1 2018-04-26 Tom de Vries <tom@codesourcery.com>
4 * config/nvptx/nvptx.c (verify_neutering_jumps)
5 (verify_neutering_labels): New function
6 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
8 2018-04-26 Tom de Vries <tom@codesourcery.com>
11 * config/nvptx/nvptx.c (needs_neutering_p): New function.
12 (nvptx_single): Use needs_neutering_p to skip over insns that do not
15 2018-04-26 Richard Biener <rguenther@suse.de>
16 Tom de Vries <tom@codesourcery.com>
19 * lto-streamer-out.c (output_function): Fixup loops if required to match
20 discovery done in the reader.
22 2018-04-26 Richard Biener <rguenther@suse.de>
24 PR tree-optimization/85116
25 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
26 have a loop exit from the single latch predecessor. Remove
27 case of header with just condition.
28 (ch_base::copy_headers): Exclude infinite loops from any
30 (pass_ch::execute): Record exits.
32 2018-04-26 Richard Biener <rguenther@suse.de>
34 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
35 prologue cost vector and pass it to vect_get_load_cost.
36 (vect_get_peeling_costs_all_drs): Likewise.
37 (vect_peeling_hash_get_lowest_cost): Likewise.
38 (vect_enhance_data_refs_alignment): Likewise.
40 2018-04-26 Richard Biener <rguenther@suse.de>
43 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
44 checking of integer<->pointer conversions.
45 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
46 sign-/zero-extending pointer types.
47 (expand_omp_for_static_chunk): Likewise.
49 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
50 Jean Lee <xiaoyur347@gmail.com>
52 * config/mips/mips.c (mips_asan_shadow_offset): New function.
53 (TARGET_ASAN_SHADOW_OFFSET): Define.
54 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
55 true for -fsanitize=address.
57 2018-04-25 Mark Wielaard <mark@klomp.org>
59 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
62 2018-04-25 Jakub Jelinek <jakub@redhat.com>
64 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
65 than "alu", remove explicit "memory" and "imm_disp" attributes.
66 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
69 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
70 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
73 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
76 * config/i386/i386.c (ix86_expand_builtin): Change memory
77 operand to XI, extend p0 to Pmode.
78 * config/i386/i386.md: Change unspec volatile and operand
79 1 mode to XI, change operand 0 mode to P.
81 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
83 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
84 GET_MODE_MASK before any checking.
85 (nds32_can_use_bset_p): Likewise.
86 (nds32_can_use_btgl_p): Likewise.
88 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
90 * config/nds32/nds32-doubleword.md: New define_split pattern for
91 illegal register number.
93 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
95 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
97 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
99 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
101 2018-04-25 Richard Biener <rguenther@suse.de>
103 * lto-streamer.h (LTO_major_version): Bump to 8.
105 2018-04-25 Jakub Jelinek <jakub@redhat.com>
107 * BASE-VER: Set to 9.0.0.
109 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
111 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
112 in __abskf2 and __powikf2.
114 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
117 * config/aarch64/constraints.md (Usg, Usj): New constraints.
118 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
119 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
120 Use the above on operand 2. Reindent.
121 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
123 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
126 * common/config/i386/i386-common.c (ix86_handle_option): Don't
128 * config/i386/i386.opt (mcet): Removed.
129 * doc/install.texi: Remove -mcet documentation.
130 * doc/invoke.texi: Likewise.
132 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
135 * doc/install.texi: Remove -mcet from bootstrap-cet.
137 2018-04-24 Jakub Jelinek <jakub@redhat.com>
140 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
141 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
145 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
146 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
147 containing a CONST_VECTOR.
149 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
151 * doc/install.texi: Update newlib dependency for nvptx.
153 2018-04-24 Jakub Jelinek <jakub@redhat.com>
156 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
157 instead of INTVAL when shifting x left.
159 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
161 PR tree-optimization/85478
162 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
163 vect_grouped_store_supported for single element vectors.
165 2018-04-24 Richard Biener <rguenther@suse.de>
168 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
169 load cost increase to the case of non-constant step.
171 2018-04-24 Jakub Jelinek <jakub@redhat.com>
174 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
175 destination if any_malformed_asm.
177 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
180 * expr.c (store_field): In the bitfield case, if the value comes from
181 a function call and is returned in registers by means of a PARALLEL,
182 do not change the mode of the temporary unless BLKmode and VOIDmode.
184 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
186 PR rtl-optimization/85423
187 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
188 dependencies to debug insns when the previous insn is non-debug.
190 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
192 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
193 enums into a single definition.
194 (fls): Fix predicates and printing.
197 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
199 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
200 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
201 and short u6 immediate.
202 (check_if_valid_sleep_operand): Remove.
203 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
205 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
207 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
208 flag_always_save_lp condition.
209 * config/nds32/nds32.opt (malways-save-lp): New option.
211 2018-04-22 Shiva Chen <shiva0217@gmail.com>
213 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
214 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
215 * config/nds32/nds32.h
216 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
217 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
219 2018-04-22 Shiva Chen <shiva0217@gmail.com>
221 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
222 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
224 2018-04-22 Shiva Chen <shiva0217@gmail.com>
225 Chung-Ju Wu <jasonwucj@gmail.com>
227 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
229 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
230 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
232 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
234 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
236 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
238 * config/nds32/nds32-protos.h (nds32_data_alignment,
239 nds32_local_alignment): Declare.
240 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
241 nds32_local_alignment): New functions.
242 (TARGET_CONSTANT_ALIGNMENT): Define.
243 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
245 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
247 * config/nds32/nds32.c
248 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
249 (TARGET_MODES_TIEABLE_P): Likewise.
251 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
253 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
256 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
257 Chung-Ju Wu <jasonwucj@gmail.com>
259 * config/nds32/constants.md (unspec_volatile_element): Add enum values
260 for unaligned access.
261 * config/nds32/nds32-intrinsic.c: Implementation of expanding
263 * config/nds32/nds32-intrinsic.md: Likewise.
264 * config/nds32/nds32_intrinsic.h: Likewise.
265 * config/nds32/nds32.h (nds32_builtins): Likewise.
266 * config/nds32/nds32.opt (munaligned-access): New option.
267 * config/nds32/nds32.c (nds32_asm_file_start): Display
268 flag_unaligned_access status.
270 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
272 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
273 -mno-relax is present.
274 * config/riscv/linux.h (LINK_SPEC): Ditto.
276 2018-04-20 Martin Sebor <msebor@redhat.com>
279 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
281 (gimple_fold_builtin_stxcpy_chk): Same.
282 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
284 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
287 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
288 __powikf2 when long double is IEEE 128-bit.
290 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
292 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
293 step to make sure stack always aligned.
295 2018-04-20 Carl Love <cel@us.ibm.com>
298 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
301 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
302 Tom de Vries <tom@codesourcery.com>
305 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
306 Emit insns for calls too.
307 (nvptx_find_par): Always look for worker-level predecessor insn.
308 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
310 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
311 (nvptx_process_pars): Propagate frames for calls.
313 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
316 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
318 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
319 (ix86_handle_option): Don't handle OPT_mibt.
320 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
322 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
324 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
325 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
326 (ix86_target_macros): Define __CET__ with flag_cf_protection
328 * config/i386/i386.c (isa2_opts): Remove -mibt.
329 * config/i386/i386.h (TARGET_IBT): Removed.
330 (TARGET_IBT_P): Likewise.
331 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
332 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
333 * config/i386/i386.opt (mcet): Update help message.
336 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
337 -mcet as an alias for -mshstk.
339 2018-04-20 Richard Biener <rguenther@suse.de>
342 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
343 complexity by forcing a single use of the multiply operand.
345 2018-04-20 Martin Jambor <mjambor@suse.cz>
348 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
349 recursion dependency to only apply to non-clones.
351 2018-04-20 Martin Jambor <mjambor@suse.cz>
354 * ipa-cp.c (create_specialized_node): Check that clones of
355 self-recursive edges exist during IPA-CP.
357 2018-04-19 Toon Moene <toon@moene.org>
359 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
362 2018-04-19 Jakub Jelinek <jakub@redhat.com>
364 PR tree-optimization/85467
365 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
366 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
367 VECTOR_CST element to type.
369 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
372 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
373 * config/i386/i386.md (builtin_setjmp_setup): Removed.
374 (builtin_longjmp): Likewise.
375 (save_stack_nonlocal): New pattern.
376 (restore_stack_nonlocal): Likewise.
378 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
381 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
382 Replace ASM_OUTPUT_LABEL with fprintf.
384 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
387 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
388 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
389 * config/i386/i386-c.c (ix86_target_macros_internal): Also
390 define __IBT__ and __SHSTK__ for -fcf-protection.
391 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
393 (ix86_trampoline_init): Likewise.
394 (x86_output_mi_thunk): Likewise.
395 (ix86_notrack_prefixed_insn_p): Likewise.
396 (ix86_option_override_internal): Don't disallow -fcf-protection.
397 * config/i386/i386.md (rdssp<mode>): Also enable for
399 (incssp<mode>): Likewise.
400 (nop_endbr): Likewise.
401 * config/i386/i386.opt (mcet): Change help message to built-in
405 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
406 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
407 enable CET built-in functions.
409 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
411 * common/config/i386/i386-common.c
412 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
413 OPTION_MASK_ISA_MOVDIRI_UNSET,
414 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
415 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
416 * config.gcc (movdirintrin.h): New header.
417 * config/i386/cpuid.h (bit_MOVDIRI,
418 bit_MOVDIR64B): New bits.
419 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
421 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
422 (VOID, PVOID, PCVOID)): New function types.
423 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
424 __builtin_ia32_directstoreu_u64,
425 __builtin_ia32_movdir64b): New builtins.
426 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
427 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
429 (ix86_valid_target_attribute_inner_p): Ditto.
430 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
431 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
432 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
433 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
434 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
435 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
436 (movdiri<mode>, movdir64b_<mode>): New.
437 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
438 * config/i386/immintrin.h: Include movdirintrin.h.
439 * config/i386/movdirintrin.h: New file.
440 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
442 2018-04-19 Richard Biener <rguenther@suse.de>
445 * cfg.c (clear_bb_flags): When loop state says we have
446 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
448 2018-04-19 Richard Biener <rguenther@suse.de>
450 PR tree-optimization/84737
451 * tree-vect-data-refs.c (vect_copy_ref_info): New function
452 copying restrict info.
453 (vect_setup_realignment): Use it.
454 * tree-vectorizer.h (vect_copy_ref_info): Declare.
455 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
456 the first DR to all generated stores.
457 (vectorizable_load): Likewise for loads.
459 2018-04-19 Jakub Jelinek <jakub@redhat.com>
461 PR tree-optimization/85446
462 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
463 the integral and pointer types to have the same precision.
465 * doc/install.texi: Document --disable-cet being the default and
468 2018-04-18 Martin Liska <mliska@suse.cz>
470 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
473 2018-04-18 Martin Liska <mliska@suse.cz>
476 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
479 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
480 arguments if they are comparable.
482 2018-04-18 Martin Liska <mliska@suse.cz>
485 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
488 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
491 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
494 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
495 ENDBR after calling __morestack.
497 2018-04-18 David Malcolm <dmalcolm@redhat.com>
500 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
501 by using gcc_base_ver to generate a gcc_driver_version, and use
502 it when generating GCC_DRIVER_NAME.
503 * configure: Regenerate.
505 2018-04-18 Jakub Jelinek <jakub@redhat.com>
508 * config.gcc: Obsolete powerpc*-*-*spe*.
510 2018-04-17 Jakub Jelinek <jakub@redhat.com>
513 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
514 (stabstr_D): Change type of unum from unsigned int to
515 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
518 2018-04-17 Jim Wilson <jimw@sifive.com>
521 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
522 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
523 Set arg_pointer_offset after using pretend_args_size.
525 2018-04-17 Jakub Jelinek <jakub@redhat.com>
527 PR rtl-optimization/85431
528 * dse.c (record_store): Ignore zero width stores.
531 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
532 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
533 __builtin_stack_restore rather than after it.
534 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
535 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
536 argument instead of virtual_dynamic_stack_rtx.
538 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
540 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
542 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
543 Add note to error message to explain internal mapping of overloaded
544 built-in function name to non-overloaded built-in function name.
545 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
548 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
551 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
552 where the inputs overlap with the output.
