/nds32/nds32-predicates.c
1812 (nds32_symbol_load_store_p): New function.
1813 * config/nds32/nds32-protos.h
1814 (nds32_symbol_load_store_p): Declare function.
1815 (make_pass_nds32_relax_opt): Declare new rtl pass function.
1816 * config/nds32/nds32.c
1817 (nds32_register_pass): New function to register pass.
1818 (nds32_register_passes): New function to register passes.
1819 * config/nds32/nds32.md (relax_group): New pattern.
1820 * config/nds32/nds32.opt (mrelax-hint): New option.
1821 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
1823 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
1825 * config/nds32/t-nds32: Modify files dependency.
1827 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1829 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
1830 (PROFILE_HOOK): Define its implementation.
1832 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1834 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
1835 type and 32-bit size.
1837 2018-04-01 Jakub Jelinek <jakub@redhat.com>
1840 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
1841 (V_128_256): New mode iterator.
1842 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
1843 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
1844 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
1846 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
1847 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
1849 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
1852 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
1853 NaN inputs correctly.
1855 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
1858 * config/rs6000/vsx.md (??r): New mode attribute.
1859 (*vsx_mov<mode>_64bit): Use it.
1860 (*vsx_mov<mode>_32bit): Likewise.
1862 2018-03-30 Martin Sebor <msebor@redhat.com>
1864 PR tree-optimization/84818
1865 * builtins.c (check_access): Use warning_n.
1867 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1870 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
1872 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
1875 2018-03-30 Julia Koval <julia.koval@intel.com>
1878 * x86-tune.def (movx, partial_reg_dependency): Enable for
1881 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
1884 * lra-constraints.c (process_alt_operands): Move setting
1885 this_alternative_matches below.
1887 2018-03-29 Martin Liska <mliska@suse.cz>
1890 * doc/invoke.texi: Document how LTO works with debug info.
1891 Describe auto-load support of binutils. Mention 'x86-64'
1892 as valid option value of -march option.
1894 2018-03-29 Jakub Jelinek <jakub@redhat.com>
1896 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
1899 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
1900 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
1901 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
1904 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1907 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
1908 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
1909 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
1910 for RS6000_BTM_POWERPC64.
1911 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
1912 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
1913 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
1918 2018-03-28 Carl Love <cel@us.ibm.com>
1921 2017-09-27 Carl Love <cel@us.ibm.com>
1923 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
1924 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
1925 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
1928 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1930 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
1931 instead of __vector bool.
1932 (_mm_max_pu8): Likewise.
1933 (_mm_min_pi16): Likewise.
1935 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1938 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
1939 (DIVWEUO): Likewise.
1941 (DIVDEUO): Likewise.
1942 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
1943 DIVWEUO and DIVDEUO.
1944 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
1945 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
1946 (div_extend): Likewise.
1947 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
1949 (__builtin_divweuo): Likewise.
1950 (__builtin_divdeo): Likewise.
1951 (__builtin_divdeuo): Likewise.
1953 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1956 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
1957 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
1959 PR tree-optimization/82004
1960 * gimple-match-head.c (optimize_pow_to_exp): New function.
1961 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
1962 Don't fold to exp if optimize_pow_to_exp is false.
1964 2018-03-28 Martin Liska <mliska@suse.cz>
1967 * calls.c (initialize_argument_information): Fix trailing space.
1968 * common.opt: Fix typo and provide better explanation for
1969 -fsanitize-coverage option.
1970 * config/i386/i386.opt: Fix typo.
1972 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1973 Martin Liska <mliska@suse.cz>
1976 * gimplify.c (asan_poison_variable): Don't do the check for
1977 gimplify_omp_ctxp here.
1978 (gimplify_decl_expr): Do it here.
1979 (gimplify_target_expr): Likewise.
1981 2018-03-28 Martin Liska <mliska@suse.cz>
1984 * config/i386/i386.c (ix86_function_arg_advance): Do not call
1985 chkp_type_bounds_count if MPX is not enabled.
1987 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
1989 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
1991 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
1994 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
1995 function to create the function decl for complex long double
1996 multiply and divide for -mabi=ieeelongdouble.
1997 (init_float128_ieee): Call it.
1999 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
2002 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
2003 -fcf-protection=branch -mibt.
2004 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
2006 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2009 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
2011 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
2014 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
2015 extern array declarations.
2017 2018-03-27 Richard Biener <rguenther@suse.de>
2020 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
2021 explicit single_use checks.
2023 2018-03-27 Richard Biener <rguenther@suse.de>
2025 PR tree-optimization/85082
2026 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
2029 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2031 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
2032 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
2033 Turn on fasynchronous-unwind-tables and funwind-tables.
2035 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
2038 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
2039 (*bmi_blsr_<mode>_ccz): Ditto.
2041 2018-03-26 Tom de Vries <tom@codesourcery.com>
2043 PR tree-optimization/85063
2044 * omp-general.c (offloading_function_p): New function. Factor out
2046 * omp-offload.c (pass_omp_target_link::gate): ... here.
2047 * omp-general.h (offloading_function_p): Declare.
2048 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
2049 with attribute omp declare target for offloading functions.
2051 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
2053 PR tree-optimization/84005
2054 * tree-data-ref.h (get_base_for_alignment): Declare.
2055 * tree-data-ref.c (get_base_for_alignment_1): New function.
2056 (get_base_for_alignment): Likewise.
2057 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
2058 get_base_for_alignment to find a suitable base object, instead
2059 of always using drb->base_address.
2061 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2064 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
2065 known size by default.
2067 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
2070 * lra-constraints.c (process_alt_operands): Don't match BLKmode
2071 and non BLKmode operands.
2073 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2076 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
2077 Clean up attributes.
2079 2018-03-23 Richard Biener <rguenther@suse.de>
2082 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
2083 we are going to emit early debug for LTO.
2085 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2088 * function.c (match_asm_constraints_1): Don't optimize if input
2089 doesn't satisfy general_operand predicate for output's mode.
2092 * alias.c (write_dependence_p): Don't require for x_canonicalized
2093 non-VOIDmode if x has VOIDmode.
2096 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
2097 just don't try to optimize it rather than assert it never happens.
2099 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2101 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
2102 macro expansions for definition of ST_INTERNAL_<mode> and
2103 LD_INTERNAL_<mode> builtins.
2104 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
2106 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
2108 (altivec_expand_st_builtin): Likewise.
2109 (altivec_expand_builtin): Remove calls to deleted functions.
2110 (rs6000_address_for_altivec): Delete this function.
2111 * config/rs6000/vector.md: Remove expands for
2112 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
2114 2018-03-22 Sudakshina Das <sudi.das@arm.com>
2117 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
2118 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
2119 re-computing once computed.
2120 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
2121 (arm_init_machine_status): Initialize
2122 machine->static_chain_stack_bytes.
2124 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2127 * doc/extend.texi: Add four new prototypes for vec_ld.
2128 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
2129 definitions for more logical presentation.
2130 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
2131 entries for V1TI variants of __builtin_altivec_ld builtin.
2132 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
2133 handling of V1TI variant of LVX icode pattern.
2134 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
2135 (rs6000_gimple_fold_builtin): Likewise.
2136 (altivec_init_builtins): Add code to define
2137 __builtin_altivec_lvx_v1ti function.
2139 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2142 * function.c (match_asm_constraints_1): Don't do the optimization
2143 if input isn't a REG, SUBREG, MEM or constant.
2145 2018-03-22 Tom de Vries <tom@codesourcery.com>
2147 PR tree-optimization/84956
2148 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
2149 bb_has_abnormal_pred.
2151 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2154 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
2155 DECL_INITIAL (decl) to decl at the end.
2156 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
2159 2018-03-21 Joseph Myers <joseph@codesourcery.com>
2161 * doc/extend.texi (__builtin_tgmath): Document when complex
2162 integer types are treated as _Complex _Float64.
2164 2018-03-21 Tom de Vries <tom@codesourcery.com>
2166 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
2168 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2170 PR tree-optimization/84960
2171 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
2172 if it is ENTRY block, move them into single succ of ENTRY in that case.
2174 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
2176 PR tree-optimization/84811
2177 * poly-int.h (poly_span_traits): Remove the T3 parameter and
2178 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
2179 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
2180 (known_subrange_p): Update accordingly. Cast each value involved
2181 in the size comparison, rather than casting the result of the
2184 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2186 PR tree-optimization/84982
2187 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
2188 by flipping the least significant bit rather than all bits from
2189 bitpos to bitpos + bitsize - 1.
2191 2018-03-21 Nathan Sidwell <nathan@acm.org>
2193 * doc/extend.texi (Deprecated Features): Remove mention of
2194 long-deleted deprecations.
2196 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2199 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
2200 * configure: Regenerate.
2202 2018-03-21 Tom de Vries <tom@codesourcery.com>
2204 PR tree-optimization/83126
2205 * tree-parloops.c (num_phis): New function.
2206 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
2208 2018-03-21 Nathan Sidwell <nathan@acm.org>
2210 * doc/extend.texi (Deprecated Features): Update deprecated flags,
2211 mention anon-struct/union members and trailing attributes.
2213 2018-03-21 Bin Cheng <bin.cheng@arm.com>
2215 PR tree-optimization/84969
2216 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
2217 builtin memset partitions if they set different rhs values.
2219 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2221 PR rtl-optimization/84989
2222 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
2223 VEC_DUPLICATE with scalar result mode.
2225 2018-03-21 Martin Liska <mliska@suse.cz>
2228 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
2229 not intended return statement.
2231 2018-03-21 Martin Liska <mliska@suse.cz>
2234 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
2235 (chkp_find_bound_slots_1): Limit number of iterations.
2237 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
2240 * Minor grammar fixes for x86 options.
2242 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2245 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
2246 holding REG_CFA_RESTORE notes, instead turn them into a USE.
2248 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
2251 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
2252 (altivec_lvx_<mode>_1op): Likewise.
2253 (altivec_stvx_<mode>_2op): Likewise.
2254 (altivec_stvx_<mode>_1op): Likewise.
2255 (altivec_lvx_<VM2:mode>): New define_expand.
2256 (altivec_stvx_<VM2:mode>): Likewise.
2257 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
2258 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2259 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
2260 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2261 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
2262 (rs6000_gen_lvx): Likewise.
2263 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
2264 (altivec_expand_stv_builtin): Likewise.
2265 (altivec_expand_builtin): Likewise.
2266 * config/rs6000/vector.md: Likewise.
2268 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2271 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
2274 2018-03-20 Richard Biener <rguenther@suse.de>
2277 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
2278 sign-conversions as zero, fall back to standard scalar_stmt
2281 2018-03-20 Martin Liska <mliska@suse.cz>
2284 * predict.c (rebuild_frequencies): Handle case when we have
2285 PROFILE_ABSENT, but flag_guess_branch_prob is false.
2287 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2290 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
2291 flag_section_anchors.
2292 * varasm.c (use_blocks_for_decl_p): Remove hack for
2293 dw2_force_const_mem.
2296 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
2298 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
2299 be created, use lowpart_subreg of operands[0] rather than operands[0]
2301 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
2302 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
2303 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
2304 and n constraint instead of aarch64_shift_imm_di and Usd.
2305 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
2306 (*aarch64_<optab>_reg_minus<mode>3): ... this.
2308 2018-03-20 Sudakshina Das <sudi.das@arm.com>
2311 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
2312 to favor GPR over NEON registers.
2313 (<shift>di3_neon): Likewise.
2315 2018-03-20 Tom de Vries <tom@codesourcery.com>
2318 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
2319 (nvptx_process_pars): Emit bar.sync asap and alap.
2321 2018-03-20 Tom de Vries <tom@codesourcery.com>
2324 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
2325 seen_label if seen_label is already set.
2327 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2330 * config/i386/i386.c (fold_builtin_cpu): For features above 31
2331 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
2332 Use 1U instead of 1. Formatting fixes.
2335 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
2336 instead of TREE_TYPE (s1) for the return value.
2338 2018-03-19 Jakub Jelinek <jakub@redhat.com>
2340 PR tree-optimization/84946
2341 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
2342 bitsize + bitsize in poly_uint64 rather than poly_int64.
2345 * dwarf2asm.c: Include fold-const.c.
2346 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
2347 of decl rather than decl itself.
2349 PR rtl-optimization/84643
2350 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
2352 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
2355 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
2356 calling assemble_variable.
2358 2018-03-19 Sudakshina Das <sudi.das@arm.com>
2361 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
2362 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
2364 2018-03-19 Jim Wilson <jimw@sifive.com>
2367 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
2368 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
2369 (riscv_first_stack_step): Likewise.
2370 (riscv_option_override): Use STACK_BOUNDARY instead of
2372 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
2374 (BIGGEST_ALIGNMENT): Set to 128.
2375 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
2376 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
2379 2018-03-19 Richard Biener <rguenther@suse.de>
2381 PR tree-optimization/84933
2382 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
2383 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
2385 2018-03-19 Richard Biener <rguenther@suse.de>
2387 PR tree-optimization/84859
2388 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
2389 (cond_if_else_store_replacement): Perform sinking operation on
2390 single-store BBs regardless of MAX_STORES_TO_SINK setting.
2391 Generalize what a BB with a single eligible store is.
2393 2018-03-19 Richard Biener <rguenther@suse.de>
2395 PR tree-optimization/84929
2396 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
2397 chrec_is_positive against non-chrec arg.
2399 2018-03-19 Tamar Christina <tamar.christina@arm.com>
2402 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
2404 2018-03-18 Martin Liska <mliska@suse.cz>
2406 PR rtl-optimization/84635
2407 * regrename.c (build_def_use): Use matches_mode only when
2410 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
2412 PR tree-optimization/84913
2413 * tree-vect-loop.c (vectorizable_reduction): Don't try to
2414 vectorize chains of COND_EXPRs.
2416 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2418 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
2420 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2422 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
2424 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2426 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
2428 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2429 Kito Cheng <kito.cheng@gmail.com>
2431 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
2432 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
2433 (nds32_adjust_reg_alloc_order): New function.
2434 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
2436 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
2438 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
2439 nds32_print_operand, nds32_print_operand_address): Use
2440 HOST_WIDE_INT_PRINT_DEC instead.
2442 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2444 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
2446 2018-03-17 Jakub Jelinek <jakub@redhat.com>
2449 * config/i386/i386.c (initial_ix86_tune_features,
2450 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
2452 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
2453 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
2454 rather than 1u << ix86_tune. Formatting fix.
2455 (ix86_option_override_internal): Change ix86_arch_mask from
2456 unsigned int to unsigned HOST_WIDE_INT, initialize to
2457 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
2458 (ix86_function_specific_restore): Likewise.
2460 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2463 * postreload.c (reload_combine_recognize_pattern): Perform
2464 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
2465 truncate_int_for_mode the result for the destination's mode.
2468 * hsa-gen.c (mem_type_for_type): Fix comment typo.
2469 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
2471 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2474 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
2477 * lra-assigns.c (lra_split_hard_reg_for): Don't use
2478 regno_allocno_class_array and sorted_pseudos.
2479 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
2480 insns where regno is used.
2482 2018-03-16 Martin Liska <mliska@suse.cz>
2485 * multiple_target.c (create_dispatcher_calls): Redirect
2486 reference in the symbol table.
2488 2018-03-16 Martin Liska <mliska@suse.cz>
2491 * multiple_target.c (create_dispatcher_calls): Redirect also
2494 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2498 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
2499 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
2500 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
2502 2018-03-16 Julia Koval <julia.koval@intel.com>
2504 * doc/invoke.texi (Skylake Server): Add CLWB.
2505 Cannonlake): Remove CLWB.
2507 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2509 PR tree-optimization/84841
2510 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
2512 (FLOAT_ONE_CONST_TYPE): Define.
2513 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
2514 (sort_by_operand_rank): Put entries with higher constant_type last
2515 rather than first to match comments.
2517 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
2519 * config/nios2/nios2.md (movsi_internal): Fix thinko in
2522 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2525 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
2526 check and use address of target always.
2528 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
2531 * config/i386/i386.c (indirect_thunk_needed): Update comments.
2532 (indirect_thunk_bnd_needed): Likewise.
2533 (indirect_thunks_used): Likewise.
2534 (indirect_thunks_bnd_used): Likewise.
2535 (indirect_return_needed): New.
2536 (indirect_return_bnd_needed): Likewise.
2537 (output_indirect_thunk_function): Add a bool argument for
2539 (output_indirect_thunk_function): Don't generate alias for
2540 function return thunk.
2541 (ix86_code_end): Call output_indirect_thunk_function to generate
2542 function return thunks.
2543 (ix86_output_function_return): Set indirect_return_bnd_needed
2544 and indirect_return_needed instead of indirect_thunk_bnd_needed
2545 and indirect_thunk_needed.
2547 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
2549 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
2550 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
2551 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
2553 2018-03-15 David Malcolm <dmalcolm@redhat.com>
2554 Paul Hua <paul.hua.gm@gmail.com>
2557 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
2559 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
2561 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
2562 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
2565 2018-03-15 Tamar Christina <tamar.christina@arm.com>
2568 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
2569 instead of GET_MODE_SIZE when comparing Units.
2571 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2574 * varasm.c (hash_section): Return an unchangeble hash value
2575 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
2576 Return !aarch64_can_use_per_function_literal_pools_p ().
2578 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2581 * optabs.c (emit_conditional_move): Pass address of cmode's copy
2582 rather than address of cmode as last argument to prepare_cmp_insn.
2584 2018-03-15 Julia Koval <julia.koval@intel.com>
2586 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
2587 F_AVX512VNNI, F_AVX512BITALG): New.
2589 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
2592 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
2593 insn for floating-point loads and stores.
2595 2018-03-14 Carl Love <cel@us.ibm.com>
2597 * config/rs6000/rs6000-c.c: Add macro definitions for
2598 ALTIVEC_BUILTIN_VEC_PERMXOR.
2599 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
2600 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
2601 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
2602 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
2604 * config/doc/extend.texi: Add prototypes for vec_permxor.
2606 2018-03-14 David Malcolm <dmalcolm@redhat.com>
2609 * diagnostic-show-locus.c (class layout_point): Convert m_line
2610 from int to linenum_type.
2611 (line_span::comparator): Use linenum "compare" function when
2612 comparing line numbers.
2613 (test_line_span): New function.
2614 (layout_range::contains_point): Convert param "row" from int to
2616 (layout_range::intersects_line_p): Likewise.
2617 (layout::will_show_line_p): Likewise.
2618 (layout::print_source_line): Likewise.
2619 (layout::should_print_annotation_line_p): Likewise.
2620 (layout::print_annotation_line): Likewise.
2621 (layout::print_leading_fixits): Likewise.
2622 (layout::annotation_line_showed_range_p): Likewise.
2623 (struct line_corrections): Likewise for field m_row.
2624 (line_corrections::line_corrections): Likewise for param "row".
2625 (layout::print_trailing_fixits): Likewise.
2626 (layout::get_state_at_point): Likewise.
2627 (layout::get_x_bound_for_row): Likewise.
2628 (layout::print_line): Likewise.
2629 (diagnostic_show_locus): Likewise for locals "last_line" and
2631 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
2632 * input.c (selftest::test_linenum_comparisons): New function.
2633 (selftest::input_c_tests): Call it.
2634 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
2635 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
2636 * selftest.h (ASSERT_GT): New macro.
2637 (ASSERT_GT_AT): New macro.
2638 (ASSERT_LT): New macro.
2639 (ASSERT_LT_AT): New macro.
2641 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
2643 PR rtl-optimization/84780
2644 * combine.c (distribute_links): Don't make a link based on pc_rtx.
2646 2018-03-14 Martin Liska <mliska@suse.cz>
2648 * tree.c (record_node_allocation_statistics): Use
2649 get_stats_node_kind.
2650 (get_stats_node_kind): New function extracted from
2651 record_node_allocation_statistics.
2652 (free_node): Use get_stats_node_kind.
2654 2018-03-14 Richard Biener <rguenther@suse.de>
2656 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
2657 that the value-set of ANTIC_IN doesn't grow.
2660 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2662 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2663 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2665 2018-03-14 Julia Koval <julia.koval@intel.com>
2667 * config.gcc (icelake-client, icelake-server): New.
2669 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
2670 (initial_ix86_arch_features): Ditto.
2671 (PTA_SKYLAKE): Add SGX.
2672 (PTA_ICELAKE): Remove.
2673 (PTA_ICELAKE_CLIENT): New.
2674 (PTA_ICELAKE_SERVER): New.
2675 (ix86_option_override_internal): Split up icelake on icelake client and
2677 (get_builtin_code_for_version): Ditto.
2678 (fold_builtin_cpu): Ditto.
2679 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
2680 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
2681 * config/i386/i386.h (processor_type): Ditto.
2682 * doc/invoke.texi: Ditto.
2684 2018-03-14 Jakub Jelinek <jakub@redhat.com>
2687 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
2688 INTEGER_CST offset, add it together with bitpos / 8 and
2689 sign extend based on POINTER_SIZE.
2693 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
2696 * config/i386/constraints.md (Yc): New register constraint.
2697 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
2698 Use Yc constraint for alternative 2 of operand 0. Remove
2699 preferred_for_speed attribute.
2701 2018-03-14 Richard Biener <rguenther@suse.de>
2703 PR tree-optimization/84830
2704 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
2705 with the old one to avoid oscillations.
2707 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
2710 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
2712 (assign_by_spills): Return a flag of reload assignment failure.
2713 Do not process the reload assignment failures. Do not spill other
2714 reload pseudos if they has the same reg class. Update n if
2716 (lra_assign): Add a return arg. Set up from the result of
2717 assign_by_spills call.
2718 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2719 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2720 usage_insns if it is not NULL.
2721 (spill_hard_reg_in_range): New function.
2722 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2723 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2724 function prototypes.
2725 (lra_assign): Change prototype.
2726 * lra.c (lra): Add code to deal with fails by splitting hard reg
2729 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
2731 * config/riscv/riscv.opt (mrelax): New option.
2732 * config/riscv/riscv.c (riscv_file_start): Emit ".option
2733 "norelax" when riscv_mrelax is disabled.
2734 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
2736 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2739 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
2740 reassociation for int modes.
2742 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2744 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
2745 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
2747 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
2748 * config/aarch64/aarch64-sve.md
2749 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
2750 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
2751 (*extend<mode><Vwide>2): Rename to...
2752 (aarch64_sve_extend<mode><Vwide>2): ...this.
2753 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
2754 renaming the old pattern to...
2755 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
2757 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
2758 define_expand, renaming the old pattern to...
2759 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
2760 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
2761 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
2762 account when deciding which SVE instruction the optab should use.
2763 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
2765 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2767 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
2768 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
2769 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
2770 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
2771 (tlsdesc_small_<mode>): Turn a define_expand and use
2772 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
2773 (tlsdesc_small_advsimd_<mode>): ...this.
2774 (tlsdesc_small_sve_<mode>): New pattern.
2776 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2778 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
2779 (UNSPEC_UMUL_HIGHPART): New constants.
2780 (MUL_HIGHPART): New int iteraor.
2781 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
2782 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
2784 (*<su>mul<mode>3_highpart): New define_insn.
2786 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2789 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2792 2018-03-13 Martin Liska <mliska@suse.cz>
2795 * (sem_item_optimizer::sem_item_optimizer): Initialize new
2797 (sem_item_optimizer::~sem_item_optimizer): Release it.
2798 (sem_item_optimizer::merge_classes): Register variable aliases.
2799 (sem_item_optimizer::fixup_pt_set): New function.
2800 (sem_item_optimizer::fixup_points_to_sets): Likewise.
2801 * ipa-icf.h: Declare new variables and functions.
2803 2018-03-13 Jakub Jelinek <jakub@redhat.com>
2806 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
2807 integer_pow2p@2 and test integer_pow2p in condition.
2808 (A < 0 ? C : 0): Similarly for @1.
2811 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
2812 characters starting at p contain '\0' character, don't look beyond
2816 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
2817 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
2820 * reg-stack.c (change_stack): Change update_end var from int to
2821 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
2822 also call set_block_for_insn on the newly added insns and rescan.
2825 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
2826 on the last operand.
2829 * tree.c (stabilize_reference_1): Return save_expr (e) for
2830 STATEMENT_LIST even if it doesn't have side-effects.
2832 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
2834 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
2836 2018-03-12 Renlin Li <renlin.li@arm.com>
2838 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
2839 aarch64_output_scalar_simd_mov_immediate.
2841 2018-03-12 Martin Sebor <msebor@redhat.com>
2843 PR tree-optimization/83456
2844 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
2845 for perfectly overlapping calls to memcpy.
2846 (gimple_fold_builtin_memory_chk): Same.
2847 (gimple_fold_builtin_strcpy): Handle no-warning.
2848 (gimple_fold_builtin_stxcpy_chk): Same.
2849 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
2851 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2853 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
2854 parameter. Use it for SFmode.
2855 (rs6000_function_arg_advance_1): Adjust.
2856 (rs6000_function_arg): Adjust.
2857 (rs6000_gimplify_va_arg): Pass false for that new parameter.
2859 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2861 PR rtl-optimization/84169
2862 PR rtl-optimization/84780
2863 * combine.c (can_combine_p): Check for a 2-insn combination whether
2864 the destination register is used between the two insns, too.
2866 2018-03-12 Richard Biener <rguenther@suse.de>
2868 PR tree-optimization/84803
2869 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
2870 for refs DR analysis didn't process.
2872 2018-03-12 Richard Biener <rguenther@suse.de>
2874 PR tree-optimization/84777
2875 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
2876 force-vectorize loops ignore whether we are optimizing for size.
2878 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
2880 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
2881 (TARGET_MD_ASM_ADJUST): Define.
2883 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
2884 Kito Cheng <kito.cheng@gmail.com>
2885 Chung-Ju Wu <jasonwucj@gmail.com>
2887 * config/nds32/nds32.c (nds32_compute_stack_frame,
2888 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
2889 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
2890 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
2891 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
2892 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
2893 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
2894 * config/nds32/nds32.md (prologue, epilogue): Use macro
2895 NDS32_V3PUSH_AVAILABLE_P to do checking.
2897 2018-03-11 Jakub Jelinek <jakub@redhat.com>
2900 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
2901 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
2902 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
2903 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
2904 addition of most attributes on !orig_type_die or the attribute not
2905 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
2907 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2908 Chung-Ju Wu <jasonwucj@gmail.com>
2910 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
2912 * config/nds32/nds32.opt (mvh): New option.
2914 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2915 Chung-Ju Wu <jasonwucj@gmail.com>
2917 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
2919 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
2920 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
2923 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2924 Chung-Ju Wu <jasonwucj@gmail.com>
2926 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
2928 * config/nds32/nds32-multiple.md (strlensi): New pattern.
2929 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
2931 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
2932 Kito Cheng <kito.cheng@gmail.com>
2933 Chung-Ju Wu <jasonwucj@gmail.com>
2935 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
2936 UNSPEC_FFMISM and UNSPEC_FLMISM.
2937 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
2938 for ffb, ffmism and flmism.
2939 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
2940 (unspec_ffmism): Ditto.
2941 (unspec_flmism): Ditto.
2942 (nds32_expand_builtin_impl): Check if string extension is available.
2943 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
2944 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
2946 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
2949 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2952 * lra-assigns.c (assign_by_spills): Return a flag of reload
2953 assignment failure. Do not process the reload assignment
2954 failures. Do not spill other reload pseudos if they has the same
2956 (lra_assign): Add a return arg. Set up from the result of
2957 assign_by_spills call.
2958 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2959 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2960 usage_insns if it is not NULL.
2961 (spill_hard_reg_in_range): New function.
2962 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2963 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2964 function prototypes.
2965 (lra_assign): Change prototype.
2966 * lra.c (lra): Add code to deal with fails by splitting hard reg
2969 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
2972 * config/i386/i386.opt: Replace Enforcment with Enforcement.
2974 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
2977 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
2978 (dw_val_node): Add val_symbolic_view.
2979 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
2980 (symview_upper_bound): New.
2981 (new_line_info_table): Initialize symviews_since_reset.
2982 (dwarf2out_source_line): Count symviews_since_reset and set
2983 symview_upper_bound.
2984 (dw_val_equal_p): Handle symview.
2985 (add_AT_symview): New.
2986 (print_dw_val): Handle symview.
2987 (attr_checksum, attr_checksum_ordered): Likewise.
2988 (same_dw_val_p, size_of_die): Likewise.
2989 (value_format, output_die): Likewise.
2990 (add_high_low_attributes): Use add_AT_symview for entry_view.
2991 (dwarf2out_finish): Reset symview_upper_bound, clear
2994 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
2997 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
2998 Add strict argument and use it.
2999 (rs6000_split_multireg_move): Update for new strict argument.
3000 (mem_operand_gpr): Disallow all non-offsettable addresses.
3001 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
3003 2018-03-09 Jakub Jelinek <jakub@redhat.com>
3006 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
3007 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
3008 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
3011 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
3012 decl, use remap_type if we want to use the type.
3014 2018-03-09 Martin Sebor <msebor@redhat.com>
3016 PR tree-optimization/84526
3017 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3019 (builtin_access::generic_overlap): Be prepared to handle non-array
3022 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
3024 PR rtl-optimization/84682
3025 * lra-constraints.c (process_address_1): Check is_address flag
3026 for address constraints.
3027 (process_alt_operands): Likewise.
3028 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
3029 preprocess_constraints.
3030 * recog.h (preprocess_constraints): Add oploc parameter.
3032 * recog.c (preprocess_constraints): Test address_operand for
3033 CT_ADDRESS constraints.
3035 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3038 * lra-assigns.c (assign_by_spills): Return a flag of reload
3039 assignment failure. Do not process the reload assignment
3040 failures. Do not spill other reload pseudos if they has the same
3042 (lra_assign): Add a return arg. Set up from the result of
3043 assign_by_spills call.
3044 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3045 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3046 usage_insns if it is not NULL.
3047 (spill_hard_reg_in_range): New function.
3048 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3049 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3050 function prototypes.
3051 (lra_assign): Change prototype.
3052 * lra.c (lra): Add code to deal with fails by splitting hard reg
3055 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3058 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
3059 Accept complain bool parameter. Only emit errors if it is true.
3060 (arm_parse_cpu_option_name): Likewise.
3061 (arm_target_thumb_only): Adjust callers of the above.
3062 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
3063 prototype to take a default true bool parameter.
3064 (arm_parse_arch_option_name): Likewise.
3066 2018-03-09 David Malcolm <dmalcolm@redhat.com>
3067 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
3071 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
3072 * configure: Regenerate.
3073 * configure.ac ("linker --version-script option"): New.
3074 ("linker soname option"): New.
3076 2018-03-09 Richard Biener <rguenther@suse.de>
3078 PR tree-optimization/84775
3079 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
3080 immediate uses of predicate stmts and mark them modified.
3083 PR tree-optimization/84178
3084 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3086 (version_loop_for_if_conversion): Delay update_ssa call.
3087 (tree_if_conversion): Delay update_ssa until after predicate
3090 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
3093 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
3094 when the function accesses prior frames.
3096 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3099 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
3100 gen_llsym, otherwise call maybe_gen_llsym.
3103 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
3104 has ',' character inside of it.
3106 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3109 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
3110 as clobbering CC_REGNUM.
3112 2018-03-08 Richard Biener <rguenther@suse.de>
3115 * tree-scalar-evolution.c: Include tree-into-ssa.h.
3116 (follow_copies_to_constant): Do not follow SSA names registered
3119 2018-03-08 Richard Biener <rguenther@suse.de>
3121 PR tree-optimization/84178
3122 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3124 (version_loop_for_if_conversion): Delay update_ssa call.
3125 (tree_if_conversion): Delay update_ssa until after predicate
3128 2018-03-08 David Malcolm <dmalcolm@redhat.com>
3130 PR tree-optimization/84178
3131 * tree-if-conv.c (release_bb_predicate): Remove the
3132 the assertion that the stmts have NULL use_ops.
3133 Discard the statements, asserting that they haven't
3134 yet been added to a BB.
3136 2018-03-08 Richard Biener <rguenther@suse.de>
3138 PR tree-optimization/84746
3139 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
3140 (phi_translate): Pass in destination ANTIC_OUT set.
3141 (phi_translate_1): Likewise. For a simplified result lookup
3142 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
3143 (phi_translate_set): Adjust.
3144 (do_pre_regular_insertion): Likewise.
3145 (do_pre_partial_partial_insertion): Likewise.
3147 2018-03-08 Martin Liska <mliska@suse.cz>
3149 PR gcov-profile/84735
3150 * doc/gcov.texi: Document usage of profile files.
3151 * gcov-io.h: Document changes in the format.
3153 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
3157 * dwarf2out.c (struct dw_line_info_table): Update comments for
3159 (FORCE_RESET_NEXT_VIEW): New.
3160 (FORCE_RESETTING_VIEW_P): New.
3161 (RESETTING_VIEW_P): Check for -1 too.
3162 (ZERO_VIEW_P): Likewise.
3163 (new_line_info_table): Force-reset next view.
3164 (dwarf2out_begin_function): Likewise.
3165 (dwarf2out_source_line): Simplify zero_view_p initialization.
3166 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
3167 view directly. Omit view when omitting .loc at line 0.
3169 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3171 PR tree-optimization/84740
3172 * tree-switch-conversion.c (process_switch): Call build_constructors
3173 only if info.phi_count is non-zero.
3175 PR tree-optimization/84739
3176 * tree-tailcall.c (find_tail_calls): Check call arguments against
3177 DECL_ARGUMENTS (current_function_decl) rather than
3178 DECL_ARGUMENTS (func) when checking for tail recursion.
3180 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3182 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
3183 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
3184 Volker Reichelt's entry and add entries for people that perform
3185 GCC fuzzy testing and report numerous bugs.
3187 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
3190 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
3191 readonly data in sdata, if that is disabled.
3192 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
3193 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3194 -mreadonly-in-sdata option.
3196 2018-03-07 Martin Sebor <msebor@redhat.com>
3198 PR tree-optimization/84468
3199 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
3200 basic block when looking for nul assignment.
3202 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
3205 * except.h (output_function_exception_table): Adjust prototype.
3206 * except.c (output_function_exception_table): Remove FNNAME parameter
3207 and add SECTION parameter. Ouput one part of the table at a time.
3208 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
3209 the first part of the exception table and emit unwind directives.
3210 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
3211 (i386_pe_seh_cold_init): Likewise.
3212 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
3213 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
3214 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
3215 (ix86_output_call_insn): Emit a nop in one more case for SEH.
3216 * config/i386/winnt.c: Include except.h.
3217 (struct seh_frame_state): Add reg_offset, after_prologue and
3218 in_cold_section fields.
3219 (i386_pe_seh_end_prologue): Set seh->after_prologue.
3220 (i386_pe_seh_cold_init): New function.
3221 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
3222 to seh->in_cold_section.
3223 (seh_emit_push): Record the offset of the push.
3224 (seh_emit_save): Record the offet of the save.
3225 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
3226 Test seh->after_prologue to disregard the epilogue.
3227 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
3228 (i386_pe_end_cold_function): New function.
3230 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3233 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
3234 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
3237 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
3239 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
3240 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
3243 * multiple_target.c: Include tree-inline.h and intl.h.
3244 (expand_target_clones): Diagnose and fail if node->definition and
3245 !tree_versionable_function_p (node->decl).
3247 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
3249 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
3251 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
3252 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
3253 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
3255 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3258 * combine.c (try_combine): Use reg_or_subregno instead of handling
3259 just paradoxical SUBREGs and REGs.