554 2018-04-17 Jakub Jelinek <jakub@redhat.com>
556 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
557 (=v, v) alternative and explicit "memory" attribute.
558 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
559 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
561 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
562 "sselog1" type instead of "sselog".
563 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
564 "sselog". Remove explicit "memory" attribute.
565 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
566 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
568 (vec_extract_hi_v32hi): Merge all alternatives into one, use
569 "sselog1" type instead of "sselog". Remove explicit "memory"
571 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
572 use "sselog1" type instead of "sselog". Remove explicit "memory"
574 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
575 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
577 (vec_extract_hi_v64qi): Merge all alternatives into one, use
578 "sselog1" type instead of "sselog". Remove explicit "memory"
580 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
581 use "sselog1" type instead of "sselog". Remove explicit "memory"
585 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
588 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
591 2018-04-17 Martin Jambor <mjambor@suse.cz>
594 * ipa-cp.c (create_specialized_node): Call
595 expand_all_artificial_thunks if necessary.
597 2018-04-17 Martin Liska <mliska@suse.cz>
600 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
601 in message, remote space in between '_G' and '('.
603 2018-04-17 Jakub Jelinek <jakub@redhat.com>
606 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
607 avx512f_vmcmp<mode>3<round_saeonly_name>,
608 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
609 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
610 avx512f_rndscale<mode><round_saeonly_name>,
611 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
612 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
613 Use %<iptr>2 instead of %2 for -masm=intel.
614 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
615 avx512f_vcvttss2usi<round_saeonly_name>,
616 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
618 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
619 avx512f_vcvttsd2usi<round_saeonly_name>,
620 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
621 Use %q1 instead of %1 for -masm=intel.
622 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
623 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
624 of %3 for -masm=intel.
625 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
627 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
629 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
631 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
632 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
634 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
636 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
637 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
638 %g1 and one with %0 and %1.
639 (avx512er_vmrcp28<mode><round_saeonly_name>,
640 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
642 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
643 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
644 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
645 of %0 and %{%4%} for -masm=intel.
646 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
647 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
648 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
649 order of %0 and %{%5%}%{z%} for -masm=intel.
651 2018-04-17 Jan Hubicka <jh@suse.cz>
654 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
656 2018-04-17 Martin Liska <mliska@suse.cz>
659 * multiple_target.c (create_dispatcher_calls): Set apostrophes
660 for target_clone error message. Make default implementation
661 clone to be a local declaration.
662 (separate_attrs): Add new argument and check for an empty
664 (expand_target_clones): Handle it.
665 (ipa_target_clone): Make redirection just for target_clones
668 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
669 Tom de Vries <tom@codesourcery.com>
672 * omp-expand.c (expand_oacc_for): Add dummy false branch for
673 tiled basic blocks without omp continue statements.
675 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
678 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
679 vec_extract expression as having side effects to make sure it gets
682 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
685 * config/i386/i386.c (get_builtin_code_for_version): Check
688 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
691 * gcc/config.gcc: Support "skylake".
692 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
694 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
695 (processor_target_table): Add "skylake".
696 (ix86_option_override_internal): Add "skylake".
697 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
698 PROCESSOR_CANNONLAKE.
699 (get_builtin_code_for_version): Fix priority for
700 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
701 PROCESSOR_SKYLAKE-AVX512.
702 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
703 (processor_type): Add PROCESSOR_SKYLAKE.
705 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
706 Jason Merrill <jason@redhat.com>
709 * convert.c (convert_to_integer_1): Use direct recursion for
710 enumeral types and types with a precision less than the number
711 of bits in their mode.
713 2018-04-16 Julia Koval <julia.koval@intel.com>
716 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
717 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
719 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
722 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
723 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
724 and -mno-direct-move.
726 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
729 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
730 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
731 Ensure negative shifts result in {0}.
733 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
735 PR rtl-optimization/79916
736 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
737 regs (if any) to define how to gnerate SD moves when LRA is in
740 2018-04-13 Jakub Jelinek <jakub@redhat.com>
742 PR rtl-optimization/85393
743 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
744 * except.c (expand_dw2_landing_pad_for_region): Make static.
745 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
746 a label and unconditional jump to old_bb, rather than
747 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
750 PR rtl-optimization/85376
751 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
752 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
753 instead of a specific value.
755 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
756 Bin Cheng <bin.cheng@arm.com>
758 PR tree-optimization/82965
759 PR tree-optimization/83991
760 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
761 by_profile_only parameter.
762 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
763 information if the loop was predicted to iterate too many times.
764 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
766 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
769 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
772 2018-04-13 Martin Liska <mliska@suse.cz>
773 Jakub Jelinek <jakub@redhat.com>
776 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
777 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
778 * builtins.c (expand_builtin_memory_copy_args): Use
779 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
780 handle dest_addr == pc_rtx.
782 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
785 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
786 asked to not generate direct moves.
787 (fix_trunc<mode>si2_stfiwx): Similar.
788 (fix_trunc<mode>si2_internal): Similar.
790 2018-04-12 Jakub Jelinek <jakub@redhat.com>
793 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
794 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
795 lookup if dest in some wider mode is known to be const0_rtx and
796 if so, record permanent equivalence for it to be ZERO_EXTEND of
797 the narrower mode destination.
799 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
801 * lto-streamer-out.c (output_function): Revert 259346.
802 * omp-expand.c (expand_oacc_for): Likewise.
804 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
806 PR rtl-optimization/85354
807 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
808 * sel-sched.c (sel_global_init): ... here.
810 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
813 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
814 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
815 mode for PE-COFF targets.
816 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
817 (i386_pe_asm_lto_end): Likewise.
818 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
819 (TARGET_ASM_LTO_END): Likewise.
820 * config/i386/winnt.c (saved_debug_info_level): New static variable.
821 (i386_pe_asm_lto_start): New function.
822 (i386_pe_asm_lto_end): Likewise.
824 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
825 Richard Biener <rguenther@suse.de>
828 * lto-streamer-out.c (output_function): Fix CFG loop state before
830 * omp-expand.c (expand_oacc_for): Handle calls to internal
831 functions like regular functions.
833 2018-04-12 Richard Biener <rguenther@suse.de>
836 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
837 for the early LTO debug to properly generate references to it
838 during DIE emission. Do not re-use that for the skeleton for
840 (dwarf2out_early_finish): Likewise.
842 2018-04-12 Jakub Jelinek <jakub@redhat.com>
846 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
847 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
848 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
849 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
850 and output is a reg, avoid creating invalid lowpart subreg, but
851 instead split into a 512-bit move. Don't split if not AVX512VL,
852 input is xmm16+ reg and output is a mem.
853 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
854 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
855 xmm16+ reg and output is a mem.
857 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
859 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
860 also for flag_dwarf2_cfi_asm.
862 2018-04-12 Jakub Jelinek <jakub@redhat.com>
864 PR rtl-optimization/85342
865 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
866 a bool scalar var inside of the loop instead. Don't try to update
867 recog_data.operand after failed apply_change_group.
869 2018-04-12 Tom de Vries <tom@codesourcery.com>
872 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
873 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
874 array with flexible array member as array without given dimension.
875 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
876 argument for undefined param to true.
878 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
881 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
882 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
883 from PowerPC section.
884 * config/rs6000/sysv4.opt (mcall-): Improve help text.
885 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
886 help text that is too long.
887 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
888 help text that is too long.
889 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
890 help text that is too long.
892 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
894 * config/alpha/alpha.md (stack_probe_internal): Rename
895 from "probe_stack". Update all callers.
897 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
899 PR rtl-optimization/84566
900 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
901 sched_macro_fuse_insns.
903 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
906 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
907 (compute_block_dependences): ... from here.
909 2018-04-11 Jakub Jelinek <jakub@redhat.com>
911 PR tree-optimization/85331
912 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
913 from int to HOST_WIDE_INT.
915 2018-04-11 Martin Jambor <mjambor@suse.cz>
918 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
919 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
920 not the same as the source val.
921 (cgraph_edge_brings_value_p): New parameter.
922 (gather_edges_for_value): Pass destination value to
923 cgraph_edge_brings_value_p.
924 (perhaps_add_new_callers): Likewise.
925 (get_info_about_necessary_edges): Likewise and exclude values brought
926 only by self-recursive edges.
927 (create_specialized_node): Redirect only clones of self-calling edges.
928 (+self_recursive_pass_through_p): New function.
929 (find_more_scalar_values_for_callers_subset): Use it.
930 (find_aggregate_values_for_callers_subset): Likewise.
931 (known_aggs_to_agg_replacement_list): Removed.
932 (decide_whether_version_node): Re-calculate known constants for all
933 remaining context clones.
935 2018-04-11 Richard Biener <rguenther@suse.de>
938 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
939 from early DWARF output.
940 (dwarf2out_early_finish): Output line info unconditionally into
941 early DWARF and add reference to it.
943 2018-04-11 Jakub Jelinek <jakub@redhat.com>
946 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
947 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
948 other than V2DFmode using iptr mode attribute.
949 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
951 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
953 PR rtl-optimization/84659
954 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
956 2018-04-11 Jakub Jelinek <jakub@redhat.com>
959 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
961 (output_loc_list): Pass address of a dummy size variable even in the
962 locview handling loop.
963 (index_location_lists): Add comment on why skip_loc_list_entry can't
966 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
969 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
972 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
975 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
976 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
977 and -mstring-compare-inline-limit.
979 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
982 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
983 for stack clash protection in a register whenever we need it to be in
986 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
988 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
989 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
991 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
994 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
996 (mlong-double-): Ditto.
997 * config/rs6000/sysv4.opt (msdata=): Ditto.
1000 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
1002 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1003 erroneous entries for
1004 "vector int vec_ldl (int, long int *)", and
1005 "vector unsigned int vec_ldl (int, unsigned long int *)".
1006 Add comments and entries for
1007 "vector bool char vec_ldl (int, bool char *)",
1008 "vector bool short vec_ldl (int, bool short *)",
1009 "vector bool int vec_ldl (int, bool int *)",
1010 "vector bool long long vec_ldl (int, bool long long *)",
1011 "vector pixel vec_ldl (int, pixel *)",
1012 "vector long long vec_ldl (int, long long *)",
1013 "vector unsigned long long vec_ldl (int, unsigned long long *)".
1014 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
1015 type tree bool_long_long_type_node and correct definition of
1016 bool_V2DI_type_node to make reference to this new type tree.
1017 (rs6000_mangle_type): Replace erroneous reference to
1018 bool_long_type_node with bool_long_long_type_node.
1019 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
1020 comments to emphasize sign distinctions for char and int types and
1021 replace RS6000_BTI_bool_long constant with
1022 RS6000_BTI_bool_long_long constant. Also add comment to restrict
1023 use of RS6000_BTI_pixel.
1024 (bool_long_type_node): Remove this macro definition.
1025 (bool_long_long_type_node): New macro definition
1027 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1029 PR rtl-optimization/85300
1030 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
1031 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
1032 simplify_unary_operation fails.
1034 2018-04-10 Martin Liska <mliska@suse.cz>
1036 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
1037 cgraph_edge and ipa_ref.
1039 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1043 * config/i386/sse.md
1044 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
1045 computation of the VEC_MERGE selector from mask.
1046 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
1047 Fix decoding of the VEC_MERGE selector into mask.
1049 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1051 PR tree-optimization/85286
1052 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
1054 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1056 * final.c (final_1): Set insn_last_address as well as
1057 insn_current_address.
1059 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1062 * explow.c (emit_stack_probe): Call validize_mem on memory location
1063 before passing it to gen_probe_stack. Create address operand and
1064 legitimize it for the probe_stack_address case.
1066 2018-04-09 Jan Hubicka <jh@suse.cz>
1069 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
1070 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
1071 * tree.c (free_lang_data_in_type): Fix handling of binfos;
1073 (free_lang_data): Rebuild type inheritance graph.
1075 2018-04-09 Martin Sebor <msebor@redhat.com>
1077 * invoke.texi (-finline-small-functions): Mention other optimization
1079 (-findirect-inlining, -fpartial-inlining): Same.
1080 (-finline-functions-called-once): Same.
1081 (-freorder-blocks-and-partition): Same.
1083 2018-04-09 Jan Hubicka <jh@suse.cz>
1086 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
1087 jumps; choose last target that matches the criteria (i.e.
1088 no partition changes for non-crossing jumps).
1089 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
1090 support for redirecting crossing jumps to non-crossing.
1092 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
1094 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
1095 also for naked functions.
1097 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
1099 * config/arc/arc.md (add_shift): New pattern.
1100 (add_shift2): Likewise.
1101 (sub_shift): Likewise.
1102 (sub_shift_cmp0_noout): Likewise.
1103 (compare_si_ashiftsi): Likewise.
1104 (xbfu_cmp0_noout): New combine pattern.
1105 (xbfu_cmp0"): Likewise.
1106 (movsi_set_cc_insn): Place the predicable variant first.
1107 (commutative_binary_cmp0_noout): Remove clobber.
1108 (commutative_binary_cmp0): New pattern.
1109 (noncommutative_binary_cmp0): Likewise.
1110 (noncommutative_binary_cmp0_noout): Likewise.
1111 (noncommutative_binary_comparison_result_used): Removed.
1112 (rsub_cmp0): New pattern.
1113 (rsub_cmp0_noout): Likewise.
1114 (extzvsi): Changed, keep only meaningful variants.
1115 (SQH, SEZ): New iterators.
1116 (SQH_postfix): New mode attribute.
1117 (SEZ_prefix): New code attribute.
1118 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
1119 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
1120 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
1122 (noncommutative_operator): Check the availability of barrel
1125 2018-04-09 Richard Biener <rguenther@suse.de>
1127 PR tree-optimization/85284
1128 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
1129 Only use the niter constraining form of simple_iv when the exit
1132 2018-04-09 Tom de Vries <tom@codesourcery.com>
1135 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1136 (define_expand "*memory_barrier"): New define_expand.
1137 (define_insn "memory_barrier"): New insn.
1139 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1141 PR rtl-optimization/80463
1142 PR rtl-optimization/83972
1143 PR rtl-optimization/83480
1145 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
1146 correct producer for the insn.
1147 (tidy_control_flow): Fixup seqnos in case of debug insns.
1149 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1151 PR rtl-optimization/83913
1153 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
1154 different sched-times when merging exprs.
1156 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1158 PR rtl-optimization/83962
1160 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
1161 tidy_fallthru_edge and tidy_control_flow.
1163 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1165 PR rtl-optimization/83530
1167 * sel-sched.c (force_next_insn): New global variable.
1168 (remove_insn_for_debug): When force_next_insn is true, also leave only
1169 next insn in the ready list.
1170 (sel_sched_region): When the region wasn't scheduled, make another pass
1171 over it with force_next_insn set to 1.
1173 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1175 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
1177 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1178 for interrupt control.
1179 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1180 functions for interrupt control.
1181 * config/nds32/nds32-intrinsic.md: Likewise.
1182 * config/nds32/nds32_intrinsic.h: Likewise.
1183 * config/nds32/nds32.h (nds32_builtins): Likewise.
1185 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1187 * config/nds32/nds32.c (nds32_init_machine_status,
1188 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
1189 strict_aligned_p field.
1190 (nds32_expand_to_rtl_hook): New function.
1191 (TARGET_EXPAND_TO_RTL_HOOK): Define.
1192 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
1194 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1195 Chung-Ju Wu <jasonwucj@gmail.com>
1197 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
1198 * config/nds32/nds32-n7.md: New file.
1199 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
1200 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
1202 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
1203 * config/nds32/nds32.md (pipeline_model): Add n7.
1204 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
1205 * config/nds32/pipelines.md: Include n7 settings.
1207 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1208 Chung-Ju Wu <jasonwucj@gmail.com>
1210 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
1211 * config/nds32/nds32-e8.md: New file.
1212 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
1213 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
1215 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
1216 * config/nds32/nds32.md (pipeline_model): Add e8.
1217 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
1218 * config/nds32/pipelines.md: Include e8 settings.
1220 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1221 Chung-Ju Wu <jasonwucj@gmail.com>
1223 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
1224 * config/nds32/nds32-n8.md: New file.
1225 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
1226 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
1228 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
1229 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
1230 * config/nds32/nds32.md (pipeline_model): Add n8.
1231 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
1232 * config/nds32/pipelines.md: Include n8 settings.
1234 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1235 Chung-Ju Wu <jasonwucj@gmail.com>
1237 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
1238 * config/nds32/nds32-n9-2r1w.md: New file.
1239 * config/nds32/nds32-n9-3r2w.md: New file.
1240 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
1241 nds32_register_ports): New or modify for cpu n9.
1242 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
1244 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
1245 * config/nds32/nds32-utils.c: New file.
1246 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
1247 TARGET_MUL_SLOW): Define.
1248 * config/nds32/nds32.md (pipeline_model): New attribute.
1249 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
1250 New options that support cpu n9.
1251 * config/nds32/pipelines.md: Include n9 settings.
1252 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
1254 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1256 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
1257 information if necessary.
1258 (output_cond_branch_compare_zero): Likewise.
1259 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
1260 (nds32_target_alignment): Refine for alignment.
1261 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
1262 (FUNCTION_BOUNDARY): Modify.
1263 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
1265 * config/nds32/nds32.opt (malways-align, malign-functions): New.
1267 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1269 * config/nds32/constants.md (unspec_volatile_element): Add values for
1270 TLB operation and data prefetch.
1271 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1272 functions for TLB operation and data prefetch.
1273 * config/nds32/nds32-intrinsic.md: Likewise.
1274 * config/nds32/nds32_intrinsic.h: Likewise.
1275 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
1276 (nds32_print_operand): Likewise.
1277 * config/nds32/nds32.h (nds32_builtins): Likewise.
1279 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
1280 Andrew Pinski <pinsika@gcc.gnu.org>
1283 * match.pd: Use constant_boolean_node of correct type instead of
1284 boolean_true_node or boolean_false_node for simplifying
1285 pointer comparisons to zero.
1287 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1289 PR tree-optimization/80021
1290 * tree.c (verify_type_variant): Make error call in verify_variant_match
1291 translatable and remove final full stop.
1293 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1295 * config/nds32/constants.md (unspec_volatile_element): Add
1296 UNSPEC_VOLATILE_EH_RETURN.
1297 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
1298 nds32_output_stack_pop): Support dwarf exception handling process.
1299 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
1300 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
1301 exception handling process.
1302 (nds32_compute_stack_frame): Likewise.
1303 (nds32_return_addr_rtx): Likewise.
1304 (nds32_initial_elimination_offset): Likewise.
1305 (nds32_expand_prologue): Likewise.
1306 (nds32_expand_epilogue): Likewise.
1307 (nds32_dynamic_chain_address): New function.
1308 * config/nds32/nds32.h (machine_function): Add fields for dwarf
1310 (DYNAMIC_CHAIN_ADDRESS): Define.
1311 (EH_RETURN_DATA_REGNO): Define.
1312 (EH_RETURN_STACKADJ_RTX): Define.
1313 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
1314 patterns for dwarf exception handling.
1316 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1318 * config/nds32/nds32.h: Clean up obsolete macros.
1320 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1322 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1323 Add enum values for particular instructions.
1324 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1325 particular intrinsic functions.
1326 * config/nds32/nds32-intrinsic.md: Likewise.
1327 * config/nds32/nds32_intrinsic.h: Likewise.
1328 * config/nds32/nds32.h (nds32_builtins): Likewise.
1329 * config/nds32/nds32.md (type): Add pbsad and pbsada.
1330 (btst, ave): New patterns for particular instructions.
1332 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1334 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1335 Add enum values for atomic load/store and memory sync.
1336 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
1338 * config/nds32/nds32-intrinsic.md: Likewise.
1339 * config/nds32/nds32_intrinsic.h: Likewise.
1340 * config/nds32/nds32.h (nds32_builtins): Likewise.
1342 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1344 PR tree-optimization/85257
1345 * fold-const.c (native_encode_vector): If not all elts could fit
1346 and off is -1, return 0 rather than offset.
1347 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
1348 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
1349 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
1350 adjust buffer in native_interpret_expr call.
1352 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1354 * config/nds32/constants.md (unspec_volatile_element): Add cache
1355 control enum values.
1356 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
1357 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
1358 * config/nds32/nds32.c (nds32_cctl_names): New.
1359 (nds32_print_operand): Handle cache control register names.
1360 * config/nds32/nds32.h (nds32_builtins): New enum values.
1361 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
1363 * config/nds32/nds32.md (type): Add mmu.
1364 * config/nds32/pipelines.md (simple_insn): Add mmu.
1366 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1368 * config/nds32/nds32.md (type): Remove call.
1369 * config/nds32/pipelines.md (simple_insn): Likewise.
1371 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1373 * config/nds32/constants.md (unspec_volatile_element): Add
1374 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
1375 UNSPEC_VOLATILE_FMFCFG.
1376 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
1377 description for fmfcfg and fmfcsr.
1378 (bdesc_1arg): Add fmtcsr.
1379 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
1380 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
1381 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
1382 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
1383 unspec_fmfcfg): New patterns.
1384 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
1385 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
1386 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
1387 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
1388 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
1389 __nds32__fmfcfg): Define.
1391 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1393 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
1394 intrinsic register names.
1395 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
1396 intrinsic register enum values and macros.
1398 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1400 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
1401 for load/store addressing form.
1402 (nds32_print_operand_address): Likewise.
1404 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
1407 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
1408 based on LABEL_REF. Remove useless assertion.
1409 (pic_address_needs_scratch): Fix formatting.
1410 (sparc_legitimize_pic_address): Minor tweaks.
1411 (sparc_delegitimize_address): Adjust assertion accordingly.
1412 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
1413 into symbolic_operand.
1414 (movsi_high_pic_label_ref): Likewise.
1415 (movsi_lo_sum_pic_label_ref): Likewise.
1416 (movdi_pic_label_ref): Likewise.
1417 (movdi_high_pic_label_ref): Likewise.
1418 (movdi_lo_sum_pic_label_ref): Likewise.
1420 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
1422 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
1423 custom LIB_SPEC setup.
1425 2018-04-06 Ruslan Bukin <br@bsdpad.com>
1426 Kito Cheng <kito.cheng@gmail.com>
1428 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
1429 * config/riscv/freebsd.h: New.
1431 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1433 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
1434 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
1437 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1438 Kito Cheng <kito.cheng@gmail.com>
1440 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
1441 nds32_output_call, nds32_symbol_binds_local_p): New functions.
1442 * config/nds32/nds32-protos.h (nds32_output_call,
1443 nds32_output_return): Declare.
1444 * config/nds32/nds32.md: Refine all the call and return patterns.
1446 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1449 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
1450 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
1452 PR rtl-optimization/84872
1453 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
1454 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
1457 2018-04-06 Tamar Christina <tamar.christina@arm.com>
1459 * expr.c (copy_blkmode_to_reg): Revert 254862.
1460 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
1462 2018-04-06 Richard Biener <rguenther@suse.de>
1465 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
1466 after seeing a component reference with an adjacent field. Treat
1467 refs to arrays at struct end of external decls similar to
1468 refs to unconstrained commons.
1470 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1473 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
1474 look through SAVE_EXPRs with non-side-effects argument. Adjust
1476 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
1479 2018-04-06 Richard Biener <rguenther@suse.de>
1482 * alias.c (find_base_term): New wrapper around find_base_term
1483 unwinding CSELIB_VAL_PTR changes.
1484 (find_base_term): Do not restore CSELIB_VAL_PTR during the
1487 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1489 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
1491 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
1492 constant definitions.
1493 ("nop"): lr 0,0 -> nopr r0
1494 ("nop_lr0", "nop_lr1"): New insn definitions.
1496 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1498 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
1499 NDS32_V3PUSH_AVAILABLE_P macro.
1501 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
1502 Chung-Ju Wu <jasonwucj@gmail.com>
1504 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
1505 (nds32*-*-*): Add float and fpu_config into supported_defaults.
1506 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
1507 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
1508 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
1509 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
1510 * config/nds32/constraints.md: New constraints and checking for hard
1511 float configuration.
1512 * config/nds32/iterators.md: New mode iterator and attribute for hard
1513 float configuration.
1514 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
1516 * config/nds32/nds32-fpu.md: New file.
1517 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
1518 deal with hard float code generation.
1519 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
1521 (abi_type, float_reg_number): New enum type.
1522 * config/nds32/nds32-predicates.c: New predicates for hard float.
1523 * config/nds32/nds32-protos.h: Declare functions for hard float.
1524 * config/nds32/nds32.c: Implementation for hard float configuration.