3261 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
3263 * config/arc/arc.c (arc_finalize_pic): Remove function.
3264 (arc_must_save_register): We use single base PIC register, remove
3265 checks to save/restore the PIC register.
3266 (arc_expand_prologue): Likewise.
3267 * config/arc/arc-protos.h (arc_set_default_type_attributes):
3269 (arc_verify_short): Likewise.
3270 (arc_attr_type): Likewise.
3271 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
3272 (walk_stores): Likewise.
3273 (arc_address_cost): Make it static.
3274 (arc_verify_short): Likewise.
3275 (branch_dest): Likewise.
3276 (arc_attr_type): Likewise.
3277 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
3278 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
3279 (arc_final_prescan_insn): Remove inserting the nops due to
3280 hardware hazards. It is done in reorg step.
3281 (insn_length_variant_t): Remove.
3282 (insn_length_parameters_t): Likewise.
3283 (arc_insn_length_parameters): Likewise.
3284 (arc_get_insn_variants): Likewise.
3285 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
3287 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3290 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
3293 PR tree-optimization/84687
3294 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
3296 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
3298 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3300 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
3301 Rename to ppc_speculation_barrier.
3302 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
3303 __builtin_ppc_speculation_barrier.
3305 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3308 * combine.c (combine_simplify_rtx): Don't try to simplify if
3309 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
3312 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
3314 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
3315 to 32 bytes when compiling for POWER9.
3317 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3320 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
3321 regparm >= 3 with no arg reg available also for calls with
3322 flag_force_indirect_call. Pass decl to ix86_function_regparm.
3325 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
3327 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
3329 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
3332 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
3334 2018-03-05 Richard Biener <rguenther@suse.de>
3336 PR tree-optimization/84486
3337 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
3338 When inserting a __builtin_assume_aligned call set the LHS
3339 SSA name alignment info accordingly.
3341 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
3343 PR tree-optimization/84114
3344 * config/aarch64/aarch64.c (aarch64_reassociation_width)
3345 Avoid reassociation of FLOAT_MODE addition.
3347 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
3349 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
3350 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
3351 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
3352 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
3353 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
3354 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
3355 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
3357 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
3358 __builtin_ia32_wbinvd): New builtins.
3359 (SPECIAL_ARGS2): New.
3360 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
3361 (SPECIAL_ARGS2): New.
3362 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
3363 (ix86_valid_target_attribute_inner_p): Ditto.
3364 (ix86_init_mmx_sse_builtins): Add special_args2.
3365 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
3366 TARGET_WBNOINVD_P): New.
3367 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
3368 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
3369 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
3370 * config/i386/immintrin.h (_wbinvd): New intrinsic.
3371 * config/i386/pconfigintrin.h: New file.
3372 * config/i386/wbnoinvdintrin.h: Ditto.
3373 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
3374 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
3376 2018-03-05 Richard Biener <rguenther@suse.de>
3378 PR tree-optimization/84670
3379 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3381 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3382 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3383 (compute_antic_aux): Only assert the number of values in ANTIC_IN
3384 doesn't grow if all successors (recursively) were visited at least
3387 2018-03-05 Richard Biener <rguenther@suse.de>
3389 PR tree-optimization/84650
3390 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
3391 if executed in the loop pipeline.
3393 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
3395 * doc/configfiles.texi (Configuration Files): Move info about
3396 conditionalizing $target-protos.h to...
3397 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
3398 differs from $target-protos.h.
3400 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
3401 Chung-Ju Wu <jasonwucj@gmail.com>
3403 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
3404 * config/nds32/nds32-multiple.md (setmemsi): Define.
3405 * config/nds32/nds32-memory-manipulation.c
3406 (nds32_gen_dup_4_byte_to_word_value): New.
3407 (emit_setmem_word_loop): New.
3408 (emit_setmem_byte_loop): New.
3409 (nds32_expand_setmem_loop): New.
3410 (nds32_expand_setmem_loop_v3m): New.
3411 (nds32_expand_setmem_unroll): New.
3412 (nds32_expand_setmem): New.
3414 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3415 Chung-Ju Wu <jasonwucj@gmail.com>
3417 * config/nds32/nds32-memory-manipulation.c
3418 (nds32_emit_load_store): New.
3419 (nds32_emit_post_inc_load_store): New.
3420 (nds32_emit_mem_move): New.
3421 (nds32_emit_mem_move_block): New.
3422 (nds32_expand_movmemsi_loop_unknown_size): New.
3423 (nds32_expand_movmemsi_loop_known_size): New.
3424 (nds32_expand_movmemsi_loop): New.
3425 (nds32_expand_movmemsi_unroll): New.
3426 (nds32_expand_movmemqi): Rename ...
3427 (nds32_expand_movmemsi): ... to this.
3428 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
3429 (movmemsi): ... to this.
3430 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
3431 (nds32_expand_movmemsi): ... to this.
3433 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3434 Monk Chiang <sh.chiang04@gmail.com>
3435 Chung-Ju Wu <jasonwucj@gmail.com>
3437 * config/nds32/nds32-protos.h
3438 (nds32_expand_load_multiple): New arguments.
3439 (nds32_expand_store_multiple): Ditto.
3440 (nds32_valid_multiple_load_store): Rename ...
3441 (nds32_valid_multiple_load_store_p): ... to this.
3442 * config/nds32/nds32-memory-manipulation.c
3443 (nds32_expand_load_multiple): Refine implementation.
3444 (nds32_expand_store_multiple): Ditto.
3445 * config/nds32/nds32-multiple.md
3446 (load_multiple): Update nds32_expand_load_multiple interface.
3447 (store_multiple): Update nds32_expand_store_multiple interface.
3448 * config/nds32/nds32-predicates.c
3449 (nds32_valid_multiple_load_store): Rename ...
3450 (nds32_valid_multiple_load_store_p): ... to this and refine
3452 * config/nds32/predicates.md
3453 (nds32_load_multiple_and_update_address_operation): New predicate.
3454 (nds32_store_multiple_and_update_address_operation): New predicate.
3456 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3457 Chung-Ju Wu <jasonwucj@gmail.com>
3459 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
3460 (combo): New attribute.
3461 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
3463 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
3465 * config/nds32/nds32.opt: Change -mcmodel= default value.
3467 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
3468 Monk Chiang <sh.chiang04@gmail.com>
3469 Chung-Ju Wu <jasonwucj@gmail.com>
3471 * config/nds32/constants.md (unspec_element): New enum.
3472 * config/nds32/constraints.md (Umw): New constraint.
3473 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
3474 * config/nds32/nds32-intrinsic.md: Likewise.
3475 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
3476 (nds32_valid_smw_lwm_base_p): New.
3477 (nds32_output_smw_single_word): New.
3478 (nds32_output_lmw_single_word): New.
3479 (nds32_expand_unaligned_load): New.
3480 (nds32_expand_unaligned_store): New.
3481 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
3482 (nds32_output_smw_single_word): Declare.
3483 (nds32_output_lmw_single_word): Declare.
3484 (nds32_expand_unaligned_load): Declare.
3485 (nds32_expand_unaligned_store): Declare.
3486 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
3487 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
3488 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
3489 NDS32_BUILTIN_UASTORE_DW.
3490 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
3493 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
3494 Kito Cheng <kito.cheng@gmail.com>
3495 Chung-Ju Wu <jasonwucj@gmail.com>
3497 * config/nds32/nds32-intrinsic.c
3498 (nds32_expand_builtin_null_ftype_reg): Delete.
3499 (nds32_expand_builtin_reg_ftype_imm): Ditto.
3500 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
3501 (nds32_read_argument): New.
3502 (nds32_legitimize_target): Ditto.
3503 (nds32_legitimize_argument): Ditto.
3504 (nds32_check_constant_argument): Ditto.
3505 (nds32_expand_unop_builtin): Ditto.
3506 (nds32_expand_unopimm_builtin): Ditto.
3507 (nds32_expand_binop_builtin): Ditto.
3508 (nds32_builtin_decl_impl): Ditto.
3509 (builtin_description): Ditto.
3510 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
3511 (nds32_init_builtins_impl): Ditto.
3512 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
3513 (nds32_builtin_decl): New.
3514 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
3515 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
3517 2018-03-02 Jeff Law <law@redhat.com>
3519 * reorg.c (stop_search_p): Handle DEBUG_INSN.
3520 (redundant_insn, fill_simple_delay_slots): Likewise.
3521 (fill_slots_from_thread): Likewise.
3522 * resource.c (mark_referenced_resources): Likewise.
3523 (mark_set_resources, find_dead_or_set_registers): Likewise.
3525 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3527 * substring-locations.h (format_warning_va): Formatting fix for
3529 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
3531 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
3532 * substring-locations.c: Include intl.h.
3533 (format_warning_va): Turned into small wrapper around
3534 format_warning_n_va, renamed to ...
3535 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
3536 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
3538 (format_warning_at_substring_n): New function.
3539 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
3540 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
3541 format_warning_at_substring with just a shorter name instead of
3542 const function pointer.
3543 (fmtwarn_n): New function.
3544 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
3545 appropriate, get rid of all the fmtstr temporaries, move conditionals
3546 with G_() wrapped string literals directly into fmtwarn arguments,
3547 cast dir.len to (int), formatting fixes.
3549 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
3551 * doc/invoke.texi: Remove "Cilk Plus" references.
3553 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3554 Richard Biener <rguenther@suse.de>
3557 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
3558 for error or warning attributes if CALL_FROM_THUNK_P is set.
3561 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3564 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
3565 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
3568 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
3569 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
3571 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
3572 instead of -1U in last predictors element's probability member.
3574 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3577 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3578 arguments if they are comparable.
3580 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3582 PR tree-optimization/84634
3583 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
3584 masks and masked_loop_p with a single loop_masks, making sure it's
3585 null for bb vectorization.
3587 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3589 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3590 (vect_analyze_data_ref_access): Use loop->safe_len rather than
3591 loop->force_vectorize to check whether there is no alias.
3593 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3596 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
3598 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
3600 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
3601 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
3602 instead of a loop around prev_real_insn.
3603 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
3607 * config/i386/i386.c (ix86_print_operand): Use conditional
3608 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
3611 2018-03-02 Richard Biener <rguenther@suse.de>
3613 PR tree-optimization/84427
3614 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
3615 (bitmap_set_subtract_values): Rewrite to handle multiple
3618 (prune_clobbered_mems): Likewise.
3619 (phi_translate): Take edge instead of pred/phiblock.
3620 (phi_translate_1): Likewise.
3621 (phi_translate_set): Likewise. Insert all translated
3622 exprs for a value into the set, keeping possibly multiple
3623 expressions per value.
3624 (compute_antic_aux): Adjust for phi_translate changes.
3625 When intersecting union the expressions and prune those
3626 not in the final value set, keeping possibly multiple
3627 expressions per value. Do not use value-insertion
3628 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
3629 all expressions. Add verification that the value-sets
3630 only shrink during iteration.
3631 (compute_partial_antic_aux): Adjust for the phi_translate changes.
3632 (do_pre_regular_insertion): Likewise.
3633 (do_pre_partial_partial_insertion): Likewise.
3635 2018-03-02 Richard Biener <rguenther@suse.de>
3638 * config/darwin.c (saved_debug_info_level): New static global.
3639 (darwin_asm_lto_start): Disable debug info generation for LTO out.
3640 (darwin_asm_lto_end): Restore debug info generation settings.
3642 2018-03-01 Martin Liska <mliska@suse.cz>
3645 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
3648 2018-03-01 Richard Biener <rguenther@suse.de>
3651 * dwarf2out.c (gen_variable_die): Properly handle late VLA
3652 type annotation with LTO when debug was disabled at compile-time.
3654 2018-03-01 Matthew Fortune <mfortune@gmail.com>
3656 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
3658 (mips_final_postscan_insn): Likewise.
3660 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
3662 PR rtl-optimization/84528
3663 * alias.c (init_alias_target): Add commentary.
3664 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
3665 a unique base value if the frame pointer is not eliminated
3666 to the stack pointer.
3668 2018-03-01 Tom de Vries <tom@codesourcery.com>
3670 PR rtl-optimization/83327
3671 * lra-int.h (hard_regs_spilled_into): Declare.
3672 * lra.c (hard_regs_spilled_into): Define.
3673 (init_reg_info): Init hard_regs_spilled_into.
3674 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
3675 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
3676 (process_bb_lives): Handle hard_regs_spilled_into.
3677 (lra_create_live_ranges_1): Before doing liveness propagation, clear
3678 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
3680 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
3682 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
3683 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
3684 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
3685 * config/rs6000/aix72.h: New file.
3687 2018-02-28 Jakub Jelinek <jakub@redhat.com>
3689 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
3690 instead of warning_at with conditional singular and plural messages
3694 * stor-layout.c (update_alignment_for_field): For
3695 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
3696 && !DECL_PACKED (field), do the alignment update, just use
3697 only desired_align instead of MAX (type_align, desired_align)
3699 (place_field): Don't do known_align < desired_align handling
3700 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
3701 is non-NULL, instead do it after rli->prev_field handling and
3702 only if not within a bitfield word. For DECL_PACKED (field)
3703 use type_align of BITS_PER_UNIT.
3705 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
3707 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
3708 superfluous parentheses and trailing spaces.
3710 2018-02-28 Richard Biener <rguenther@suse.de>
3712 PR tree-optimization/84584
3713 * graphite-scop-detection.c (scop_detection::add_scop): Discard
3714 SCoPs with fake exit edge.
3716 2018-02-28 Martin Liska <mliska@suse.cz>
3719 * timevar.c (timer::print): Fix format to properly print 100%
3722 2018-02-28 Richard Biener <rguenther@suse.de>
3725 * genmatch.c (capture_info::walk_match): Do not mark
3726 captured expressions without operands as expr_p given
3727 they act more like predicates and should be subject to
3728 "lost tail" side-effect preserving.
3730 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
3732 PR rtl-optimization/81611
3733 * auto-inc-dec.c (attempt_change): Move dead note from
3734 mem_insn if it's the next use of regno
3735 (find_address): Take address use of reg holding
3736 non-incremented value. Add parm to limit search to the named
3738 (merge_in_block): Attempt to use a mem insn that is the next
3739 use of the original regno.
3741 2018-02-27 Martin Sebor <msebor@redhat.com>
3744 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
3745 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
3747 2018-02-27 Martin Sebor <msebor@redhat.com>
3749 PR translation/84207
3750 * diagnostic-core.h (warning_n, error_n, inform_n): Change
3751 n argument to unsigned HOST_WIDE_INT.
3752 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
3753 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
3754 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
3755 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
3757 2018-02-27 Richard Biener <rguenther@suse.de>
3759 PR tree-optimization/84512
3760 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
3761 Do not use the estimate returned from record_stmt_cost for
3762 the scalar iteration cost but sum properly using add_stmt_cost.
3764 2018-02-27 Richard Biener <rguenther@suse.de>
3766 PR tree-optimization/84466
3767 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3768 Adjust last change to less strictly validate use operands.
3770 2018-02-27 Martin Liska <mliska@suse.cz>
3772 PR gcov-profile/84548
3773 * gcov.c (process_file): Allow partial overlap and consider it
3774 also as group functions.
3775 (output_lines): Properly calculate range of lines for a group.
3777 2018-02-27 Martin Liska <mliska@suse.cz>
3779 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
3780 'ggc' suffixes. Change first column width.
3781 (timer::print): Fix formatting of the column.
3783 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
3785 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
3786 preserve inline entry blocks for the sake of debug inline
3787 entry point markers alone.
3788 (remove_unused_locals): Suggest in comments a better place to
3789 force the preservation of inline entry blocks that are
3790 otherwise unused, but do not preserve them.
3792 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3794 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
3796 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3799 * config/i386/constraints.md (Bs): Replace
3800 ix86_indirect_branch_register with
3801 TARGET_INDIRECT_BRANCH_REGISTER.
3803 * config/i386/i386.md (indirect_jump): Likewise.
3804 (tablejump): Likewise.
3805 (*sibcall_memory): Likewise.
3806 (*sibcall_value_memory): Likewise.
3807 Peepholes of indirect call and jump via memory: Likewise.
3808 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
3809 (*sibcall_value_GOT_32): Likewise.
3810 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3811 (GOT_memory_operand): Likewise.
3812 (call_insn_operand): Likewise.
3813 (sibcall_insn_operand): Likewise.
3814 (GOT32_symbol_operand): Likewise.
3815 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
3817 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3819 PR rtl-optimization/83496
3820 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
3821 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
3822 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
3823 redundant insn, if any.
3824 (relax_delay_slots): Likewise.
3825 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
3827 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
3829 PR tree-optimization/83965
3830 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
3831 that grouped statements are part of a reduction chain. Return
3832 true if the statement is not marked as a reduction itself but
3834 (vect_recog_dot_prod_pattern): Don't check whether the statement
3835 is part of a group here.
3836 (vect_recog_sad_pattern): Likewise.
3837 (vect_recog_widen_sum_pattern): Likewise.
3839 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3842 * final.c (rest_of_clean_state): Also look for calls inside sequences.
3844 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3847 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
3849 (ix86_output_indirect_function_return): New prototype.
3850 (ix86_split_simple_return_pop_internal): Likewise.
3851 * config/i386/i386.c (indirect_return_via_cx): New.
3852 (indirect_return_via_cx_bnd): Likewise.
3853 (indirect_thunk_name): Handle return va CX_REG.
3854 (output_indirect_thunk_function): Create alias for
3855 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
3856 (ix86_output_indirect_jmp): Remove the bool argument.
3857 (ix86_output_indirect_function_return): New function.
3858 (ix86_split_simple_return_pop_internal): Likewise.
3859 * config/i386/i386.md (*indirect_jump): Don't pass false
3860 to ix86_output_indirect_jmp.
3861 (*tablejump_1): Likewise.
3862 (simple_return_pop_internal): Change it to define_insn_and_split.
3863 Call ix86_split_simple_return_pop_internal to split it for
3865 (simple_return_indirect_internal): Call
3866 ix86_output_indirect_function_return instead of
3867 ix86_output_indirect_jmp.
3869 2018-02-26 Jakub Jelinek <jakub@redhat.com>
3872 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
3873 memset and value initialization afterwards.
3875 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
3877 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
3879 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3882 * common/config/aarch64/aarch64-common.c
3883 (aarch_option_optimization_table[]): Switch
3884 off fomit-frame-pointer
3886 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3887 Chung-Ju Wu <jasonwucj@gmail.com>
3889 * config/nds32/nds32-multiple.md (load_multiple): Disallow
3891 (store_multiple): Ditto.
3893 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3895 * config.gcc: Add --with-cpu support for nds32 target.
3896 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
3897 * config/nds32/nds32.opt: Add -mcpu= option.
3899 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
3901 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
3902 isel=yes): Warn for these deprecated options.
3904 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
3906 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
3907 ISA_2_5_MASKS_EMBEDDED.
3909 2018-02-23 Jakub Jelinek <jakub@redhat.com>
3911 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
3912 p->max as pointers rather than using iterative_hash_expr.
3914 2018-02-23 Carl Love <cel@us.ibm.com>
3916 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
3917 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
3919 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
3920 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
3921 P8V_BUILTIN_VEC_VUNSIGNED2.
3923 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
3926 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
3927 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
3929 * lra-constraints.c (curr_insn_transform): Set up
3930 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
3932 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
3933 * lra-eliminations.c (spill_pseudos): Ditto.
3934 (process_insn_for_elimination): Ditto.
3935 * lra-lives.c (reg_early_clobber_p): Use the new macros.
3936 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
3937 LRA_NON_CLOBBERED_ALT.
3939 2018-02-22 Martin Sebor <msebor@redhat.com>
3941 PR tree-optimization/84480
3942 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
3943 to maybe_diag_stxncpy_trunc. Call it.
3944 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
3945 from gimple_fold_builtin_strcpy. Print inlining stack.
3946 (handle_builtin_stxncpy): Print inlining stack.
3947 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
3949 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
3952 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
3953 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
3954 and -fcheck-pointer-bounds are used together.
3955 (indirect_thunk_prefix): New enum.
3956 (indirect_thunk_need_prefix): New function.
3957 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
3958 "_nt" instead of "_bnd" for NOTRACK prefix.
3959 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
3960 (output_indirect_thunk_function): Likewise.
3962 (ix86_code_end): Update output_indirect_thunk_function calls.
3963 (ix86_output_indirect_branch_via_reg): Replace
3964 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
3965 (ix86_output_indirect_branch_via_push): Likewise.
3966 (ix86_output_function_return): Likewise.
3967 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
3968 incompatible with -fcf-protection=branch and
3969 -fcheck-pointer-bounds.
3971 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3974 * config/aarch64/aarch64.c (aarch64_print_address_internal):
3975 Change gcc_assert call to output_operand_lossage.
3977 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3979 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
3981 2018-02-22 DJ Delorie <dj@redhat.com>
3982 Sebastian Perta <sebastian.perta@renesas.com>
3983 Oleg Endo <olegendo@gcc.gnu.org>
3985 * config/rx/rx.c (rx_rtx_costs): New function.
3986 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
3988 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
3990 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
3992 2018-02-22 Martin Liska <mliska@suse.cz>
3995 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
3996 Add "native" as a possible value.
3998 2018-02-22 Martin Liska <mliska@suse.cz>
4001 * config/i386/i386.c (ix86_option_override_internal):
4002 Add "native" as a possible value for -march and -mtune.
4004 2018-02-22 Jakub Jelinek <jakub@redhat.com>
4007 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
4008 to all type variants.
4010 PR tree-optimization/84503
4011 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
4012 width as info->bitpos + info->bitsize - start.
4013 (merged_store_group::merge_overlapping): Simplify width computation.
4014 (check_no_overlap): New function.
4015 (imm_store_chain_info::try_coalesce_bswap): Compute expected
4016 start + width and last_order of the group, fail if check_no_overlap
4018 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
4019 to group if check_no_overlap fails.
4021 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4023 * config/rs6000/altivec.md: Delete contraint arguments to
4024 define_expand, define_split, and define_peephole2, and in
4025 define_insn_and_split if always unused.
4026 * config/rs6000/darwin.md: Ditto.
4027 * config/rs6000/dfp.md: Ditto.
4028 * config/rs6000/rs6000.md: Ditto.
4029 * config/rs6000/sync.md: Ditto.
4030 * config/rs6000/vector.md: Ditto.
4031 * config/rs6000/vsx.md: Ditto.
4033 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4035 * config/rs6000/altivec.md: Write output control strings as braced
4036 blocks instead of double-quoted strings.
4037 * config/rs6000/darwin.md: Ditto.
4038 * config/rs6000/rs6000.md: Ditto.
4039 * config/rs6000/vector.md: Ditto.
4040 * config/rs6000/vsx.md: Ditto.
4042 2018-02-21 Jason Merrill <jason@redhat.com>
4044 PR c++/84314 - ICE with templates and fastcall attribute.
4045 * attribs.c (build_type_attribute_qual_variant): Remove assert.
4047 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4049 * ipa-cp.c (determine_versionability): Fix comment typos.
4051 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4054 * ipa-cp.c (determine_versionability): Do not version functions caling
4057 2018-02-21 Martin Liska <mliska@suse.cz>
4060 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
4061 Add "native" as a possible value.
4062 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
4063 the macro when native cpu detection is available.
4065 2018-02-21 Martin Liska <mliska@suse.cz>
4068 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
4069 Add "native" as a possible value.
4070 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
4071 when native cpu detection is available.
4073 2018-02-21 Jakub Jelinek <jakub@redhat.com>
4074 Martin Sebor <msebor@redhat.com>
4076 PR tree-optimization/84478
4077 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
4079 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
4080 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
4081 support which is conservatively correct, for 2 only stay conservative
4082 for maxlen. Formatting and comment capitalization fixes. Add STRICT
4083 argument to the 2 argument get_range_strlen, adjust 6 arg
4084 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
4086 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
4087 (gimple_fold_builtin_strlen): Pass true as last argument to
4090 2018-02-20 Martin Sebor <msebor@redhat.com>
4093 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
4094 (builtin_memref::set_base_and_offset): Same. Handle inner references.
4095 (builtin_memref::builtin_memref): Factor out parts into
4096 set_base_and_offset and call it.
4098 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4101 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
4102 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
4103 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
4104 search at the associated MODE_INT.
4106 2018-02-20 Jeff Law <law@redhat.com>
4109 PR tree-optimization/81592
4111 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
4112 for range data rather than using global data.
4113 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
4114 range data rather than using global data.
4115 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
4116 pass it to children as needed.
4117 (struct directive::fmtresult): Similarly.
4118 (struct directive::set_width): Similarly.
4119 (struct directive::set_precision): Similarly.
4120 (format_integer, format_directive, parse_directive): Similarly.
4121 (format_none): Accept unnamed vr_values parameter.
4122 (format_percent, format_floating, format_character): Similarly.
4123 (format_string, format_plain): Similarly.
4124 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
4125 the EVRP range analyzer for range data rather than using global data.
4126 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
4127 gimple-ssa-evrp-analyze.h
4128 (class sprintf_dom_walker): Add after_dom_children member function.
4129 Add evrp_range_analyzer member.
4130 (sprintf_dom_walker::before_dom_children): Call into the EVRP
4131 range analyzer as needed.
4132 (sprintf_dom_walker::after_dom_children): New member function.
4133 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
4135 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
4136 (evrp_range_analyzer::pop_to_marker): Likewise.
4138 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4140 PR tree-optimization/84419
4141 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
4142 with the required type if its current type is compatible but
4145 2018-02-20 Jakub Jelinek <jakub@redhat.com>
4148 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
4149 after vectorization.
4151 2018-02-20 Martin Liska <mliska@suse.cz>
4154 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
4155 possible values if we don't have a hint.
4157 2018-02-20 Martin Liska <mliska@suse.cz>
4161 * final.c (shorten_branches): Build align_tab array with one
4163 * opts.c (finish_options): Add alignment option limit check.
4164 (MAX_CODE_ALIGN): Likewise.
4165 (MAX_CODE_ALIGN_VALUE): Likewise.
4166 * doc/invoke.texi: Document maximum allowed option value for
4167 all -falign-* options.
4169 2018-02-19 Jakub Jelinek <jakub@redhat.com>
4172 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
4173 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
4174 * var-tracking.c (emit_note_insn_var_location): Remove all references
4175 to NOTE_INSN_CALL_ARG_LOCATION.
4176 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
4177 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
4178 Use copy_rtx_if_shared.
4179 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
4180 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
4181 (dwarf2out_var_location): Remove handling of
4182 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
4184 * final.c (final_scan_insn): Remove all references to
4185 NOTE_INSN_CALL_ARG_LOCATION.
4186 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
4187 before dumping final insns.
4188 * except.c (emit_note_eh_region_end): Remove all references to
4189 NOTE_INSN_CALL_ARG_LOCATION.
4190 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
4191 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
4192 * config/arc/arc.c (hwloop_optimize): Likewise.
4193 * config/arm/arm.c (create_fix_barrier): Likewise.
4194 * config/s390/s390.c (s390_chunkify_start): Likewise.
4195 * config/sh/sh.c (find_barrier): Likewise.
4196 * config/i386/i386.c (rest_of_insert_endbranch,
4197 ix86_seh_fixup_eh_fallthru): Likewise.
4198 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
4199 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
4200 * config/frv/frv.c (frv_function_prologue): Likewise.
4201 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
4203 (note_outside_basic_block_p): Remove all references to
4204 NOTE_INSN_CALL_ARG_LOCATION.
4205 * gengtype.c (adjust_field_rtx_def): Likewise.
4206 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
4208 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
4209 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
4212 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
4215 PR tree-optimization/84452
4216 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
4217 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
4220 2018-02-19 Martin Liska <mliska@suse.cz>
4223 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
4225 2018-02-19 Martin Liska <mliska@suse.cz>
4226 Richard Sandiford <richard.sandiford@linaro.org>
4228 PR tree-optimization/82491
4229 * gimple-fold.c (get_base_constructor): Make earlier bail out
4232 2018-02-19 Carl Love <cel@us.ibm.com>
4234 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
4235 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
4237 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
4238 P8V_BUILTIN_VEC_NEG.
4240 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
4242 * config/rl78/rl78.md (movdf): New define expand.
4244 2018-02-19 Martin Liska <mliska@suse.cz>
4247 * doc/invoke.texi: Fix typo.
4248 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
4250 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
4252 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
4253 handle rs6000_single_float and rs6000_double_float specially for
4256 2018-02-16 Jeff Law <law@redhat.com>
4258 * config/rx/rx.c (add_pop_cfi_notes): New function.;
4261 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4264 * ipa-inline.c (inline_small_functions): Fix a typo.
4266 2018-02-16 Nathan Sidwell <nathan@acm.org>
4268 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
4270 2018-02-16 Carl Love <cel@us.ibm.com>
4272 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
4273 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
4274 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
4275 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
4276 expansion to P8V_BUILTIN_VEC_FLOAT2.
4278 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
4280 PR rtl-optimization/70023
4281 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
4282 src_regno into account.
4284 2018-02-16 Carl Love <cel@us.ibm.com>
4286 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
4287 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
4288 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
4289 * config/rs6000/rs6000.c: Remove case statements for
4290 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
4291 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
4292 and P9V_BUILTIN_VEC_VINSERT4B.
4293 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
4294 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
4295 * config/rs6000/vsx.md:
4296 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
4299 2018-02-16 Carl Love <cel@us.ibm.com>
4301 * config/rs6000/altivec.h: Add builtin names vec_extract4b
4303 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
4305 * config/rs6000/rs6000-c.c: Add the definitions for
4306 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
4307 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
4308 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
4309 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
4310 definition for insert4b and define insn *insert3b_internal.
4311 * doc/extend.texi: Add documentation for vec_extract4b.
4313 2018-02-16 Nathan Sidwell <nathan@acm.org>
4315 * doc/extend.texi (Backwards Compatibility): Mention friend
4316 injection. Note for-scope is deprecated.
4317 * doc/invoke.texi (-ffriend-injection): Deprecate.
4319 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
4321 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
4322 that moved to I2, also allow destinations that are a paradoxical
4323 subreg (instead of a normal reg).
4325 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
4328 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
4331 2018-02-16 Richard Biener <rguenther@suse.de>
4333 PR tree-optimization/84037
4334 PR tree-optimization/84016
4336 * config/i386/i386.c (ix86_builtin_vectorization_cost):
4337 Adjust vec_construct for the fact we need additional higher latency
4338 128bit inserts for AVX256 and AVX512 vector builds.
4339 (ix86_add_stmt_cost): Scale vector construction cost for
4342 2018-02-16 Richard Biener <rguenther@suse.de>
4344 PR tree-optimization/84417
4345 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
4346 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
4347 (non_rewritable_lvalue_p): Likewise, use poly-ints.
4349 2018-02-16 Martin Liska <mliska@suse.cz>
4352 * internal-fn.def (ASAN_CHECK): Set proper flags.
4353 (ASAN_MARK): Likewise.
4355 2018-02-16 Julia Koval <julia.koval@intel.com>
4357 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
4358 from PTA_CANNONLAKE.
4360 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4363 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
4364 Use ++iter rather than iter++ for std::list iterators.
4365 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
4366 defer deleting them until all nodes in the forest are processed. Do
4367 free even leaf nodes. Change to_process into auto_vec.
4370 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
4371 * vec.h (vec_default_construct): Use memset instead of placement new
4372 if BROKEN_VALUE_INITIALIZATION is defined.
4373 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
4374 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
4377 PR rtl-optimization/83723
4378 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
4379 * lra.c (lra_substitute_pseudo): Likewise. If true, use
4380 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
4382 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
4384 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
4386 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
4388 PR rtl-optimization/81443
4389 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
4390 from inner REGs to paradoxical SUBREGs.
4392 2018-02-16 Richard Biener <rguenther@suse.de>
4394 PR tree-optimization/84399
4395 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4396 For operands we can analyze at their definition make sure we can
4397 analyze them at each use as well.
4399 2018-02-16 Richard Biener <rguenther@suse.de>
4401 PR tree-optimization/84190
4402 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
4403 volatile accesses if the decl isn't volatile.
4405 2018-02-15 Jason Merrill <jason@redhat.com>
4407 PR c++/84314 - ICE with templates and fastcall attribute.
4408 * attribs.c (build_type_attribute_qual_variant): Don't clobber
4409 TYPE_CANONICAL on an existing type.
4411 2018-02-15 Jakub Jelinek <jakub@redhat.com>
4413 PR tree-optimization/84383
4414 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
4415 dstoff nor call operand_equal_p if dstbase is NULL.
4417 PR tree-optimization/84334
4418 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
4419 also a CONSTANT_CLASS_P, punt.
4421 2018-02-14 Jim Wilson <jimw@sifive.com>
4423 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
4424 first SMALL_OPERAND check. New local min_second_step. Move assert
4425 to where locals are set. Add TARGET_RVC support.
4426 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
4428 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
4430 * doc/invoke.texi: Correct -Wformat-overflow code sample.
4432 2018-02-14 Martin Sebor <msebor@redhat.com>
4434 PR tree-optimization/83698
4435 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
4436 arrays constrain the offset range to their bounds.
4437 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
4438 (builtin_access::overlap): Avoid setting the size of overlap if it's
4440 (maybe_diag_overlap): Also consider arrays when deciding what values
4441 of offsets to include in diagnostics.
4443 2018-02-14 Martin Sebor <msebor@redhat.com>
4446 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
4447 that correspond to the kind of a declaration.
4449 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
4452 * config/pa/pa.md: Load address of PIC label using the linkage table
4453 if the label is nonlocal.
4455 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
4457 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
4458 warning message if user requests -maltivec=be.
4459 * doc/invoke.texi: Document deprecation of -maltivec=be.
4461 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
4464 * config/rs6000/rs6000-c.c: Update definitions for
4465 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
4466 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
4468 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4471 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
4472 add _get_ssp intrinsics. Remove argument from
4473 __builtin_ia32_rdssp[d|q].
4474 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
4475 * config/i386/i386-builtin.def: Remove argument from
4476 __builtin_ia32_rdssp[d|q].
4477 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
4478 ix86_expand_special_args_builtin for _rdssp[d|q].
4479 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
4480 Clear register before usage.
4481 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
4482 Add documentation for new _get_ssp and _inc_ssp intrinsics.
4484 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
4486 PR tree-optimization/84357
4487 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
4488 operand 1 of an ARRAY_REF too.
4490 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
4493 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
4494 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4496 (set_of_reg): New struct.
4497 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
4498 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
4499 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4501 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
4502 Split into bitclr, bitset, bitinvert patterns if appropriate.
4503 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
4504 use rx_fuse_in_memory_bitop.
4505 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
4506 to named insn, correct maximum insn length.
4508 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
4511 * machmode.def: Define a complex mode for PARTIAL_INT.
4512 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
4514 * doc/rtl.texi: Document CSPImode.
4515 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
4517 (msp430_hard_regno_nregs_with_padding): Likewise.
4519 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
4522 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
4524 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
4526 PR rtl-optimization/84169
4527 * combine.c (try_combine): New variable split_i2i3. Set it to true if
4528 we generated a parallel as new i3 and we split that to new i2 and i3
4529 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
4530 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
4531 those to i2, not i1. Partially rewrite this scan code.
4533 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4536 * stor-layout.c (place_field): For variable length fields, adjust
4537 offset_align afterwards not just based on the field's alignment,
4538 but also on the size.
4541 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
4542 of exps and logs in the use_exp2 case.
4544 2018-02-13 Jeff Law <law@redhat.com>
4546 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
4549 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
4552 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
4556 * common.opt (gas-loc-support, gas-locview-support): New.