1525 * config/nds32/nds32.h: Definitions for hard float configuration.
1526 * config/nds32/nds32.md: Include hard float machine description and
1527 modify patterns for hard float configuration.
1528 * config/nds32/nds32.opt: New options for hard float configuration.
1529 * config/nds32/predicates.md: New predicates for hard float
1532 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1534 * common/config/nds32/nds32-common.c
1535 (nds32_option_optimization_table): Enable -mreleax-hint by default.
1537 2018-04-05 Jakub Jelinek <jakub@redhat.com>
1540 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
1541 CONSTRUCTOR_ELT (ctor, ...)->value.
1543 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
1546 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
1548 2018-04-05 Tom de Vries <tom@codesourcery.com>
1551 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
1554 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1555 Kito Cheng <kito.cheng@gmail.com>
1557 * config/nds32/constraints.md (U33): Fine-tune checking condition.
1558 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
1559 * config/nds32/nds32.h (nds32_16bit_address_type): Add
1560 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
1562 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1563 Kito Cheng <kito.cheng@gmail.com>
1565 * config/nds32/constraints.md (Ufe): New memory constraint.
1566 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
1567 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
1568 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
1570 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
1571 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
1573 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1575 * config/nds32/nds32.md: Use optimize_size in the condition for
1576 alu-shift instructions.
1578 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1580 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
1582 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1584 * config/nds32/nds32.md (negsi2): Refine pattern.
1586 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1587 Chung-Ju Wu <jasonwucj@gmail.com>
1589 * config/nds32/iterators.md (shift_rotate): New code iterator.
1590 (shift): New code attribute.
1591 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
1592 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
1593 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
1594 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
1595 bit-wise operations.
1596 (andsi3, *andsi3): Ditto.
1597 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
1598 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
1599 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
1600 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
1601 nds32_ior_operand, nds32_xor_operand): New predicates.
1603 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1605 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
1606 (addsi3, subsi3): ... this.
1608 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1610 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
1612 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1614 * config/nds32/nds32.md: Adjust indention.
1616 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1618 * config/nds32/nds32.md (feature): New attribute.
1620 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1622 * config/nds32/nds32.md (subtype): New attribute.
1624 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
1627 * config/arm/arm-builtins.c (arm_expand_builtin): Change
1628 expansion to perform a bitwise AND of the argument followed by a
1629 boolean negation of the result.
1631 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
1633 PR rtl-optimization/84878
1634 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
1635 the basic block. Assert the use reference is not artificial and that
1636 it has an associated insn.
1638 2018-04-04 Michael Matz <matz@suse.de>
1640 * builtins.c (compute_objsize): Pass correct operand
1641 to array_at_struct_end_p.
1643 2018-04-04 Richard Biener <rguenther@suse.de>
1646 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
1647 from contexts for DINFO_LEVEL_TERSE and below.
1649 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1651 * config/nds32/nds32-doubleword.md (move_<mode>): Require
1652 resiter_operand condition.
1653 * config/nds32/nds32.md (*move<mode>): Ditto.
1655 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1656 Monk Chiang <sh.chiang04@gmail.com>
1658 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
1660 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1662 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
1664 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1665 Kito Cheng <kito.cheng@gmail.com>
1667 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
1668 nds32_cond_code_str, output_cond_branch,
1669 output_cond_branch_compare_zero, nds32_expand_cbranch,
1670 nds32_expand_cstore, nds32_expand_movcc,
1671 nds32_output_cbranchsi4_equality_zero,
1672 nds32_output_cbranchsi4_equality_reg,
1673 nds32_output_cbranchsi4_equality_reg_or_const_int,
1674 nds32_output_cbranchsi4_greater_less_zero: New functions.
1675 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
1676 nds32_expand_cstore, nds32_expand_movcc,
1677 nds32_output_cbranchsi4_equality_zero,
1678 nds32_output_cbranchsi4_equality_reg,
1679 nds32_output_cbranchsi4_equality_reg_or_const_int,
1680 nds32_output_cbranchsi4_greater_less_zero): Declare.
1681 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
1682 nds32_rimm11s_operand): New predicates.
1683 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
1684 * config/nds32/nds32.md: Rewrite all the branch and conditional move
1687 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1689 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
1690 * config/nds32/nds32.md: Ditto.
1691 * config/nds32/pipelines.md: Ditto.
1693 2018-04-04 Richard Biener <rguenther@suse.de>
1695 PR tree-optimization/85168
1696 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
1697 propagating abnormals.
1699 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1701 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
1703 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1704 Kito Cheng <kito.cheng@gmail.com>
1706 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
1707 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
1708 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
1709 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
1710 * config/nds32/nds32.md (sibcall_internal): New.
1711 (sibcall_register): Remove.
1712 (sibcall_immediate): Remove.
1713 (sibcall_value_internal): New.
1714 (sibcall_value_register): Remove.
1715 (sibcall_value_immediate): Remove.
1716 * config/nds32/predicates.md (nds32_general_register_operand): New.
1717 (nds32_call_address_operand): New.
1719 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1721 PR rtl-optimization/85167
1722 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
1723 bb_defs if *split_p, instead preinitialize it to NULL.
1725 PR tree-optimization/85156
1726 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
1727 evaluating the argument multiple times.
1729 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
1731 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
1733 (_mm_cvtpd_ps): Likewise.
1734 (_mm_cvttpd_epi32): Likewise.
1735 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
1736 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
1737 vector, pixel, and bool following altivec.h include.
1739 2018-04-03 Martin Sebor <msebor@redhat.com>
1741 * doc/extend.texi (Common Function Attributes): Clarify.
1742 (const attribute): Likewise.
1743 (pure attribute): Likewise.
1745 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1748 * config/i386/i386.c (ix86_expand_vector_set): Use
1749 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
1751 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
1753 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
1754 instructions when changing rounding bits to preserve precision bits
1755 in the x87 control word.
1757 2018-04-03 Martin Liska <mliska@suse.cz>
1759 PR tree-optimization/82491
1760 * rtl.h (strip_offset_and_add): Replace += suboffset with
1761 poly_uint64 () + suboffset.
1763 2018-03-29 Martin Liska <mliska@suse.cz>
1764 Martin Jambor <mjambor@suse.cz>
1767 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
1768 param_type is not an integral or pointer type.
1770 2018-04-03 Richard Biener <rguenther@suse.de>
1772 * sese.h (recompute_all_dominators): Remove.
1774 2018-04-02 Martin Sebor <msebor@redhat.com>
1776 * doc/invoke.texi (-Wrestrict): Fix typos.
1778 2018-04-02 Jim Wilson <jimw@sifive.com>
1780 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
1781 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
1782 (<optab>di3, <optab>si3_extend): Likewise.
1783 (<optab>si3_mask, <optab>si3_mask_1): New.
1784 (<optab>di3_mask, <optab>di3_mask_1): New.
1785 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
1786 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
1787 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
1789 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
1791 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
1794 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
1796 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
1797 (nds32_canonicalize_comparison): New function.
1799 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1800 Kito Cheng <kito.cheng@gmail.com>
1801 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1803 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
1804 * config/nds32/constants.md (unspec_volatile_element): Add
1805 UNSPEC_VOLATILE_RELAX_GROUP.
1806 * config/nds32/nds32-relax-opt.c: New file.
1807 * config/nds32/nds32-predicates.c
1808 (nds32_symbol_load_store_p): New function.
1809 * config/nds32/nds32-protos.h
1810 (nds32_symbol_load_store_p): Declare function.
1811 (make_pass_nds32_relax_opt): Declare new rtl pass function.
1812 * config/nds32/nds32.c
1813 (nds32_register_pass): New function to register pass.
1814 (nds32_register_passes): New function to register passes.
1815 * config/nds32/nds32.md (relax_group): New pattern.
1816 * config/nds32/nds32.opt (mrelax-hint): New option.
1817 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
1819 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
1821 * config/nds32/t-nds32: Modify files dependency.
1823 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1825 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
1826 (PROFILE_HOOK): Define its implementation.
1828 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1830 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
1831 type and 32-bit size.
1833 2018-04-01 Jakub Jelinek <jakub@redhat.com>
1836 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
1837 (V_128_256): New mode iterator.
1838 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
1839 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
1840 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
1842 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
1843 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
1845 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
1848 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
1849 NaN inputs correctly.
1851 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
1854 * config/rs6000/vsx.md (??r): New mode attribute.
1855 (*vsx_mov<mode>_64bit): Use it.
1856 (*vsx_mov<mode>_32bit): Likewise.
1858 2018-03-30 Martin Sebor <msebor@redhat.com>
1860 PR tree-optimization/84818
1861 * builtins.c (check_access): Use warning_n.
1863 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1866 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
1868 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
1871 2018-03-30 Julia Koval <julia.koval@intel.com>
1874 * x86-tune.def (movx, partial_reg_dependency): Enable for
1877 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
1880 * lra-constraints.c (process_alt_operands): Move setting
1881 this_alternative_matches below.
1883 2018-03-29 Martin Liska <mliska@suse.cz>
1886 * doc/invoke.texi: Document how LTO works with debug info.
1887 Describe auto-load support of binutils. Mention 'x86-64'
1888 as valid option value of -march option.
1890 2018-03-29 Jakub Jelinek <jakub@redhat.com>
1892 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
1895 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
1896 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
1897 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
1900 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1903 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
1904 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
1905 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
1906 for RS6000_BTM_POWERPC64.
1907 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
1908 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
1909 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
1914 2018-03-28 Carl Love <cel@us.ibm.com>
1917 2017-09-27 Carl Love <cel@us.ibm.com>
1919 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
1920 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
1921 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
1924 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1926 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
1927 instead of __vector bool.
1928 (_mm_max_pu8): Likewise.
1929 (_mm_min_pi16): Likewise.
1931 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1934 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
1935 (DIVWEUO): Likewise.
1937 (DIVDEUO): Likewise.
1938 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
1939 DIVWEUO and DIVDEUO.
1940 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
1941 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
1942 (div_extend): Likewise.
1943 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
1945 (__builtin_divweuo): Likewise.
1946 (__builtin_divdeo): Likewise.
1947 (__builtin_divdeuo): Likewise.
1949 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1952 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
1953 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
1955 PR tree-optimization/82004
1956 * gimple-match-head.c (optimize_pow_to_exp): New function.
1957 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
1958 Don't fold to exp if optimize_pow_to_exp is false.
1960 2018-03-28 Martin Liska <mliska@suse.cz>
1963 * calls.c (initialize_argument_information): Fix trailing space.
1964 * common.opt: Fix typo and provide better explanation for
1965 -fsanitize-coverage option.
1966 * config/i386/i386.opt: Fix typo.
1968 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1969 Martin Liska <mliska@suse.cz>
1972 * gimplify.c (asan_poison_variable): Don't do the check for
1973 gimplify_omp_ctxp here.
1974 (gimplify_decl_expr): Do it here.
1975 (gimplify_target_expr): Likewise.
1977 2018-03-28 Martin Liska <mliska@suse.cz>
1980 * config/i386/i386.c (ix86_function_arg_advance): Do not call
1981 chkp_type_bounds_count if MPX is not enabled.
1983 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
1985 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
1987 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
1990 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
1991 function to create the function decl for complex long double
1992 multiply and divide for -mabi=ieeelongdouble.
1993 (init_float128_ieee): Call it.
1995 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
1998 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
1999 -fcf-protection=branch -mibt.
2000 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
2002 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2005 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
2007 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
2010 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
2011 extern array declarations.
2013 2018-03-27 Richard Biener <rguenther@suse.de>
2016 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
2017 explicit single_use checks.
2019 2018-03-27 Richard Biener <rguenther@suse.de>
2021 PR tree-optimization/85082
2022 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
2025 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2027 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
2028 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
2029 Turn on fasynchronous-unwind-tables and funwind-tables.
2031 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
2034 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
2035 (*bmi_blsr_<mode>_ccz): Ditto.
2037 2018-03-26 Tom de Vries <tom@codesourcery.com>
2039 PR tree-optimization/85063
2040 * omp-general.c (offloading_function_p): New function. Factor out
2042 * omp-offload.c (pass_omp_target_link::gate): ... here.
2043 * omp-general.h (offloading_function_p): Declare.
2044 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
2045 with attribute omp declare target for offloading functions.
2047 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
2049 PR tree-optimization/84005
2050 * tree-data-ref.h (get_base_for_alignment): Declare.
2051 * tree-data-ref.c (get_base_for_alignment_1): New function.
2052 (get_base_for_alignment): Likewise.
2053 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
2054 get_base_for_alignment to find a suitable base object, instead
2055 of always using drb->base_address.
2057 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2060 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
2061 known size by default.
2063 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
2066 * lra-constraints.c (process_alt_operands): Don't match BLKmode
2067 and non BLKmode operands.
2069 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2072 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
2073 Clean up attributes.
2075 2018-03-23 Richard Biener <rguenther@suse.de>
2078 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
2079 we are going to emit early debug for LTO.
2081 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2084 * function.c (match_asm_constraints_1): Don't optimize if input
2085 doesn't satisfy general_operand predicate for output's mode.
2088 * alias.c (write_dependence_p): Don't require for x_canonicalized
2089 non-VOIDmode if x has VOIDmode.
2092 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
2093 just don't try to optimize it rather than assert it never happens.
2095 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2097 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
2098 macro expansions for definition of ST_INTERNAL_<mode> and
2099 LD_INTERNAL_<mode> builtins.
2100 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
2102 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
2104 (altivec_expand_st_builtin): Likewise.
2105 (altivec_expand_builtin): Remove calls to deleted functions.
2106 (rs6000_address_for_altivec): Delete this function.
2107 * config/rs6000/vector.md: Remove expands for
2108 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
2110 2018-03-22 Sudakshina Das <sudi.das@arm.com>
2113 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
2114 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
2115 re-computing once computed.
2116 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
2117 (arm_init_machine_status): Initialize
2118 machine->static_chain_stack_bytes.
2120 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2123 * doc/extend.texi: Add four new prototypes for vec_ld.
2124 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
2125 definitions for more logical presentation.
2126 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
2127 entries for V1TI variants of __builtin_altivec_ld builtin.
2128 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
2129 handling of V1TI variant of LVX icode pattern.
2130 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
2131 (rs6000_gimple_fold_builtin): Likewise.
2132 (altivec_init_builtins): Add code to define
2133 __builtin_altivec_lvx_v1ti function.
2135 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2138 * function.c (match_asm_constraints_1): Don't do the optimization
2139 if input isn't a REG, SUBREG, MEM or constant.
2141 2018-03-22 Tom de Vries <tom@codesourcery.com>
2143 PR tree-optimization/84956
2144 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
2145 bb_has_abnormal_pred.
2147 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2150 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
2151 DECL_INITIAL (decl) to decl at the end.
2152 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
2155 2018-03-21 Joseph Myers <joseph@codesourcery.com>
2157 * doc/extend.texi (__builtin_tgmath): Document when complex
2158 integer types are treated as _Complex _Float64.
2160 2018-03-21 Tom de Vries <tom@codesourcery.com>
2162 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
2164 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2166 PR tree-optimization/84960
2167 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
2168 if it is ENTRY block, move them into single succ of ENTRY in that case.
2170 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
2172 PR tree-optimization/84811
2173 * poly-int.h (poly_span_traits): Remove the T3 parameter and
2174 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
2175 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
2176 (known_subrange_p): Update accordingly. Cast each value involved
2177 in the size comparison, rather than casting the result of the
2180 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2182 PR tree-optimization/84982
2183 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
2184 by flipping the least significant bit rather than all bits from
2185 bitpos to bitpos + bitsize - 1.
2187 2018-03-21 Nathan Sidwell <nathan@acm.org>
2189 * doc/extend.texi (Deprecated Features): Remove mention of
2190 long-deleted deprecations.
2192 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2195 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
2196 * configure: Regenerate.
2198 2018-03-21 Tom de Vries <tom@codesourcery.com>
2200 PR tree-optimization/83126
2201 * tree-parloops.c (num_phis): New function.
2202 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
2204 2018-03-21 Nathan Sidwell <nathan@acm.org>
2206 * doc/extend.texi (Deprecated Features): Update deprecated flags,
2207 mention anon-struct/union members and trailing attributes.
2209 2018-03-21 Bin Cheng <bin.cheng@arm.com>
2211 PR tree-optimization/84969
2212 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
2213 builtin memset partitions if they set different rhs values.
2215 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2217 PR rtl-optimization/84989
2218 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
2219 VEC_DUPLICATE with scalar result mode.
2221 2018-03-21 Martin Liska <mliska@suse.cz>
2224 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
2225 not intended return statement.
2227 2018-03-21 Martin Liska <mliska@suse.cz>
2230 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
2231 (chkp_find_bound_slots_1): Limit number of iterations.
2233 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
2236 * Minor grammar fixes for x86 options.
2238 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2241 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
2242 holding REG_CFA_RESTORE notes, instead turn them into a USE.
2244 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
2247 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
2248 (altivec_lvx_<mode>_1op): Likewise.
2249 (altivec_stvx_<mode>_2op): Likewise.
2250 (altivec_stvx_<mode>_1op): Likewise.
2251 (altivec_lvx_<VM2:mode>): New define_expand.
2252 (altivec_stvx_<VM2:mode>): Likewise.
2253 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
2254 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2255 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
2256 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2257 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
2258 (rs6000_gen_lvx): Likewise.
2259 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
2260 (altivec_expand_stv_builtin): Likewise.
2261 (altivec_expand_builtin): Likewise.
2262 * config/rs6000/vector.md: Likewise.
2264 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2267 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
2270 2018-03-20 Richard Biener <rguenther@suse.de>
2273 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
2274 sign-conversions as zero, fall back to standard scalar_stmt
2277 2018-03-20 Martin Liska <mliska@suse.cz>
2280 * predict.c (rebuild_frequencies): Handle case when we have
2281 PROFILE_ABSENT, but flag_guess_branch_prob is false.
2283 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2286 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
2287 flag_section_anchors.
2288 * varasm.c (use_blocks_for_decl_p): Remove hack for
2289 dw2_force_const_mem.
2292 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
2294 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
2295 be created, use lowpart_subreg of operands[0] rather than operands[0]
2297 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
2298 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
2299 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
2300 and n constraint instead of aarch64_shift_imm_di and Usd.
2301 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
2302 (*aarch64_<optab>_reg_minus<mode>3): ... this.
2304 2018-03-20 Sudakshina Das <sudi.das@arm.com>
2307 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
2308 to favor GPR over NEON registers.
2309 (<shift>di3_neon): Likewise.
2311 2018-03-20 Tom de Vries <tom@codesourcery.com>
2314 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
2315 (nvptx_process_pars): Emit bar.sync asap and alap.
2317 2018-03-20 Tom de Vries <tom@codesourcery.com>
2320 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
2321 seen_label if seen_label is already set.
2323 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2326 * config/i386/i386.c (fold_builtin_cpu): For features above 31
2327 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
2328 Use 1U instead of 1. Formatting fixes.
2331 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
2332 instead of TREE_TYPE (s1) for the return value.
2334 2018-03-19 Jakub Jelinek <jakub@redhat.com>
2336 PR tree-optimization/84946
2337 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
2338 bitsize + bitsize in poly_uint64 rather than poly_int64.
2341 * dwarf2asm.c: Include fold-const.c.
2342 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
2343 of decl rather than decl itself.
2345 PR rtl-optimization/84643
2346 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
2348 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
2351 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
2352 calling assemble_variable.
2354 2018-03-19 Sudakshina Das <sudi.das@arm.com>
2357 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
2358 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
2360 2018-03-19 Jim Wilson <jimw@sifive.com>
2363 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
2364 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
2365 (riscv_first_stack_step): Likewise.
2366 (riscv_option_override): Use STACK_BOUNDARY instead of
2368 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
2370 (BIGGEST_ALIGNMENT): Set to 128.
2371 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
2372 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
2375 2018-03-19 Richard Biener <rguenther@suse.de>
2377 PR tree-optimization/84933
2378 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
2379 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
2381 2018-03-19 Richard Biener <rguenther@suse.de>
2383 PR tree-optimization/84859
2384 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
2385 (cond_if_else_store_replacement): Perform sinking operation on
2386 single-store BBs regardless of MAX_STORES_TO_SINK setting.
2387 Generalize what a BB with a single eligible store is.
2389 2018-03-19 Richard Biener <rguenther@suse.de>
2391 PR tree-optimization/84929
2392 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
2393 chrec_is_positive against non-chrec arg.
2395 2018-03-19 Tamar Christina <tamar.christina@arm.com>
2398 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
2400 2018-03-18 Martin Liska <mliska@suse.cz>
2402 PR rtl-optimization/84635
2403 * regrename.c (build_def_use): Use matches_mode only when
2406 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
2408 PR tree-optimization/84913
2409 * tree-vect-loop.c (vectorizable_reduction): Don't try to
2410 vectorize chains of COND_EXPRs.
2412 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2414 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
2416 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2418 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
2420 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2422 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
2424 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2425 Kito Cheng <kito.cheng@gmail.com>
2427 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
2428 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
2429 (nds32_adjust_reg_alloc_order): New function.
2430 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
2432 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
2434 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
2435 nds32_print_operand, nds32_print_operand_address): Use
2436 HOST_WIDE_INT_PRINT_DEC instead.
2438 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2440 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
2442 2018-03-17 Jakub Jelinek <jakub@redhat.com>
2445 * config/i386/i386.c (initial_ix86_tune_features,
2446 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
2448 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
2449 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
2450 rather than 1u << ix86_tune. Formatting fix.
2451 (ix86_option_override_internal): Change ix86_arch_mask from
2452 unsigned int to unsigned HOST_WIDE_INT, initialize to
2453 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
2454 (ix86_function_specific_restore): Likewise.
2456 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2459 * postreload.c (reload_combine_recognize_pattern): Perform
2460 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
2461 truncate_int_for_mode the result for the destination's mode.
2464 * hsa-gen.c (mem_type_for_type): Fix comment typo.
2465 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
2467 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2470 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
2473 * lra-assigns.c (lra_split_hard_reg_for): Don't use
2474 regno_allocno_class_array and sorted_pseudos.
2475 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
2476 insns where regno is used.
2478 2018-03-16 Martin Liska <mliska@suse.cz>
2481 * multiple_target.c (create_dispatcher_calls): Redirect
2482 reference in the symbol table.
2484 2018-03-16 Martin Liska <mliska@suse.cz>
2487 * multiple_target.c (create_dispatcher_calls): Redirect also
2490 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2494 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
2495 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
2496 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
2498 2018-03-16 Julia Koval <julia.koval@intel.com>
2500 * doc/invoke.texi (Skylake Server): Add CLWB.
2501 Cannonlake): Remove CLWB.
2503 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2505 PR tree-optimization/84841
2506 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
2508 (FLOAT_ONE_CONST_TYPE): Define.
2509 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
2510 (sort_by_operand_rank): Put entries with higher constant_type last
2511 rather than first to match comments.
2513 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
2515 * config/nios2/nios2.md (movsi_internal): Fix thinko in
2518 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2521 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
2522 check and use address of target always.
2524 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
2527 * config/i386/i386.c (indirect_thunk_needed): Update comments.
2528 (indirect_thunk_bnd_needed): Likewise.
2529 (indirect_thunks_used): Likewise.
2530 (indirect_thunks_bnd_used): Likewise.
2531 (indirect_return_needed): New.
2532 (indirect_return_bnd_needed): Likewise.
2533 (output_indirect_thunk_function): Add a bool argument for
2535 (output_indirect_thunk_function): Don't generate alias for
2536 function return thunk.
2537 (ix86_code_end): Call output_indirect_thunk_function to generate
2538 function return thunks.
2539 (ix86_output_function_return): Set indirect_return_bnd_needed
2540 and indirect_return_needed instead of indirect_thunk_bnd_needed
2541 and indirect_thunk_needed.
2543 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
2545 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
2546 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
2547 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
2549 2018-03-15 David Malcolm <dmalcolm@redhat.com>
2550 Paul Hua <paul.hua.gm@gmail.com>
2553 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
2555 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
2557 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
2558 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
2561 2018-03-15 Tamar Christina <tamar.christina@arm.com>
2564 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
2565 instead of GET_MODE_SIZE when comparing Units.
2567 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2570 * varasm.c (hash_section): Return an unchangeble hash value
2571 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
2572 Return !aarch64_can_use_per_function_literal_pools_p ().
2574 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2577 * optabs.c (emit_conditional_move): Pass address of cmode's copy
2578 rather than address of cmode as last argument to prepare_cmp_insn.