4557 (ginline-points, ginternal-reset-location-views): New.
4558 * doc/invoke.texi: Document them. Use @itemx where intended.
4559 (gvariable-location-views): Adjust.
4560 * target.def (reset_location_view): New.
4561 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
4562 (TARGET_RESET_LOCATION_VIEW): New.
4563 * doc/tm.texi: Rebuilt.
4564 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
4565 (dwarf2out_default_as_locview_support): New.
4566 (output_asm_line_debug_info): Use option variables.
4567 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
4568 (output_loc_list): Likewise.
4569 (add_high_low_attributes): Check option variables.
4570 Don't output entry view attribute in strict mode.
4571 (gen_inlined_subroutine_die): Check option variables.
4572 (dwarf2out_inline_entry): Likewise.
4573 (init_sections_and_labels): Likewise.
4574 (dwarf2out_early_finish): Likewise.
4575 (maybe_reset_location_view): New, from...
4576 (dwarf2out_var_location): ... here. Call it.
4577 * debug.h (dwarf2out_default_as_loc_support): Declare.
4578 (dwarf2out_default_as_locview_support): Declare.
4579 * hooks.c (hook_int_rtx_insn_0): New.
4580 * hooks.h (hook_int_rtx_insn_0): Declare.
4581 * toplev.c (process_options): Take -gas-loc-support and
4582 -gas-locview-support from dwarf2out. Enable
4583 -gvariable-location-views by default only with locview
4584 assembler support. Enable -ginternal-reset-location-views by
4585 default only if the target defines the corresponding hook.
4586 Enable -ginline-points by default if location views are
4587 enabled; force it disabled if statement frontiers are
4589 * tree-inline.c (expand_call_inline): Check option variables.
4590 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
4592 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
4594 PR tree-optimization/84321
4595 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
4596 handling. Also check whether the anti-range contains any values
4597 that satisfy the mask; switch to a VR_RANGE if not.
4599 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
4602 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
4604 2018-02-13 Martin Jambor <mjambor@suse.cz>
4607 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
4608 of call statements, also set location of a load to a temporary.
4610 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
4612 * config/rl78/rl78.c (add_vector_labels): New function.
4613 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
4614 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
4615 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
4616 which checks that no arguments are passed.
4617 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
4618 * doc/extend.texi: Documentation for the new attribute.
4620 2018-02-13 Andreas Schwab <schwab@suse.de>
4622 * config/riscv/linux.h (CPP_SPEC): Define.
4624 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4627 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
4628 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
4629 OPTION_MASK_ISA_AES as first argument to def_builtin_const
4630 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
4631 instead of OPTION_MASK_ISA_PCLMUL as first argument to
4632 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
4633 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
4634 temporarily for AES and PCLMUL builtins.
4636 PR tree-optimization/84339
4637 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
4638 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
4642 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
4643 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
4644 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
4646 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
4648 * omp-simd-clone.h: New file.
4649 * omp-simd-clone.c: Include omp-simd-clone.h.
4650 (expand_simd_clones): No longer static.
4651 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
4652 cgraph.h and omp-simd-clone.h.
4653 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
4654 (vect_recog_widen_shift_pattern): Formatting fix.
4655 (vect_pattern_recog_1): Don't check optab for calls.
4658 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
4659 operands[2] into a REG before using gen_lowpart on it.
4661 2018-02-12 Jeff Law <law@redhat.com>
4664 * config/sh/sh.c (find_barrier): Consider a sibling call
4667 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
4668 successfully back substituting a reg.
4670 2018-02-12 Richard Biener <rguenther@suse.de>
4672 PR tree-optimization/84037
4673 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
4674 parameter, move visited init to caller.
4675 (vect_slp_analyze_operations): Separate cost from validity
4676 check, initialize visited once for all instances.
4677 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
4679 * tree-vect-stmts.c (vect_model_simple_cost): Make early
4681 (vect_model_promotion_demotion_cost): Likewise.
4682 (vectorizable_bswap): Guard cost modeling with !slp_node
4683 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
4685 (vectorizable_call): Likewise.
4686 (vectorizable_conversion): Likewise.
4687 (vectorizable_assignment): Likewise.
4688 (vectorizable_shift): Likewise.
4689 (vectorizable_operation): Likewise.
4690 (vectorizable_store): Likewise.
4691 (vectorizable_load): Likewise.
4692 (vectorizable_condition): Likewise.
4693 (vectorizable_comparison): Likewise.
4695 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
4698 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
4699 (ASAN_MARK): Fix fnspec to account for return value, change pointer
4700 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
4702 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4705 * params.def (inline-min-speedup): Increase from 8 to 15.
4706 (max-inline-insns-auto): Decrease from 40 to 30.
4707 * ipa-split.c (consider_split): Add some buffer for function to
4708 be considered inlining candidate.
4709 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
4712 2018-02-12 Richard Biener <rguenther@suse.de>
4714 PR tree-optimization/84037
4715 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
4716 matched stmts if we cannot swap the non-matched ones.
4718 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4720 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
4721 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
4722 _mm_maskz_scalef_round_ss): New intrinsics.
4723 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
4724 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
4725 __builtin_ia32_scalefss_round): Remove.
4726 (__builtin_ia32_scalefsd_mask_round,
4727 __builtin_ia32_scalefss_mask_round): New intrinsics.
4728 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
4729 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
4730 ((match_operand:VF_128 2 "<round_nimm_predicate>"
4731 "<round_constraint>")): Changed to ...
4732 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
4733 "<round_scalar_constraint>")): ... this.
4734 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
4735 %0, %1, %2<round_op3>}"): Changed to ...
4736 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
4737 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
4738 %2<round_scalar_mask_op3>}"): ... this.
4739 * config/i386/subst.md (round_scalar_nimm_predicate): New.
4741 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4743 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
4744 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
4745 (_mm_maskz_sqrt_round_ss): New intrinsics.
4746 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
4747 (__builtin_ia32_sqrtsd_mask_round)
4748 (__builtin_ia32_sqrtss_mask_round): New builtins.
4749 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
4750 (__builtin_ia32_sqrtss_round): Remove.
4751 (__builtin_ia32_sqrtsd_mask_round)
4752 (__builtin_ia32_sqrtss_mask_round): New builtins.
4753 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
4754 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
4755 ((match_operand:VF_128 1 "vector_operand"
4756 "xBm,<round_constraint>")): Changed to ...
4757 ((match_operand:VF_128 1 "vector_operand"
4758 "xBm,<round_scalar_constraint>")): ... this.
4759 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
4760 %0, %2, %<iptr>1<round_op3>}): Changed to ...
4761 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
4762 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
4763 %<iptr>1<round_scalar_mask_op3>}): ... this.
4764 ((set_attr "prefix" "<round_prefix>")): Changed to ...
4765 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
4767 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
4770 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
4771 Cast vec_cmpeq result to correct type.
4772 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
4773 Cast vec_cmpgt result to correct type.
4775 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
4777 * final.c (final_scan_insn_1): Renamed from...
4778 (final_scan_insn): ... this. New wrapper, to recover
4779 seen from the outermost call in recursive ones.
4780 * config/sparc/sparc.c (output_return): Drop seen from call.
4781 (output_sibcall): Likewise.
4782 * config/visium/visium.c (output_branch): Likewise.
4784 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
4786 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
4789 2018-02-10 Alan Modra <amodra@gmail.com>
4792 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
4793 Specify LR as an input.
4795 2018-02-10 Jakub Jelinek <jakub@redhat.com>
4798 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
4799 remove_member_access_dummy_vars): New functions.
4800 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
4801 lower_omp_1, execute_lower_omp): Use them.
4803 PR rtl-optimization/84308
4804 * shrink-wrap.c (spread_components): Release todo vector.
4806 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
4808 PR rtl-optimization/57193
4809 * ira-color.c (struct allocno_color_data): Add member
4810 conflict_allocno_hard_prefs.
4811 (update_conflict_allocno_hard_prefs): New.
4812 (bucket_allocno_compare_func): Add a preference based on
4813 conflict_allocno_hard_prefs.
4814 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
4815 (color_allocnos): Remove a dead code. Initiate
4816 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
4818 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4821 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
4822 constraint from =wa to wa. Avoid a subreg on the output operand,
4823 instead use a pseudo and subreg it in a move.
4824 (p9_xxbrd_<mode>): Changed to ...
4825 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
4826 (p9_xxbrd_v2df): New expander.
4827 (p9_xxbrw_<mode>): Changed to ...
4828 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
4829 (p9_xxbrw_v4sf): New expander.
4831 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4833 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
4835 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
4838 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
4839 multiply in 32-bit mode.
4840 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
4841 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
4844 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4846 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
4847 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
4848 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
4849 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
4851 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
4854 * dwarf2out.c (is_trivial_indirect_ref): New function.
4855 (dwarf2out_late_global_decl): Do not generate a location
4856 attribute for variables that have a non-trivial DECL_VALUE_EXPR
4857 and that are not defined in the current unit.
4859 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4861 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
4862 instead of a libcall for UNORDERED.
4864 2018-02-09 Tamar Christina <tamar.christina@arm.com>
4867 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
4868 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
4870 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4873 * config/s390/s390.c (s390_set_current_function): Invoke
4874 s390_indirect_branch_settings also if fndecl didn't change.
4876 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4878 * config/rs6000/rs6000.md (blockage): Set length to zero.
4880 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4882 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
4884 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4887 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
4888 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
4892 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
4893 PARALLEL incoming that failed vt_get_decl_and_offset check.
4896 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
4897 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
4899 (get_variable_section): For decls in named .bss* sections pass true as
4900 second argument to bss_initializer_p.
4902 2018-02-09 Marek Polacek <polacek@redhat.com>
4903 Jakub Jelinek <jakub@redhat.com>
4906 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
4907 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
4908 Sync some changes from cxx_fold_indirect_ref.
4910 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4912 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
4914 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
4915 (BLOCK_INLINE_ENTRY_LABEL): New.
4916 (dwarf2out_var_location): Disregard inline entry markers.
4917 (inline_entry_data): New struct.
4918 (inline_entry_data_hasher): New hashtable type.
4919 (inline_entry_data_hasher::hash): New.
4920 (inline_entry_data_hasher::equal): New.
4921 (inline_entry_data_table): New variable.
4922 (add_high_low_attributes): Add DW_AT_entry_pc and
4923 DW_AT_GNU_entry_view attributes if a pending entry is found
4924 in inline_entry_data_table. Add old entry_pc attribute only
4925 if debug nonbinding markers are disabled.
4926 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
4927 markers are enabled.
4928 (block_within_block_p, dwarf2out_inline_entry): New.
4929 (dwarf2out_finish): Check that no entries remained in
4930 inline_entry_data_table.
4931 * final.c (reemit_insn_block_notes): Handle inline entry notes.
4932 (final_scan_insn, notice_source_line): Likewise.
4933 (rest_of_clean_state): Skip inline entry markers.
4934 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
4936 * gimple.c (gimple_build_debug_inline_entry): New.
4937 * gimple.h (enum gimple_debug_subcode): Add
4938 GIMPLE_DEBUG_INLINE_ENTRY.
4939 (gimple_build_debug_inline_entry): Declare.
4940 (gimple_debug_inline_entry_p): New.
4941 (gimple_debug_nonbind_marker_p): Adjust.
4942 * insn-notes.def (INLINE_ENTRY): New.
4943 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
4944 inline entry marker notes.
4945 (print_insn): Likewise.
4946 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
4947 (INSN_DEBUG_MARKER_KIND): Likewise.
4948 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
4949 * tree-inline.c (expand_call_inline): Build and insert
4950 debug_inline_entry stmt.
4951 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
4952 inline entry blocks early, if nonbind markers are enabled.
4953 (dump_scope_block): Dump fragment info.
4954 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
4955 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
4956 (gimple_build_debug_inline_entry): New.
4957 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
4958 Enable/disable inline entry points too.
4959 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
4960 (DEBUG_INSN): Describe inline entry markers.
4962 * common.opt (gvariable-location-views): New.
4963 (gvariable-location-views=incompat5): New.
4964 * config.in: Rebuilt.
4965 * configure: Rebuilt.
4966 * configure.ac: Test assembler for view support.
4967 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
4968 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
4969 * dwarf2out.c (var_loc_view): New typedef.
4970 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
4971 (dwarf2out_locviews_in_attribute): New.
4972 (dwarf2out_locviews_in_loclist): New.
4973 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
4974 (enum dw_line_info_opcode): Add LI_adv_address.
4975 (struct dw_line_info_table): Add view.
4976 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
4977 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
4978 (zero_view_p): New variable.
4979 (ZERO_VIEW_P): New macro.
4980 (output_asm_line_debug_info): New.
4981 (struct var_loc_node): Add view.
4982 (add_AT_view_list, AT_loc_list): New.
4983 (add_var_loc_to_decl): Add view param. Test it against last.
4984 (new_loc_list): Add view params. Record them.
4985 (AT_loc_list_ptr): Handle loc and view lists.
4986 (view_list_to_loc_list_val_node): New.
4987 (print_dw_val): Handle dw_val_class_view_list.
4988 (size_of_die): Likewise.
4989 (value_format): Likewise.
4990 (loc_list_has_views): New.
4991 (gen_llsym): Set vl_symbol too.
4992 (maybe_gen_llsym, skip_loc_list_entry): New.
4993 (dwarf2out_maybe_output_loclist_view_pair): New.
4994 (output_loc_list): Output view list or entries too.
4995 (output_view_list_offset): New.
4996 (output_die): Handle dw_val_class_view_list.
4997 (output_dwarf_version): New.
4998 (output_compilation_unit_header): Use it.
4999 (output_skeleton_debug_sections): Likewise.
5000 (output_rnglists, output_line_info): Likewise.
5001 (output_pubnames, output_aranges): Update version comments.
5002 (output_one_line_info_table): Output view numbers in asm comments.
5003 (dw_loc_list): Determine current endview, pass it to new_loc_list.
5004 Call maybe_gen_llsym.
5005 (loc_list_from_tree_1): Adjust.
5006 (add_AT_location_description): Create view list attribute if
5007 needed, check it's absent otherwise.
5008 (convert_cfa_to_fb_loc_list): Adjust.
5009 (maybe_emit_file): Call output_asm_line_debug_info for test.
5010 (dwarf2out_var_location): Reset views as needed. Precompute
5011 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
5012 attribute. Set view.
5013 (new_line_info_table): Reset next view.
5014 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
5015 (dwarf2out_source_line): Likewise. Output view resets and labels to
5016 the assembler, or select appropriate line info opcodes.
5017 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
5018 (optimize_string_length): Catch it. Adjust.
5019 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
5020 dw_val_class_view_list, and remove it if no longer needed.
5021 (hash_loc_list): Hash view numbers.
5022 (loc_list_hasher::equal): Compare them.
5023 (optimize_location_lists): Check whether a view list symbol is
5024 needed, and whether the locview attribute is present, and
5025 whether they match. Remove the locview attribute if no longer
5027 (index_location_lists): Call skip_loc_list_entry for test.
5028 (dwarf2out_finish): Call output_asm_line_debug_info for test.
5029 Use output_dwarf_version.
5030 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
5031 (struct dw_val_node): Add val_view_list.
5032 * final.c (SEEN_NEXT_VIEW): New.
5033 (set_next_view_needed): New.
5034 (clear_next_view_needed): New.
5035 (maybe_output_next_view): New.
5036 (final_start_function): Rename to...
5037 (final_start_function_1): ... this. Take pointer to FIRST,
5038 add SEEN parameter. Emit param bindings in the initial view.
5039 (final_start_function): Reintroduce SEEN-less interface.
5040 (final): Rename to...
5041 (final_1): ... this. Take SEEN parameter. Output final pending
5042 next view at the end.
5043 (final): Reintroduce seen-less interface.
5044 (final_scan_insn): Output pending next view before switching
5045 sections or ending a block. Mark the next view as needed when
5046 outputting variable locations. Notify debug backend of section
5047 changes, and of location view changes.
5048 (rest_of_handle_final): Adjust.
5049 * toplev.c (process_options): Autodetect value for debug variable
5050 location views option. Warn on incompat5 without -gdwarf-5.
5051 * doc/invoke.texi (gvariable-location-views): New.
5052 (gvariable-location-views=incompat5): New.
5053 (gno-variable-location-views): New.
5055 2018-02-08 David Malcolm <dmalcolm@redhat.com>
5057 PR tree-optimization/84136
5058 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
5059 that the result of find_edge is non-NULL.
5061 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5064 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
5065 storing integer register in SImode. Fix cost of 256 and 512
5066 byte aligned SSE register store.
5068 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5070 * config/i386/i386.c (ix86_multiplication_cost): Fix
5071 multiplication cost for TARGET_AVX512DQ.
5073 2018-02-08 Marek Polacek <polacek@redhat.com>
5075 PR tree-optimization/84238
5076 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
5079 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5081 PR tree-optimization/84265
5082 * tree-vect-stmts.c (vectorizable_store): Don't treat
5083 VMAT_CONTIGUOUS accesses as grouped.
5084 (vectorizable_load): Likewise.
5086 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5088 PR tree-optimization/81635
5089 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
5090 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
5091 (test_round_for_mask): New functions.
5092 (wide_int_cc_tests): Call test_round_for_mask.
5093 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
5094 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
5095 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
5096 range returned by get_range_info.
5098 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5101 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
5102 * symtab.c: Include builtins.h
5103 (symtab_node::output_to_lto_symbol_table_p): Move here
5104 from lto-streamer-out.c:output_symbol_p.
5105 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
5106 (output_symbol_p): Move all logic to symtab.c
5107 (produce_symtab): Update.
5109 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5111 * config/s390/s390-opts.h (enum indirect_branch): Define.
5112 * config/s390/s390-protos.h (s390_return_addr_from_memory)
5113 (s390_indirect_branch_via_thunk)
5114 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
5115 (enum s390_indirect_branch_type): Define.
5116 * config/s390/s390.c (struct s390_frame_layout, struct
5117 machine_function): Remove.
5118 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
5119 (indirect_branch_table_label_no, indirect_branch_table_name):
5121 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
5122 (enum s390_indirect_branch_option): Define.
5123 (s390_return_addr_from_memory): New function.
5124 (s390_handle_string_attribute): New function.
5125 (s390_attribute_table): Add new attribute handler.
5126 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
5127 (s390_indirect_branch_via_thunk): New function.
5128 (s390_indirect_branch_via_inline_thunk): New function.
5129 (s390_function_ok_for_sibcall): When jumping via thunk disallow
5130 sibling call optimization for non z10 compiles.
5131 (s390_emit_call): Force indirect branch target to be a single
5132 register. Add r1 clobber for non-z10 compiles.
5133 (s390_emit_epilogue): Emit return jump via return_use expander.
5134 (s390_reorg): Handle JUMP_INSNs as execute targets.
5135 (s390_option_override_internal): Perform validity checks for the
5136 new command line options.
5137 (s390_indirect_branch_attrvalue): New function.
5138 (s390_indirect_branch_settings): New function.
5139 (s390_set_current_function): Invoke s390_indirect_branch_settings.
5140 (s390_output_indirect_thunk_function): New function.
5141 (s390_code_end): Implement target hook.
5142 (s390_case_values_threshold): Implement target hook.
5143 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
5145 * config/s390/s390.h (struct s390_frame_layout)
5146 (struct machine_function): Move here from s390.c.
5147 (TARGET_INDIRECT_BRANCH_NOBP_RET)
5148 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
5149 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
5150 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
5151 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
5152 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
5153 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
5154 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
5155 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
5156 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
5157 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
5158 (mnemonic attribute): Add values which aren't recognized
5160 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
5161 pattern for branch conversion. Fix mnemonic attribute.
5162 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
5163 indirect branch via thunk if requested.
5164 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
5165 ("*indirect_jump"): Disable for branch conversion using out of
5167 ("indirect_jump_via_thunk<mode>_z10")
5168 ("indirect_jump_via_thunk<mode>")
5169 ("indirect_jump_via_inlinethunk<mode>_z10")
5170 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
5171 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
5172 ("casesi_jump_via_inlinethunk<mode>_z10")
5173 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
5174 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
5175 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
5176 ("*indirect2_jump"): Disable for branch conversion.
5177 ("casesi_jump"): Turn into expander and expand patterns for branch
5179 ("return_use"): New expander.
5180 ("*return"): Emit return via thunk and rename it to ...
5181 ("*return<mode>"): ... this one.
5182 * config/s390/s390.opt: Add new options and and enum for the
5185 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5187 * lra-constraints.c (match_reload): Unconditionally use
5188 gen_lowpart_SUBREG, rather than selecting between that
5189 and equivalent gen_rtx_SUBREG code.
5191 2018-02-08 Richard Biener <rguenther@suse.de>
5193 PR tree-optimization/84233
5194 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
5195 changed flag instead of boguously re-using phi_inserted.
5197 2018-02-08 Martin Jambor <mjambor@suse.cz>
5199 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
5200 static local variables.
5202 2018-02-08 Richard Biener <rguenther@suse.de>
5204 PR tree-optimization/84278
5205 * tree-vect-stmts.c (vectorizable_store): When looking for
5206 smaller vector types to perform grouped strided loads/stores
5207 make sure the mode is supported by the target.
5208 (vectorizable_load): Likewise.
5210 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5212 * config/aarch64/aarch64.c (aarch64_components_for_bb):
5213 Increase LDP/STP opportunities by adding adjacent callee-saves.
5215 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5217 PR rtl-optimization/84068
5218 PR rtl-optimization/83459
5219 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
5221 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
5223 PR tree-optimization/84224
5224 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
5225 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
5228 2018-02-07 Iain Sandoe <iain@codesourcery.com>
5231 * config/rs6000/altivec.md (*restore_world): Remove LR use.
5232 * config/rs6000/predicates.md (restore_world_operation): Adjust op
5233 count, remove one USE.
5235 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
5237 * doc/install.texi (Configuration): Document the
5238 --with-long-double-format={ibm,ieee} PowerPC configuration
5242 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
5243 Convert from define_expand to be define_insn_and_split. Rework
5244 float/double/_Float128 conversions to QI/HI/SImode to work with
5245 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
5246 conversions to QI/HImode types did a store and then a load to
5247 truncate the value. For conversions to VSX registers, don't split
5248 the insn, instead emit the code directly. Use the code iterator
5249 any_fix to combine signed and unsigned conversions.
5250 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
5251 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
5252 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5253 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
5254 (fix_<mode>di2_hw): Likewise.
5255 (fixuns_<mode>di2_hw): Likewise.
5256 (fix_<mode>si2_hw): Likewise.
5257 (fixuns_<mode>si2_hw): Likewise.
5258 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
5259 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5260 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
5261 fix<uns>_trunc<SFDF:mode>si2_p8.
5262 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
5264 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
5265 (fix<uns>_<mode>_mem): Likewise.
5266 (fctiw<u>z_<mode>_mem): Likewise.
5267 (fix<uns>_<mode>_mem): Likewise.
5268 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
5269 the register allocator from doing a direct move to the GPRs to do
5270 a store, and instead use the ISA 3.0 store byte/half-word from
5271 vector register instruction. For IEEE 128-bit floating point,
5272 also optimize stores of 32-bit ints.
5273 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
5275 2018-02-07 Alan Hayward <alan.hayward@arm.com>
5277 * genextract.c (push_pathstr_operand): New function to support
5279 (walk_rtx): Call push_pathstr_operand.
5280 (print_path): Support [a-zA-Z].
5282 2018-02-07 Richard Biener <rguenther@suse.de>
5284 PR tree-optimization/84037
5285 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
5286 (cse_and_gimplify_to_preheader): Declare.
5287 (vect_get_place_in_interleaving_chain): Likewise.
5288 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5290 (_loop_vec_info::~_loop_vec_info): Delete it.
5291 (cse_and_gimplify_to_preheader): New function.
5292 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
5293 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
5294 (vectorizable_load): Likewise. For grouped stores always base
5295 the IV on the first element.
5296 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
5297 condition before gimplifying.
5299 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5301 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
5302 *DIV_EXPR and *MOD_EXPR.
5304 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
5307 * config/i386/i386.c (ix86_option_override_internal): Mask out
5308 the CF_SET bit when checking -fcf-protection.
5310 2018-02-07 Tom de Vries <tom@codesourcery.com>
5313 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
5316 2018-02-07 Richard Biener <rguenther@suse.de>
5318 PR tree-optimization/84204
5319 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
5322 PR tree-optimization/84205
5323 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
5324 special-case isl_ast_op_zdiv_r.
5326 PR tree-optimization/84223
5327 * graphite-scop-detection.c (gather_bbs::before_dom_children):
5328 Only add conditions from within the region.
5329 (gather_bbs::after_dom_children): Adjust.
5331 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
5334 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
5335 * config/avr/avr.md: Only post-reload split REG-REG moves if
5336 either register is GENERAL_REG_P.
5338 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5340 PR tree-optimization/84235
5341 * tree-ssa-scopedtables.c
5342 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
5343 if the subtraction is performed in floating point type where NaNs are
5344 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
5345 build 1. Formatting fix.
5347 2018-02-06 Jakub Jelinek <jakub@redhat.com>
5350 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
5351 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
5352 and skip it regardless of bb boundaries. Use CALL_P macro,
5353 don't test INSN_P (insn) together with CALL_P or JUMP_P check
5354 unnecessarily, formatting fix.
5356 2018-02-06 Michael Collison <michael.collison@arm.com>
5358 * config/arm/thumb2.md:
5359 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
5360 (*thumb_mov_notscc): Ditto.
5362 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
5365 * config/rs6000/rs6000.md (su code attribute): Use "u" for
5366 unsigned_fix, not "s".
5368 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5370 * configure.ac (gcc_fn_eh_frame_ro): New function.
5371 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
5372 correct .eh_frame permissions.
5373 * configure: Regenerate.
5375 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
5377 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
5380 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5382 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5383 Display warning message for -mno-speculate-indirect-jumps.
5385 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
5387 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
5389 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
5391 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
5393 PR tree-optimization/84225
5394 * tree-eh.c (find_trapping_overflow): Only call
5395 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
5397 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5400 * config/i386/i386.c: Reimplement the check of possible options
5401 -mibt/-mshstk conbination. Change error messages.
5402 * doc/invoke.texi: Fix a typo: remove extra '='.
5404 2018-02-06 Marek Polacek <polacek@redhat.com>
5406 PR tree-optimization/84228
5407 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
5409 2018-02-06 Tamar Christina <tamar.christina@arm.com>
5412 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
5413 emitted arch directives.
5414 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
5415 __ARM_FEATURE_COPROC before changing architectures.
5417 2018-02-06 Richard Biener <rguenther@suse.de>
5419 * config/i386/i386.c (print_reg): Fix typo.
5420 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
5422 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
5424 * configure: Regenerate.
5426 2018-02-05 Martin Sebor <msebor@redhat.com>
5428 PR tree-optimization/83369
5429 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
5432 2018-02-05 Martin Liska <mliska@suse.cz>
5434 * doc/invoke.texi: Cherry-pick upstream r323995.
5436 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
5438 * ira.c (ira_init_register_move_cost): Adjust comment.
5440 2018-02-05 Martin Liska <mliska@suse.cz>
5442 PR gcov-profile/84137
5443 * doc/gcov.texi: Fix typo in documentation.
5445 2018-02-05 Martin Liska <mliska@suse.cz>
5447 PR gcov-profile/83879
5448 * doc/gcov.texi: Document necessity of --dynamic-list-data when
5449 using dlopen functionality.
5451 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
5453 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
5454 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
5455 _mm_maskz_range_ss, _mm_mask_range_round_ss,
5456 _mm_maskz_range_round_ss): New intrinsics.
5457 (__builtin_ia32_rangesd128_round)
5458 (__builtin_ia32_rangess128_round): Remove.
5459 (__builtin_ia32_rangesd128_mask_round,
5460 __builtin_ia32_rangess128_mask_round): New builtins.
5461 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
5462 __builtin_ia32_rangess128_round): Remove.
5463 (__builtin_ia32_rangesd128_mask_round,
5464 __builtin_ia32_rangess128_mask_round): New builtins.
5465 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
5466 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
5467 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
5468 "<round_saeonly_constraint>")): Changed to ...
5469 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
5470 "<round_saeonly_scalar_constraint>")): ... this.
5471 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
5472 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
5473 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
5474 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
5475 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
5477 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
5479 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
5481 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
5482 Remove all values except native, 8540 and 8548.
5484 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
5486 * config/i386/i386.c (ix86_output_function_return): Pass
5487 INVALID_REGNUM, instead of -1, as invalid register number to
5488 indirect_thunk_name and output_indirect_thunk.
5490 2018-02-02 Julia Koval <julia.koval@intel.com>
5492 * config.gcc: Add -march=icelake.
5493 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
5494 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
5495 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
5496 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
5497 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
5498 (processor_target_table): Add icelake.
5499 (ix86_option_override_internal): Handle new PTAs.
5500 (get_builtin_code_for_version): Handle icelake.
5501 (M_INTEL_COREI7_ICELAKE): New.
5502 (fold_builtin_cpu): Handle icelake.
5503 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
5504 * doc/invoke.texi: Add -march=icelake.
5506 2018-02-02 Julia Koval <julia.koval@intel.com>
5508 * config/i386/i386.c (ix86_option_override_internal): Change flags type
5509 to wide_int_bitmask.
5510 * wide-int-bitmask.h: New.
5512 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5515 * config/i386/i386.md: Replace Pmode with word_mode in
5516 builtin_setjmp_setup and builtin_longjmp to support x32.
5518 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
5522 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
5524 (rs6000_supported_cpu_names): New static variable.
5525 (linux_cpu_translation_table): Likewise.
5526 (elf_platform) <cpu>: Define new static variable and use it.
5527 Translate kernel AT_PLATFORM name to canonical name if needed.
5528 Error if platform name is unknown.
5530 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
5533 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
5535 2018-02-01 Jeff Law <law@redhat.com>
5538 * config/i386/i386.c (release_scratch_register_on_entry): Add new
5539 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
5540 the scratch if RELEASE_VIA_POP is false.
5541 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
5542 If we have to save a temporary register, decrement SIZE appropriately.
5543 Pass new arguments to release_scratch_register_on_entry.
5544 (ix86_adjust_stack_and_probe): Likewise.
5545 (ix86_emit_probe_stack_range): Pass new arguments to
5546 release_scratch_register_on_entry.
5548 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
5550 PR rtl-optimization/84157
5551 * combine.c (change_zero_ext): Use REG_P predicate in
5552 front of HARD_REGISTER_P predicate.
5554 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
5556 * config/avr/avr.c (avr_option_override): Move disabling of
5557 -fdelete-null-pointer-checks to...
5558 * common/config/avr/avr-common.c (avr_option_optimization_table):
5561 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5563 PR tree-optimization/81635
5564 * tree-data-ref.c (split_constant_offset_1): For types that
5565 wrap on overflow, try to use range info to prove that wrapping
5568 2018-02-01 Renlin Li <renlin.li@arm.com>
5571 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
5573 (aarch64_register_move_cost): Likewise.
5574 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
5576 (REG_CLASS_NAMES): Likewise.
5577 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
5578 TAILCALL_ADDR_REGS. Remove IP registers.
5579 * config/aarch64/aarch64.md (Ucs): Update register constraint.
5581 2018-02-01 Richard Biener <rguenther@suse.de>
5583 * domwalk.h (dom_walker::dom_walker): Add additional constructor
5584 for specifying RPO order and allow NULL for that.
5585 * domwalk.c (dom_walker::dom_walker): Likewise.
5586 (dom_walker::walk): Handle NULL RPO order.
5587 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
5589 (rewrite_update_dom_walker): Likewise.
5590 (mark_def_dom_walker): Likewise.
5592 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5594 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
5595 (aarch64_maybe_expand_sve_subreg_move): Declare.
5596 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
5597 * config/aarch64/predicates.md (aarch64_any_register_operand): New
5599 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
5600 that are semantically a reverse operation.
5601 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
5602 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
5603 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
5605 (aarch64_can_change_mode_class): For big-endian, forbid changes
5606 between two SVE modes if they have different element sizes.
5608 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5610 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
5611 the TImode handling for big-endian targets.
5613 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5615 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
5616 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
5618 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
5619 Remove BSWAP handing for big-endian targets and use the form of
5620 LD1RQ appropariate for the mode.
5622 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5624 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
5625 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
5628 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5631 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
5632 check for operands that need to go through aarch64_sve_reload_be.
5634 2018-02-01 Jakub Jelinek <jakub@redhat.com>
5636 PR tree-optimization/81661
5637 PR tree-optimization/84117
5638 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
5639 * tree-eh.c: Include gimplify.h.
5640 (find_trapping_overflow, replace_trapping_overflow,
5641 rewrite_to_non_trapping_overflow): New functions.
5642 * tree-vect-loop.c: Include tree-eh.h.
5643 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
5644 * tree-data-ref.c: Include tree-eh.h.
5645 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
5647 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
5649 PR rtl-optimization/84123
5650 * combine.c (change_zero_ext): Check if hard register satisfies
5651 can_change_dest_mode before calling gen_lowpart_SUBREG.
5653 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
5656 * ira.c (ira_init_register_move_cost): Remove assert.
5658 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5660 PR rtl-optimization/84071
5661 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
5662 * doc/tm.texi: Regenerate.
5664 2018-01-31 Richard Biener <rguenther@suse.de>
5666 PR tree-optimization/84132
5667 * tree-data-ref.c (analyze_miv_subscript): Properly
5668 check whether evolution_function_is_affine_multivariate_p
5669 before calling gcd_of_steps_may_divide_p.
5671 2018-01-31 Julia Koval <julia.koval@intel.com>
5674 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
5675 * config/i386/i386.md (rdpid_rex64) New.
5676 (rdpid): Make 32bit only.
5678 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
5681 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
5682 an IDENTIFIER_NODE for FUNCTION_TYPE's.
5684 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5687 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5689 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5691 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5693 PR rtl-optimization/84071
5694 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
5695 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
5697 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5699 * config/arc/arc.c (arc_handle_aux_attribute): New function.
5700 (arc_attribute_table): Add 'aux' attribute.
5701 (arc_in_small_data_p): Consider aux like variables.
5702 (arc_is_aux_reg_p): New function.
5703 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
5704 (arc_get_aux_arg): New function.
5705 (prepare_move_operands): Handle aux-register access.
5706 (arc_handle_aux_attribute): New function.
5707 * doc/extend.texi (ARC Variable attributes): Add subsection.
5709 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5711 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
5712 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
5713 (arc_attribute_table): Add 'uncached' attribute.
5714 (arc_print_operand): Print '.di' flag for uncached memory
5716 (arc_in_small_data_p): Do not consider for small data the uncached
5718 (arc_is_uncached_mem_p): New function.
5719 * config/arc/predicates.md (compact_store_memory_operand): Check
5720 for uncached memory accesses.
5721 (nonvol_nonimm_operand): Likewise.
5722 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
5724 2018-01-31 Jakub Jelinek <jakub@redhat.com>
5727 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
5728 falign-loops=): Add Optimization flag.
5730 2018-01-30 Jeff Law <law@redhat.com>
5733 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
5734 INT_REGISTERS_SAVED. Check it prior to calling
5735 get_scratch_register_on_entry.
5736 (ix86_adjust_stack_and_probe): Similarly.
5737 (ix86_emit_probe_stack_range): Similarly.
5738 (ix86_expand_prologue): Corresponding changes.