2580 2018-03-15 Julia Koval <julia.koval@intel.com>
2582 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
2583 F_AVX512VNNI, F_AVX512BITALG): New.
2585 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
2588 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
2589 insn for floating-point loads and stores.
2591 2018-03-14 Carl Love <cel@us.ibm.com>
2593 * config/rs6000/rs6000-c.c: Add macro definitions for
2594 ALTIVEC_BUILTIN_VEC_PERMXOR.
2595 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
2596 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
2597 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
2598 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
2600 * config/doc/extend.texi: Add prototypes for vec_permxor.
2602 2018-03-14 David Malcolm <dmalcolm@redhat.com>
2605 * diagnostic-show-locus.c (class layout_point): Convert m_line
2606 from int to linenum_type.
2607 (line_span::comparator): Use linenum "compare" function when
2608 comparing line numbers.
2609 (test_line_span): New function.
2610 (layout_range::contains_point): Convert param "row" from int to
2612 (layout_range::intersects_line_p): Likewise.
2613 (layout::will_show_line_p): Likewise.
2614 (layout::print_source_line): Likewise.
2615 (layout::should_print_annotation_line_p): Likewise.
2616 (layout::print_annotation_line): Likewise.
2617 (layout::print_leading_fixits): Likewise.
2618 (layout::annotation_line_showed_range_p): Likewise.
2619 (struct line_corrections): Likewise for field m_row.
2620 (line_corrections::line_corrections): Likewise for param "row".
2621 (layout::print_trailing_fixits): Likewise.
2622 (layout::get_state_at_point): Likewise.
2623 (layout::get_x_bound_for_row): Likewise.
2624 (layout::print_line): Likewise.
2625 (diagnostic_show_locus): Likewise for locals "last_line" and
2627 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
2628 * input.c (selftest::test_linenum_comparisons): New function.
2629 (selftest::input_c_tests): Call it.
2630 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
2631 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
2632 * selftest.h (ASSERT_GT): New macro.
2633 (ASSERT_GT_AT): New macro.
2634 (ASSERT_LT): New macro.
2635 (ASSERT_LT_AT): New macro.
2637 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
2639 PR rtl-optimization/84780
2640 * combine.c (distribute_links): Don't make a link based on pc_rtx.
2642 2018-03-14 Martin Liska <mliska@suse.cz>
2644 * tree.c (record_node_allocation_statistics): Use
2645 get_stats_node_kind.
2646 (get_stats_node_kind): New function extracted from
2647 record_node_allocation_statistics.
2648 (free_node): Use get_stats_node_kind.
2650 2018-03-14 Richard Biener <rguenther@suse.de>
2652 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
2653 that the value-set of ANTIC_IN doesn't grow.
2656 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2658 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2659 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2661 2018-03-14 Julia Koval <julia.koval@intel.com>
2663 * config.gcc (icelake-client, icelake-server): New.
2665 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
2666 (initial_ix86_arch_features): Ditto.
2667 (PTA_SKYLAKE): Add SGX.
2668 (PTA_ICELAKE): Remove.
2669 (PTA_ICELAKE_CLIENT): New.
2670 (PTA_ICELAKE_SERVER): New.
2671 (ix86_option_override_internal): Split up icelake on icelake client and
2673 (get_builtin_code_for_version): Ditto.
2674 (fold_builtin_cpu): Ditto.
2675 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
2676 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
2677 * config/i386/i386.h (processor_type): Ditto.
2678 * doc/invoke.texi: Ditto.
2680 2018-03-14 Jakub Jelinek <jakub@redhat.com>
2683 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
2684 INTEGER_CST offset, add it together with bitpos / 8 and
2685 sign extend based on POINTER_SIZE.
2689 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
2692 * config/i386/constraints.md (Yc): New register constraint.
2693 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
2694 Use Yc constraint for alternative 2 of operand 0. Remove
2695 preferred_for_speed attribute.
2697 2018-03-14 Richard Biener <rguenther@suse.de>
2699 PR tree-optimization/84830
2700 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
2701 with the old one to avoid oscillations.
2703 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
2706 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
2708 (assign_by_spills): Return a flag of reload assignment failure.
2709 Do not process the reload assignment failures. Do not spill other
2710 reload pseudos if they has the same reg class. Update n if
2712 (lra_assign): Add a return arg. Set up from the result of
2713 assign_by_spills call.
2714 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2715 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2716 usage_insns if it is not NULL.
2717 (spill_hard_reg_in_range): New function.
2718 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2719 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2720 function prototypes.
2721 (lra_assign): Change prototype.
2722 * lra.c (lra): Add code to deal with fails by splitting hard reg
2725 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
2727 * config/riscv/riscv.opt (mrelax): New option.
2728 * config/riscv/riscv.c (riscv_file_start): Emit ".option
2729 "norelax" when riscv_mrelax is disabled.
2730 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
2732 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2735 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
2736 reassociation for int modes.
2738 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2740 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
2741 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
2743 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
2744 * config/aarch64/aarch64-sve.md
2745 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
2746 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
2747 (*extend<mode><Vwide>2): Rename to...
2748 (aarch64_sve_extend<mode><Vwide>2): ...this.
2749 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
2750 renaming the old pattern to...
2751 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
2753 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
2754 define_expand, renaming the old pattern to...
2755 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
2756 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
2757 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
2758 account when deciding which SVE instruction the optab should use.
2759 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
2761 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2763 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
2764 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
2765 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
2766 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
2767 (tlsdesc_small_<mode>): Turn a define_expand and use
2768 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
2769 (tlsdesc_small_advsimd_<mode>): ...this.
2770 (tlsdesc_small_sve_<mode>): New pattern.
2772 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2774 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
2775 (UNSPEC_UMUL_HIGHPART): New constants.
2776 (MUL_HIGHPART): New int iteraor.
2777 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
2778 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
2780 (*<su>mul<mode>3_highpart): New define_insn.
2782 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2785 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2788 2018-03-13 Martin Liska <mliska@suse.cz>
2791 * (sem_item_optimizer::sem_item_optimizer): Initialize new
2793 (sem_item_optimizer::~sem_item_optimizer): Release it.
2794 (sem_item_optimizer::merge_classes): Register variable aliases.
2795 (sem_item_optimizer::fixup_pt_set): New function.
2796 (sem_item_optimizer::fixup_points_to_sets): Likewise.
2797 * ipa-icf.h: Declare new variables and functions.
2799 2018-03-13 Jakub Jelinek <jakub@redhat.com>
2802 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
2803 integer_pow2p@2 and test integer_pow2p in condition.
2804 (A < 0 ? C : 0): Similarly for @1.
2807 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
2808 characters starting at p contain '\0' character, don't look beyond
2812 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
2813 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
2816 * reg-stack.c (change_stack): Change update_end var from int to
2817 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
2818 also call set_block_for_insn on the newly added insns and rescan.
2821 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
2822 on the last operand.
2825 * tree.c (stabilize_reference_1): Return save_expr (e) for
2826 STATEMENT_LIST even if it doesn't have side-effects.
2828 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
2830 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
2832 2018-03-12 Renlin Li <renlin.li@arm.com>
2834 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
2835 aarch64_output_scalar_simd_mov_immediate.
2837 2018-03-12 Martin Sebor <msebor@redhat.com>
2839 PR tree-optimization/83456
2840 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
2841 for perfectly overlapping calls to memcpy.
2842 (gimple_fold_builtin_memory_chk): Same.
2843 (gimple_fold_builtin_strcpy): Handle no-warning.
2844 (gimple_fold_builtin_stxcpy_chk): Same.
2845 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
2847 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2849 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
2850 parameter. Use it for SFmode.
2851 (rs6000_function_arg_advance_1): Adjust.
2852 (rs6000_function_arg): Adjust.
2853 (rs6000_gimplify_va_arg): Pass false for that new parameter.
2855 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2857 PR rtl-optimization/84169
2858 PR rtl-optimization/84780
2859 * combine.c (can_combine_p): Check for a 2-insn combination whether
2860 the destination register is used between the two insns, too.
2862 2018-03-12 Richard Biener <rguenther@suse.de>
2864 PR tree-optimization/84803
2865 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
2866 for refs DR analysis didn't process.
2868 2018-03-12 Richard Biener <rguenther@suse.de>
2870 PR tree-optimization/84777
2871 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
2872 force-vectorize loops ignore whether we are optimizing for size.
2874 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
2876 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
2877 (TARGET_MD_ASM_ADJUST): Define.
2879 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
2880 Kito Cheng <kito.cheng@gmail.com>
2881 Chung-Ju Wu <jasonwucj@gmail.com>
2883 * config/nds32/nds32.c (nds32_compute_stack_frame,
2884 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
2885 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
2886 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
2887 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
2888 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
2889 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
2890 * config/nds32/nds32.md (prologue, epilogue): Use macro
2891 NDS32_V3PUSH_AVAILABLE_P to do checking.
2893 2018-03-11 Jakub Jelinek <jakub@redhat.com>
2896 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
2897 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
2898 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
2899 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
2900 addition of most attributes on !orig_type_die or the attribute not
2901 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
2903 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2904 Chung-Ju Wu <jasonwucj@gmail.com>
2906 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
2908 * config/nds32/nds32.opt (mvh): New option.
2910 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2911 Chung-Ju Wu <jasonwucj@gmail.com>
2913 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
2915 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
2916 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
2919 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2920 Chung-Ju Wu <jasonwucj@gmail.com>
2922 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
2924 * config/nds32/nds32-multiple.md (strlensi): New pattern.
2925 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
2927 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
2928 Kito Cheng <kito.cheng@gmail.com>
2929 Chung-Ju Wu <jasonwucj@gmail.com>
2931 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
2932 UNSPEC_FFMISM and UNSPEC_FLMISM.
2933 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
2934 for ffb, ffmism and flmism.
2935 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
2936 (unspec_ffmism): Ditto.
2937 (unspec_flmism): Ditto.
2938 (nds32_expand_builtin_impl): Check if string extension is available.
2939 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
2940 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
2942 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
2945 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2948 * lra-assigns.c (assign_by_spills): Return a flag of reload
2949 assignment failure. Do not process the reload assignment
2950 failures. Do not spill other reload pseudos if they has the same
2952 (lra_assign): Add a return arg. Set up from the result of
2953 assign_by_spills call.
2954 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2955 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2956 usage_insns if it is not NULL.
2957 (spill_hard_reg_in_range): New function.
2958 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2959 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2960 function prototypes.
2961 (lra_assign): Change prototype.
2962 * lra.c (lra): Add code to deal with fails by splitting hard reg
2965 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
2968 * config/i386/i386.opt: Replace Enforcment with Enforcement.
2970 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
2973 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
2974 (dw_val_node): Add val_symbolic_view.
2975 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
2976 (symview_upper_bound): New.
2977 (new_line_info_table): Initialize symviews_since_reset.
2978 (dwarf2out_source_line): Count symviews_since_reset and set
2979 symview_upper_bound.
2980 (dw_val_equal_p): Handle symview.
2981 (add_AT_symview): New.
2982 (print_dw_val): Handle symview.
2983 (attr_checksum, attr_checksum_ordered): Likewise.
2984 (same_dw_val_p, size_of_die): Likewise.
2985 (value_format, output_die): Likewise.
2986 (add_high_low_attributes): Use add_AT_symview for entry_view.
2987 (dwarf2out_finish): Reset symview_upper_bound, clear
2990 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
2993 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
2994 Add strict argument and use it.
2995 (rs6000_split_multireg_move): Update for new strict argument.
2996 (mem_operand_gpr): Disallow all non-offsettable addresses.
2997 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
2999 2018-03-09 Jakub Jelinek <jakub@redhat.com>
3002 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
3003 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
3004 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
3007 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
3008 decl, use remap_type if we want to use the type.
3010 2018-03-09 Martin Sebor <msebor@redhat.com>
3012 PR tree-optimization/84526
3013 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3015 (builtin_access::generic_overlap): Be prepared to handle non-array
3018 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
3020 PR rtl-optimization/84682
3021 * lra-constraints.c (process_address_1): Check is_address flag
3022 for address constraints.
3023 (process_alt_operands): Likewise.
3024 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
3025 preprocess_constraints.
3026 * recog.h (preprocess_constraints): Add oploc parameter.
3028 * recog.c (preprocess_constraints): Test address_operand for
3029 CT_ADDRESS constraints.
3031 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3034 * lra-assigns.c (assign_by_spills): Return a flag of reload
3035 assignment failure. Do not process the reload assignment
3036 failures. Do not spill other reload pseudos if they has the same
3038 (lra_assign): Add a return arg. Set up from the result of
3039 assign_by_spills call.
3040 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3041 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3042 usage_insns if it is not NULL.
3043 (spill_hard_reg_in_range): New function.
3044 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3045 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3046 function prototypes.
3047 (lra_assign): Change prototype.
3048 * lra.c (lra): Add code to deal with fails by splitting hard reg
3051 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3054 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
3055 Accept complain bool parameter. Only emit errors if it is true.
3056 (arm_parse_cpu_option_name): Likewise.
3057 (arm_target_thumb_only): Adjust callers of the above.
3058 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
3059 prototype to take a default true bool parameter.
3060 (arm_parse_arch_option_name): Likewise.
3062 2018-03-09 David Malcolm <dmalcolm@redhat.com>
3063 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
3067 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
3068 * configure: Regenerate.
3069 * configure.ac ("linker --version-script option"): New.
3070 ("linker soname option"): New.
3072 2018-03-09 Richard Biener <rguenther@suse.de>
3074 PR tree-optimization/84775
3075 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
3076 immediate uses of predicate stmts and mark them modified.
3079 PR tree-optimization/84178
3080 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3082 (version_loop_for_if_conversion): Delay update_ssa call.
3083 (tree_if_conversion): Delay update_ssa until after predicate
3086 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
3089 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
3090 when the function accesses prior frames.
3092 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3095 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
3096 gen_llsym, otherwise call maybe_gen_llsym.
3099 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
3100 has ',' character inside of it.
3102 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3105 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
3106 as clobbering CC_REGNUM.
3108 2018-03-08 Richard Biener <rguenther@suse.de>
3111 * tree-scalar-evolution.c: Include tree-into-ssa.h.
3112 (follow_copies_to_constant): Do not follow SSA names registered
3115 2018-03-08 Richard Biener <rguenther@suse.de>
3117 PR tree-optimization/84178
3118 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3120 (version_loop_for_if_conversion): Delay update_ssa call.
3121 (tree_if_conversion): Delay update_ssa until after predicate
3124 2018-03-08 David Malcolm <dmalcolm@redhat.com>
3126 PR tree-optimization/84178
3127 * tree-if-conv.c (release_bb_predicate): Remove the
3128 the assertion that the stmts have NULL use_ops.
3129 Discard the statements, asserting that they haven't
3130 yet been added to a BB.
3132 2018-03-08 Richard Biener <rguenther@suse.de>
3134 PR tree-optimization/84746
3135 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
3136 (phi_translate): Pass in destination ANTIC_OUT set.
3137 (phi_translate_1): Likewise. For a simplified result lookup
3138 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
3139 (phi_translate_set): Adjust.
3140 (do_pre_regular_insertion): Likewise.
3141 (do_pre_partial_partial_insertion): Likewise.
3143 2018-03-08 Martin Liska <mliska@suse.cz>
3145 PR gcov-profile/84735
3146 * doc/gcov.texi: Document usage of profile files.
3147 * gcov-io.h: Document changes in the format.
3149 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
3153 * dwarf2out.c (struct dw_line_info_table): Update comments for
3155 (FORCE_RESET_NEXT_VIEW): New.
3156 (FORCE_RESETTING_VIEW_P): New.
3157 (RESETTING_VIEW_P): Check for -1 too.
3158 (ZERO_VIEW_P): Likewise.
3159 (new_line_info_table): Force-reset next view.
3160 (dwarf2out_begin_function): Likewise.
3161 (dwarf2out_source_line): Simplify zero_view_p initialization.
3162 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
3163 view directly. Omit view when omitting .loc at line 0.
3165 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3167 PR tree-optimization/84740
3168 * tree-switch-conversion.c (process_switch): Call build_constructors
3169 only if info.phi_count is non-zero.
3171 PR tree-optimization/84739
3172 * tree-tailcall.c (find_tail_calls): Check call arguments against
3173 DECL_ARGUMENTS (current_function_decl) rather than
3174 DECL_ARGUMENTS (func) when checking for tail recursion.
3176 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3178 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
3179 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
3180 Volker Reichelt's entry and add entries for people that perform
3181 GCC fuzzy testing and report numerous bugs.
3183 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
3186 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
3187 readonly data in sdata, if that is disabled.
3188 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
3189 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3190 -mreadonly-in-sdata option.
3192 2018-03-07 Martin Sebor <msebor@redhat.com>
3194 PR tree-optimization/84468
3195 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
3196 basic block when looking for nul assignment.
3198 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
3201 * except.h (output_function_exception_table): Adjust prototype.
3202 * except.c (output_function_exception_table): Remove FNNAME parameter
3203 and add SECTION parameter. Ouput one part of the table at a time.
3204 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
3205 the first part of the exception table and emit unwind directives.
3206 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
3207 (i386_pe_seh_cold_init): Likewise.
3208 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
3209 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
3210 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
3211 (ix86_output_call_insn): Emit a nop in one more case for SEH.
3212 * config/i386/winnt.c: Include except.h.
3213 (struct seh_frame_state): Add reg_offset, after_prologue and
3214 in_cold_section fields.
3215 (i386_pe_seh_end_prologue): Set seh->after_prologue.
3216 (i386_pe_seh_cold_init): New function.
3217 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
3218 to seh->in_cold_section.
3219 (seh_emit_push): Record the offset of the push.
3220 (seh_emit_save): Record the offet of the save.
3221 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
3222 Test seh->after_prologue to disregard the epilogue.
3223 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
3224 (i386_pe_end_cold_function): New function.
3226 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3229 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
3230 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
3233 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
3235 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
3236 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
3239 * multiple_target.c: Include tree-inline.h and intl.h.
3240 (expand_target_clones): Diagnose and fail if node->definition and
3241 !tree_versionable_function_p (node->decl).
3243 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
3245 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
3247 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
3248 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
3249 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
3251 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3254 * combine.c (try_combine): Use reg_or_subregno instead of handling
3255 just paradoxical SUBREGs and REGs.
3257 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
3259 * config/arc/arc.c (arc_finalize_pic): Remove function.
3260 (arc_must_save_register): We use single base PIC register, remove
3261 checks to save/restore the PIC register.
3262 (arc_expand_prologue): Likewise.
3263 * config/arc/arc-protos.h (arc_set_default_type_attributes):
3265 (arc_verify_short): Likewise.
3266 (arc_attr_type): Likewise.
3267 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
3268 (walk_stores): Likewise.
3269 (arc_address_cost): Make it static.
3270 (arc_verify_short): Likewise.
3271 (branch_dest): Likewise.
3272 (arc_attr_type): Likewise.
3273 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
3274 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
3275 (arc_final_prescan_insn): Remove inserting the nops due to
3276 hardware hazards. It is done in reorg step.
3277 (insn_length_variant_t): Remove.
3278 (insn_length_parameters_t): Likewise.
3279 (arc_insn_length_parameters): Likewise.
3280 (arc_get_insn_variants): Likewise.
3281 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
3283 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3286 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
3289 PR tree-optimization/84687
3290 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
3292 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
3294 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3296 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
3297 Rename to ppc_speculation_barrier.
3298 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
3299 __builtin_ppc_speculation_barrier.
3301 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3304 * combine.c (combine_simplify_rtx): Don't try to simplify if
3305 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
3308 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
3310 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
3311 to 32 bytes when compiling for POWER9.
3313 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3316 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
3317 regparm >= 3 with no arg reg available also for calls with
3318 flag_force_indirect_call. Pass decl to ix86_function_regparm.
3321 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
3323 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
3325 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
3328 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
3330 2018-03-05 Richard Biener <rguenther@suse.de>
3332 PR tree-optimization/84486
3333 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
3334 When inserting a __builtin_assume_aligned call set the LHS
3335 SSA name alignment info accordingly.
3337 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
3339 PR tree-optimization/84114
3340 * config/aarch64/aarch64.c (aarch64_reassociation_width)
3341 Avoid reassociation of FLOAT_MODE addition.
3343 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
3345 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
3346 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
3347 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
3348 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
3349 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
3350 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
3351 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
3353 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
3354 __builtin_ia32_wbinvd): New builtins.
3355 (SPECIAL_ARGS2): New.
3356 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
3357 (SPECIAL_ARGS2): New.
3358 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
3359 (ix86_valid_target_attribute_inner_p): Ditto.
3360 (ix86_init_mmx_sse_builtins): Add special_args2.
3361 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
3362 TARGET_WBNOINVD_P): New.
3363 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
3364 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
3365 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
3366 * config/i386/immintrin.h (_wbinvd): New intrinsic.
3367 * config/i386/pconfigintrin.h: New file.
3368 * config/i386/wbnoinvdintrin.h: Ditto.
3369 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
3370 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
3372 2018-03-05 Richard Biener <rguenther@suse.de>
3374 PR tree-optimization/84670
3375 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3377 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3378 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3379 (compute_antic_aux): Only assert the number of values in ANTIC_IN
3380 doesn't grow if all successors (recursively) were visited at least
3383 2018-03-05 Richard Biener <rguenther@suse.de>
3385 PR tree-optimization/84650
3386 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
3387 if executed in the loop pipeline.
3389 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
3391 * doc/configfiles.texi (Configuration Files): Move info about
3392 conditionalizing $target-protos.h to...
3393 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
3394 differs from $target-protos.h.
3396 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
3397 Chung-Ju Wu <jasonwucj@gmail.com>
3399 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
3400 * config/nds32/nds32-multiple.md (setmemsi): Define.
3401 * config/nds32/nds32-memory-manipulation.c
3402 (nds32_gen_dup_4_byte_to_word_value): New.
3403 (emit_setmem_word_loop): New.
3404 (emit_setmem_byte_loop): New.
3405 (nds32_expand_setmem_loop): New.
3406 (nds32_expand_setmem_loop_v3m): New.
3407 (nds32_expand_setmem_unroll): New.
3408 (nds32_expand_setmem): New.
3410 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3411 Chung-Ju Wu <jasonwucj@gmail.com>
3413 * config/nds32/nds32-memory-manipulation.c
3414 (nds32_emit_load_store): New.
3415 (nds32_emit_post_inc_load_store): New.
3416 (nds32_emit_mem_move): New.
3417 (nds32_emit_mem_move_block): New.
3418 (nds32_expand_movmemsi_loop_unknown_size): New.
3419 (nds32_expand_movmemsi_loop_known_size): New.
3420 (nds32_expand_movmemsi_loop): New.
3421 (nds32_expand_movmemsi_unroll): New.
3422 (nds32_expand_movmemqi): Rename ...
3423 (nds32_expand_movmemsi): ... to this.
3424 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
3425 (movmemsi): ... to this.
3426 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
3427 (nds32_expand_movmemsi): ... to this.
3429 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3430 Monk Chiang <sh.chiang04@gmail.com>
3431 Chung-Ju Wu <jasonwucj@gmail.com>
3433 * config/nds32/nds32-protos.h
3434 (nds32_expand_load_multiple): New arguments.
3435 (nds32_expand_store_multiple): Ditto.
3436 (nds32_valid_multiple_load_store): Rename ...
3437 (nds32_valid_multiple_load_store_p): ... to this.
3438 * config/nds32/nds32-memory-manipulation.c
3439 (nds32_expand_load_multiple): Refine implementation.
3440 (nds32_expand_store_multiple): Ditto.
3441 * config/nds32/nds32-multiple.md
3442 (load_multiple): Update nds32_expand_load_multiple interface.
3443 (store_multiple): Update nds32_expand_store_multiple interface.
3444 * config/nds32/nds32-predicates.c
3445 (nds32_valid_multiple_load_store): Rename ...
3446 (nds32_valid_multiple_load_store_p): ... to this and refine
3448 * config/nds32/predicates.md
3449 (nds32_load_multiple_and_update_address_operation): New predicate.
3450 (nds32_store_multiple_and_update_address_operation): New predicate.
3452 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3453 Chung-Ju Wu <jasonwucj@gmail.com>
3455 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
3456 (combo): New attribute.
3457 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
3459 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
3461 * config/nds32/nds32.opt: Change -mcmodel= default value.