5740 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5743 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
5744 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
5746 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
5749 * lra-constraints.c (curr_insn_transform): Process AND in the
5752 2018-01-30 Jakub Jelinek <jakub@redhat.com>
5754 PR rtl-optimization/83986
5755 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
5756 dependence against last_pending_memory_flush in addition to
5759 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
5761 PR tree-optimization/81611
5762 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
5765 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5768 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
5771 2018-01-30 Richard Biener <rguenther@suse.de>
5772 Jakub Jelinek <jakub@redhat.com>
5774 PR tree-optimization/84111
5775 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
5776 inner loops added during recursion, as they don't have up-to-date
5779 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5782 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
5783 (can_inline_edge_by_limits_p): ... here.
5784 (can_early_inline_edge_p, check_callers,
5785 update_caller_keys, update_callee_keys, recursive_inlining,
5786 add_new_edges_to_heap, speculation_useful_p,
5787 inline_small_functions,
5788 inline_small_functions, flatten_function,
5789 inline_to_all_callers_1): Update.
5791 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5793 * profile-count.c (profile_count::combine_with_ipa_count): Handle
5796 2018-01-30 Richard Biener <rguenther@suse.de>
5798 PR tree-optimization/83008
5799 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
5800 invariant and constant vector uses in stmts when they need
5803 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5806 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
5807 * configure: Regenerate.
5809 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5811 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
5813 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
5814 Use gen_rtx_REG rather than gen_lowpart.
5816 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5818 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
5819 rather than 0 when creating partial subregs.
5821 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5823 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
5826 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
5829 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
5830 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
5831 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
5832 flags. This restores the settings used before the 2017-07-24.
5833 Turning off pre increment/decrement/modify allows IVOPTS to
5834 optimize DF/SF loops where the index is an int.
5836 2018-01-29 Richard Biener <rguenther@suse.de>
5837 Kelvin Nilsen <kelvin@gcc.gnu.org>
5840 * tree-vect-stmts.c (vectorizable_call): Don't call
5841 targetm.vectorize_builtin_md_vectorized_function if callee is
5844 2018-01-22 Carl Love <cel@us.ibm.com>
5846 * doc/extend.tex: Fix typo in second arg in
5847 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
5849 2018-01-29 Richard Biener <rguenther@suse.de>
5851 PR tree-optimization/84086
5852 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
5853 (flush_ssaname_freelist): When SSA names were released reset
5854 the SCEV hash table.
5856 2018-01-29 Richard Biener <rguenther@suse.de>
5858 PR tree-optimization/84057
5859 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
5860 removed paths when removing edges.
5862 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
5864 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
5865 -mfunction-return=@var{choice}.
5867 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
5870 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
5872 (layout::print_source_line): Likewise.
5873 (test_get_line_width_without_trailing_whitespace): Add test cases.
5875 2018-01-27 Jakub Jelinek <jakub@redhat.com>
5878 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
5881 2018-01-26 Jim Wilson <jimw@sifive.com>
5883 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
5885 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
5888 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5890 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
5891 and CMP + SUB-immediate -> SUBS.
5893 2018-01-26 Martin Sebor <msebor@redhat.com>
5895 PR tree-optimization/83896
5896 * tree-ssa-strlen.c (get_string_len): Rename...
5897 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
5898 Avoid assuming length is constant.
5899 (handle_char_store): Use HOST_WIDE_INT for string length.
5901 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
5904 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
5905 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
5907 2018-01-26 Richard Biener <rguenther@suse.de>
5909 PR rtl-optimization/84003
5910 * dse.c (record_store): Only record redundant stores when
5911 the earlier store aliases at least all accesses the later one does.
5913 2018-01-26 Jakub Jelinek <jakub@redhat.com>
5915 PR rtl-optimization/83985
5916 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
5917 REG_CFA_RESTORE insns.
5918 (delete_unmarked_insns): Don't ignore separate shrink wrapping
5919 REG_CFA_RESTORE insns here.
5922 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
5923 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
5925 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5927 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
5928 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
5929 (arc_init): Likewise.
5930 (arc_override_options): Likewise.
5931 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
5933 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
5935 * config/arc/arc.h (TARGET_DBNZ): Define.
5936 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
5937 properly set the tune attribute.
5938 (dbnz): Use TARGET_DBNZ guard.
5939 * config/arc/arc.opt (mtune): Add core3 option.
5941 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5943 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
5944 recognize new pic like addresses.
5945 (arc_delegitimize_address): Clean up.
5947 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5949 * config/arc/arc-arches.def: Option mrf16 valid for all
5951 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
5952 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
5953 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
5954 * config/arc/arc-tables.opt: Regenerate.
5955 * config/arc/arc.c (arc_conditional_register_usage): Handle
5956 reduced register file case.
5957 (arc_file_start): Set must have build attributes.
5958 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
5960 * config/arc/arc.opt (mrf16): Add new option.
5961 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
5962 * config/arc/genmultilib.awk: Handle new mrf16 option.
5963 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
5964 * config/arc/t-multilib: Regenerate.
5965 * doc/invoke.texi (ARC Options): Document mrf16 option.
5967 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5969 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
5970 * config/arc/arc.c (arc_handle_secure_attribute): New function.
5971 (arc_attribute_table): Add 'secure_call' attribute.
5972 (arc_print_operand): Print secure call operand.
5973 (arc_function_ok_for_sibcall): Don't optimize tail calls when
5975 (arc_is_secure_call_p): New function. * config/arc/arc.md
5976 (call_i): Add support for sjli instruction.
5977 (call_value_i): Likewise.
5978 * config/arc/constraints.md (Csc): New constraint.
5980 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5981 John Eric Martin <John.Martin@emmicro-us.com>
5983 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
5984 * config/arc/arc.c (_arc_jli_section): New struct.
5985 (arc_jli_section): New type.
5986 (rc_jli_sections): New static variable.
5987 (arc_handle_jli_attribute): New function.
5988 (arc_attribute_table): Add jli_always and jli_fixed attribute.
5989 (arc_file_end): New function.
5990 (TARGET_ASM_FILE_END): Define.
5991 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
5992 (arc_add_jli_section): New function.
5993 (jli_call_scan): Likewise.
5994 (arc_reorg): Call jli_call_scan.
5995 (arc_output_addsi): Remove 'S' from printing asm operand.
5996 (arc_is_jli_call_p): New function.
5997 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
5999 (movhi_insn): Likewise.
6000 (movsi_insn): Likewise.
6001 (movsi_set_cc_insn): Likewise.
6002 (loadqi_update): Likewise.
6003 (load_zeroextendqisi_update): Likewise.
6004 (load_signextendqisi_update): Likewise.
6005 (loadhi_update): Likewise.
6006 (load_zeroextendhisi_update): Likewise.
6007 (load_signextendhisi_update): Likewise.
6008 (loadsi_update): Likewise.
6009 (loadsf_update): Likewise.
6010 (movsicc_insn): Likewise.
6011 (bset_insn): Likewise.
6012 (bxor_insn): Likewise.
6013 (bclr_insn): Likewise.
6014 (bmsk_insn): Likewise.
6015 (bicsi3_insn): Likewise.
6016 (cmpsi_cc_c_insn): Likewise.
6017 (movsi_ne): Likewise.
6018 (movsi_cond_exec): Likewise.
6019 (clrsbsi2): Likewise.
6030 (call_i): Remove 'S' asm letter, add jli instruction.
6031 (call_value_i): Likewise.
6032 * config/arc/arc.op (mjli-always): New option.
6033 * config/arc/constraints.md (Cji): New constraint.
6034 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
6036 (subsf3_fpx): Likewise.
6037 (mulsf3_fpx): Likewise.
6038 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
6040 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
6042 * doc/invoke.texi (ARC): Document mjli-always option.
6044 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
6046 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
6047 avoid addition with 0 and use incw and decw where possible.
6049 2018-01-26 Richard Biener <rguenther@suse.de>
6051 PR tree-optimization/81082
6052 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
6053 association if it requires casting to unsigned.
6054 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
6055 from fold_plusminus_mult_expr to catch important cases late when
6056 range info is available.
6058 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6060 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
6061 * configure.ac (hidden_linkonce): New test.
6062 * configure: Regenerate.
6063 * config.in: Regenerate.
6065 2018-01-26 Julia Koval <julia.koval@intel.com>
6067 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
6068 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
6069 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
6070 _mm_mask_bitshuffle_epi64_mask): Fix type.
6071 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
6072 USI_FTYPE_V4DI_V4DI_USI): Remove.
6073 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
6074 __builtin_ia32_vpshufbitqmb256_mask,
6075 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
6076 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
6077 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
6079 2018-01-26 Alan Modra <amodra@gmail.com>
6082 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
6083 UNSPEC_VBPERMQ. Sort other unspecs.
6085 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
6087 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
6089 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
6092 * predict.c (drop_profile): Do not push/pop cfun; update also
6094 (handle_missing_profiles): Fix logic looking for zero profiles.
6096 2018-01-25 Jakub Jelinek <jakub@redhat.com>
6099 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
6100 on functions with #pragma omp declare simd or functions with simd
6102 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
6103 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
6104 Remove trailing \n from warning_at calls.
6106 2018-01-25 Tom de Vries <tom@codesourcery.com>
6109 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6110 for neutered workers.
6112 2018-01-24 Joseph Myers <joseph@codesourcery.com>
6115 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
6116 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
6118 2018-01-24 Jeff Law <law@redhat.com>
6121 * i386.c (get_probe_interval): Move to earlier point.
6122 (ix86_compute_frame_layout): If -fstack-clash-protection and
6123 the frame is larger than the probe interval, then use pushes
6124 to save registers rather than reg->mem moves.
6125 (ix86_expand_prologue): Remove conditional for int_registers_saved
6128 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
6131 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
6132 min/max for never referenced object.
6134 2018-01-24 Jakub Jelinek <jakub@redhat.com>
6137 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
6139 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
6140 attributes from DECL_ATTRIBUTES (decl) without affecting
6141 DECL_ATTRIBUTES (current_function_decl).
6142 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
6143 functions with non-NULL DECL_ABSTRACT_ORIGIN.
6145 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
6147 PR tree-optimization/83979
6148 * fold-const.c (fold_comparison): Use constant_boolean_node
6149 instead of boolean_{true,false}_node.
6151 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
6153 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
6156 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6158 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6159 Simplify the clause that sets the length attribute.
6160 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6161 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
6162 clause that sets the length attribute.
6163 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6165 2018-01-24 Tom de Vries <tom@codesourcery.com>
6168 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
6169 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
6170 Add strict parameter.
6171 (prevent_branch_around_nothing): Insert dummy insn between branch to
6172 label and label with no ptx insn inbetween.
6173 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
6175 2018-01-24 Tom de Vries <tom@codesourcery.com>
6178 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6179 for neutered threads in warp.
6180 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
6182 2018-01-24 Richard Biener <rguenther@suse.de>
6184 PR tree-optimization/83176
6185 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
6188 2018-01-24 Richard Biener <rguenther@suse.de>
6190 PR tree-optimization/82819
6191 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
6192 code generating pluses that are no-ops in the target precision.
6194 2018-01-24 Richard Biener <rguenther@suse.de>
6197 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
6199 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6201 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
6202 to merge probabilities.
6203 * predict.c (probably_never_executed): Also mark as cold functions
6204 with global 0 profile and guessed local profile.
6205 * profile-count.c (profile_probability::combine_with_count): New
6207 * profile-count.h (profile_probability::operator*,
6208 profile_probability::operator*=, profile_probability::operator/,
6209 profile_probability::operator/=): Reduce precision to adjusted
6210 and set value to guessed on contradictory divisions.
6211 (profile_probability::combine_with_freq): Remove.
6212 (profile_probability::combine_wiht_count): Declare.
6213 (profile_count::force_nonzero):: Set to adjusted.
6214 (profile_count::probability_in):: Set quality to adjusted.
6215 * tree-ssa-tail-merge.c (replace_block_by): Use
6218 2018-01-23 Andrew Waterman <andrew@sifive.com>
6219 Jim Wilson <jimw@sifive.com>
6221 * config/riscv/riscv.c (riscv_stack_boundary): New.
6222 (riscv_option_override): Set riscv_stack_boundary. Handle
6223 riscv_preferred_stack_boundary_arg.
6224 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
6225 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
6226 (STACK_BOUNDARY): Set to riscv_stack_boundary.
6227 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
6228 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
6229 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
6231 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
6234 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
6235 of struct ix86_frame.
6236 (ix86_expand_epilogue): Likewise. Add a local variable for
6237 the reg_save_offset field in struct ix86_frame.
6239 2018-01-23 Bin Cheng <bin.cheng@arm.com>
6241 PR tree-optimization/82604
6242 * tree-loop-distribution.c (enum partition_kind): New enum item
6243 PKIND_PARTIAL_MEMSET.
6244 (partition_builtin_p): Support above new enum item.
6245 (generate_code_for_partition): Ditto.
6246 (compute_access_range): Differentiate cases that equality can be
6247 proven at all loops, the innermost loops or no loops.
6248 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
6249 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
6250 (finalize_partitions, distribute_loop): Don't fuse partition of
6251 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
6252 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
6255 2018-01-23 Martin Liska <mliska@suse.cz>
6257 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
6258 order to ignore the predictor.
6259 (PRED_POLYMORPHIC_CALL): Likewise.
6260 (PRED_RECURSIVE_CALL): Likewise.
6262 2018-01-23 Martin Liska <mliska@suse.cz>
6264 * tree-profile.c (tree_profiling): Print function header to
6265 aware reader which function we are working on.
6266 * value-prof.c (gimple_find_values_to_profile): Do not print
6267 not interesting value histograms.
6269 2018-01-23 Martin Liska <mliska@suse.cz>
6271 * profile-count.h (enum profile_quality): Add
6272 profile_uninitialized as the first value. Do not number values
6273 as they are zero based.
6274 (profile_count::verify): Update sanity check.
6275 (profile_probability::verify): Likewise.
6277 2018-01-23 Nathan Sidwell <nathan@acm.org>
6279 * doc/invoke.texi (ffor-scope): Deprecate.
6281 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6283 PR tree-optimization/83510
6284 * domwalk.c (set_all_edges_as_executable): New function.
6285 (dom_walker::dom_walker): Convert bool param
6286 "skip_unreachable_blocks" to enum reachability. Move setup of
6287 edge flags to set_all_edges_as_executable and only do it when
6288 reachability is REACHABLE_BLOCKS.
6289 * domwalk.h (enum dom_walker::reachability): New enum.
6290 (dom_walker::dom_walker): Convert bool param
6291 "skip_unreachable_blocks" to enum reachability.
6292 (set_all_edges_as_executable): New decl.
6293 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
6294 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
6296 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
6297 but converting true to REACHABLE_BLOCKS.
6298 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
6300 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
6301 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
6302 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
6304 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
6305 if check_all_array_refs will be called.
6307 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6309 * tree.c (selftest::test_location_wrappers): Add more test
6312 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6314 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
6315 (selftest::test_bit_in_range): Likewise.
6317 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6320 * doc/sourcebuild.texi (vect_float): Say that the selector
6321 only describes the situation when -funsafe-math-optimizations is on.
6322 (vect_float_strict): Document.
6324 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6326 PR tree-optimization/83965
6327 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
6328 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
6329 instead of checking only for a reduction.
6330 (vect_recog_widen_sum_pattern): Likewise.
6332 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6334 * predict.c (probably_never_executed): Only use precise profile info.
6335 (compute_function_frequency): Skip after inlining hack since we now
6336 have quality checking.
6338 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6340 * profile-count.h (profile_probability::very_unlikely,
6341 profile_probability::unlikely, profile_probability::even): Set
6342 precision to guessed.
6344 2018-01-23 Richard Biener <rguenther@suse.de>
6346 PR tree-optimization/83963
6347 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
6348 Properly terminate dominator walk when crossing the exit edge not
6349 when visiting its source block.
6351 2018-01-23 Jakub Jelinek <jakub@redhat.com>
6354 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
6355 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
6357 2018-01-22 Jakub Jelinek <jakub@redhat.com>
6359 PR tree-optimization/83957
6360 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
6361 semicolon after for body surrounded by braces.
6363 PR tree-optimization/83081
6364 * profile-count.h (profile_probability::split): New method.
6365 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
6366 Use profile_probability::split.
6367 (do_compare_rtx_and_jump): Fix adjustment of probabilities
6368 when splitting a single conditional jump into 2.
6370 2018-01-22 David Malcolm <dmalcolm@redhat.com>
6372 PR tree-optimization/69452
6373 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
6376 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6378 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
6379 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
6380 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
6382 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6384 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
6385 * config/rl78/rl78.md (movdi): New define_expand.
6386 * config/rl78/rl78.c (rl78_split_movdi): New function.
6388 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
6391 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
6393 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
6394 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
6395 128-bit to produce an UNSPEC move to get the double word with the
6396 signbit and then a shift directly to do signbit.
6397 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
6398 implementation with a new version that just does either a direct
6399 move or a regular move. Move memory interface to separate insns.
6400 Move insns so they are next to the expander.
6401 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
6402 with signbit move. Split big and little endian case.
6403 (signbit<mode>2_dm_mem_le): Likewise.
6404 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
6405 (signbit<mode>2_dm2): Likewise.
6407 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6409 * config/rl78/rl78.md (anddi3): New define_expand.
6411 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6413 * config/rl78/rl78.md (umindi3): New define_expand.
6415 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6417 * config/rl78/rl78.md (smindi3): New define_expand.
6419 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6421 * config/rl78/rl78.md (smaxdi3): New define_expand.
6423 2018-01-22 Carl Love <cel@us.ibm.com>
6425 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
6426 LVX_V1TI): Add macro expansion.
6427 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
6428 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
6429 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
6430 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
6431 Change check to determine if the instruction is a byte reversing
6432 entry. Fix typo in comment.
6433 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
6434 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
6435 Add def_builtin calls for new builtins.
6436 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
6437 Add define_insn expansion.
6439 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6441 * config/rl78/rl78.md (umaxdi3): New define_expand.
6443 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6445 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
6446 for non-QImode registers.
6448 2018-01-22 Richard Biener <rguenther@suse.de>
6450 PR tree-optimization/83963
6451 * graphite-scop-detection.c (scop_detection::get_sese): Delay
6452 including the loop exit block.
6453 (scop_detection::merge_sese): Likewise.
6454 (scop_detection::add_scop): Do it here instead.
6456 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6458 * doc/sourcebuild.texi (arm_softfloat): Document.
6460 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
6463 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
6464 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
6465 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
6467 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6468 David Edelsohn <dje.gcc@gmail.com>
6471 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6472 Change "crset eq" to "crset 2".
6473 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6474 (*call_indirect_aix<mode>_nospec): Likewise.
6475 (*call_value_indirect_aix<mode>_nospec): Likewise.
6476 (*call_indirect_elfv2<mode>_nospec): Likewise.
6477 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
6478 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
6479 change assembly output from . to $.
6480 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6481 (indirect_jump<mode>_nospec): Change assembly output from . to $.
6482 (*tablejump<mode>_internal1_nospec): Likewise.
6484 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
6487 * config/sh/sh_optimize_sett_clrt.cc:
6488 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
6490 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
6492 PR tree-optimization/83940
6493 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
6494 offset_dt to vect_constant_def rather than vect_unknown_def_type.
6495 (vect_check_load_store_mask): Add a mask_dt_out parameter and
6496 use it to pass back the definition type.
6497 (vect_check_store_rhs): Likewise rhs_dt_out.
6498 (vect_build_gather_load_calls): Add a mask_dt argument and use
6499 it instead of a call to vect_is_simple_use.
6500 (vectorizable_store): Update calls to vect_check_load_store_mask
6501 and vect_check_store_rhs. Use the dt returned by the latter instead
6502 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
6503 instead of calls to vect_is_simple_use. Pass the scalar rather
6504 than the vector operand to vect_is_simple_use when handling
6505 second and subsequent copies of an rhs value.
6506 (vectorizable_load): Update calls to vect_check_load_store_mask
6507 and vect_build_gather_load_calls. Use the cached mask_dt and
6508 gs_info.offset_dt instead of calls to vect_is_simple_use.
6510 2018-01-20 Jakub Jelinek <jakub@redhat.com>
6513 * tree-emutls.c: Include gimplify.h.
6514 (lower_emutls_2): New function.
6515 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
6516 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
6517 it before further processing.
6520 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
6521 UINTVAL (trueop1) instead of INTVAL (op1).
6523 2018-01-19 Jakub Jelinek <jakub@redhat.com>
6527 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
6528 INCOMING_FRAME_SP_OFFSET if not defined.
6529 (scan_trace): Add ENTRY argument. If true and
6530 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
6531 emit a note to adjust the CFA offset.
6532 (create_cfi_notes): Adjust scan_trace callers.
6533 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
6534 INCOMING_FRAME_SP_OFFSET in the CIE.
6535 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
6536 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
6538 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
6539 * doc/tm.texi: Regenerated.
6541 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6543 PR rtl-optimization/83147
6544 * lra-constraints.c (remove_inheritance_pseudos): Use
6545 lra_substitute_pseudo_within_insn.
6547 2018-01-19 Tom de Vries <tom@codesourcery.com>
6548 Cesar Philippidis <cesar@codesourcery.com>
6551 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
6553 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
6556 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
6557 spaces for function labels.
6559 2018-01-19 Martin Liska <mliska@suse.cz>
6561 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
6562 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
6563 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
6564 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
6565 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
6566 (PRED_CONST_RETURN): Change from 69 to 65.
6567 (PRED_NULL_RETURN): Change from 91 to 71.
6568 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
6569 (PRED_LOOP_GUARD): Change from 66 to 73.
6571 2018-01-19 Martin Liska <mliska@suse.cz>
6573 * predict.c (predict_insn_def): Add new assert.
6574 (struct branch_predictor): Change type to signed integer.
6575 (test_prediction_value_range): Amend test to cover
6577 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
6578 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
6579 (PRED_LOOP_ITERATIONS_MAX): Likewise.
6580 (PRED_LOOP_IV_COMPARE): Likewise.
6581 * predict.h (PROB_UNINITIALIZED): Define new constant.
6583 2018-01-19 Martin Liska <mliska@suse.cz>
6585 * predict.c (dump_prediction): Add new format for
6586 analyze_brprob.py script which is enabled with -details
6588 * profile-count.h (precise_p): New function.
6590 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6592 PR tree-optimization/83922
6593 * tree-vect-loop.c (vect_verify_full_masking): Return false if
6594 there are no statements that need masking.
6595 (vect_active_double_reduction_p): New function.
6596 (vect_analyze_loop_operations): Use it when handling phis that
6597 are not in the loop header.
6599 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6601 PR tree-optimization/83914
6602 * tree-vect-loop.c (vectorizable_induction): Don't convert
6603 init_expr or apply the peeling adjustment for inductions
6604 that are nested within the vectorized loop.
6606 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6608 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
6611 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6615 * function.h (gimplify_parameters): Add gimple_seq * argument.
6616 * function.c: Include gimple.h and options.h.
6617 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
6618 for the added local temporaries if needed.
6619 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
6620 if there are any parameter cleanups, wrap whole body into a
6621 try/finally with the cleanups.
6623 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
6626 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
6627 Use GET_MODE_CLASS for scalar floating point.
6629 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6633 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
6634 Fix call of call_cgraph_insertion_hooks.
6636 2018-01-18 Martin Sebor <msebor@redhat.com>
6638 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
6640 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6643 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
6646 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
6649 * common.opt: (-ffile-prefix-map): New option.
6650 * opts.c (common_handle_option): Defer it.
6651 * opts-global.c (handle_common_deferred_options): Handle it.
6652 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
6653 * file-prefix-map.h: New file.
6654 (remap_debug_filename, add_debug_prefix_map): ...here.
6655 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
6656 * final.c (debug_prefix_map, add_debug_prefix_map
6657 remap_debug_filename): Move to...
6658 * file-prefix-map.c: New file.
6659 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
6660 generalize, get rid of alloca(), use strrchr() instead of strchr().
6661 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
6662 Implement in terms of add_prefix_map().
6663 (remap_macro_filename, remap_debug_filename): Implement in term of
6665 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
6666 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
6667 * dbxout.c: Include file-prefix-map.h.
6668 * varasm.c: Likewise.
6669 * vmsdbgout.c: Likewise.
6670 * xcoffout.c: Likewise.
6671 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
6672 * doc/cppopts.texi (-fmacro-prefix-map): Document.
6673 * doc/invoke.texi (-ffile-prefix-map): Document.
6674 (-fdebug-prefix-map): Update description.
6676 2018-01-18 Martin Liska <mliska@suse.cz>
6678 * config/i386/i386.c (indirect_thunk_name): Document that also
6680 (output_indirect_thunk): Document why both instructions
6681 (pause and lfence) are generated.
6683 2018-01-18 Richard Biener <rguenther@suse.de>
6685 PR tree-optimization/83887
6686 * graphite-scop-detection.c
6687 (scop_detection::get_nearest_dom_with_single_entry): Remove.
6688 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
6689 (scop_detection::merge_sese): Re-implement with a flood-fill
6690 algorithm that properly finds a SESE region if it exists.
6692 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6695 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
6696 pointer_diff optimizations use view_convert instead of convert.
6698 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6700 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6701 Generate different code for -mno-speculate-indirect-jumps.
6702 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6703 (*call_indirect_aix<mode>): Disable for
6704 -mno-speculate-indirect-jumps.
6705 (*call_indirect_aix<mode>_nospec): New define_insn.
6706 (*call_value_indirect_aix<mode>): Disable for
6707 -mno-speculate-indirect-jumps.
6708 (*call_value_indirect_aix<mode>_nospec): New define_insn.
6709 (*sibcall_nonlocal_sysv<mode>): Generate different code for
6710 -mno-speculate-indirect-jumps.
6711 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6713 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6715 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
6716 long double type, set the flags for noting the default long double
6717 type, even if we don't pass or return a long double type.
6719 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
6722 * ipa-inline.c (flatten_function): Do not overwrite final inlining
6725 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
6727 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
6728 support for merge[hl].
6729 (fold_mergehl_helper): New helper function.
6730 (tree-vector-builder.h): New #include for tree_vector_builder usage.
6731 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
6732 (altivec_vmrglw_direct): Add xxmrglw insn.
6734 2018-01-17 Andrew Waterman <andrew@sifive.com>
6736 * config/riscv/riscv.c (riscv_conditional_register_usage): If
6737 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
6739 2018-01-17 David Malcolm <dmalcolm@redhat.com>
6742 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
6743 call the lto_location_cache before reading the
6744 DECL_SOURCE_LOCATION of the types.
6746 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
6747 Richard Sandiford <richard.sandiford@linaro.org>
6749 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
6750 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
6751 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
6752 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
6753 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
6755 * config/aarch64/constraints.md (aarch64_movti_operand):
6757 * config/aarch64/predicates.md (Uti): Add new constraint.
6759 2018-01-17 Carl Love <cel@us.ibm.com>
6761 * config/rs6000/vsx.md (define_expand xl_len_r,
6762 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
6763 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
6765 (define_expand, define_insn): Move the shift left from the
6766 define_insn to the define_expand for lxvl and stxvl instructions.
6767 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
6768 and XL_LEN_R definitions to PURE.
6770 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
6772 * config/i386/i386.c (indirect_thunk_name): Declare regno
6773 as unsigned int. Compare regno with INVALID_REGNUM.
6774 (output_indirect_thunk): Ditto.
6775 (output_indirect_thunk_function): Ditto.
6776 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
6777 in the call to output_indirect_thunk_function.
6779 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
6782 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
6783 rather than the size of inner_type to determine the stack slot size
6784 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
6786 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
6789 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
6792 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
6794 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
6795 endian Linux systems to optionally enable multilibs for selecting
6796 the long double type if the user configured an explicit type.
6797 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
6798 have no long double multilibs if not defined.
6799 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
6800 warn if the user used -mabi={ieee,ibm}longdouble and we built
6801 multilibs for long double.
6802 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
6803 appropriate multilib option.
6804 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
6806 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
6807 for building long double multilibs.
6808 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
6810 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
6812 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
6815 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
6817 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
6820 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
6823 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
6826 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6828 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
6829 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
6831 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
6833 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
6834 different rtl trees depending on TARGET_64BIT.
6835 (rs6000_gen_lvx): Likewise.
6837 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6839 * config/visium/visium.md (nop): Tweak comment.
6840 (hazard_nop): Likewise.
6842 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6844 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
6845 -mspeculate-indirect-jumps.
6846 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
6847 for -mno-speculate-indirect-jumps.
6848 (*call_indirect_elfv2<mode>_nospec): New define_insn.
6849 (*call_value_indirect_elfv2<mode>): Disable for
6850 -mno-speculate-indirect-jumps.
6851 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
6852 (indirect_jump): Emit different RTL for
6853 -mno-speculate-indirect-jumps.
6854 (*indirect_jump<mode>): Disable for
6855 -mno-speculate-indirect-jumps.
6856 (*indirect_jump<mode>_nospec): New define_insn.
6857 (tablejump): Emit different RTL for
6858 -mno-speculate-indirect-jumps.
6859 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
6860 (tablejumpsi_nospec): New define_expand.
6861 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
6862 (tablejumpdi_nospec): New define_expand.
6863 (*tablejump<mode>_internal1): Disable for
6864 -mno-speculate-indirect-jumps.
6865 (*tablejump<mode>_internal1_nospec): New define_insn.
6866 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
6869 2018-01-16 Artyom Skrobov tyomitch@gmail.com
6871 * caller-save.c (insert_save): Drop unnecessary parameter. All
6874 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6875 Richard Biener <rguenth@suse.de>
6878 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
6879 return early, inline manually is_gimple_sizepos. Make sure if we
6880 call gimplify_expr we don't end up with a gimple constant.
6881 * tree.c (variably_modified_type_p): Don't return true for
6882 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
6883 * gimplify.h (is_gimple_sizepos): Remove.
6885 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6887 PR tree-optimization/83857
6888 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
6889 vectorizable_live_operation for pure SLP statements.
6890 (vectorizable_live_operation): Handle PHIs.
6892 2018-01-16 Richard Biener <rguenther@suse.de>
6894 PR tree-optimization/83867
6895 * tree-vect-stmts.c (vect_transform_stmt): Precompute
6896 nested_in_vect_loop_p since the scalar stmt may get invalidated.
6898 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6901 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
6902 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
6903 If off is not INTEGER_CST, issue a may not be aligned warning
6904 rather than isn't aligned. Use isn%'t rather than isn't.
6905 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
6907 <case MULT_EXPR>: Improve the case when bottom and one of the
6908 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
6909 operand, in that case check if the other operand is multiple of
6910 bottom divided by the INTEGER_CST operand.
6912 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6915 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
6916 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
6917 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
6918 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
6919 * config/pa/pa.c (pa_function_arg_advance): Likewise.
6920 (pa_function_arg, pa_arg_partial_bytes): Likewise.
6921 (pa_function_arg_size): New function.
6923 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6925 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
6926 in a separate statement.
6928 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6930 PR tree-optimization/83847
6931 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
6932 group gathers and scatters.
6934 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6936 PR rtl-optimization/86620
6937 * params.def (max-sched-ready-insns): Bump minimum value to 1.
6939 PR rtl-optimization/83213
6940 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
6941 to last if both are JUMP_INSNs.
6943 PR tree-optimization/83843
6944 * gimple-ssa-store-merging.c
6945 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
6946 store_immediate_info for bswap/nop orig_stores.
6948 2018-01-15 Andrew Waterman <andrew@sifive.com>
6950 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
6952 <UDIV>: Increase cost if !TARGET_DIV.
6954 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
6956 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
6957 (define_attr "cr_logical_3op"): New.
6958 (cceq_ior_compare): Adjust.
6959 (cceq_ior_compare_complement): Adjust.
6960 (*cceq_rev_compare): Adjust.
6961 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
6962 (is_cracked_insn): Adjust.
6963 (insn_must_be_first_in_group): Adjust.
6964 * config/rs6000/40x.md: Adjust.
6965 * config/rs6000/440.md: Adjust.
6966 * config/rs6000/476.md: Adjust.
6967 * config/rs6000/601.md: Adjust.
6968 * config/rs6000/603.md: Adjust.
6969 * config/rs6000/6xx.md: Adjust.
6970 * config/rs6000/7450.md: Adjust.
6971 * config/rs6000/7xx.md: Adjust.
6972 * config/rs6000/8540.md: Adjust.
6973 * config/rs6000/cell.md: Adjust.
6974 * config/rs6000/e300c2c3.md: Adjust.
6975 * config/rs6000/e500mc.md: Adjust.
6976 * config/rs6000/e500mc64.md: Adjust.
6977 * config/rs6000/e5500.md: Adjust.
6978 * config/rs6000/e6500.md: Adjust.
6979 * config/rs6000/mpc.md: Adjust.
6980 * config/rs6000/power4.md: Adjust.
6981 * config/rs6000/power5.md: Adjust.
6982 * config/rs6000/power6.md: Adjust.
6983 * config/rs6000/power7.md: Adjust.
6984 * config/rs6000/power8.md: Adjust.
6985 * config/rs6000/power9.md: Adjust.
6986 * config/rs6000/rs64.md: Adjust.
6987 * config/rs6000/titan.md: Adjust.
6989 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6991 * config/i386/predicates.md (indirect_branch_operand): Rewrite
6992 ix86_indirect_branch_register logic.
6994 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6996 * config/i386/constraints.md (Bs): Update
6997 ix86_indirect_branch_register check. Don't check
6998 ix86_indirect_branch_register with GOT_memory_operand.
7000 * config/i386/predicates.md (GOT_memory_operand): Don't check
7001 ix86_indirect_branch_register here.
7002 (GOT32_symbol_operand): Likewise.
7004 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7006 * config/i386/predicates.md (constant_call_address_operand):
7007 Rewrite ix86_indirect_branch_register logic.
7008 (sibcall_insn_operand): Likewise.
7010 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7012 * config/i386/constraints.md (Bs): Replace
7013 ix86_indirect_branch_thunk_register with
7014 ix86_indirect_branch_register.
7016 * config/i386/i386.md (indirect_jump): Likewise.
7017 (tablejump): Likewise.
7018 (*sibcall_memory): Likewise.
7019 (*sibcall_value_memory): Likewise.
7020 Peepholes of indirect call and jump via memory: Likewise.
7021 * config/i386/i386.opt: Likewise.
7022 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7023 (GOT_memory_operand): Likewise.
7024 (call_insn_operand): Likewise.
7025 (sibcall_insn_operand): Likewise.
7026 (GOT32_symbol_operand): Likewise.
7028 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7031 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
7032 type rather than type addr's type points to.
7033 (expand_omp_atomic_mutex): Likewise.
7034 (expand_omp_atomic): Likewise.
7036 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7039 * config/i386/i386.c (output_indirect_thunk_function): Use
7040 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
7041 for __x86_return_thunk.
7043 2018-01-15 Richard Biener <rguenther@suse.de>
7046 * expmed.c (extract_bit_field_1): Fix typo.
7048 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7051 * config/arm/iterators.md (VF): New mode iterator.
7052 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
7053 Remove integer-related logic from pattern.
7054 (neon_vabd<mode>_3): Likewise.
7056 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7059 * common.opt (fstrict-overflow): No longer an alias.
7060 (fwrapv-pointer): New option.
7061 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
7062 also for pointer types based on flag_wrapv_pointer.
7063 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
7064 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
7065 opts->x_flag_wrapv got set.
7066 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
7067 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
7068 POINTER_TYPE_OVERFLOW_UNDEFINED.
7069 * match.pd: Likewise in address comparison pattern.
7070 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
7072 2018-01-15 Richard Biener <rguenther@suse.de>
7075 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
7076 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
7077 Reset type names to their identifier if their TYPE_DECL doesn't
7078 have linkage (and thus is used for ODR and devirt).
7079 (save_debug_info_for_decl): Remove.
7080 (save_debug_info_for_type): Likewise.
7081 (add_tree_to_fld_list): Adjust.
7082 * tree-pretty-print.c (dump_generic_node): Make dumping of
7083 type names more robust.
7085 2018-01-15 Richard Biener <rguenther@suse.de>
7087 * BASE-VER: Bump to 8.0.1.
7089 2018-01-14 Martin Sebor <msebor@redhat.com>
7092 * builtins.c (check_access): Avoid warning when the no-warning bit
7095 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
7097 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
7098 * ira-color (allocno_hard_regs_compare): Likewise.
7100 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
7103 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
7104 Use .pushsection/.popsection.
7106 2018-01-14 Martin Sebor <msebor@redhat.com>
7109 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
7111 2018-01-14 Jakub Jelinek <jakub@redhat.com>
7113 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
7114 entry from extra_headers.
7115 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
7116 extra_headers, make the list bitwise identical to the i?86-*-* one.
7118 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7120 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
7121 -mcmodel=large with -mindirect-branch=thunk,
7122 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
7123 -mfunction-return=thunk-extern.
7124 * doc/invoke.texi: Document -mcmodel=large is incompatible with
7125 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
7126 -mfunction-return=thunk and -mfunction-return=thunk-extern.
7128 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7130 * config/i386/i386.c (print_reg): Print the name of the full
7131 integer register without '%'.
7132 (ix86_print_operand): Handle 'V'.
7133 * doc/extend.texi: Document 'V' modifier.
7135 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7137 * config/i386/constraints.md (Bs): Disallow memory operand for
7138 -mindirect-branch-register.
7140 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7141 (GOT_memory_operand): Likewise.
7142 (call_insn_operand): Likewise.
7143 (sibcall_insn_operand): Likewise.
7144 (GOT32_symbol_operand): Likewise.
7145 * config/i386/i386.md (indirect_jump): Call convert_memory_address
7146 for -mindirect-branch-register.
7147 (tablejump): Likewise.
7148 (*sibcall_memory): Likewise.
7149 (*sibcall_value_memory): Likewise.
7150 Disallow peepholes of indirect call and jump via memory for
7151 -mindirect-branch-register.
7152 (*call_pop): Replace m with Bw.
7153 (*call_value_pop): Likewise.
7154 (*sibcall_pop_memory): Replace m with Bs.
7155 * config/i386/i386.opt (mindirect-branch-register): New option.
7156 * doc/invoke.texi: Document -mindirect-branch-register option.
7158 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7160 * config/i386/i386-protos.h (ix86_output_function_return): New.
7161 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
7162 set function_return_type.
7163 (indirect_thunk_name): Add ret_p to indicate thunk for function
7165 (output_indirect_thunk_function): Pass false to
7166 indirect_thunk_name.
7167 (ix86_output_indirect_branch_via_reg): Likewise.
7168 (ix86_output_indirect_branch_via_push): Likewise.
7169 (output_indirect_thunk_function): Create alias for function
7170 return thunk if regno < 0.
7171 (ix86_output_function_return): New function.
7172 (ix86_handle_fndecl_attribute): Handle function_return.
7173 (ix86_attribute_table): Add function_return.
7174 * config/i386/i386.h (machine_function): Add
7175 function_return_type.
7176 * config/i386/i386.md (simple_return_internal): Use
7177 ix86_output_function_return.
7178 (simple_return_internal_long): Likewise.
7179 * config/i386/i386.opt (mfunction-return=): New option.
7180 (indirect_branch): Mention -mfunction-return=.
7181 * doc/extend.texi: Document function_return function attribute.
7182 * doc/invoke.texi: Document -mfunction-return= option.
7184 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7186 * config/i386/i386-opts.h (indirect_branch): New.
7187 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
7188 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
7189 with local indirect jump when converting indirect call and jump.
7190 (ix86_set_indirect_branch_type): New.
7191 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
7192 (indirectlabelno): New.
7193 (indirect_thunk_needed): Likewise.
7194 (indirect_thunk_bnd_needed): Likewise.
7195 (indirect_thunks_used): Likewise.
7196 (indirect_thunks_bnd_used): Likewise.
7197 (INDIRECT_LABEL): Likewise.
7198 (indirect_thunk_name): Likewise.
7199 (output_indirect_thunk): Likewise.
7200 (output_indirect_thunk_function): Likewise.
7201 (ix86_output_indirect_branch_via_reg): Likewise.
7202 (ix86_output_indirect_branch_via_push): Likewise.
7203 (ix86_output_indirect_branch): Likewise.
7204 (ix86_output_indirect_jmp): Likewise.
7205 (ix86_code_end): Call output_indirect_thunk_function if needed.
7206 (ix86_output_call_insn): Call ix86_output_indirect_branch if
7208 (ix86_handle_fndecl_attribute): Handle indirect_branch.
7209 (ix86_attribute_table): Add indirect_branch.
7210 * config/i386/i386.h (machine_function): Add indirect_branch_type
7211 and has_local_indirect_jump.
7212 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
7214 (tablejump): Likewise.
7215 (*indirect_jump): Use ix86_output_indirect_jmp.
7216 (*tablejump_1): Likewise.
7217 (simple_return_indirect_internal): Likewise.
7218 * config/i386/i386.opt (mindirect-branch=): New option.
7219 (indirect_branch): New.
7222 (thunk-inline): Likewise.
7223 (thunk-extern): Likewise.
7224 * doc/extend.texi: Document indirect_branch function attribute.
7225 * doc/invoke.texi: Document -mindirect-branch= option.
7227 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
7230 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
7232 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
7234 * ipa-inline.c (want_inline_small_function_p): Return false if
7235 inlining has already failed with CIF_FINAL_ERROR.
7236 (update_caller_keys): Call want_inline_small_function_p before
7238 (update_callee_keys): Likewise.
7240 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7242 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
7244 (rs6000_quadword_masked_address_p): Likewise.
7245 (quad_aligned_load_p): Likewise.
7246 (quad_aligned_store_p): Likewise.
7247 (const_load_sequence_p): Add comment to describe the outer-most loop.
7248 (mimic_memory_attributes_and_flags): New function.
7249 (rs6000_gen_stvx): Likewise.
7250 (replace_swapped_aligned_store): Likewise.
7251 (rs6000_gen_lvx): Likewise.
7252 (replace_swapped_aligned_load): Likewise.
7253 (replace_swapped_load_constant): Capitalize argument name in
7254 comment describing this function.
7255 (rs6000_analyze_swaps): Add a third pass to search for vector loads
7256 and stores that access quad-word aligned addresses and replace
7257 with stvx or lvx instructions when appropriate.
7258 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
7259 New function prototype.
7260 (rs6000_quadword_masked_address_p): Likewise.
7261 (rs6000_gen_lvx): Likewise.
7262 (rs6000_gen_stvx): Likewise.
7263 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
7264 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
7265 when memory address is aligned.
7266 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
7267 this split to select lvx instruction when memory address is aligned.
7268 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
7269 instruction when memory address is aligned.
7270 (*vsx_le_perm_load_v16qi): Likewise.
7271 (four unnamed splitters): Modify to select the stvx instruction
7272 when memory is aligned.
7274 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
7276 * predict.c (determine_unlikely_bbs): Handle correctly BBs
7277 which appears in the queue multiple times.
7279 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7280 Alan Hayward <alan.hayward@arm.com>
7281 David Sherwood <david.sherwood@arm.com>
7283 * tree-vectorizer.h (vec_lower_bound): New structure.
7284 (_loop_vec_info): Add check_nonzero and lower_bounds.
7285 (LOOP_VINFO_CHECK_NONZERO): New macro.
7286 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
7287 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
7288 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
7289 fields. Make seg_len the distance travelled, not including the
7291 (dr_direction_indicator): Declare.
7292 (dr_zero_step_indicator): Likewise.
7293 (dr_known_forward_stride_p): Likewise.
7294 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
7296 (runtime_alias_check_p): Allow runtime alias checks with
7298 (operator ==): Compare access_size and align.
7299 (prune_runtime_alias_test_list): Rework for new distinction between
7300 the access_size and seg_len.
7301 (create_intersect_range_checks_index): Likewise. Cope with polynomial
7303 (get_segment_min_max): New function.
7304 (create_intersect_range_checks): Use it.
7305 (dr_step_indicator): New function.
7306 (dr_direction_indicator): Likewise.
7307 (dr_zero_step_indicator): Likewise.
7308 (dr_known_forward_stride_p): Likewise.
7309 * tree-loop-distribution.c (data_ref_segment_size): Return
7310 DR_STEP * (niters - 1).
7311 (compute_alias_check_pairs): Update call to the dr_with_seg_len
7313 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
7314 (vect_preserves_scalar_order_p): New function, split out from...
7315 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
7316 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
7317 (vect_vfa_access_size): New function.
7318 (vect_vfa_align): Likewise.
7319 (vect_compile_time_alias): Take access_size_a and access_b arguments.
7320 (dump_lower_bound): New function.
7321 (vect_check_lower_bound): Likewise.
7322 (vect_small_gap_p): Likewise.
7323 (vectorizable_with_step_bound_p): Likewise.
7324 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
7325 depencies if the vectorization factor is 1. Convert the checks
7326 for nonzero steps into checks on the bounds of DR_STEP. Try using
7327 a bunds check for variable steps if the minimum required step is
7328 relatively small. Update calls to the dr_with_seg_len
7329 constructor and to vect_compile_time_alias.
7330 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
7332 (vect_loop_versioning): Call it.
7333 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
7335 (vect_estimate_min_profitable_iters): Account for any bounds checks.
7337 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7338 Alan Hayward <alan.hayward@arm.com>
7339 David Sherwood <david.sherwood@arm.com>
7341 * doc/sourcebuild.texi (vect_scatter_store): Document.
7342 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
7344 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
7346 * genopinit.c (main): Add supports_vec_scatter_store and
7347 supports_vec_scatter_store_cached to target_optabs.
7348 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
7349 IFN_MASK_SCATTER_STORE.
7350 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
7352 * internal-fn.h (internal_store_fn_p): Declare.
7353 (internal_fn_stored_value_index): Likewise.
7354 * internal-fn.c (scatter_store_direct): New macro.
7355 (expand_scatter_store_optab_fn): New function.
7356 (direct_scatter_store_optab_supported_p): New macro.
7357 (internal_store_fn_p): New function.
7358 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
7359 IFN_MASK_SCATTER_STORE.
7360 (internal_fn_mask_index): Likewise.
7361 (internal_fn_stored_value_index): New function.
7362 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
7364 * optabs-query.h (supports_vec_scatter_store_p): Declare.
7365 * optabs-query.c (supports_vec_scatter_store_p): New function.
7366 * tree-vectorizer.h (vect_get_store_rhs): Declare.
7367 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
7368 true for scatter stores.
7369 (vect_gather_scatter_fn_p): Handle scatter stores too.
7370 (vect_check_gather_scatter): Consider using scatter stores if
7371 supports_vec_scatter_store_p.
7372 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
7374 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7375 internal_fn_stored_value_index.
7376 (check_load_store_masking): Handle scatter stores too.
7377 (vect_get_store_rhs): Make public.
7378 (vectorizable_call): Use internal_store_fn_p.
7379 (vectorizable_store): Handle scatter store internal functions.
7380 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
7381 when deciding whether the end of the group has been reached.
7382 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
7383 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
7384 (mask_scatter_store<mode>): New insns.
7386 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7387 Alan Hayward <alan.hayward@arm.com>
7388 David Sherwood <david.sherwood@arm.com>
7390 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
7391 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
7392 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
7394 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
7395 Use vect_truncate_gather_scatter_offset if we can't treat the
7396 operation as a normal gather load or scatter store.
7397 (get_group_load_store_type): Take the gather_scatter_info
7398 as argument. Try using a gather load or scatter store for
7399 single-element groups.
7400 (get_load_store_type): Update calls to get_group_load_store_type
7401 and vect_use_strided_gather_scatters_p.
7403 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7404 Alan Hayward <alan.hayward@arm.com>
7405 David Sherwood <david.sherwood@arm.com>
7407 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
7408 optional tree argument.
7409 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
7411 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
7412 but continue to use the current value as a fallback.
7413 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
7414 to compare the updates.
7415 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
7416 (get_load_store_type): Use it when handling a strided access.
7417 (vect_get_strided_load_store_ops): New function.
7418 (vect_get_data_ptr_increment): Likewise.
7419 (vectorizable_load): Handle strided gather loads. Always pass
7420 a step to vect_create_data_ref_ptr and bump_vector_ptr.
7422 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7423 Alan Hayward <alan.hayward@arm.com>
7424 David Sherwood <david.sherwood@arm.com>
7426 * doc/md.texi (gather_load@var{m}): Document.
7427 (mask_gather_load@var{m}): Likewise.
7428 * genopinit.c (main): Add supports_vec_gather_load and
7429 supports_vec_gather_load_cached to target_optabs.
7430 * optabs-tree.c (init_tree_optimization_optabs): Use
7431 ggc_cleared_alloc to allocate target_optabs.
7432 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
7433 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
7435 * internal-fn.h (internal_load_fn_p): Declare.
7436 (internal_gather_scatter_fn_p): Likewise.
7437 (internal_fn_mask_index): Likewise.
7438 (internal_gather_scatter_fn_supported_p): Likewise.
7439 * internal-fn.c (gather_load_direct): New macro.
7440 (expand_gather_load_optab_fn): New function.
7441 (direct_gather_load_optab_supported_p): New macro.
7442 (direct_internal_fn_optab): New function.
7443 (internal_load_fn_p): Likewise.
7444 (internal_gather_scatter_fn_p): Likewise.
7445 (internal_fn_mask_index): Likewise.
7446 (internal_gather_scatter_fn_supported_p): Likewise.
7447 * optabs-query.c (supports_at_least_one_mode_p): New function.
7448 (supports_vec_gather_load_p): Likewise.
7449 * optabs-query.h (supports_vec_gather_load_p): Declare.
7450 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
7451 and memory_type field.
7452 (NUM_PATTERNS): Bump to 15.
7453 * tree-vect-data-refs.c: Include internal-fn.h.
7454 (vect_gather_scatter_fn_p): New function.
7455 (vect_describe_gather_scatter_call): Likewise.
7456 (vect_check_gather_scatter): Try using internal functions for
7457 gather loads. Recognize existing calls to a gather load function.
7458 (vect_analyze_data_refs): Consider using gather loads if
7459 supports_vec_gather_load_p.
7460 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
7461 (vect_get_gather_scatter_offset_type): Likewise.
7462 (vect_convert_mask_for_vectype): Likewise.
7463 (vect_add_conversion_to_patterm): Likewise.
7464 (vect_try_gather_scatter_pattern): Likewise.
7465 (vect_recog_gather_scatter_pattern): New pattern recognizer.
7466 (vect_vect_recog_func_ptrs): Add it.
7467 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7468 internal_fn_mask_index and internal_gather_scatter_fn_p.
7469 (check_load_store_masking): Take the gather_scatter_info as an
7470 argument and handle gather loads.
7471 (vect_get_gather_scatter_ops): New function.
7472 (vectorizable_call): Check internal_load_fn_p.
7473 (vectorizable_load): Likewise. Handle gather load internal
7475 (vectorizable_store): Update call to check_load_store_masking.
7476 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
7477 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
7478 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
7479 (aarch64_gather_scale_operand_d): New predicates.
7480 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
7481 (mask_gather_load<mode>): New insns.
7483 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7484 Alan Hayward <alan.hayward@arm.com>
7485 David Sherwood <david.sherwood@arm.com>
7487 * optabs.def (fold_left_plus_optab): New optab.
7488 * doc/md.texi (fold_left_plus_@var{m}): Document.
7489 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
7490 * internal-fn.c (fold_left_direct): Define.
7491 (expand_fold_left_optab_fn): Likewise.
7492 (direct_fold_left_optab_supported_p): Likewise.
7493 * fold-const-call.c (fold_const_fold_left): New function.
7494 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
7495 * tree-parloops.c (valid_reduction_p): New function.
7496 (gather_scalar_reductions): Use it.
7497 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
7498 (vect_finish_replace_stmt): Declare.
7499 * tree-vect-loop.c (fold_left_reduction_fn): New function.
7500 (needs_fold_left_reduction_p): New function, split out from...
7501 (vect_is_simple_reduction): ...here. Accept reductions that
7502 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
7503 (vect_force_simple_reduction): Also store the reduction type in
7504 the assignment's STMT_VINFO_REDUC_TYPE.
7505 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
7506 (merge_with_identity): New function.
7507 (vect_expand_fold_left): Likewise.
7508 (vectorize_fold_left_reduction): Likewise.
7509 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
7510 scalar phi in place for it. Check for target support and reject
7511 cases that would reassociate the operation. Defer the transform
7512 phase to vectorize_fold_left_reduction.
7513 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
7514 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
7515 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
7517 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7519 * tree-if-conv.c (predicate_mem_writes): Remove redundant
7520 call to ifc_temp_var.
7522 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7523 Alan Hayward <alan.hayward@arm.com>
7524 David Sherwood <david.sherwood@arm.com>
7526 * target.def (legitimize_address_displacement): Take the original
7527 offset as a poly_int.
7528 * targhooks.h (default_legitimize_address_displacement): Update
7530 * targhooks.c (default_legitimize_address_displacement): Likewise.
7531 * doc/tm.texi: Regenerate.
7532 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
7533 as an argument, moving assert of ad->disp == ad->disp_term to...
7534 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
7535 Try calling targetm.legitimize_address_displacement before expanding
7536 the address rather than afterwards, and adjust for the new interface.
7537 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
7538 Match the new hook interface. Handle SVE addresses.
7539 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
7542 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7544 * Makefile.in (OBJS): Add early-remat.o.
7545 * target.def (select_early_remat_modes): New hook.
7546 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
7547 * doc/tm.texi: Regenerate.
7548 * targhooks.h (default_select_early_remat_modes): Declare.
7549 * targhooks.c (default_select_early_remat_modes): New function.
7550 * timevar.def (TV_EARLY_REMAT): New timevar.
7551 * passes.def (pass_early_remat): New pass.
7552 * tree-pass.h (make_pass_early_remat): Declare.
7553 * early-remat.c: New file.
7554 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
7556 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
7558 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7559 Alan Hayward <alan.hayward@arm.com>
7560 David Sherwood <david.sherwood@arm.com>
7562 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
7563 vfm1 with a bound_epilog parameter.
7564 (vect_do_peeling): Update calls accordingly, and move the prologue
7565 call earlier in the function. Treat the base bound_epilog as 0 for
7566 fully-masked loops and retain vf - 1 for other loops. Add 1 to
7567 this base when peeling for gaps.
7568 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
7569 with fully-masked loops.
7570 (vect_estimate_min_profitable_iters): Handle the single peeled
7571 iteration in that case.
7573 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7574 Alan Hayward <alan.hayward@arm.com>
7575 David Sherwood <david.sherwood@arm.com>
7577 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
7578 single-element interleaving even if the size is not a power of 2.
7579 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
7580 accesses for single-element interleaving if the group size is
7583 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7584 Alan Hayward <alan.hayward@arm.com>
7585 David Sherwood <david.sherwood@arm.com>
7587 * doc/md.texi (fold_extract_last_@var{m}): Document.
7588 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
7589 * optabs.def (fold_extract_last_optab): New optab.
7590 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
7591 * internal-fn.c (fold_extract_direct): New macro.
7592 (expand_fold_extract_optab_fn): Likewise.
7593 (direct_fold_extract_optab_supported_p): Likewise.
7594 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
7595 * tree-vect-loop.c (vect_model_reduction_cost): Handle
7596 EXTRACT_LAST_REDUCTION.
7597 (get_initial_def_for_reduction): Do not create an initial vector
7598 for EXTRACT_LAST_REDUCTION reductions.
7599 (vectorizable_reduction): Leave the scalar phi in place for
7600 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
7601 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
7602 epilogue code for EXTRACT_LAST_REDUCTION and defer the
7603 transform phase to vectorizable_condition.
7604 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
7606 (vect_finish_stmt_generation): ...here.
7607 (vect_finish_replace_stmt): New function.
7608 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
7609 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
7611 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
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 (extract_last_@var{m}): Document.
7618 * optabs.def (extract_last_optab): New optab.
7619 * internal-fn.def (EXTRACT_LAST): New internal function.
7620 * internal-fn.c (cond_unary_direct): New macro.
7621 (expand_cond_unary_optab_fn): Likewise.
7622 (direct_cond_unary_optab_supported_p): Likewise.
7623 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
7624 loops using EXTRACT_LAST.
7625 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
7626 (extract_last_<mode>): ...this optab.
7627 (vec_extract<mode><Vel>): Update accordingly.
7629 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7630 Alan Hayward <alan.hayward@arm.com>
7631 David Sherwood <david.sherwood@arm.com>
7633 * target.def (empty_mask_is_expensive): New hook.
7634 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
7635 * doc/tm.texi: Regenerate.
7636 * targhooks.h (default_empty_mask_is_expensive): Declare.
7637 * targhooks.c (default_empty_mask_is_expensive): New function.
7638 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
7639 if the target says that empty masks are expensive.
7640 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
7642 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
7644 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7645 Alan Hayward <alan.hayward@arm.com>
7646 David Sherwood <david.sherwood@arm.com>
7648 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
7649 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
7650 (vect_use_loop_mask_for_alignment_p): New function.
7651 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
7652 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
7653 niters_skip argument. Make sure that the first niters_skip elements
7654 of the first iteration are inactive.
7655 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
7656 Update call to vect_set_loop_masks_directly.
7657 (get_misalign_in_elems): New function, split out from...
7658 (vect_gen_prolog_loop_niters): ...here.
7659 (vect_update_init_of_dr): Take a code argument that specifies whether
7660 the adjustment should be added or subtracted.
7661 (vect_update_init_of_drs): Likewise.
7662 (vect_prepare_for_masked_peels): New function.
7663 (vect_do_peeling): Skip prologue peeling if we're using a mask
7664 instead. Update call to vect_update_inits_of_drs.
7665 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7667 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
7668 alignment. Do not include the number of peeled iterations in
7669 the minimum threshold in that case.
7670 (vectorizable_induction): Adjust the start value down by
7671 LOOP_VINFO_MASK_SKIP_NITERS iterations.
7672 (vect_transform_loop): Call vect_prepare_for_masked_peels.
7673 Take the number of skipped iterations into account when calculating
7675 * tree-vect-stmts.c (vect_gen_while_not): New function.
7677 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7678 Alan Hayward <alan.hayward@arm.com>
7679 David Sherwood <david.sherwood@arm.com>
7681 * doc/sourcebuild.texi (vect_fully_masked): Document.
7682 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
7684 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
7686 (vect_analyze_loop_2): ...here. Don't check the vectorization
7687 factor against the number of loop iterations if the loop is
7690 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7691 Alan Hayward <alan.hayward@arm.com>
7692 David Sherwood <david.sherwood@arm.com>
7694 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
7695 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
7696 (dump_groups): Update accordingly.
7697 (iv_use::mem_type): New member variable.
7698 (address_p): New function.
7699 (record_use): Add a mem_type argument and initialize the new
7701 (record_group_use): Add a mem_type argument. Use address_p.
7702 Remove obsolete null checks of base_object. Update call to record_use.
7703 (find_interesting_uses_op): Update call to record_group_use.
7704 (find_interesting_uses_cond): Likewise.
7705 (find_interesting_uses_address): Likewise.
7706 (get_mem_type_for_internal_fn): New function.
7707 (find_address_like_use): Likewise.
7708 (find_interesting_uses_stmt): Try find_address_like_use before
7709 calling find_interesting_uses_op.
7710 (addr_offset_valid_p): Use the iv mem_type field as the type
7711 of the addressed memory.
7712 (add_autoinc_candidates): Likewise.
7713 (get_address_cost): Likewise.
7714 (split_small_address_groups_p): Use address_p.
7715 (split_address_groups): Likewise.
7716 (add_iv_candidate_for_use): Likewise.
7717 (autoinc_possible_for_pair): Likewise.
7718 (rewrite_groups): Likewise.
7719 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
7720 (determine_group_iv_cost): Update after split of USE_ADDRESS.
7721 (get_alias_ptr_type_for_ptr_address): New function.
7722 (rewrite_use_address): Rewrite address uses in calls that were
7723 identified by find_address_like_use.
7725 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7726 Alan Hayward <alan.hayward@arm.com>
7727 David Sherwood <david.sherwood@arm.com>
7729 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
7731 * gimple-expr.h (is_gimple_addressable: Likewise.
7732 * gimple-expr.c (is_gimple_address): Likewise.
7733 * internal-fn.c (expand_call_mem_ref): New function.
7734 (expand_mask_load_optab_fn): Use it.
7735 (expand_mask_store_optab_fn): Likewise.
7737 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7738 Alan Hayward <alan.hayward@arm.com>
7739 David Sherwood <david.sherwood@arm.com>
7741 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
7742 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
7743 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
7744 (cond_umax@var{mode}): Document.
7745 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
7746 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
7747 (cond_umin_optab, cond_umax_optab): New optabs.
7748 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
7749 (COND_IOR, COND_XOR): New internal functions.
7750 * internal-fn.h (get_conditional_internal_fn): Declare.
7751 * internal-fn.c (cond_binary_direct): New macro.
7752 (expand_cond_binary_optab_fn): Likewise.
7753 (direct_cond_binary_optab_supported_p): Likewise.
7754 (get_conditional_internal_fn): New function.
7755 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
7756 Cope with reduction statements that are vectorized as calls rather
7758 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
7759 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
7760 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
7761 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
7762 (UNSPEC_COND_EOR): New unspecs.
7763 (optab): Add mappings for them.
7764 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
7765 (sve_int_op, sve_fp_op): New int attributes.
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 * optabs.def (while_ult_optab): New optab.
7772 * doc/md.texi (while_ult@var{m}@var{n}): Document.
7773 * internal-fn.def (WHILE_ULT): New internal function.
7774 * internal-fn.h (direct_internal_fn_supported_p): New override
7775 that takes two types as argument.
7776 * internal-fn.c (while_direct): New macro.
7777 (expand_while_optab_fn): New function.
7778 (convert_optab_supported_p): Likewise.
7779 (direct_while_optab_supported_p): New macro.
7780 * wide-int.h (wi::udiv_ceil): New function.
7781 * tree-vectorizer.h (rgroup_masks): New structure.
7782 (vec_loop_masks): New typedef.
7783 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
7785 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
7786 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
7787 (vect_max_vf): New function.
7788 (slpeel_make_loop_iterate_ntimes): Delete.
7789 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
7790 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
7791 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
7792 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
7793 internal-fn.h, stor-layout.h and optabs-query.h.
7794 (vect_set_loop_mask): New function.
7795 (add_preheader_seq): Likewise.
7796 (add_header_seq): Likewise.
7797 (interleave_supported_p): Likewise.
7798 (vect_maybe_permute_loop_masks): Likewise.
7799 (vect_set_loop_masks_directly): Likewise.
7800 (vect_set_loop_condition_masked): Likewise.
7801 (vect_set_loop_condition_unmasked): New function, split out from
7802 slpeel_make_loop_iterate_ntimes.
7803 (slpeel_make_loop_iterate_ntimes): Rename to..
7804 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
7805 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
7806 (vect_do_peeling): Update call accordingly.
7807 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
7809 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7810 mask_compare_type, can_fully_mask_p and fully_masked_p.
7811 (release_vec_loop_masks): New function.
7812 (_loop_vec_info): Use it to free the loop masks.
7813 (can_produce_all_loop_masks_p): New function.
7814 (vect_get_max_nscalars_per_iter): Likewise.
7815 (vect_verify_full_masking): Likewise.
7816 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
7817 retries, and free the mask rgroups before retrying. Check loop-wide
7818 reasons for disallowing fully-masked loops. Make the final decision
7819 about whether use a fully-masked loop or not.
7820 (vect_estimate_min_profitable_iters): Do not assume that peeling
7821 for the number of iterations will be needed for fully-masked loops.
7822 (vectorizable_reduction): Disable fully-masked loops.
7823 (vectorizable_live_operation): Likewise.
7824 (vect_halve_mask_nunits): New function.
7825 (vect_double_mask_nunits): Likewise.
7826 (vect_record_loop_mask): Likewise.
7827 (vect_get_loop_mask): Likewise.
7828 (vect_transform_loop): Handle the case in which the final loop
7829 iteration might handle a partial vector. Call vect_set_loop_condition
7830 instead of slpeel_make_loop_iterate_ntimes.
7831 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
7832 (check_load_store_masking): New function.
7833 (prepare_load_store_mask): Likewise.
7834 (vectorizable_store): Handle fully-masked loops.
7835 (vectorizable_load): Likewise.
7836 (supportable_widening_operation): Use vect_halve_mask_nunits for
7838 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
7839 (vect_gen_while): New function.
7840 * config/aarch64/aarch64.md (umax<mode>3): New expander.
7841 (aarch64_uqdec<mode>): New insn.
7843 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7844 Alan Hayward <alan.hayward@arm.com>
7845 David Sherwood <david.sherwood@arm.com>
7847 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
7848 (reduc_xor_scal_optab): New optabs.
7849 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
7850 (reduc_xor_scal_@var{m}): Document.
7851 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
7852 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
7854 * fold-const-call.c (fold_const_call): Handle them.
7855 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
7856 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
7857 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
7858 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
7859 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
7860 (UNSPEC_XORV): New unspecs.
7861 (optab): Add entries for them.
7862 (BITWISEV): New int iterator.
7863 (bit_reduc_op): New int attributes.
7865 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7866 Alan Hayward <alan.hayward@arm.com>
7867 David Sherwood <david.sherwood@arm.com>
7869 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
7870 * internal-fn.def (VEC_SHL_INSERT): New internal function.
7871 * optabs.def (vec_shl_insert_optab): New optab.
7872 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
7873 (duplicate_and_interleave): Likewise.
7874 * tree-vect-loop.c: Include internal-fn.h.
7875 (neutral_op_for_slp_reduction): New function, split out from
7876 get_initial_defs_for_reduction.
7877 (get_initial_def_for_reduction): Handle option 2 for variable-length
7878 vectors by loading the neutral value into a vector and then shifting
7879 the initial value into element 0.
7880 (get_initial_defs_for_reduction): Replace the code argument with
7881 the neutral value calculated by neutral_op_for_slp_reduction.
7882 Use gimple_build_vector for constant-length vectors.
7883 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
7884 but the first group_size elements have a neutral value.
7885 Use duplicate_and_interleave otherwise.
7886 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
7887 Update call to get_initial_defs_for_reduction. Handle SLP
7888 reductions for variable-length vectors by creating one vector
7889 result for each scalar result, with the elements associated
7890 with other scalar results stubbed out with the neutral value.
7891 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
7892 Require IFN_VEC_SHL_INSERT for double reductions on
7893 variable-length vectors, or SLP reductions that have
7894 a neutral value. Require can_duplicate_and_interleave_p
7895 support for variable-length unchained SLP reductions if there
7896 is no neutral value, such as for MIN/MAX reductions. Also require
7897 the number of vector elements to be a multiple of the number of
7898 SLP statements when doing variable-length unchained SLP reductions.
7899 Update call to vect_create_epilog_for_reduction.
7900 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
7901 and remove initial values.
7902 (duplicate_and_interleave): Make public.
7903 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
7904 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
7906 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7907 Alan Hayward <alan.hayward@arm.com>
7908 David Sherwood <david.sherwood@arm.com>
7910 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
7911 (can_duplicate_and_interleave_p): New function.
7912 (vect_get_and_check_slp_defs): Take the vector of statements
7913 rather than just the current one. Remove excess parentheses.
7914 Restriction rejectinon of vect_constant_def and vect_external_def
7915 for variable-length vectors to boolean types, or types for which
7916 can_duplicate_and_interleave_p is false.
7917 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
7918 (duplicate_and_interleave): New function.
7919 (vect_get_constant_vectors): Use gimple_build_vector for
7920 constant-length vectors and suitable variable-length constant
7921 vectors. Use duplicate_and_interleave for other variable-length
7922 vectors. Don't defer the update when inserting new statements.
7924 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7925 Alan Hayward <alan.hayward@arm.com>
7926 David Sherwood <david.sherwood@arm.com>
7928 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
7929 min_profitable_iters doesn't go negative.
7931 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7932 Alan Hayward <alan.hayward@arm.com>
7933 David Sherwood <david.sherwood@arm.com>
7935 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
7936 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
7937 * optabs.def (vec_mask_load_lanes_optab): New optab.
7938 (vec_mask_store_lanes_optab): Likewise.
7939 * internal-fn.def (MASK_LOAD_LANES): New internal function.
7940 (MASK_STORE_LANES): Likewise.
7941 * internal-fn.c (mask_load_lanes_direct): New macro.
7942 (mask_store_lanes_direct): Likewise.
7943 (expand_mask_load_optab_fn): Handle masked operations.
7944 (expand_mask_load_lanes_optab_fn): New macro.
7945 (expand_mask_store_optab_fn): Handle masked operations.
7946 (expand_mask_store_lanes_optab_fn): New macro.
7947 (direct_mask_load_lanes_optab_supported_p): Likewise.
7948 (direct_mask_store_lanes_optab_supported_p): Likewise.
7949 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
7951 (vect_load_lanes_supported): Likewise.
7952 * tree-vect-data-refs.c (strip_conversion): New function.
7953 (can_group_stmts_p): Likewise.
7954 (vect_analyze_data_ref_accesses): Use it instead of checking
7955 for a pair of assignments.
7956 (vect_store_lanes_supported): Take a masked_p parameter.
7957 (vect_load_lanes_supported): Likewise.
7958 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
7959 vect_store_lanes_supported and vect_load_lanes_supported.
7960 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7961 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
7962 parameter. Don't allow gaps for masked accesses.
7963 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
7964 and vect_load_lanes_supported.
7965 (get_load_store_type): Take a masked_p parameter and update
7966 call to get_group_load_store_type.
7967 (vectorizable_store): Update call to get_load_store_type.
7968 Handle IFN_MASK_STORE_LANES.
7969 (vectorizable_load): Update call to get_load_store_type.
7970 Handle IFN_MASK_LOAD_LANES.
7972 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7973 Alan Hayward <alan.hayward@arm.com>
7974 David Sherwood <david.sherwood@arm.com>
7976 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
7978 * config/aarch64/aarch64-protos.h
7979 (aarch64_sve_struct_memory_operand_p): Declare.
7980 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
7981 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
7982 (VPRED, vpred): Handle SVE structure modes.
7983 * config/aarch64/constraints.md (Utx): New constraint.
7984 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
7985 (aarch64_sve_struct_nonimmediate_operand): New predicates.
7986 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
7987 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
7988 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
7989 structure modes. Split into pieces after RA.
7990 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
7991 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
7993 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
7994 SVE structure modes.
7995 (aarch64_classify_address): Likewise.
7996 (sizetochar): Move earlier in file.
7997 (aarch64_print_operand): Handle SVE register lists.
7998 (aarch64_array_mode): New function.
7999 (aarch64_sve_struct_memory_operand_p): Likewise.
8000 (TARGET_ARRAY_MODE): Redefine.
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 * target.def (array_mode): New target hook.
8007 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
8008 * doc/tm.texi: Regenerate.
8009 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
8010 * hooks.c (hook_optmode_mode_uhwi_none): New function.
8011 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
8013 * stor-layout.c (mode_for_array): Likewise. Support polynomial
8016 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8017 Alan Hayward <alan.hayward@arm.com>
8018 David Sherwood <david.sherwood@arm.com>
8020 * fold-const.c (fold_binary_loc): Check the argument types
8021 rather than the result type when testing for a vector operation.
8023 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8025 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
8026 * doc/tm.texi: Regenerate.
8028 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8029 Alan Hayward <alan.hayward@arm.com>
8030 David Sherwood <david.sherwood@arm.com>
8032 * doc/invoke.texi (-msve-vector-bits=): Document new option.
8033 (sve): Document new AArch64 extension.
8034 * doc/md.texi (w): Extend the description of the AArch64
8035 constraint to include SVE vectors.
8036 (Upl, Upa): Document new AArch64 predicate constraints.
8037 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
8039 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
8040 (msve-vector-bits=): New option.
8041 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
8042 SVE when these are disabled.
8043 (sve): New extension.
8044 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
8045 modes. Adjust their number of units based on aarch64_sve_vg.
8046 (MAX_BITSIZE_MODE_ANY_MODE): Define.
8047 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
8048 aarch64_addr_query_type.
8049 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
8050 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
8051 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
8052 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
8053 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
8054 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
8055 (aarch64_simd_imm_zero_p): Delete.
8056 (aarch64_check_zero_based_sve_index_immediate): Declare.
8057 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8058 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8059 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8060 (aarch64_sve_float_mul_immediate_p): Likewise.
8061 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8063 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
8064 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
8065 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
8066 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
8067 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
8068 (aarch64_regmode_natural_size): Likewise.
8069 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
8070 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
8072 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
8073 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
8074 for VG and the SVE predicate registers.
8075 (V_ALIASES): Add a "z"-prefixed alias.
8076 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
8077 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
8078 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
8079 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
8080 (REG_CLASS_NAMES): Add entries for them.
8081 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
8082 and the predicate registers.
8083 (aarch64_sve_vg): Declare.
8084 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
8085 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
8086 (REGMODE_NATURAL_SIZE): Define.
8087 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
8089 * config/aarch64/aarch64.c: Include cfgrtl.h.
8090 (simd_immediate_info): Add a constructor for series vectors,
8091 and an associated step field.
8092 (aarch64_sve_vg): New variable.
8093 (aarch64_dbx_register_number): Handle VG and the predicate registers.
8094 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
8095 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
8096 (VEC_ANY_DATA, VEC_STRUCT): New constants.
8097 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
8098 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
8099 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
8100 (aarch64_get_mask_mode): New functions.
8101 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
8102 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
8103 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
8104 predicate modes and predicate registers. Explicitly restrict
8105 GPRs to modes of 16 bytes or smaller. Only allow FP registers
8106 to store a vector mode if it is recognized by
8107 aarch64_classify_vector_mode.
8108 (aarch64_regmode_natural_size): New function.
8109 (aarch64_hard_regno_caller_save_mode): Return the original mode
8111 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
8112 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
8113 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
8114 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
8116 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
8117 does not overlap dest if the function is frame-related. Handle
8119 (aarch64_split_add_offset): New function.
8120 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
8121 them aarch64_add_offset.
8122 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
8123 and update call to aarch64_sub_sp.
8124 (aarch64_add_cfa_expression): New function.
8125 (aarch64_expand_prologue): Pass extra temporary registers to the
8126 functions above. Handle the case in which we need to emit new
8127 DW_CFA_expressions for registers that were originally saved
8128 relative to the stack pointer, but now have to be expressed
8129 relative to the frame pointer.
8130 (aarch64_output_mi_thunk): Pass extra temporary registers to the
8132 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
8133 IP0 and IP1 values for SVE frames.
8134 (aarch64_expand_vec_series): New function.
8135 (aarch64_expand_sve_widened_duplicate): Likewise.
8136 (aarch64_expand_sve_const_vector): Likewise.
8137 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
8138 Handle SVE constants. Use emit_move_insn to move a force_const_mem
8139 into the register, rather than emitting a SET directly.
8140 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
8141 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
8142 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
8143 (offset_9bit_signed_scaled_p): New functions.
8144 (aarch64_replicate_bitmask_imm): New function.
8145 (aarch64_bitmask_imm): Use it.
8146 (aarch64_cannot_force_const_mem): Reject expressions involving
8147 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
8148 (aarch64_classify_index): Handle SVE indices, by requiring
8149 a plain register index with a scale that matches the element size.
8150 (aarch64_classify_address): Handle SVE addresses. Assert that
8151 the mode of the address is VOIDmode or an integer mode.
8152 Update call to aarch64_classify_symbol.
8153 (aarch64_classify_symbolic_expression): Update call to
8154 aarch64_classify_symbol.
8155 (aarch64_const_vec_all_in_range_p): New function.
8156 (aarch64_print_vector_float_operand): Likewise.
8157 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
8158 "vN" for FP registers with SVE modes. Handle (const ...) vectors
8159 and the FP immediates 1.0 and 0.5.
8160 (aarch64_print_address_internal): Handle SVE addresses.
8161 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
8162 (aarch64_regno_regclass): Handle predicate registers.
8163 (aarch64_secondary_reload): Handle big-endian reloads of SVE
8165 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
8166 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
8167 (aarch64_convert_sve_vector_bits): New function.
8168 (aarch64_override_options): Use it to handle -msve-vector-bits=.
8169 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8171 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
8172 Handle SVE vector and predicate modes. Accept VL-based constants
8173 that need only one temporary register, and VL offsets that require
8174 no temporary registers.
8175 (aarch64_conditional_register_usage): Mark the predicate registers
8176 as fixed if SVE isn't available.
8177 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
8178 Return true for SVE vector and predicate modes.
8179 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
8180 rather than an unsigned int. Handle SVE modes.
8181 (aarch64_preferred_simd_mode): Update call accordingly. Handle
8183 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
8185 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8186 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8187 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8188 (aarch64_sve_float_mul_immediate_p): New functions.
8189 (aarch64_sve_valid_immediate): New function.
8190 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
8191 Explicitly reject structure modes. Check for INDEX constants.
8192 Handle PTRUE and PFALSE constants.
8193 (aarch64_check_zero_based_sve_index_immediate): New function.
8194 (aarch64_simd_imm_zero_p): Delete.
8195 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
8196 vector modes. Accept constants in the range of CNT[BHWD].
8197 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
8198 ask for an Advanced SIMD mode.
8199 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
8200 (aarch64_simd_vector_alignment): Handle SVE predicates.
8201 (aarch64_vectorize_preferred_vector_alignment): New function.
8202 (aarch64_simd_vector_alignment_reachable): Use it instead of
8204 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
8205 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
8207 (MAX_VECT_LEN): Delete.
8208 (expand_vec_perm_d): Add a vec_flags field.
8209 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
8210 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
8211 (aarch64_evpc_ext): Don't apply a big-endian lane correction
8213 (aarch64_evpc_rev): Rename to...
8214 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
8215 (aarch64_evpc_rev_global): New function.
8216 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
8217 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
8219 (aarch64_evpc_sve_tbl): New function.
8220 (aarch64_expand_vec_perm_const_1): Update after rename of
8221 aarch64_evpc_rev. Handle SVE permutes too, trying
8222 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
8223 than aarch64_evpc_tbl.
8224 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
8225 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
8226 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
8227 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
8228 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
8229 (aarch64_expand_sve_vcond): New functions.
8230 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
8231 of aarch64_vector_mode_p.
8232 (aarch64_dwarf_poly_indeterminate_value): New function.
8233 (aarch64_compute_pressure_classes): Likewise.
8234 (aarch64_can_change_mode_class): Likewise.
8235 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
8236 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
8237 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
8238 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
8239 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
8240 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
8241 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
8242 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
8244 (Dn, Dl, Dr): Accept const as well as const_vector.
8245 (Dz): Likewise. Compare against CONST0_RTX.
8246 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
8247 of "vector" where appropriate.
8248 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
8249 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
8250 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
8251 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
8252 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
8253 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
8254 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
8255 (v_int_equiv): Extend to SVE modes.
8256 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
8258 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
8259 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
8260 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
8261 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
8262 (SVE_COND_FP_CMP): New int iterators.
8263 (perm_hilo): Handle the new unpack unspecs.
8264 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
8266 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
8267 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
8268 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
8269 (aarch64_equality_operator, aarch64_constant_vector_operand)
8270 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
8271 (aarch64_sve_nonimmediate_operand): Likewise.
8272 (aarch64_sve_general_operand): Likewise.
8273 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
8274 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
8275 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
8276 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
8277 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
8278 (aarch64_sve_float_arith_immediate): Likewise.
8279 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
8280 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
8281 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
8282 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
8283 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
8284 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
8285 (aarch64_sve_float_arith_operand): Likewise.
8286 (aarch64_sve_float_arith_with_sub_operand): Likewise.
8287 (aarch64_sve_float_mul_operand): Likewise.
8288 (aarch64_sve_vec_perm_operand): Likewise.
8289 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
8290 (aarch64_mov_operand): Accept const_poly_int and const_vector.
8291 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
8292 as well as const_vector.
8293 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
8294 in file. Use CONST0_RTX and CONSTM1_RTX.
8295 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
8296 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
8297 Use aarch64_simd_imm_zero.
8298 * config/aarch64/aarch64-sve.md: New file.
8299 * config/aarch64/aarch64.md: Include it.
8300 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
8301 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
8302 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
8303 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
8304 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
8305 (sve): New attribute.
8306 (enabled): Disable instructions with the sve attribute unless
8308 (movqi, movhi): Pass CONST_POLY_INT operaneds through
8309 aarch64_expand_mov_immediate.
8310 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
8311 CNT[BHSD] immediates.
8312 (movti): Split CONST_POLY_INT moves into two halves.
8313 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
8314 Split additions that need a temporary here if the destination
8315 is the stack pointer.
8316 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
8317 (*add<mode>3_poly_1): New instruction.
8318 (set_clobber_cc): New expander.
8320 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8322 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
8323 parameter and use it instead of GET_MODE_SIZE (innermode). Use
8324 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
8325 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
8326 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
8327 Change innermode from fixed_mode_size to machine_mode.
8328 (simplify_subreg): Update call accordingly. Handle a constant-sized
8329 subreg of a variable-length CONST_VECTOR.
8331 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8332 Alan Hayward <alan.hayward@arm.com>
8333 David Sherwood <david.sherwood@arm.com>
8335 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
8336 (add_offset_to_base): New function, split out from...
8337 (create_mem_ref): ...here. When handling a scale other than 1,
8338 check first whether the address is valid without the offset.
8339 Add it into the base if so, leaving the index and scale as-is.
8341 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8344 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
8345 fold_for_warn before checking if arg2 is INTEGER_CST.
8347 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
8349 * config/rs6000/predicates.md (load_multiple_operation): Delete.
8350 (store_multiple_operation): Delete.
8351 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
8352 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
8353 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
8354 guarded by TARGET_STRING.
8355 (rs6000_output_load_multiple): Delete.
8356 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
8357 OPTION_MASK_STRING / TARGET_STRING handling.
8358 (print_operand) <'N', 'O'>: Add comment that these are unused now.
8359 (const rs6000_opt_masks) <"string">: Change mask to 0.
8360 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
8361 (MASK_STRING): Delete.
8362 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
8364 (load_multiple): Delete.
8371 (store_multiple): Delete.
8378 (movmemsi_8reg): Delete.
8379 (corresponding unnamed define_insn): Delete.
8380 (movmemsi_6reg): Delete.
8381 (corresponding unnamed define_insn): Delete.
8382 (movmemsi_4reg): Delete.
8383 (corresponding unnamed define_insn): Delete.
8384 (movmemsi_2reg): Delete.
8385 (corresponding unnamed define_insn): Delete.
8386 (movmemsi_1reg): Delete.
8387 (corresponding unnamed define_insn): Delete.
8388 * config/rs6000/rs6000.opt (mno-string): New.
8389 (mstring): Replace by deprecation warning stub.
8390 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
8392 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8394 * regrename.c (regrename_do_replace): If replacing the same
8395 reg multiple times, try to reuse last created gen_raw_REG.
8398 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
8399 main to workaround a bug in GDB.
8401 2018-01-12 Tom de Vries <tom@codesourcery.com>
8404 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
8406 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
8408 PR rtl-optimization/80481
8409 * ira-color.c (get_cap_member): New function.
8410 (allocnos_conflict_by_live_ranges_p): Use it.
8411 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
8412 (setup_slot_coalesced_allocno_live_ranges): Ditto.
8414 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
8417 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
8418 (*saddl_se_1): Ditto.
8420 (*ssubl_se_1): Ditto.
8422 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8424 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
8425 rather than wi::to_widest for DR_INITs.
8426 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
8427 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
8428 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
8430 (vect_analyze_group_access_1): Note that here.
8432 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8434 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
8435 polynomial type sizes.
8437 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8439 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
8440 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
8441 (gimple_add_tmp_var): Likewise.
8443 2018-01-12 Martin Liska <mliska@suse.cz>
8445 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
8446 (gimple_alloc_sizes): Likewise.
8447 (dump_gimple_statistics): Use PRIu64 in printf format.
8448 * gimple.h: Change uint64_t to int.
8450 2018-01-12 Martin Liska <mliska@suse.cz>
8452 * tree-core.h: Use uint64_t instead of int.
8453 * tree.c (tree_node_counts): Likewise.
8454 (tree_node_sizes): Likewise.
8455 (dump_tree_statistics): Use PRIu64 in printf format.
8457 2018-01-12 Martin Liska <mliska@suse.cz>
8459 * Makefile.in: As qsort_chk is implemented in vec.c, add
8460 vec.o to linkage of gencfn-macros.
8461 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
8462 passing the info to record_node_allocation_statistics.
8463 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
8465 * ggc-common.c (struct ggc_usage): Add operator== and use
8466 it in operator< and compare function.
8467 * mem-stats.h (struct mem_usage): Likewise.
8468 * vec.c (struct vec_usage): Remove operator< and compare
8469 function. Can be simply inherited.
8471 2018-01-12 Martin Jambor <mjambor@suse.cz>
8474 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
8475 * tree-ssa-math-opts.c: Include domwalk.h.
8476 (convert_mult_to_fma_1): New function.
8477 (fma_transformation_info): New type.
8478 (fma_deferring_state): Likewise.
8479 (cancel_fma_deferring): New function.
8480 (result_of_phi): Likewise.
8481 (last_fma_candidate_feeds_initial_phi): Likewise.
8482 (convert_mult_to_fma): Added deferring logic, split actual
8483 transformation to convert_mult_to_fma_1.
8484 (math_opts_dom_walker): New type.
8485 (math_opts_dom_walker::after_dom_children): New method, body moved
8486 here from pass_optimize_widening_mul::execute, added deferring logic
8488 (pass_optimize_widening_mul::execute): Moved most of code to
8489 math_opts_dom_walker::after_dom_children.
8490 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
8491 * config/i386/i386.c (ix86_option_override_internal): Added
8492 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
8494 2018-01-12 Richard Biener <rguenther@suse.de>
8497 * dwarf2out.c (gen_variable_die): Do not reset old_die for
8498 inline instance vars.
8500 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
8503 * config/rx/rx.c (rx_is_restricted_memory_address):
8506 2018-01-12 Richard Biener <rguenther@suse.de>
8508 PR tree-optimization/80846
8509 * target.def (split_reduction): New target hook.
8510 * targhooks.c (default_split_reduction): New function.
8511 * targhooks.h (default_split_reduction): Declare.
8512 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
8513 target requests first reduce vectors by combining low and high
8515 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
8516 (get_vectype_for_scalar_type_and_size): Export.
8517 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
8518 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
8519 * doc/tm.texi: Regenerate.
8520 * config/i386/i386.c (ix86_split_reduction): Implement
8521 TARGET_VECTORIZE_SPLIT_REDUCTION.
8523 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8526 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
8527 in PIC mode except for TARGET_VXWORKS_RTP.
8528 * config/sparc/sparc.c: Include cfgrtl.h.
8529 (TARGET_INIT_PIC_REG): Define.
8530 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
8531 (sparc_pic_register_p): New predicate.
8532 (sparc_legitimate_address_p): Use it.
8533 (sparc_legitimize_pic_address): Likewise.
8534 (sparc_delegitimize_address): Likewise.
8535 (sparc_mode_dependent_address_p): Likewise.
8536 (gen_load_pcrel_sym): Remove 4th parameter.
8537 (load_got_register): Adjust call to above. Remove obsolete stuff.
8538 (sparc_expand_prologue): Do not call load_got_register here.
8539 (sparc_flat_expand_prologue): Likewise.
8540 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
8541 (sparc_use_pseudo_pic_reg): New function.
8542 (sparc_init_pic_reg): Likewise.
8543 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8544 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
8546 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
8548 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
8549 Add item for branch_cost.
8551 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8553 PR rtl-optimization/83565
8554 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
8555 not extend the result to a larger mode for rotate operations.
8556 (num_sign_bit_copies1): Likewise.
8558 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8561 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
8563 Use values-Xc.o for -pedantic.
8564 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
8566 2018-01-12 Martin Liska <mliska@suse.cz>
8569 * ipa-devirt.c (final_warning_record::grow_type_warnings):
8571 (possible_polymorphic_call_targets): Use it.
8572 (ipa_devirt): Likewise.
8574 2018-01-12 Martin Liska <mliska@suse.cz>
8576 * profile-count.h (enum profile_quality): Use 0 as invalid
8577 enum value of profile_quality.
8579 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
8581 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
8582 -mext-string options.
8584 2018-01-12 Richard Biener <rguenther@suse.de>
8586 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
8587 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
8588 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
8590 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
8592 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
8594 * configure.ac (--with-long-double-format): Add support for the
8595 configuration option to change the default long double format on
8597 * config.gcc (powerpc*-linux*-*): Likewise.
8598 * configure: Regenerate.
8599 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
8600 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
8601 used without modification.
8603 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8605 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
8606 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
8607 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
8608 MISC_BUILTIN_SPEC_BARRIER.
8609 (rs6000_init_builtins): Likewise.
8610 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
8612 (speculation_barrier): New define_insn.
8613 * doc/extend.texi: Document __builtin_speculation_barrier.
8615 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8618 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
8619 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
8620 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
8622 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
8623 integral modes instead of "ss" and "sd".
8624 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
8625 vectors with 32-bit and 64-bit elements.
8626 (vecdupssescalarmodesuffix): New mode attribute.
8627 (vec_dup<mode>): Use it.
8629 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
8632 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
8633 frame if argument is passed on stack.
8635 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8638 * ree.c (combine_reaching_defs): Optimize also
8639 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
8640 reg2=any_extend(exp); reg1=reg2;, formatting fix.
8642 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8645 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
8647 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8650 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
8651 after they are computed.
8653 2018-01-11 Bin Cheng <bin.cheng@arm.com>
8655 PR tree-optimization/83695
8656 * gimple-loop-linterchange.cc
8657 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
8658 reset cached scev information after interchange.
8659 (pass_linterchange::execute): Remove call to scev_reset_htab.
8661 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8663 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
8664 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
8665 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
8666 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
8667 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
8668 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
8669 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
8670 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
8671 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
8672 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
8673 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
8674 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
8675 (V_lane_reg): Likewise.
8676 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
8678 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
8679 (vfmal_lane_low<mode>_intrinsic,
8680 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
8681 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
8682 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
8683 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
8684 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
8685 vfmsl_lane_high<mode>_intrinsic): New define_insns.
8687 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8689 * config/arm/arm-cpus.in (fp16fml): New feature.
8690 (ALL_SIMD): Add fp16fml.
8691 (armv8.2-a): Add fp16fml as an option.
8692 (armv8.3-a): Likewise.
8693 (armv8.4-a): Add fp16fml as part of fp16.
8694 * config/arm/arm.h (TARGET_FP16FML): Define.
8695 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
8697 * config/arm/arm-modes.def (V2HF): Define.
8698 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
8699 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
8700 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
8701 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
8702 vfmsl_low, vfmsl_high): New set of builtins.
8703 * config/arm/iterators.md (PLUSMINUS): New code iterator.
8704 (vfml_op): New code attribute.
8705 (VFMLHALVES): New int iterator.
8706 (VFML, VFMLSEL): New mode attributes.
8707 (V_reg): Define mapping for V2HF.
8708 (V_hi, V_lo): New mode attributes.
8709 (VF_constraint): Likewise.
8710 (vfml_half, vfml_half_selector): New int attributes.
8711 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
8713 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
8714 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
8716 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
8717 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
8718 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
8719 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
8721 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
8722 Document new effective target and option set.
8724 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8726 * config/arm/arm-cpus.in (armv8_4): New feature.
8727 (ARMv8_4a): New fgroup.
8728 (armv8.4-a): New arch.
8729 * config/arm/arm-tables.opt: Regenerate.
8730 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
8731 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
8732 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
8733 Add matching rules for -march=armv8.4-a and extensions.
8734 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
8736 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
8739 * config/rx/rx.md (BW): New mode attribute.
8740 (sync_lock_test_and_setsi): Add mode suffix to insn output.
8742 2018-01-11 Richard Biener <rguenther@suse.de>
8744 PR tree-optimization/83435
8745 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
8746 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
8747 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
8749 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8750 Alan Hayward <alan.hayward@arm.com>
8751 David Sherwood <david.sherwood@arm.com>
8753 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
8755 (aarch64_classify_address): Initialize it. Track polynomial offsets.
8756 (aarch64_print_address_internal): Use it to check for a zero offset.
8758 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8759 Alan Hayward <alan.hayward@arm.com>
8760 David Sherwood <david.sherwood@arm.com>
8762 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
8763 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
8764 Return a poly_int64 rather than a HOST_WIDE_INT.
8765 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
8766 rather than a HOST_WIDE_INT.
8767 * config/aarch64/aarch64.h (aarch64_frame): Protect with
8768 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
8769 hard_fp_offset, frame_size, initial_adjust, callee_offset and
8770 final_offset from HOST_WIDE_INT to poly_int64.
8771 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
8772 to_constant when getting the number of units in an Advanced SIMD
8774 (aarch64_builtin_vectorized_function): Check for a constant number
8776 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
8778 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
8779 attribute instead of GET_MODE_NUNITS.
8780 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
8781 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
8782 GET_MODE_SIZE for fixed-size registers.
8783 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
8784 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
8785 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
8786 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
8787 (aarch64_print_operand, aarch64_print_address_internal)
8788 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
8789 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
8790 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
8791 Handle polynomial GET_MODE_SIZE.
8792 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
8793 wider than SImode without modification.
8794 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
8795 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
8796 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
8797 passing and returning SVE modes.
8798 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
8799 rather than GEN_INT.
8800 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
8801 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
8802 (aarch64_allocate_and_probe_stack_space): Likewise.
8803 (aarch64_layout_frame): Cope with polynomial offsets.
8804 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
8805 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
8807 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
8808 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
8809 poly_int64 rather than a HOST_WIDE_INT.
8810 (aarch64_get_separate_components, aarch64_process_components)
8811 (aarch64_expand_prologue, aarch64_expand_epilogue)
8812 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
8813 (aarch64_anchor_offset): New function, split out from...
8814 (aarch64_legitimize_address): ...here.
8815 (aarch64_builtin_vectorization_cost): Handle polynomial
8816 TYPE_VECTOR_SUBPARTS.
8817 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
8819 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
8820 number of elements from the PARALLEL rather than the mode.
8821 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
8822 rather than GET_MODE_BITSIZE.
8823 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
8824 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
8825 (aarch64_expand_vec_perm_const_1): Handle polynomial
8826 d->perm.length () and d->perm elements.
8827 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
8828 Apply to_constant to d->perm elements.
8829 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
8830 polynomial CONST_VECTOR_NUNITS.
8831 (aarch64_move_pointer): Take amount as a poly_int64 rather
8833 (aarch64_progress_pointer): Avoid temporary variable.
8834 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
8835 the mode attribute instead of GET_MODE.
8837 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8838 Alan Hayward <alan.hayward@arm.com>
8839 David Sherwood <david.sherwood@arm.com>
8841 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
8842 x exists before using it.
8843 (aarch64_add_constant_internal): Rename to...
8844 (aarch64_add_offset_1): ...this. Replace regnum with separate
8845 src and dest rtxes. Handle the case in which they're different,
8846 including when the offset is zero. Replace scratchreg with an rtx.
8847 Use 2 additions if there is no spare register into which we can
8848 move a 16-bit constant.
8849 (aarch64_add_constant): Delete.
8850 (aarch64_add_offset): Replace reg with separate src and dest
8851 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
8852 Use aarch64_add_offset_1.
8853 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
8854 an rtx rather than an int. Take the delta as a poly_int64
8855 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
8856 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
8857 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
8858 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
8859 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
8861 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
8862 aarch64_add_constant.
8864 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8866 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
8867 Use scalar_float_mode.
8869 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8871 * config/aarch64/aarch64-simd.md
8872 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
8873 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
8874 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
8875 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
8876 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
8877 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
8878 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
8879 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
8880 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
8881 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
8883 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8886 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
8887 targ_options->x_arm_arch_string is non NULL.
8889 2018-01-11 Tamar Christina <tamar.christina@arm.com>
8891 * config/aarch64/aarch64.h
8892 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
8894 2018-01-11 Sudakshina Das <sudi.das@arm.com>
8897 * expmed.c (emit_store_flag_force): Swap if const op0
8898 and change VOIDmode to mode of op0.
8900 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8902 PR rtl-optimization/83761
8903 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
8904 than bytes to mode_for_size.
8906 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8909 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
8910 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
8913 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8916 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
8917 when in layout mode.
8918 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
8919 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
8922 2018-01-10 Michael Collison <michael.collison@arm.com>
8924 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
8925 * config/aarch64/aarch64-option-extension.def: Add
8926 AARCH64_OPT_EXTENSION of 'fp16fml'.
8927 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8928 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
8929 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
8930 * config/aarch64/constraints.md (Ui7): New constraint.
8931 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
8932 (VFMLA_SEL_W): Ditto.
8935 (VFMLA16_LOW): New int iterator.
8936 (VFMLA16_HIGH): Ditto.
8937 (UNSPEC_FMLAL): New unspec.
8938 (UNSPEC_FMLSL): Ditto.
8939 (UNSPEC_FMLAL2): Ditto.
8940 (UNSPEC_FMLSL2): Ditto.
8941 (f16mac): New code attribute.
8942 * config/aarch64/aarch64-simd-builtins.def
8943 (aarch64_fmlal_lowv2sf): Ditto.
8944 (aarch64_fmlsl_lowv2sf): Ditto.
8945 (aarch64_fmlalq_lowv4sf): Ditto.
8946 (aarch64_fmlslq_lowv4sf): Ditto.
8947 (aarch64_fmlal_highv2sf): Ditto.
8948 (aarch64_fmlsl_highv2sf): Ditto.
8949 (aarch64_fmlalq_highv4sf): Ditto.
8950 (aarch64_fmlslq_highv4sf): Ditto.
8951 (aarch64_fmlal_lane_lowv2sf): Ditto.
8952 (aarch64_fmlsl_lane_lowv2sf): Ditto.
8953 (aarch64_fmlal_laneq_lowv2sf): Ditto.
8954 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
8955 (aarch64_fmlalq_lane_lowv4sf): Ditto.
8956 (aarch64_fmlsl_lane_lowv4sf): Ditto.
8957 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
8958 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
8959 (aarch64_fmlal_lane_highv2sf): Ditto.
8960 (aarch64_fmlsl_lane_highv2sf): Ditto.
8961 (aarch64_fmlal_laneq_highv2sf): Ditto.
8962 (aarch64_fmlsl_laneq_highv2sf): Ditto.
8963 (aarch64_fmlalq_lane_highv4sf): Ditto.
8964 (aarch64_fmlsl_lane_highv4sf): Ditto.
8965 (aarch64_fmlalq_laneq_highv4sf): Ditto.
8966 (aarch64_fmlsl_laneq_highv4sf): Ditto.
8967 * config/aarch64/aarch64-simd.md:
8968 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
8969 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8970 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
8971 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8972 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
8973 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
8974 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
8975 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
8976 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
8977 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
8978 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
8979 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
8980 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
8981 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
8982 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
8983 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
8984 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
8985 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
8986 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
8987 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
8988 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
8989 (vfmlsl_low_u32): Ditto.
8990 (vfmlalq_low_u32): Ditto.
8991 (vfmlslq_low_u32): Ditto.
8992 (vfmlal_high_u32): Ditto.
8993 (vfmlsl_high_u32): Ditto.
8994 (vfmlalq_high_u32): Ditto.
8995 (vfmlslq_high_u32): Ditto.
8996 (vfmlal_lane_low_u32): Ditto.
8997 (vfmlsl_lane_low_u32): Ditto.
8998 (vfmlal_laneq_low_u32): Ditto.
8999 (vfmlsl_laneq_low_u32): Ditto.
9000 (vfmlalq_lane_low_u32): Ditto.
9001 (vfmlslq_lane_low_u32): Ditto.
9002 (vfmlalq_laneq_low_u32): Ditto.
9003 (vfmlslq_laneq_low_u32): Ditto.
9004 (vfmlal_lane_high_u32): Ditto.
9005 (vfmlsl_lane_high_u32): Ditto.
9006 (vfmlal_laneq_high_u32): Ditto.
9007 (vfmlsl_laneq_high_u32): Ditto.
9008 (vfmlalq_lane_high_u32): Ditto.
9009 (vfmlslq_lane_high_u32): Ditto.
9010 (vfmlalq_laneq_high_u32): Ditto.
9011 (vfmlslq_laneq_high_u32): Ditto.
9012 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
9013 (AARCH64_FL_FOR_ARCH8_4): New.
9014 (AARCH64_ISA_F16FML): New ISA flag.
9015 (TARGET_F16FML): New feature flag for fp16fml.
9016 (doc/invoke.texi): Document new fp16fml option.
9018 2018-01-10 Michael Collison <michael.collison@arm.com>
9020 * config/aarch64/aarch64-builtins.c:
9021 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
9022 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9023 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
9024 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
9025 (AARCH64_ISA_SHA3): New ISA flag.
9026 (TARGET_SHA3): New feature flag for sha3.
9027 * config/aarch64/iterators.md (sha512_op): New int attribute.
9028 (CRYPTO_SHA512): New int iterator.
9029 (UNSPEC_SHA512H): New unspec.
9030 (UNSPEC_SHA512H2): Ditto.
9031 (UNSPEC_SHA512SU0): Ditto.
9032 (UNSPEC_SHA512SU1): Ditto.
9033 * config/aarch64/aarch64-simd-builtins.def
9034 (aarch64_crypto_sha512hqv2di): New builtin.
9035 (aarch64_crypto_sha512h2qv2di): Ditto.
9036 (aarch64_crypto_sha512su0qv2di): Ditto.
9037 (aarch64_crypto_sha512su1qv2di): Ditto.
9038 (aarch64_eor3qv8hi): Ditto.
9039 (aarch64_rax1qv2di): Ditto.
9040 (aarch64_xarqv2di): Ditto.
9041 (aarch64_bcaxqv8hi): Ditto.
9042 * config/aarch64/aarch64-simd.md:
9043 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
9044 (aarch64_crypto_sha512su0qv2di): Ditto.
9045 (aarch64_crypto_sha512su1qv2di): Ditto.
9046 (aarch64_eor3qv8hi): Ditto.
9047 (aarch64_rax1qv2di): Ditto.
9048 (aarch64_xarqv2di): Ditto.
9049 (aarch64_bcaxqv8hi): Ditto.
9050 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
9051 (vsha512h2q_u64): Ditto.
9052 (vsha512su0q_u64): Ditto.
9053 (vsha512su1q_u64): Ditto.
9054 (veor3q_u16): Ditto.
9055 (vrax1q_u64): Ditto.
9057 (vbcaxq_u16): Ditto.
9058 * config/arm/types.md (crypto_sha512): New type attribute.
9059 (crypto_sha3): Ditto.
9060 (doc/invoke.texi): Document new sha3 option.
9062 2018-01-10 Michael Collison <michael.collison@arm.com>
9064 * config/aarch64/aarch64-builtins.c:
9065 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
9066 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9067 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
9068 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
9069 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
9070 (AARCH64_ISA_SM4): New ISA flag.
9071 (TARGET_SM4): New feature flag for sm4.
9072 * config/aarch64/aarch64-simd-builtins.def
9073 (aarch64_sm3ss1qv4si): Ditto.
9074 (aarch64_sm3tt1aq4si): Ditto.
9075 (aarch64_sm3tt1bq4si): Ditto.
9076 (aarch64_sm3tt2aq4si): Ditto.
9077 (aarch64_sm3tt2bq4si): Ditto.
9078 (aarch64_sm3partw1qv4si): Ditto.
9079 (aarch64_sm3partw2qv4si): Ditto.
9080 (aarch64_sm4eqv4si): Ditto.
9081 (aarch64_sm4ekeyqv4si): Ditto.
9082 * config/aarch64/aarch64-simd.md:
9083 (aarch64_sm3ss1qv4si): Ditto.
9084 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
9085 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
9086 (aarch64_sm4eqv4si): Ditto.
9087 (aarch64_sm4ekeyqv4si): Ditto.
9088 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
9089 (sm3part_op): Ditto.
9090 (CRYPTO_SM3TT): Ditto.
9091 (CRYPTO_SM3PART): Ditto.
9092 (UNSPEC_SM3SS1): New unspec.
9093 (UNSPEC_SM3TT1A): Ditto.
9094 (UNSPEC_SM3TT1B): Ditto.
9095 (UNSPEC_SM3TT2A): Ditto.
9096 (UNSPEC_SM3TT2B): Ditto.
9097 (UNSPEC_SM3PARTW1): Ditto.
9098 (UNSPEC_SM3PARTW2): Ditto.
9099 (UNSPEC_SM4E): Ditto.
9100 (UNSPEC_SM4EKEY): Ditto.
9101 * config/aarch64/constraints.md (Ui2): New constraint.
9102 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
9103 * config/arm/types.md (crypto_sm3): New type attribute.
9104 (crypto_sm4): Ditto.
9105 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
9106 (vsm3tt1aq_u32): Ditto.
9107 (vsm3tt1bq_u32): Ditto.
9108 (vsm3tt2aq_u32): Ditto.
9109 (vsm3tt2bq_u32): Ditto.
9110 (vsm3partw1q_u32): Ditto.
9111 (vsm3partw2q_u32): Ditto.
9112 (vsm4eq_u32): Ditto.
9113 (vsm4ekeyq_u32): Ditto.
9114 (doc/invoke.texi): Document new sm4 option.
9116 2018-01-10 Michael Collison <michael.collison@arm.com>
9118 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
9119 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
9120 (AARCH64_FL_FOR_ARCH8_4): New.
9121 (AARCH64_FL_V8_4): New flag.
9122 (doc/invoke.texi): Document new armv8.4-a option.
9124 2018-01-10 Michael Collison <michael.collison@arm.com>
9126 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9127 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
9128 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
9129 * config/aarch64/aarch64-option-extension.def: Add
9130 AARCH64_OPT_EXTENSION of 'sha2'.
9131 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
9132 (crypto): Disable sha2 and aes if crypto disabled.
9133 (crypto): Enable aes and sha2 if enabled.
9134 (simd): Disable sha2 and aes if simd disabled.
9135 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
9137 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
9138 (TARGET_SHA2): New feature flag for sha2.
9139 (TARGET_AES): New feature flag for aes.
9140 * config/aarch64/aarch64-simd.md:
9141 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
9142 conditional on TARGET_AES.
9143 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
9144 (aarch64_crypto_sha1hsi): Make pattern conditional
9146 (aarch64_crypto_sha1hv4si): Ditto.
9147 (aarch64_be_crypto_sha1hv4si): Ditto.
9148 (aarch64_crypto_sha1su1v4si): Ditto.
9149 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
9150 (aarch64_crypto_sha1su0v4si): Ditto.
9151 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
9152 (aarch64_crypto_sha256su0v4si): Ditto.
9153 (aarch64_crypto_sha256su1v4si): Ditto.
9154 (doc/invoke.texi): Document new aes and sha2 options.
9156 2018-01-10 Martin Sebor <msebor@redhat.com>
9158 PR tree-optimization/83781
9159 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
9162 2018-01-11 Martin Sebor <msebor@gmail.com>
9163 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9165 PR tree-optimization/83501
9166 PR tree-optimization/81703
9168 * tree-ssa-strlen.c (get_string_cst): Rename...
9169 (get_string_len): ...to this. Handle global constants.
9170 (handle_char_store): Adjust.
9172 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
9173 Jim Wilson <jimw@sifive.com>
9175 * config/riscv/riscv-protos.h (riscv_output_return): New.
9176 * config/riscv/riscv.c (struct machine_function): New naked_p field.
9177 (riscv_attribute_table, riscv_output_return),
9178 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
9179 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
9180 (riscv_compute_frame_info): Only compute frame->mask if not a naked
9182 (riscv_expand_prologue): Add early return for naked function.
9183 (riscv_expand_epilogue): Likewise.
9184 (riscv_function_ok_for_sibcall): Return false for naked function.
9185 (riscv_set_current_function): New.
9186 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
9187 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
9188 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
9189 * doc/extend.texi (RISC-V Function Attributes): New.
9191 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
9193 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
9194 check for 128-bit long double before checking TCmode.
9195 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
9196 128-bit long doubles before checking TFmode or TCmode.
9197 (FLOAT128_IBM_P): Likewise.
9199 2018-01-10 Martin Sebor <msebor@redhat.com>
9201 PR tree-optimization/83671
9202 * builtins.c (c_strlen): Unconditionally return zero for the empty
9204 Use -Warray-bounds for warnings.
9205 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
9206 for non-constant array indices with COMPONENT_REF, arrays of
9207 arrays, and pointers to arrays.
9208 (gimple_fold_builtin_strlen): Determine and set length range for
9209 non-constant character arrays.
9211 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
9214 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
9217 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
9219 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
9221 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9224 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
9225 VECTOR_MEM_ALTIVEC_OR_VSX_P.
9226 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
9227 indexed_or_indirect_operand predicate.
9228 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
9229 (*vsx_le_perm_load_v8hi): Likewise.
9230 (*vsx_le_perm_load_v16qi): Likewise.
9231 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
9232 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
9233 (*vsx_le_perm_store_v8hi): Likewise.
9234 (*vsx_le_perm_store_v16qi): Likewise.
9235 (eight unnamed splitters): Likewise.
9237 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9239 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
9240 * config/rs6000/emmintrin.h: Likewise.
9241 * config/rs6000/mmintrin.h: Likewise.
9242 * config/rs6000/xmmintrin.h: Likewise.
9244 2018-01-10 David Malcolm <dmalcolm@redhat.com>
9247 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
9249 * tree.c (tree_nop_conversion): Return true for location wrapper
9251 (maybe_wrap_with_location): New function.
9252 (selftest::check_strip_nops): New function.
9253 (selftest::test_location_wrappers): New function.
9254 (selftest::tree_c_tests): Call it.
9255 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
9256 (maybe_wrap_with_location): New decl.
9257 (EXPR_LOCATION_WRAPPER_P): New macro.
9258 (location_wrapper_p): New inline function.
9259 (tree_strip_any_location_wrapper): New inline function.
9261 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
9264 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
9265 stack_realign_offset for the largest alignment of stack slot
9267 (ix86_find_max_used_stack_alignment): New function.
9268 (ix86_finalize_stack_frame_flags): Use it. Set
9269 max_used_stack_alignment if we don't realign stack.
9270 * config/i386/i386.h (machine_function): Add
9271 max_used_stack_alignment.
9273 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
9275 * config/arm/arm.opt (-mbranch-cost): New option.
9276 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
9279 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
9282 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
9283 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
9285 2018-01-10 Richard Biener <rguenther@suse.de>
9288 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
9289 early out so it also covers the case where we have a non-NULL
9292 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9294 PR tree-optimization/83753
9295 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
9296 for non-strided grouped accesses if the number of elements is 1.
9298 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9301 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
9302 * i386.h (TARGET_USE_GATHER): Define.
9303 * x86-tune.def (X86_TUNE_USE_GATHER): New.
9305 2018-01-10 Martin Liska <mliska@suse.cz>
9308 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
9309 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
9311 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
9312 CLEANUP_NO_PARTITIONING is not set.
9314 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9316 * doc/rtl.texi: Remove documentation of (const ...) wrappers
9317 for vectors, as a partial revert of r254296.
9318 * rtl.h (const_vec_p): Delete.
9319 (const_vec_duplicate_p): Don't test for vector CONSTs.
9320 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
9321 * expmed.c (make_tree): Likewise.
9324 * common.md (E, F): Use CONSTANT_P instead of checking for
9326 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9327 checking for CONST_VECTOR.
9329 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9332 * predict.c (force_edge_cold): Handle in more sane way edges
9335 2018-01-09 Carl Love <cel@us.ibm.com>
9337 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
9339 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
9340 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
9341 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
9342 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
9343 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
9344 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
9345 * config/rs6000/rs6000-protos.h: Add extern defition for
9346 rs6000_generate_float2_double_code.
9347 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
9349 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
9350 (float2_v2df): Add define_expand.
9352 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
9355 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
9356 op_mode in the force_to_mode call.
9358 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9360 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
9361 instead of checking each element individually.
9362 (aarch64_evpc_uzp): Likewise.
9363 (aarch64_evpc_zip): Likewise.
9364 (aarch64_evpc_ext): Likewise.
9365 (aarch64_evpc_rev): Likewise.
9366 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
9367 instead of checking each element individually. Return true without
9369 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
9370 whether all selected elements come from the same input, instead of
9371 checking each element individually. Remove calls to gen_rtx_REG,
9372 start_sequence and end_sequence and instead assert that no rtl is
9375 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9377 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
9378 order of HIGH and CONST checks.
9380 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9382 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
9383 if the destination isn't an SSA_NAME.
9385 2018-01-09 Richard Biener <rguenther@suse.de>
9387 PR tree-optimization/83668
9388 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
9390 (canonicalize_loop_form): ... here, renamed from ...
9391 (canonicalize_loop_closed_ssa_form): ... this and amended to
9392 swap successor edges for loop exit blocks to make us use
9393 the RPO order we need for initial schedule generation.
9395 2018-01-09 Joseph Myers <joseph@codesourcery.com>
9397 PR tree-optimization/64811
9398 * match.pd: When optimizing comparisons with Inf, avoid
9399 introducing or losing exceptions from comparisons with NaN.
9401 2018-01-09 Martin Liska <mliska@suse.cz>
9404 * asan.c (shadow_mem_size): Add gcc_assert.
9406 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
9408 Don't save registers in main().
9411 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
9412 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
9413 * config/avr/avr.c (avr_set_current_function): Don't error if
9414 naked, OS_task or OS_main are specified at the same time.
9415 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
9417 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
9419 * common/config/avr/avr-common.c (avr_option_optimization_table):
9420 Switch on -mmain-is-OS_task for optimizing compilations.
9422 2018-01-09 Richard Biener <rguenther@suse.de>
9424 PR tree-optimization/83572
9425 * graphite.c: Include cfganal.h.
9426 (graphite_transform_loops): Connect infinite loops to exit
9427 and remove fake edges at the end.
9429 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9431 * ipa-inline.c (edge_badness): Revert accidental checkin.
9433 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9436 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
9437 symbols; not inline clones.
9439 2018-01-09 Jakub Jelinek <jakub@redhat.com>
9442 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
9443 hard registers. Formatting fixes.
9445 PR preprocessor/83722
9446 * gcc.c (try_generate_repro): Pass
9447 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
9448 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
9451 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
9452 Kito Cheng <kito.cheng@gmail.com>
9454 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
9455 (riscv_leaf_function_p): Delete.
9456 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
9458 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9460 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
9462 (do_ifelse): New function.
9463 (do_isel): New function.
9464 (do_sub3): New function.
9465 (do_add3): New function.
9466 (do_load_mask_compare): New function.
9467 (do_overlap_load_compare): New function.
9468 (expand_compare_loop): New function.
9469 (expand_block_compare): Call expand_compare_loop() when appropriate.
9470 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
9472 (-mblock-compare-inline-loop-limit): New option.
9474 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9477 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
9478 Reverse order of second and third operands in first alternative.
9479 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
9480 of first and second elements in UNSPEC_VPERMR vector.
9481 (altivec_expand_vec_perm_le): Likewise.
9483 2018-01-08 Jeff Law <law@redhat.com>
9485 PR rtl-optimizatin/81308
9486 * tree-switch-conversion.c (cfg_altered): New file scoped static.
9487 (process_switch): If group_case_labels makes a change, then set
9489 (pass_convert_switch::execute): If a switch is converted, then
9490 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
9492 PR rtl-optimization/81308
9493 * recog.c (split_all_insns): Conditionally cleanup the CFG after
9496 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
9498 PR target/83663 - Revert r255946
9499 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
9500 generation for cases where splatting a value is not useful.
9501 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
9502 across a vec_duplicate and a paradoxical subreg forming a vector
9503 mode to a vec_concat.
9505 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9507 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
9508 -march=armv8.3-a variants.
9509 * config/arm/t-multilib: Likewise.
9510 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
9512 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9514 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
9516 (cceq_ior_compare_complement): Give it a name so I can use it, and
9517 change boolean_or_operator predicate to boolean_operator so it can
9518 be used to generate a crand.
9519 (eqne): New code iterator.
9520 (bd/bd_neg): New code_attrs.
9521 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
9522 a single define_insn.
9523 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
9524 decrement (bdnzt/bdnzf/bdzt/bdzf).
9525 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
9526 with the new names of the branch decrement patterns, and added the
9527 names of the branch decrement conditional patterns.
9529 2018-01-08 Richard Biener <rguenther@suse.de>
9531 PR tree-optimization/83563
9532 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
9535 2018-01-08 Richard Biener <rguenther@suse.de>
9538 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
9540 2018-01-08 Richard Biener <rguenther@suse.de>
9542 PR tree-optimization/83685
9543 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
9544 references to abnormals.
9546 2018-01-08 Richard Biener <rguenther@suse.de>
9549 * dwarf2out.c (output_indirect_strings): Handle empty
9550 skeleton_debug_str_hash.
9551 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
9553 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9555 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
9556 (emit_store_direct): Likewise.
9557 (arc_trampoline_adjust_address): Likewise.
9558 (arc_asm_trampoline_template): New function.
9559 (arc_initialize_trampoline): Use asm_trampoline_template.
9560 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
9561 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
9562 * config/arc/arc.md (flush_icache): Delete pattern.
9564 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9566 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
9567 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
9570 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9573 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
9574 by not USED_FOR_TARGET.
9575 (make_pass_resolve_sw_modes): Likewise.
9577 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9579 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
9582 2018-01-08 Richard Biener <rguenther@suse.de>
9585 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
9587 2018-01-08 Richard Biener <rguenther@suse.de>
9590 * match.pd ((t * 2) / 2) -> t): Add missing :c.
9592 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
9595 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
9596 basic blocks with a small number of successors.
9597 (convert_control_dep_chain_into_preds): Improve handling of
9599 (dump_predicates): Split apart into...
9600 (dump_pred_chain): ...here...
9601 (dump_pred_info): ...and here.
9602 (can_one_predicate_be_invalidated_p): Add debugging printfs.
9603 (can_chain_union_be_invalidated_p): Improve check for invalidation
9605 (uninit_uses_cannot_happen): Avoid unnecessary if
9606 convert_control_dep_chain_into_preds yielded nothing.
9608 2018-01-06 Martin Sebor <msebor@redhat.com>
9610 PR tree-optimization/83640
9611 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
9612 subtracting negative offset from size.
9613 (builtin_access::overlap): Adjust offset bounds of the access to fall
9614 within the size of the object if possible.
9616 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
9618 PR rtl-optimization/83699
9619 * expmed.c (extract_bit_field_1): Restrict the vector usage of
9620 extract_bit_field_as_subreg to cases in which the extracted
9621 value is also a vector.
9623 * lra-constraints.c (process_alt_operands): Test for the equivalence
9624 substitutions when detecting a possible reload cycle.
9626 2018-01-06 Jakub Jelinek <jakub@redhat.com>
9629 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
9630 by default if flag_selective_schedling{,2}. Formatting fixes.
9632 PR rtl-optimization/83682
9633 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
9634 if it has non-VECTOR_MODE element mode.
9635 (vec_duplicate_p): Likewise.
9638 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
9639 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
9641 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9644 * config/i386/i386-builtin.def
9645 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
9646 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
9647 Require also OPTION_MASK_ISA_AVX512F in addition to
9648 OPTION_MASK_ISA_GFNI.
9649 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
9650 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
9651 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
9652 to OPTION_MASK_ISA_GFNI.
9653 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
9654 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
9655 OPTION_MASK_ISA_AVX512BW.
9656 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
9657 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
9658 addition to OPTION_MASK_ISA_GFNI.
9659 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
9660 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
9661 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
9662 to OPTION_MASK_ISA_GFNI.
9663 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
9664 a requirement for all ISAs rather than any of them with a few
9666 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
9668 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
9669 bitmasks to be enabled with 3 exceptions, instead of requiring any
9670 enabled ISA with lots of exceptions.
9671 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
9672 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
9673 Change avx512bw in isa attribute to avx512f.
9674 * config/i386/sgxintrin.h: Add license boilerplate.
9675 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
9676 to __AVX512F__ and __AVX512VL to __AVX512VL__.
9677 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
9678 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
9680 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
9681 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
9682 temporarily sse2 rather than sse if not enabled already.
9685 * config/i386/sse.md (VI248_VLBW): Rename to ...
9686 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
9687 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
9688 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
9689 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
9690 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
9691 mode iterator instead of VI248_VLBW.
9693 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
9695 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
9696 (record_modified): Skip clobbers; add debug output.
9697 (param_change_prob): Use sreal frequencies.
9699 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9701 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
9702 punt for user-aligned variables.
9704 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9706 * tree-chrec.c (chrec_contains_symbols): Return true for
9709 2018-01-05 Sudakshina Das <sudi.das@arm.com>
9712 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
9713 of (x|y) == x for BICS pattern.
9715 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9717 PR tree-optimization/83605
9718 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
9719 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
9722 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
9724 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
9725 * config/epiphany/rtems.h: New file.
9727 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9728 Uros Bizjak <ubizjak@gmail.com>
9731 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
9732 QIreg_operand instead of register_operand predicate.
9733 * config/i386/i386.c (ix86_rop_should_change_byte_p,
9734 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
9735 comments instead of -fmitigate[-_]rop.
9737 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9740 * cgraphunit.c (symbol_table::compile): Switch to text_section
9741 before calling assembly_start debug hook.
9742 * run-rtl-passes.c (run_rtl_passes): Likewise.
9745 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9747 * tree-vrp.c (extract_range_from_binary_expr_1): Check
9748 range_int_cst_p rather than !symbolic_range_p before calling
9749 extract_range_from_multiplicative_op_1.
9751 2018-01-04 Jeff Law <law@redhat.com>
9753 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
9754 redundant test in assertion.
9756 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9758 * doc/rtl.texi: Document machine_mode wrapper classes.
9760 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9762 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
9765 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9767 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
9768 the VEC_PERM_EXPR fold to fail.
9770 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9773 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
9774 to switched_sections.
9776 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9779 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
9782 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
9785 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
9786 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
9788 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9791 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
9792 is BLKmode and bitpos not zero or mode change is needed.
9794 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9797 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
9800 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
9803 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
9804 instead of MULT rtx. Update all corresponding splitters.
9806 (*ssub<modesuffix>): Ditto.
9808 (*cmp_sadd_di): Update split patterns.
9809 (*cmp_sadd_si): Ditto.
9810 (*cmp_sadd_sidi): Ditto.
9811 (*cmp_ssub_di): Ditto.
9812 (*cmp_ssub_si): Ditto.
9813 (*cmp_ssub_sidi): Ditto.
9814 * config/alpha/predicates.md (const23_operand): New predicate.
9815 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
9816 Look for ASHIFT, not MULT inner operand.
9817 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
9819 2018-01-04 Martin Liska <mliska@suse.cz>
9821 PR gcov-profile/83669
9822 * gcov.c (output_intermediate_file): Add version to intermediate
9824 * doc/gcov.texi: Document new field 'version' in intermediate
9825 file format. Fix location of '-k' option of gcov command.
9827 2018-01-04 Martin Liska <mliska@suse.cz>
9830 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
9832 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9834 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
9836 2018-01-03 Martin Sebor <msebor@redhat.com>
9838 PR tree-optimization/83655
9839 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
9840 checking calls with invalid arguments.
9842 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9844 * tree-vect-stmts.c (vect_get_store_rhs): New function.
9845 (vectorizable_mask_load_store): Delete.
9846 (vectorizable_call): Return false for masked loads and stores.
9847 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
9848 instead of gimple_assign_rhs1.
9849 (vectorizable_load): Handle IFN_MASK_LOAD.
9850 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
9852 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9854 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
9856 (vectorizable_mask_load_store): ...here.
9857 (vectorizable_load): ...and here.
9859 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9861 * tree-vect-stmts.c (vect_build_all_ones_mask)
9862 (vect_build_zero_merge_argument): New functions, split out from...
9863 (vectorizable_load): ...here.
9865 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9867 * tree-vect-stmts.c (vect_check_store_rhs): New function,
9869 (vectorizable_mask_load_store): ...here.
9870 (vectorizable_store): ...and here.
9872 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9874 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
9876 (vectorizable_mask_load_store): ...here.
9878 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9880 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
9881 (vect_model_store_cost): Take a vec_load_store_type instead of a
9883 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
9884 (vect_model_store_cost): Take a vec_load_store_type instead of a
9886 (vectorizable_mask_load_store): Update accordingly.
9887 (vectorizable_store): Likewise.
9888 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
9890 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9892 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
9893 IFN_MASK_LOAD calls here rather than...
9894 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
9896 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9897 Alan Hayward <alan.hayward@arm.com>
9898 David Sherwood <david.sherwood@arm.com>
9900 * expmed.c (extract_bit_field_1): For vector extracts,
9901 fall back to extract_bit_field_as_subreg if vec_extract
9904 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9905 Alan Hayward <alan.hayward@arm.com>
9906 David Sherwood <david.sherwood@arm.com>
9908 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
9909 they are variable or constant sized.
9910 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
9911 slots for constant-sized data.
9913 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9914 Alan Hayward <alan.hayward@arm.com>
9915 David Sherwood <david.sherwood@arm.com>
9917 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
9918 handling COND_EXPRs with boolean comparisons, try to find a better
9919 basis for the mask type than the boolean itself.
9921 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9923 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
9924 is calculated and how it can be overridden.
9925 * genmodes.c (max_bitsize_mode_any_mode): New variable.
9926 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
9928 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
9931 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9932 Alan Hayward <alan.hayward@arm.com>
9933 David Sherwood <david.sherwood@arm.com>
9935 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
9936 Remove the mode argument.
9937 (aarch64_simd_valid_immediate): Remove the mode and inverse
9939 * config/aarch64/iterators.md (bitsize): New iterator.
9940 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
9941 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
9942 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
9943 aarch64_simd_valid_immediate.
9944 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
9945 (aarch64_reg_or_bic_imm): Likewise.
9946 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
9947 with an insn_type enum and msl with a modifier_type enum.
9948 Replace element_width with a scalar_mode. Change the shift
9949 to unsigned int. Add constructors for scalar_float_mode and
9950 scalar_int_mode elements.
9951 (aarch64_vect_float_const_representable_p): Delete.
9952 (aarch64_can_const_movi_rtx_p)
9953 (aarch64_simd_scalar_immediate_valid_for_move)
9954 (aarch64_simd_make_constant): Update call to
9955 aarch64_simd_valid_immediate.
9956 (aarch64_advsimd_valid_immediate_hs): New function.
9957 (aarch64_advsimd_valid_immediate): Likewise.
9958 (aarch64_simd_valid_immediate): Remove mode and inverse
9959 arguments. Rewrite to use the above. Use const_vec_duplicate_p
9960 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
9961 and aarch64_float_const_representable_p on the result.
9962 (aarch64_output_simd_mov_immediate): Remove mode argument.
9963 Update call to aarch64_simd_valid_immediate and use of
9964 simd_immediate_info.
9965 (aarch64_output_scalar_simd_mov_immediate): Update call
9968 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9969 Alan Hayward <alan.hayward@arm.com>
9970 David Sherwood <david.sherwood@arm.com>
9972 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
9973 (mode_nunits): Likewise CONST_MODE_NUNITS.
9974 * machmode.def (ADJUST_NUNITS): Document.
9975 * genmodes.c (mode_data::need_nunits_adj): New field.
9976 (blank_mode): Update accordingly.
9977 (adj_nunits): New variable.
9978 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
9980 (emit_mode_size_inline): Set need_bytesize_adj for all modes
9981 listed in adj_nunits.
9982 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
9983 listed in adj_nunits. Don't emit case statements for such modes.
9984 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
9985 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
9986 nothing if adj_nunits is nonnull.
9987 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
9988 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
9989 (emit_mode_fbit): Update use of print_maybe_const_decl.
9990 (emit_move_size): Likewise. Treat the array as non-const
9992 (emit_mode_adjustments): Handle adj_nunits.
9994 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9996 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
9997 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
9998 (VECTOR_MODES): Use it.
9999 (make_vector_modes): Take the prefix as an argument.
10001 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10002 Alan Hayward <alan.hayward@arm.com>
10003 David Sherwood <david.sherwood@arm.com>
10005 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
10006 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
10007 for MODE_VECTOR_BOOL.
10008 * machmode.def (VECTOR_BOOL_MODE): Document.
10009 * genmodes.c (VECTOR_BOOL_MODE): New macro.
10010 (make_vector_bool_mode): New function.
10011 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
10013 * lto-streamer-in.c (lto_input_mode_table): Likewise.
10014 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
10016 * stor-layout.c (int_mode_for_mode): Likewise.
10017 * tree.c (build_vector_type_for_mode): Likewise.
10018 * varasm.c (output_constant_pool_2): Likewise.
10019 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
10020 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
10021 for MODE_VECTOR_BOOL.
10022 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
10023 of mode class checks.
10024 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
10025 instead of a list of mode class checks.
10026 (expand_vector_scalar_condition): Likewise.
10027 (type_for_widest_vector_mode): Handle BImode as an inner mode.
10029 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10030 Alan Hayward <alan.hayward@arm.com>
10031 David Sherwood <david.sherwood@arm.com>
10033 * machmode.h (mode_size): Change from unsigned short to
10035 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
10036 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10037 or if measurement_type is not polynomial.
10038 (fixed_size_mode::includes_p): Check for constant-sized modes.
10039 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
10040 return a poly_uint16 rather than an unsigned short.
10041 (emit_mode_size): Change the type of mode_size from unsigned short
10042 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
10043 (emit_mode_adjustments): Cope with polynomial vector sizes.
10044 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10046 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10048 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
10049 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
10050 * caller-save.c (setup_save_areas): Likewise.
10051 (replace_reg_with_saved_mem): Likewise.
10052 * calls.c (emit_library_call_value_1): Likewise.
10053 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
10054 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
10055 (gen_lowpart_for_combine): Likewise.
10056 * convert.c (convert_to_integer_1): Likewise.
10057 * cse.c (equiv_constant, cse_insn): Likewise.
10058 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
10059 (cselib_subst_to_values): Likewise.
10060 * dce.c (word_dce_process_block): Likewise.
10061 * df-problems.c (df_word_lr_mark_ref): Likewise.
10062 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
10063 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
10064 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
10065 (rtl_for_decl_location): Likewise.
10066 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
10067 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
10068 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
10069 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
10070 (expand_expr_real_1): Likewise.
10071 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
10072 (pad_below): Likewise.
10073 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10074 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
10075 * ira.c (get_subreg_tracking_sizes): Likewise.
10076 * ira-build.c (ira_create_allocno_objects): Likewise.
10077 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
10078 (ira_sort_regnos_for_alter_reg): Likewise.
10079 * ira-costs.c (record_operand_costs): Likewise.
10080 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
10081 (resolve_simple_move): Likewise.
10082 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
10083 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
10084 (lra_constraints): Likewise.
10085 (CONST_POOL_OK_P): Reject variable-sized modes.
10086 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
10087 (add_pseudo_to_slot, lra_spill): Likewise.
10088 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10089 * optabs-query.c (get_best_extraction_insn): Likewise.
10090 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10091 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
10092 (expand_mult_highpart, valid_multiword_target_p): Likewise.
10093 * recog.c (offsettable_address_addr_space_p): Likewise.
10094 * regcprop.c (maybe_mode_change): Likewise.
10095 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
10096 * regrename.c (build_def_use): Likewise.
10097 * regstat.c (dump_reg_info): Likewise.
10098 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
10099 (find_reloads, find_reloads_subreg_address): Likewise.
10100 * reload1.c (eliminate_regs_1): Likewise.
10101 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
10102 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
10103 (simplify_binary_operation_1, simplify_subreg): Likewise.
10104 * targhooks.c (default_function_arg_padding): Likewise.
10105 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
10106 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
10107 (verify_gimple_assign_ternary): Likewise.
10108 * tree-inline.c (estimate_move_cost): Likewise.
10109 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10110 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
10111 (get_address_cost_ainc): Likewise.
10112 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
10113 (vect_supportable_dr_alignment): Likewise.
10114 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10115 (vectorizable_reduction): Likewise.
10116 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
10117 (vectorizable_operation, vectorizable_load): Likewise.
10118 * tree.c (build_same_sized_truth_vector_type): Likewise.
10119 * valtrack.c (cleanup_auto_inc_dec): Likewise.
10120 * var-tracking.c (emit_note_insn_var_location): Likewise.
10121 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
10122 (ADDR_VEC_ALIGN): Likewise.
10124 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10125 Alan Hayward <alan.hayward@arm.com>
10126 David Sherwood <david.sherwood@arm.com>
10128 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
10130 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10131 or if measurement_type is polynomial.
10132 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
10133 * combine.c (make_extraction): Likewise.
10134 * dse.c (find_shift_sequence): Likewise.
10135 * dwarf2out.c (mem_loc_descriptor): Likewise.
10136 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
10137 (extract_bit_field, extract_low_bits): Likewise.
10138 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
10139 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
10140 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
10141 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
10142 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10143 * reload.c (find_reloads): Likewise.
10144 * reload1.c (alter_reg): Likewise.
10145 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
10146 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
10147 * tree-if-conv.c (predicate_mem_writes): Likewise.
10148 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
10149 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
10150 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
10151 * valtrack.c (dead_debug_insert_temp): Likewise.
10152 * varasm.c (mergeable_constant_section): Likewise.
10153 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
10155 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10156 Alan Hayward <alan.hayward@arm.com>
10157 David Sherwood <david.sherwood@arm.com>
10159 * expr.c (expand_assignment): Cope with polynomial mode sizes
10160 when assigning to a CONCAT.
10162 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10163 Alan Hayward <alan.hayward@arm.com>
10164 David Sherwood <david.sherwood@arm.com>
10166 * machmode.h (mode_precision): Change from unsigned short to
10168 (mode_to_precision): Return a poly_uint16 rather than an unsigned
10170 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
10171 or if measurement_type is not polynomial.
10172 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
10173 in which the mode is already known to be a scalar_int_mode.
10174 * genmodes.c (emit_mode_precision): Change the type of mode_precision
10175 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
10177 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10178 for GET_MODE_PRECISION.
10179 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10180 for GET_MODE_PRECISION.
10181 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
10183 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
10184 (expand_field_assignment, make_extraction): Likewise.
10185 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
10186 (get_last_value): Likewise.
10187 * convert.c (convert_to_integer_1): Likewise.
10188 * cse.c (cse_insn): Likewise.
10189 * expr.c (expand_expr_real_1): Likewise.
10190 * lra-constraints.c (simplify_operand_subreg): Likewise.
10191 * optabs-query.c (can_atomic_load_p): Likewise.
10192 * optabs.c (expand_atomic_load): Likewise.
10193 (expand_atomic_store): Likewise.
10194 * ree.c (combine_reaching_defs): Likewise.
10195 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
10196 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
10197 * tree.h (type_has_mode_precision_p): Likewise.
10198 * ubsan.c (instrument_si_overflow): Likewise.
10200 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10201 Alan Hayward <alan.hayward@arm.com>
10202 David Sherwood <david.sherwood@arm.com>
10204 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
10205 polynomial numbers of units.
10206 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
10207 (valid_vector_subparts_p): New function.
10208 (build_vector_type): Remove temporary shim and take the number
10209 of units as a poly_uint64 rather than an int.
10210 (build_opaque_vector_type): Take the number of units as a
10211 poly_uint64 rather than an int.
10212 * tree.c (build_vector_from_ctor): Handle polynomial
10213 TYPE_VECTOR_SUBPARTS.
10214 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
10215 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
10216 (build_vector_from_val): If the number of units is variable,
10217 use build_vec_duplicate_cst for constant operands and
10218 VEC_DUPLICATE_EXPR otherwise.
10219 (make_vector_type): Remove temporary is_constant ().
10220 (build_vector_type, build_opaque_vector_type): Take the number of
10221 units as a poly_uint64 rather than an int.
10222 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
10224 * cfgexpand.c (expand_debug_expr): Likewise.
10225 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
10226 (store_constructor, expand_expr_real_1): Likewise.
10227 (const_scalar_mask_from_tree): Likewise.
10228 * fold-const-call.c (fold_const_reduction): Likewise.
10229 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
10230 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
10231 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
10232 (fold_relational_const): Likewise.
10233 (native_interpret_vector): Likewise. Change the size from an
10234 int to an unsigned int.
10235 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
10236 TYPE_VECTOR_SUBPARTS.
10237 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
10238 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
10239 duplicating a non-constant operand into a variable-length vector.
10240 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
10241 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
10242 * ipa-icf.c (sem_variable::equals): Likewise.
10243 * match.pd: Likewise.
10244 * omp-simd-clone.c (simd_clone_subparts): Likewise.
10245 * print-tree.c (print_node): Likewise.
10246 * stor-layout.c (layout_type): Likewise.
10247 * targhooks.c (default_builtin_vectorization_cost): Likewise.
10248 * tree-cfg.c (verify_gimple_comparison): Likewise.
10249 (verify_gimple_assign_binary): Likewise.
10250 (verify_gimple_assign_ternary): Likewise.
10251 (verify_gimple_assign_single): Likewise.
10252 * tree-pretty-print.c (dump_generic_node): Likewise.
10253 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10254 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
10255 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
10256 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
10257 (vect_shift_permute_load_chain): Likewise.
10258 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
10259 (expand_vector_condition, optimize_vector_constructor): Likewise.
10260 (lower_vec_perm, get_compute_type): Likewise.
10261 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10262 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
10263 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
10264 (vect_recog_mask_conversion_pattern): Likewise.
10265 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
10266 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
10267 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10268 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
10269 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
10270 (vectorizable_shift, vectorizable_operation, vectorizable_store)
10271 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
10272 (supportable_widening_operation): Likewise.
10273 (supportable_narrowing_operation): Likewise.
10274 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
10276 * varasm.c (output_constant): Likewise.
10278 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10279 Alan Hayward <alan.hayward@arm.com>
10280 David Sherwood <david.sherwood@arm.com>
10282 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
10283 so that both the length == 3 and length != 3 cases set up their
10284 own permute vectors. Add comments explaining why we know the
10285 number of elements is constant.
10286 (vect_permute_load_chain): Likewise.
10288 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10289 Alan Hayward <alan.hayward@arm.com>
10290 David Sherwood <david.sherwood@arm.com>
10292 * machmode.h (mode_nunits): Change from unsigned char to
10294 (ONLY_FIXED_SIZE_MODES): New macro.
10295 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
10296 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
10297 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
10299 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
10300 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
10301 or if measurement_type is not polynomial.
10302 * genmodes.c (ZERO_COEFFS): New macro.
10303 (emit_mode_nunits_inline): Make mode_nunits_inline return a
10305 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
10306 Use ZERO_COEFFS when emitting initializers.
10307 * data-streamer.h (bp_pack_poly_value): New function.
10308 (bp_unpack_poly_value): Likewise.
10309 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10310 for GET_MODE_NUNITS.
10311 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10312 for GET_MODE_NUNITS.
10313 * tree.c (make_vector_type): Remove temporary shim and make
10314 the real function take the number of units as a poly_uint64
10315 rather than an int.
10316 (build_vector_type_for_mode): Handle polynomial nunits.
10317 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
10318 * emit-rtl.c (const_vec_series_p_1): Likewise.
10319 (gen_rtx_CONST_VECTOR): Likewise.
10320 * fold-const.c (test_vec_duplicate_folding): Likewise.
10321 * genrecog.c (validate_pattern): Likewise.
10322 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
10323 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10324 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
10325 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
10326 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
10327 * rtlanal.c (subreg_get_info): Likewise.
10328 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10329 (vect_grouped_load_supported): Likewise.
10330 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
10331 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
10332 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
10333 (simplify_const_unary_operation, simplify_binary_operation_1)
10334 (simplify_const_binary_operation, simplify_ternary_operation)
10335 (test_vector_ops_duplicate, test_vector_ops): Likewise.
10336 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
10337 instead of CONST_VECTOR_NUNITS.
10338 * varasm.c (output_constant_pool_2): Likewise.
10339 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
10340 explicit-encoded elements in the XVEC for variable-length vectors.
10342 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10344 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
10346 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10347 Alan Hayward <alan.hayward@arm.com>
10348 David Sherwood <david.sherwood@arm.com>
10350 * coretypes.h (fixed_size_mode): Declare.
10351 (fixed_size_mode_pod): New typedef.
10352 * builtins.h (target_builtins::x_apply_args_mode)
10353 (target_builtins::x_apply_result_mode): Change type to
10354 fixed_size_mode_pod.
10355 * builtins.c (apply_args_size, apply_result_size, result_vector)
10356 (expand_builtin_apply_args_1, expand_builtin_apply)
10357 (expand_builtin_return): Update accordingly.
10359 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10361 * cse.c (hash_rtx_cb): Hash only the encoded elements.
10362 * cselib.c (cselib_hash_rtx): Likewise.
10363 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
10364 CONST_VECTOR encoding.
10366 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10367 Jeff Law <law@redhat.com>
10370 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
10371 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
10372 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
10373 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
10376 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
10377 explicitly probe *sp in a noreturn function if there were any callee
10378 register saves or frame pointer is needed.
10380 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10383 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
10384 BLKmode for ternary, binary or unary expressions.
10387 * var-tracking.c (delete_vta_debug_insn): New inline function.
10388 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
10389 insns from get_insns () to NULL instead of each bb separately.
10390 Use delete_vta_debug_insn. No longer static.
10391 (vt_debug_insns_local, variable_tracking_main_1): Adjust
10392 delete_vta_debug_insns callers.
10393 * rtl.h (delete_vta_debug_insns): Declare.
10394 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
10395 instead of variable_tracking_main.
10397 2018-01-03 Martin Sebor <msebor@redhat.com>
10399 PR tree-optimization/83603
10400 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
10401 arguments past the endof the argument list in functions declared
10402 without a prototype.
10403 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
10404 Avoid checking when arguments are null.
10406 2018-01-03 Martin Sebor <msebor@redhat.com>
10409 * doc/extend.texi (attribute const): Fix a typo.
10410 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
10411 issuing -Wsuggest-attribute for void functions.
10413 2018-01-03 Martin Sebor <msebor@redhat.com>
10415 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
10416 offset_int::from instead of wide_int::to_shwi.
10417 (maybe_diag_overlap): Remove assertion.
10418 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
10419 * gimple-ssa-sprintf.c (format_directive): Same.
10420 (parse_directive): Same.
10421 (sprintf_dom_walker::compute_format_length): Same.
10422 (try_substitute_return_value): Same.
10424 2018-01-03 Jeff Law <law@redhat.com>
10426 PR middle-end/83654
10427 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
10428 non-constant residual for zero at runtime and avoid probing in
10429 that case. Reorganize code for trailing problem to mirror handling
10432 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10434 PR tree-optimization/83501
10435 * tree-ssa-strlen.c (get_string_cst): New.
10436 (handle_char_store): Call get_string_cst.
10438 2018-01-03 Martin Liska <mliska@suse.cz>
10440 PR tree-optimization/83593
10441 * tree-ssa-strlen.c: Include tree-cfg.h.
10442 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
10443 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
10444 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
10446 (strlen_dom_walker::before_dom_children): Call
10447 gimple_purge_dead_eh_edges. Dump tranformation with details
10449 (strlen_dom_walker::before_dom_children): Update call by adding
10450 new argument cleanup_eh.
10451 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
10453 2018-01-03 Martin Liska <mliska@suse.cz>
10456 * cif-code.def (VARIADIC_THUNK): New enum value.
10457 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
10460 2018-01-03 Jan Beulich <jbeulich@suse.com>
10462 * sse.md (mov<mode>_internal): Tighten condition for when to use
10463 vmovdqu<ssescalarsize> for TI and OI modes.
10465 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10467 Update copyright years.
10469 2018-01-03 Martin Liska <mliska@suse.cz>
10472 * ipa-visibility.c (function_and_variable_visibility): Skip
10473 functions with noipa attribure.
10475 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10477 * gcc.c (process_command): Update copyright notice dates.
10478 * gcov-dump.c (print_version): Ditto.
10479 * gcov.c (print_version): Ditto.
10480 * gcov-tool.c (print_version): Ditto.
10481 * gengtype.c (create_file): Ditto.
10482 * doc/cpp.texi: Bump @copying's copyright year.
10483 * doc/cppinternals.texi: Ditto.
10484 * doc/gcc.texi: Ditto.
10485 * doc/gccint.texi: Ditto.
10486 * doc/gcov.texi: Ditto.
10487 * doc/install.texi: Ditto.
10488 * doc/invoke.texi: Ditto.
10490 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10492 * vector-builder.h (vector_builder::m_full_nelts): Change from
10493 unsigned int to poly_uint64.
10494 (vector_builder::full_nelts): Update prototype accordingly.
10495 (vector_builder::new_vector): Likewise.
10496 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
10497 (vector_builder::operator ==): Likewise.
10498 (vector_builder::finalize): Likewise.
10499 * int-vector-builder.h (int_vector_builder::int_vector_builder):
10500 Take the number of elements as a poly_uint64 rather than an
10502 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
10503 from unsigned int to poly_uint64.
10504 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
10505 (vec_perm_indices::new_vector): Likewise.
10506 (vec_perm_indices::length): Likewise.
10507 (vec_perm_indices::nelts_per_input): Likewise.
10508 (vec_perm_indices::input_nelts): Likewise.
10509 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
10510 number of elements per input as a poly_uint64 rather than an
10511 unsigned int. Use the original encoding for variable-length
10512 vectors, rather than clamping each individual element.
10513 For the second and subsequent elements in each pattern,
10514 clamp the step and base before clamping their sum.
10515 (vec_perm_indices::series_p): Handle polynomial element counts.
10516 (vec_perm_indices::all_in_range_p): Likewise.
10517 (vec_perm_indices_to_tree): Likewise.
10518 (vec_perm_indices_to_rtx): Likewise.
10519 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
10520 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
10521 (tree_vector_builder::new_binary_operation): Handle polynomial
10522 element counts. Return false if we need to know the number
10523 of elements at compile time.
10524 * fold-const.c (fold_vec_perm): Punt if the number of elements
10525 isn't known at compile time.
10527 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10529 * vec-perm-indices.h (vec_perm_builder): Change element type
10530 from HOST_WIDE_INT to poly_int64.
10531 (vec_perm_indices::element_type): Update accordingly.
10532 (vec_perm_indices::clamp): Handle polynomial element_types.
10533 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10534 (vec_perm_indices::all_in_range_p): Likewise.
10535 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
10537 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
10538 polynomial vec_perm_indices element types.
10539 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
10540 * fold-const.c (fold_vec_perm): Likewise.
10541 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
10542 * tree-vect-generic.c (lower_vec_perm): Likewise.
10543 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10544 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
10545 element type to HOST_WIDE_INT.
10547 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10548 Alan Hayward <alan.hayward@arm.com>
10549 David Sherwood <david.sherwood@arm.com>
10551 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
10552 rather than an int. Use plus_constant.
10553 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
10554 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
10556 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10557 Alan Hayward <alan.hayward@arm.com>
10558 David Sherwood <david.sherwood@arm.com>
10560 * calls.c (emit_call_1, expand_call): Change struct_value_size from
10561 a HOST_WIDE_INT to a poly_int64.
10563 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10564 Alan Hayward <alan.hayward@arm.com>
10565 David Sherwood <david.sherwood@arm.com>
10567 * calls.c (load_register_parameters): Cope with polynomial
10568 mode sizes. Require a constant size for BLKmode parameters
10569 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
10570 forces a parameter to be padded at the lsb end in order to
10571 fill a complete number of words, require the parameter size
10572 to be ordered wrt UNITS_PER_WORD.
10574 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10575 Alan Hayward <alan.hayward@arm.com>
10576 David Sherwood <david.sherwood@arm.com>
10578 * reload1.c (spill_stack_slot_width): Change element type
10579 from unsigned int to poly_uint64_pod.
10580 (alter_reg): Treat mode sizes as polynomial.
10582 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10583 Alan Hayward <alan.hayward@arm.com>
10584 David Sherwood <david.sherwood@arm.com>
10586 * reload.c (complex_word_subreg_p): New function.
10587 (reload_inner_reg_of_subreg, push_reload): Use it.
10589 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10590 Alan Hayward <alan.hayward@arm.com>
10591 David Sherwood <david.sherwood@arm.com>
10593 * lra-constraints.c (process_alt_operands): Reject matched
10594 operands whose sizes aren't ordered.
10595 (match_reload): Refer to this check here.
10597 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10598 Alan Hayward <alan.hayward@arm.com>
10599 David Sherwood <david.sherwood@arm.com>
10601 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
10602 that the mode size is in the set {1, 2, 4, 8, 16}.
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 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
10609 Use plus_constant instead of gen_rtx_PLUS.
10611 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10612 Alan Hayward <alan.hayward@arm.com>
10613 David Sherwood <david.sherwood@arm.com>
10615 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
10616 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
10617 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
10618 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
10619 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
10620 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
10621 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
10622 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
10623 * config/i386/i386.c (ix86_push_rounding): ...this new function.
10624 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
10626 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
10627 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
10628 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
10629 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
10630 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
10631 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
10632 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
10633 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
10634 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
10635 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
10637 * expr.c (emit_move_resolve_push): Treat the input and result
10638 of PUSH_ROUNDING as a poly_int64.
10639 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
10640 (emit_push_insn): Likewise.
10641 * lra-eliminations.c (mark_not_eliminable): Likewise.
10642 * recog.c (push_operand): Likewise.
10643 * reload1.c (elimination_effects): Likewise.
10644 * rtlanal.c (nonzero_bits1): Likewise.
10645 * calls.c (store_one_arg): Likewise. Require the padding to be
10646 known at compile time.
10648 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10649 Alan Hayward <alan.hayward@arm.com>
10650 David Sherwood <david.sherwood@arm.com>
10652 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
10653 Use plus_constant instead of gen_rtx_PLUS.
10655 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10656 Alan Hayward <alan.hayward@arm.com>
10657 David Sherwood <david.sherwood@arm.com>
10659 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
10660 rather than an int.
10662 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10663 Alan Hayward <alan.hayward@arm.com>
10664 David Sherwood <david.sherwood@arm.com>
10666 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
10667 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
10668 via stack temporaries. Treat the mode size as polynomial too.
10670 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10671 Alan Hayward <alan.hayward@arm.com>
10672 David Sherwood <david.sherwood@arm.com>
10674 * expr.c (expand_expr_real_2): When handling conversions involving
10675 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
10676 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
10677 as a poly_uint64 too.
10679 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10680 Alan Hayward <alan.hayward@arm.com>
10681 David Sherwood <david.sherwood@arm.com>
10683 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
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 * combine.c (can_change_dest_mode): Handle polynomial
10690 REGMODE_NATURAL_SIZE.
10691 * expmed.c (store_bit_field_1): Likewise.
10692 * expr.c (store_constructor): Likewise.
10693 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
10694 and polynomial REGMODE_NATURAL_SIZE.
10695 (gen_lowpart_common): Likewise.
10696 * reginfo.c (record_subregs_of_mode): Likewise.
10697 * rtlanal.c (read_modify_subreg_p): Likewise.
10699 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10700 Alan Hayward <alan.hayward@arm.com>
10701 David Sherwood <david.sherwood@arm.com>
10703 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
10704 numbers of elements.
10706 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10707 Alan Hayward <alan.hayward@arm.com>
10708 David Sherwood <david.sherwood@arm.com>
10710 * match.pd: Cope with polynomial numbers of vector elements.
10712 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10713 Alan Hayward <alan.hayward@arm.com>
10714 David Sherwood <david.sherwood@arm.com>
10716 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
10717 in a POINTER_PLUS_EXPR.
10719 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10720 Alan Hayward <alan.hayward@arm.com>
10721 David Sherwood <david.sherwood@arm.com>
10723 * omp-simd-clone.c (simd_clone_subparts): New function.
10724 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
10725 (ipa_simd_modify_function_body): Likewise.
10727 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10728 Alan Hayward <alan.hayward@arm.com>
10729 David Sherwood <david.sherwood@arm.com>
10731 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
10732 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
10733 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
10734 (expand_vector_condition, vector_element): Likewise.
10735 (subparts_gt): New function.
10736 (get_compute_type): Use subparts_gt.
10737 (count_type_subparts): Delete.
10738 (expand_vector_operations_1): Use subparts_gt instead of
10739 count_type_subparts.
10741 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10742 Alan Hayward <alan.hayward@arm.com>
10743 David Sherwood <david.sherwood@arm.com>
10745 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
10746 (vect_compile_time_alias): ...this new function. Do the calculation
10747 on poly_ints rather than trees.
10748 (vect_prune_runtime_alias_test_list): Update call accordingly.
10750 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10751 Alan Hayward <alan.hayward@arm.com>
10752 David Sherwood <david.sherwood@arm.com>
10754 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
10756 (vect_schedule_slp_instance): Likewise.
10758 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10759 Alan Hayward <alan.hayward@arm.com>
10760 David Sherwood <david.sherwood@arm.com>
10762 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
10763 constant and extern definitions for variable-length vectors.
10764 (vect_get_constant_vectors): Note that the number of units
10765 is known to be constant.
10767 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10768 Alan Hayward <alan.hayward@arm.com>
10769 David Sherwood <david.sherwood@arm.com>
10771 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
10772 of units as polynomial. Choose between WIDE and NARROW based
10775 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10776 Alan Hayward <alan.hayward@arm.com>
10777 David Sherwood <david.sherwood@arm.com>
10779 * tree-vect-stmts.c (simd_clone_subparts): New function.
10780 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
10782 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10783 Alan Hayward <alan.hayward@arm.com>
10784 David Sherwood <david.sherwood@arm.com>
10786 * tree-vect-stmts.c (vectorizable_call): Treat the number of
10787 vectors as polynomial. Use build_index_vector for
10788 IFN_GOMP_SIMD_LANE.
10790 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10791 Alan Hayward <alan.hayward@arm.com>
10792 David Sherwood <david.sherwood@arm.com>
10794 * tree-vect-stmts.c (get_load_store_type): Treat the number of
10795 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
10796 for variable-length vectors.
10797 (vectorizable_mask_load_store): Treat the number of units as
10798 polynomial, asserting that it is constant if the condition has
10799 already been enforced.
10800 (vectorizable_store, vectorizable_load): Likewise.
10802 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10803 Alan Hayward <alan.hayward@arm.com>
10804 David Sherwood <david.sherwood@arm.com>
10806 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
10807 of units as polynomial. Punt if we can't tell at compile time
10808 which vector contains the final result.
10810 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10811 Alan Hayward <alan.hayward@arm.com>
10812 David Sherwood <david.sherwood@arm.com>
10814 * tree-vect-loop.c (vectorizable_induction): Treat the number
10815 of units as polynomial. Punt on SLP inductions. Use an integer
10816 VEC_SERIES_EXPR for variable-length integer reductions. Use a
10817 cast of such a series for variable-length floating-point
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.h (build_index_vector): Declare.
10825 * tree.c (build_index_vector): New function.
10826 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
10827 of units as polynomial, forcibly converting it to a constant if
10828 vectorizable_reduction has already enforced the condition.
10829 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
10830 to create a {1,2,3,...} vector.
10831 (vectorizable_reduction): Treat the number of units as polynomial.
10832 Choose vectype_in based on the largest scalar element size rather
10833 than the smallest number of units. Enforce the restrictions
10836 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10837 Alan Hayward <alan.hayward@arm.com>
10838 David Sherwood <david.sherwood@arm.com>
10840 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
10841 number of units as polynomial.
10843 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10844 Alan Hayward <alan.hayward@arm.com>
10845 David Sherwood <david.sherwood@arm.com>
10847 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
10848 * target.def (autovectorize_vector_sizes): Return the vector sizes
10849 by pointer, using vector_sizes rather than a bitmask.
10850 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
10851 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
10852 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
10854 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
10855 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
10856 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
10857 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
10858 * omp-general.c (omp_max_vf): Likewise.
10859 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10860 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
10861 * tree-vect-loop.c (vect_analyze_loop): Likewise.
10862 * tree-vect-slp.c (vect_slp_bb): Likewise.
10863 * doc/tm.texi: Regenerate.
10864 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
10866 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
10867 the vector size as a poly_uint64 rather than an unsigned int.
10868 (current_vector_size): Change from an unsigned int to a poly_uint64.
10869 (get_vectype_for_scalar_type): Update accordingly.
10870 * tree.h (build_truth_vector_type): Take the size and number of
10871 units as a poly_uint64 rather than an unsigned int.
10872 (build_vector_type): Add a temporary overload that takes
10873 the number of units as a poly_uint64 rather than an unsigned int.
10874 * tree.c (make_vector_type): Likewise.
10875 (build_truth_vector_type): Take the number of units as a poly_uint64
10876 rather than an unsigned int.
10878 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10879 Alan Hayward <alan.hayward@arm.com>
10880 David Sherwood <david.sherwood@arm.com>
10882 * target.def (get_mask_mode): Take the number of units and length
10883 as poly_uint64s rather than unsigned ints.
10884 * targhooks.h (default_get_mask_mode): Update accordingly.
10885 * targhooks.c (default_get_mask_mode): Likewise.
10886 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
10887 * doc/tm.texi: Regenerate.
10889 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10890 Alan Hayward <alan.hayward@arm.com>
10891 David Sherwood <david.sherwood@arm.com>
10893 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
10894 * omp-general.c (omp_max_vf): Likewise.
10895 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
10896 (expand_omp_simd): Handle polynomial safelen.
10897 * omp-low.c (omplow_simd_context): Add a default constructor.
10898 (omplow_simd_context::max_vf): Change from int to poly_uint64.
10899 (lower_rec_simd_input_clauses): Update accordingly.
10900 (lower_rec_input_clauses): Likewise.
10902 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10903 Alan Hayward <alan.hayward@arm.com>
10904 David Sherwood <david.sherwood@arm.com>
10906 * tree-vectorizer.h (vect_nunits_for_cost): New function.
10907 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
10908 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
10909 (vect_analyze_slp_cost): Likewise.
10910 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
10911 (vect_model_load_cost): Likewise.
10913 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10914 Alan Hayward <alan.hayward@arm.com>
10915 David Sherwood <david.sherwood@arm.com>
10917 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
10918 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
10919 from an unsigned int * to a poly_uint64_pod *.
10920 (calculate_unrolling_factor): New function.
10921 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
10923 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10924 Alan Hayward <alan.hayward@arm.com>
10925 David Sherwood <david.sherwood@arm.com>
10927 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
10928 from an unsigned int to a poly_uint64.
10929 (_loop_vec_info::slp_unrolling_factor): Likewise.
10930 (_loop_vec_info::vectorization_factor): Change from an int
10932 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
10933 (vect_get_num_vectors): New function.
10934 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
10935 (vect_get_num_copies): Use vect_get_num_vectors.
10936 (vect_analyze_data_ref_dependences): Change max_vf from an int *
10937 to an unsigned int *.
10938 (vect_analyze_data_refs): Change min_vf from an int * to a
10940 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10941 than an unsigned HOST_WIDE_INT.
10942 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
10943 (vect_analyze_data_ref_dependence): Change max_vf from an int *
10944 to an unsigned int *.
10945 (vect_analyze_data_ref_dependences): Likewise.
10946 (vect_compute_data_ref_alignment): Handle polynomial vf.
10947 (vect_enhance_data_refs_alignment): Likewise.
10948 (vect_prune_runtime_alias_test_list): Likewise.
10949 (vect_shift_permute_load_chain): Likewise.
10950 (vect_supportable_dr_alignment): Likewise.
10951 (dependence_distance_ge_vf): Take the vectorization factor as a
10952 poly_uint64 rather than an unsigned HOST_WIDE_INT.
10953 (vect_analyze_data_refs): Change min_vf from an int * to a
10955 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
10956 vfm1 as a poly_uint64 rather than an int. Make the same change
10957 for the returned bound_scalar.
10958 (vect_gen_vector_loop_niters): Handle polynomial vf.
10959 (vect_do_peeling): Likewise. Update call to
10960 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
10961 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
10963 * tree-vect-loop.c (vect_determine_vectorization_factor)
10964 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
10965 (vect_get_known_peeling_cost): Likewise.
10966 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
10967 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
10968 (vect_transform_loop): Likewise. Use the lowest possible VF when
10969 updating the upper bounds of the loop.
10970 (vect_min_worthwhile_factor): Make static. Return an unsigned int
10971 rather than an int.
10972 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
10973 polynomial unroll factors.
10974 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
10975 (vect_make_slp_decision): Likewise.
10976 (vect_supported_load_permutation_p): Likewise, and polynomial
10978 (vect_analyze_slp_cost): Handle polynomial vf.
10979 (vect_slp_analyze_node_operations): Likewise.
10980 (vect_slp_analyze_bb_1): Likewise.
10981 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10982 than an unsigned HOST_WIDE_INT.
10983 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
10984 (vectorizable_load): Handle polynomial vf.
10985 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
10987 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
10989 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10990 Alan Hayward <alan.hayward@arm.com>
10991 David Sherwood <david.sherwood@arm.com>
10993 * match.pd: Handle bit operations involving three constants
10994 and try to fold one pair.
10996 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10998 * tree-vect-loop-manip.c: Include gimple-fold.h.
10999 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
11000 niters_maybe_zero parameters. Handle other cases besides a step of 1.
11001 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
11002 Add a path that uses a step of VF instead of 1, but disable it
11004 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
11005 and niters_no_overflow parameters. Update calls to
11006 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
11007 Create a new SSA name if the latter choses to use a ste other
11008 than zero, and return it via niters_vector_mult_vf_var.
11009 * tree-vect-loop.c (vect_transform_loop): Update calls to
11010 vect_do_peeling, vect_gen_vector_loop_niters and
11011 slpeel_make_loop_iterate_ntimes.
11012 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
11013 (vect_gen_vector_loop_niters): Update declarations after above changes.
11015 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
11017 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
11018 128-bit round to integer instructions.
11019 (ceil<mode>2): Likewise.
11020 (btrunc<mode>2): Likewise.
11021 (round<mode>2): Likewise.
11023 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11025 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
11026 unaligned VSX load/store on P8/P9.
11027 (expand_block_clear): Allow the use of unaligned VSX
11028 load/store on P8/P9.
11030 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11032 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
11034 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
11035 swap associated with both a load and a store.
11037 2018-01-02 Andrew Waterman <andrew@sifive.com>
11039 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
11040 * config/riscv/riscv.md (clear_cache): Use it.
11042 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
11044 * web.c: Remove out-of-date comment.
11046 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11048 * expr.c (fixup_args_size_notes): Check that any existing
11049 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
11050 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
11051 (emit_single_push_insn): ...here.
11053 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11055 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
11056 (const_vector_encoded_nelts): New function.
11057 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
11058 (const_vector_int_elt, const_vector_elt): Declare.
11059 * emit-rtl.c (const_vector_int_elt_1): New function.
11060 (const_vector_elt): Likewise.
11061 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
11062 of CONST_VECTOR_ELT.
11064 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11066 * expr.c: Include rtx-vector-builder.h.
11067 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
11068 directly on the tree encoding.
11069 (const_vector_from_tree): Likewise.
11070 * optabs.c: Include rtx-vector-builder.h.
11071 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
11072 sequence of "u" values.
11073 * vec-perm-indices.c: Include rtx-vector-builder.h.
11074 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
11075 directly on the vec_perm_indices encoding.
11077 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11079 * doc/rtl.texi (const_vector): Describe new encoding scheme.
11080 * Makefile.in (OBJS): Add rtx-vector-builder.o.
11081 * rtx-vector-builder.h: New file.
11082 * rtx-vector-builder.c: Likewise.
11083 * rtl.h (rtx_def::u2): Add a const_vector field.
11084 (CONST_VECTOR_NPATTERNS): New macro.
11085 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
11086 (CONST_VECTOR_DUPLICATE_P): Likewise.
11087 (CONST_VECTOR_STEPPED_P): Likewise.
11088 (CONST_VECTOR_ENCODED_ELT): Likewise.
11089 (const_vec_duplicate_p): Check for a duplicated vector encoding.
11090 (unwrap_const_vec_duplicate): Likewise.
11091 (const_vec_series_p): Check for a non-duplicated vector encoding.
11092 Say that the function only returns true for integer vectors.
11093 * emit-rtl.c: Include rtx-vector-builder.h.
11094 (gen_const_vec_duplicate_1): Delete.
11095 (gen_const_vector): Call gen_const_vec_duplicate instead of
11096 gen_const_vec_duplicate_1.
11097 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
11098 (gen_const_vec_duplicate): Use rtx_vector_builder.
11099 (gen_const_vec_series): Likewise.
11100 (gen_rtx_CONST_VECTOR): Likewise.
11101 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
11102 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11103 Build a new vector rather than modifying a CONST_VECTOR in-place.
11104 (handle_special_swappables): Update call accordingly.
11105 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
11106 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11107 Build a new vector rather than modifying a CONST_VECTOR in-place.
11108 (handle_special_swappables): Update call accordingly.
11110 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11112 * simplify-rtx.c (simplify_const_binary_operation): Use
11113 CONST_VECTOR_ELT instead of XVECEXP.
11115 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11117 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
11118 the selector elements to be different from the data elements
11119 if the selector is a VECTOR_CST.
11120 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
11121 ssizetype for the selector.
11123 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11125 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
11126 before testing each element individually.
11127 * tree-vect-generic.c (lower_vec_perm): Likewise.
11129 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11131 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
11132 * selftest-run-tests.c (selftest::run_tests): Call it.
11133 * vector-builder.h (vector_builder::operator ==): New function.
11134 (vector_builder::operator !=): Likewise.
11135 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
11136 (vec_perm_indices::all_from_input_p): New function.
11137 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11138 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
11139 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
11140 instead of reading the VECTOR_CST directly. Detect whether both
11141 vector inputs are the same before constructing the vec_perm_indices,
11142 and update the number of inputs argument accordingly. Use the
11143 utility functions added above. Only construct sel2 if we need to.
11145 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11147 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
11148 the broadcast of the low byte.
11149 (expand_mult_highpart): Use an explicit encoding for the permutes.
11150 * optabs-query.c (can_mult_highpart_p): Likewise.
11151 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
11152 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11153 (vectorizable_bswap): Likewise.
11154 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
11155 explicit encoding for the power-of-2 permutes.
11156 (vect_permute_store_chain): Likewise.
11157 (vect_grouped_load_supported): Likewise.
11158 (vect_permute_load_chain): Likewise.
11160 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11162 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
11163 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
11164 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
11165 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11166 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
11167 (vect_gen_perm_mask_any): Likewise.
11169 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11171 * int-vector-builder.h: New file.
11172 * vec-perm-indices.h: Include int-vector-builder.h.
11173 (vec_perm_indices): Redefine as an int_vector_builder.
11174 (auto_vec_perm_indices): Delete.
11175 (vec_perm_builder): Redefine as a stand-alone class.
11176 (vec_perm_indices::vec_perm_indices): New function.
11177 (vec_perm_indices::clamp): Likewise.
11178 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
11179 (vec_perm_indices::new_vector): New function.
11180 (vec_perm_indices::new_expanded_vector): Update for new
11181 vec_perm_indices class.
11182 (vec_perm_indices::rotate_inputs): New function.
11183 (vec_perm_indices::all_in_range_p): Operate directly on the
11184 encoded form, without computing elided elements.
11185 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
11186 encoding. Update for new vec_perm_indices class.
11187 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
11188 the given vec_perm_builder.
11189 (expand_vec_perm_var): Update vec_perm_builder constructor.
11190 (expand_mult_highpart): Use vec_perm_builder instead of
11191 auto_vec_perm_indices.
11192 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
11193 vec_perm_indices instead of auto_vec_perm_indices. Use a single
11194 or double series encoding as appropriate.
11195 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
11196 vec_perm_indices instead of auto_vec_perm_indices.
11197 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11198 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11199 (vect_permute_store_chain): Likewise.
11200 (vect_grouped_load_supported): Likewise.
11201 (vect_permute_load_chain): Likewise.
11202 (vect_shift_permute_load_chain): Likewise.
11203 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11204 (vect_transform_slp_perm_load): Likewise.
11205 (vect_schedule_slp_instance): Likewise.
11206 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11207 (vectorizable_mask_load_store): Likewise.
11208 (vectorizable_bswap): Likewise.
11209 (vectorizable_store): Likewise.
11210 (vectorizable_load): Likewise.
11211 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
11212 vec_perm_indices instead of auto_vec_perm_indices. Use
11213 tree_to_vec_perm_builder to read the vector from a tree.
11214 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
11215 vec_perm_builder instead of a vec_perm_indices.
11216 (have_whole_vector_shift): Use vec_perm_builder and
11217 vec_perm_indices instead of auto_vec_perm_indices. Leave the
11218 truncation to calc_vec_perm_mask_for_shift.
11219 (vect_create_epilog_for_reduction): Likewise.
11220 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
11221 from auto_vec_perm_indices to vec_perm_indices.
11222 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11223 instead of changing individual elements.
11224 (aarch64_vectorize_vec_perm_const): Use new_vector to install
11225 the vector in d.perm.
11226 * config/arm/arm.c (expand_vec_perm_d::perm): Change
11227 from auto_vec_perm_indices to vec_perm_indices.
11228 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11229 instead of changing individual elements.
11230 (arm_vectorize_vec_perm_const): Use new_vector to install
11231 the vector in d.perm.
11232 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
11233 Update vec_perm_builder constructor.
11234 (rs6000_expand_interleave): Likewise.
11235 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
11236 (rs6000_expand_interleave): Likewise.
11238 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11240 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
11241 to qimode could truncate the indices.
11242 * optabs.c (expand_vec_perm_var): Likewise.
11244 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11246 * Makefile.in (OBJS): Add vec-perm-indices.o.
11247 * vec-perm-indices.h: New file.
11248 * vec-perm-indices.c: Likewise.
11249 * target.h (vec_perm_indices): Replace with a forward class
11251 (auto_vec_perm_indices): Move to vec-perm-indices.h.
11252 * optabs.h: Include vec-perm-indices.h.
11253 (expand_vec_perm): Delete.
11254 (selector_fits_mode_p, expand_vec_perm_var): Declare.
11255 (expand_vec_perm_const): Declare.
11256 * target.def (vec_perm_const_ok): Replace with...
11257 (vec_perm_const): ...this new hook.
11258 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
11259 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
11260 * doc/tm.texi: Regenerate.
11261 * optabs.def (vec_perm_const): Delete.
11262 * doc/md.texi (vec_perm_const): Likewise.
11263 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
11264 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
11265 expand_vec_perm for constant permutation vectors. Assert that
11266 the mode of variable permutation vectors is the integer equivalent
11267 of the mode that is being permuted.
11268 * optabs-query.h (selector_fits_mode_p): Declare.
11269 * optabs-query.c: Include vec-perm-indices.h.
11270 (selector_fits_mode_p): New function.
11271 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
11272 is defined, instead of checking whether the vec_perm_const_optab
11273 exists. Use targetm.vectorize.vec_perm_const instead of
11274 targetm.vectorize.vec_perm_const_ok. Check whether the indices
11275 fit in the vector mode before using a variable permute.
11276 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
11277 vec_perm_indices instead of an rtx.
11278 (expand_vec_perm): Replace with...
11279 (expand_vec_perm_const): ...this new function. Take the selector
11280 as a vec_perm_indices rather than an rtx. Also take the mode of
11281 the selector. Update call to shift_amt_for_vec_perm_mask.
11282 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
11283 Use vec_perm_indices::new_expanded_vector to expand the original
11284 selector into bytes. Check whether the indices fit in the vector
11285 mode before using a variable permute.
11286 (expand_vec_perm_var): Make global.
11287 (expand_mult_highpart): Use expand_vec_perm_const.
11288 * fold-const.c: Includes vec-perm-indices.h.
11289 * tree-ssa-forwprop.c: Likewise.
11290 * tree-vect-data-refs.c: Likewise.
11291 * tree-vect-generic.c: Likewise.
11292 * tree-vect-loop.c: Likewise.
11293 * tree-vect-slp.c: Likewise.
11294 * tree-vect-stmts.c: Likewise.
11295 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
11297 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
11298 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
11299 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
11300 (aarch64_vectorize_vec_perm_const): ...this new function.
11301 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11302 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11303 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
11304 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
11305 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11306 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11307 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
11309 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
11310 check for NEON modes.
11311 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
11312 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
11313 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
11314 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
11316 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
11317 the old VEC_PERM_CONST conditions.
11318 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
11319 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
11320 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
11321 (ia64_vectorize_vec_perm_const_ok): Merge into...
11322 (ia64_vectorize_vec_perm_const): ...this new function.
11323 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
11324 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
11325 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
11326 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
11327 * config/mips/mips.c (mips_expand_vec_perm_const)
11328 (mips_vectorize_vec_perm_const_ok): Merge into...
11329 (mips_vectorize_vec_perm_const): ...this new function.
11330 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
11331 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
11332 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
11333 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
11334 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
11335 (rs6000_expand_vec_perm_const): Delete.
11336 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
11338 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11339 (altivec_expand_vec_perm_const_le): Take each operand individually.
11340 Operate on constant selectors rather than rtxes.
11341 (altivec_expand_vec_perm_const): Likewise. Update call to
11342 altivec_expand_vec_perm_const_le.
11343 (rs6000_expand_vec_perm_const): Delete.
11344 (rs6000_vectorize_vec_perm_const_ok): Delete.
11345 (rs6000_vectorize_vec_perm_const): New function.
11346 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11347 an element count and rtx array.
11348 (rs6000_expand_extract_even): Update call accordingly.
11349 (rs6000_expand_interleave): Likewise.
11350 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
11351 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
11352 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
11353 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
11354 (rs6000_expand_vec_perm_const): Delete.
11355 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11356 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11357 (altivec_expand_vec_perm_const_le): Take each operand individually.
11358 Operate on constant selectors rather than rtxes.
11359 (altivec_expand_vec_perm_const): Likewise. Update call to
11360 altivec_expand_vec_perm_const_le.
11361 (rs6000_expand_vec_perm_const): Delete.
11362 (rs6000_vectorize_vec_perm_const_ok): Delete.
11363 (rs6000_vectorize_vec_perm_const): New function. Remove stray
11364 reference to the SPE evmerge intructions.
11365 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11366 an element count and rtx array.
11367 (rs6000_expand_extract_even): Update call accordingly.
11368 (rs6000_expand_interleave): Likewise.
11369 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
11370 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
11372 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11374 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11376 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
11377 vector mode and that that mode matches the mode of the data
11379 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
11380 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
11381 directly using expand_vec_perm_1 when forcing selectors into
11383 (expand_vec_perm_var): New function, split out from expand_vec_perm.
11385 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11387 * optabs-query.h (can_vec_perm_p): Delete.
11388 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
11389 * optabs-query.c (can_vec_perm_p): Split into...
11390 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
11391 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
11392 particular selector is valid.
11393 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11394 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11395 (vect_grouped_load_supported): Likewise.
11396 (vect_shift_permute_load_chain): Likewise.
11397 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11398 (vect_transform_slp_perm_load): Likewise.
11399 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11400 (vectorizable_bswap): Likewise.
11401 (vect_gen_perm_mask_checked): Likewise.
11402 * fold-const.c (fold_ternary_loc): Likewise. Don't take
11403 implementations of variable permutation vectors into account
11404 when deciding which selector to use.
11405 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
11406 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
11407 with a false third argument.
11408 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
11409 to test whether the constant selector is valid and can_vec_perm_var_p
11410 to test whether a variable selector is valid.
11412 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11414 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
11415 * optabs-query.c (can_vec_perm_p): Likewise.
11416 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
11417 instead of vec_perm_indices.
11418 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
11419 (vect_gen_perm_mask_checked): Likewise,
11420 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
11421 (vect_gen_perm_mask_checked): Likewise,
11423 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11425 * optabs-query.h (qimode_for_vec_perm): Declare.
11426 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
11427 (qimode_for_vec_perm): ...this new function.
11428 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
11430 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11432 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
11433 does not have a conditional at the top.
11435 2018-01-02 Richard Biener <rguenther@suse.de>
11437 * ipa-inline.c (big_speedup_p): Fix expression.
11439 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11442 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
11445 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11449 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
11450 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
11451 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
11452 cond_taken_branch_cost 3->4.
11454 2018-01-01 Jakub Jelinek <jakub@redhat.com>
11456 PR tree-optimization/83581
11457 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
11458 TODO_cleanup_cfg if any changes have been made.
11460 PR middle-end/83608
11461 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
11462 convert_modes if target mode has the right side, but different mode
11465 PR middle-end/83609
11466 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
11467 last argument when extracting from CONCAT. If either from_real or
11468 from_imag is NULL, use expansion through memory. If result is not
11469 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
11470 the parts directly to inner mode, if even that fails, use expansion
11473 PR middle-end/83623
11474 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
11475 check for bswap in mode rather than HImode and use that in expand_unop
11478 Copyright (C) 2018 Free Software Foundation, Inc.
11480 Copying and distribution of this file, with or without modification,
11481 are permitted in any medium without royalty provided the copyright
11482 notice and this notice are preserved.