3463 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
3464 Monk Chiang <sh.chiang04@gmail.com>
3465 Chung-Ju Wu <jasonwucj@gmail.com>
3467 * config/nds32/constants.md (unspec_element): New enum.
3468 * config/nds32/constraints.md (Umw): New constraint.
3469 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
3470 * config/nds32/nds32-intrinsic.md: Likewise.
3471 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
3472 (nds32_valid_smw_lwm_base_p): New.
3473 (nds32_output_smw_single_word): New.
3474 (nds32_output_lmw_single_word): New.
3475 (nds32_expand_unaligned_load): New.
3476 (nds32_expand_unaligned_store): New.
3477 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
3478 (nds32_output_smw_single_word): Declare.
3479 (nds32_output_lmw_single_word): Declare.
3480 (nds32_expand_unaligned_load): Declare.
3481 (nds32_expand_unaligned_store): Declare.
3482 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
3483 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
3484 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
3485 NDS32_BUILTIN_UASTORE_DW.
3486 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
3489 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
3490 Kito Cheng <kito.cheng@gmail.com>
3491 Chung-Ju Wu <jasonwucj@gmail.com>
3493 * config/nds32/nds32-intrinsic.c
3494 (nds32_expand_builtin_null_ftype_reg): Delete.
3495 (nds32_expand_builtin_reg_ftype_imm): Ditto.
3496 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
3497 (nds32_read_argument): New.
3498 (nds32_legitimize_target): Ditto.
3499 (nds32_legitimize_argument): Ditto.
3500 (nds32_check_constant_argument): Ditto.
3501 (nds32_expand_unop_builtin): Ditto.
3502 (nds32_expand_unopimm_builtin): Ditto.
3503 (nds32_expand_binop_builtin): Ditto.
3504 (nds32_builtin_decl_impl): Ditto.
3505 (builtin_description): Ditto.
3506 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
3507 (nds32_init_builtins_impl): Ditto.
3508 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
3509 (nds32_builtin_decl): New.
3510 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
3511 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
3513 2018-03-02 Jeff Law <law@redhat.com>
3515 * reorg.c (stop_search_p): Handle DEBUG_INSN.
3516 (redundant_insn, fill_simple_delay_slots): Likewise.
3517 (fill_slots_from_thread): Likewise.
3518 * resource.c (mark_referenced_resources): Likewise.
3519 (mark_set_resources, find_dead_or_set_registers): Likewise.
3521 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3523 * substring-locations.h (format_warning_va): Formatting fix for
3525 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
3527 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
3528 * substring-locations.c: Include intl.h.
3529 (format_warning_va): Turned into small wrapper around
3530 format_warning_n_va, renamed to ...
3531 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
3532 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
3534 (format_warning_at_substring_n): New function.
3535 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
3536 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
3537 format_warning_at_substring with just a shorter name instead of
3538 const function pointer.
3539 (fmtwarn_n): New function.
3540 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
3541 appropriate, get rid of all the fmtstr temporaries, move conditionals
3542 with G_() wrapped string literals directly into fmtwarn arguments,
3543 cast dir.len to (int), formatting fixes.
3545 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
3547 * doc/invoke.texi: Remove "Cilk Plus" references.
3549 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3550 Richard Biener <rguenther@suse.de>
3553 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
3554 for error or warning attributes if CALL_FROM_THUNK_P is set.
3557 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3560 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
3561 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
3564 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
3565 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
3567 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
3568 instead of -1U in last predictors element's probability member.
3570 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3573 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3574 arguments if they are comparable.
3576 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3578 PR tree-optimization/84634
3579 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
3580 masks and masked_loop_p with a single loop_masks, making sure it's
3581 null for bb vectorization.
3583 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3585 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3586 (vect_analyze_data_ref_access): Use loop->safe_len rather than
3587 loop->force_vectorize to check whether there is no alias.
3589 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3592 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
3594 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
3596 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
3597 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
3598 instead of a loop around prev_real_insn.
3599 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
3603 * config/i386/i386.c (ix86_print_operand): Use conditional
3604 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
3607 2018-03-02 Richard Biener <rguenther@suse.de>
3609 PR tree-optimization/84427
3610 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
3611 (bitmap_set_subtract_values): Rewrite to handle multiple
3614 (prune_clobbered_mems): Likewise.
3615 (phi_translate): Take edge instead of pred/phiblock.
3616 (phi_translate_1): Likewise.
3617 (phi_translate_set): Likewise. Insert all translated
3618 exprs for a value into the set, keeping possibly multiple
3619 expressions per value.
3620 (compute_antic_aux): Adjust for phi_translate changes.
3621 When intersecting union the expressions and prune those
3622 not in the final value set, keeping possibly multiple
3623 expressions per value. Do not use value-insertion
3624 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
3625 all expressions. Add verification that the value-sets
3626 only shrink during iteration.
3627 (compute_partial_antic_aux): Adjust for the phi_translate changes.
3628 (do_pre_regular_insertion): Likewise.
3629 (do_pre_partial_partial_insertion): Likewise.
3631 2018-03-02 Richard Biener <rguenther@suse.de>
3634 * config/darwin.c (saved_debug_info_level): New static global.
3635 (darwin_asm_lto_start): Disable debug info generation for LTO out.
3636 (darwin_asm_lto_end): Restore debug info generation settings.
3638 2018-03-01 Martin Liska <mliska@suse.cz>
3641 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
3644 2018-03-01 Richard Biener <rguenther@suse.de>
3647 * dwarf2out.c (gen_variable_die): Properly handle late VLA
3648 type annotation with LTO when debug was disabled at compile-time.
3650 2018-03-01 Matthew Fortune <mfortune@gmail.com>
3652 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
3654 (mips_final_postscan_insn): Likewise.
3656 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
3658 PR rtl-optimization/84528
3659 * alias.c (init_alias_target): Add commentary.
3660 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
3661 a unique base value if the frame pointer is not eliminated
3662 to the stack pointer.
3664 2018-03-01 Tom de Vries <tom@codesourcery.com>
3666 PR rtl-optimization/83327
3667 * lra-int.h (hard_regs_spilled_into): Declare.
3668 * lra.c (hard_regs_spilled_into): Define.
3669 (init_reg_info): Init hard_regs_spilled_into.
3670 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
3671 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
3672 (process_bb_lives): Handle hard_regs_spilled_into.
3673 (lra_create_live_ranges_1): Before doing liveness propagation, clear
3674 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
3676 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
3678 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
3679 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
3680 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
3681 * config/rs6000/aix72.h: New file.
3683 2018-02-28 Jakub Jelinek <jakub@redhat.com>
3685 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
3686 instead of warning_at with conditional singular and plural messages
3690 * stor-layout.c (update_alignment_for_field): For
3691 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
3692 && !DECL_PACKED (field), do the alignment update, just use
3693 only desired_align instead of MAX (type_align, desired_align)
3695 (place_field): Don't do known_align < desired_align handling
3696 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
3697 is non-NULL, instead do it after rli->prev_field handling and
3698 only if not within a bitfield word. For DECL_PACKED (field)
3699 use type_align of BITS_PER_UNIT.
3701 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
3703 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
3704 superfluous parentheses and trailing spaces.
3706 2018-02-28 Richard Biener <rguenther@suse.de>
3708 PR tree-optimization/84584
3709 * graphite-scop-detection.c (scop_detection::add_scop): Discard
3710 SCoPs with fake exit edge.
3712 2018-02-28 Martin Liska <mliska@suse.cz>
3715 * timevar.c (timer::print): Fix format to properly print 100%
3718 2018-02-28 Richard Biener <rguenther@suse.de>
3721 * genmatch.c (capture_info::walk_match): Do not mark
3722 captured expressions without operands as expr_p given
3723 they act more like predicates and should be subject to
3724 "lost tail" side-effect preserving.
3726 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
3728 PR rtl-optimization/81611
3729 * auto-inc-dec.c (attempt_change): Move dead note from
3730 mem_insn if it's the next use of regno
3731 (find_address): Take address use of reg holding
3732 non-incremented value. Add parm to limit search to the named
3734 (merge_in_block): Attempt to use a mem insn that is the next
3735 use of the original regno.
3737 2018-02-27 Martin Sebor <msebor@redhat.com>
3740 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
3741 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
3743 2018-02-27 Martin Sebor <msebor@redhat.com>
3745 PR translation/84207
3746 * diagnostic-core.h (warning_n, error_n, inform_n): Change
3747 n argument to unsigned HOST_WIDE_INT.
3748 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
3749 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
3750 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
3751 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
3753 2018-02-27 Richard Biener <rguenther@suse.de>
3755 PR tree-optimization/84512
3756 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
3757 Do not use the estimate returned from record_stmt_cost for
3758 the scalar iteration cost but sum properly using add_stmt_cost.
3760 2018-02-27 Richard Biener <rguenther@suse.de>
3762 PR tree-optimization/84466
3763 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3764 Adjust last change to less strictly validate use operands.
3766 2018-02-27 Martin Liska <mliska@suse.cz>
3768 PR gcov-profile/84548
3769 * gcov.c (process_file): Allow partial overlap and consider it
3770 also as group functions.
3771 (output_lines): Properly calculate range of lines for a group.
3773 2018-02-27 Martin Liska <mliska@suse.cz>
3775 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
3776 'ggc' suffixes. Change first column width.
3777 (timer::print): Fix formatting of the column.
3779 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
3781 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
3782 preserve inline entry blocks for the sake of debug inline
3783 entry point markers alone.
3784 (remove_unused_locals): Suggest in comments a better place to
3785 force the preservation of inline entry blocks that are
3786 otherwise unused, but do not preserve them.
3788 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3790 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
3792 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3795 * config/i386/constraints.md (Bs): Replace
3796 ix86_indirect_branch_register with
3797 TARGET_INDIRECT_BRANCH_REGISTER.
3799 * config/i386/i386.md (indirect_jump): Likewise.
3800 (tablejump): Likewise.
3801 (*sibcall_memory): Likewise.
3802 (*sibcall_value_memory): Likewise.
3803 Peepholes of indirect call and jump via memory: Likewise.
3804 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
3805 (*sibcall_value_GOT_32): Likewise.
3806 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3807 (GOT_memory_operand): Likewise.
3808 (call_insn_operand): Likewise.
3809 (sibcall_insn_operand): Likewise.
3810 (GOT32_symbol_operand): Likewise.
3811 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
3813 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3815 PR rtl-optimization/83496
3816 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
3817 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
3818 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
3819 redundant insn, if any.
3820 (relax_delay_slots): Likewise.
3821 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
3823 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
3825 PR tree-optimization/83965
3826 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
3827 that grouped statements are part of a reduction chain. Return
3828 true if the statement is not marked as a reduction itself but
3830 (vect_recog_dot_prod_pattern): Don't check whether the statement
3831 is part of a group here.
3832 (vect_recog_sad_pattern): Likewise.
3833 (vect_recog_widen_sum_pattern): Likewise.
3835 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3838 * final.c (rest_of_clean_state): Also look for calls inside sequences.
3840 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3843 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
3845 (ix86_output_indirect_function_return): New prototype.
3846 (ix86_split_simple_return_pop_internal): Likewise.
3847 * config/i386/i386.c (indirect_return_via_cx): New.
3848 (indirect_return_via_cx_bnd): Likewise.
3849 (indirect_thunk_name): Handle return va CX_REG.
3850 (output_indirect_thunk_function): Create alias for
3851 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
3852 (ix86_output_indirect_jmp): Remove the bool argument.
3853 (ix86_output_indirect_function_return): New function.
3854 (ix86_split_simple_return_pop_internal): Likewise.
3855 * config/i386/i386.md (*indirect_jump): Don't pass false
3856 to ix86_output_indirect_jmp.
3857 (*tablejump_1): Likewise.
3858 (simple_return_pop_internal): Change it to define_insn_and_split.
3859 Call ix86_split_simple_return_pop_internal to split it for
3861 (simple_return_indirect_internal): Call
3862 ix86_output_indirect_function_return instead of
3863 ix86_output_indirect_jmp.
3865 2018-02-26 Jakub Jelinek <jakub@redhat.com>
3868 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
3869 memset and value initialization afterwards.
3871 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
3873 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
3875 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3878 * common/config/aarch64/aarch64-common.c
3879 (aarch_option_optimization_table[]): Switch
3880 off fomit-frame-pointer
3882 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3883 Chung-Ju Wu <jasonwucj@gmail.com>
3885 * config/nds32/nds32-multiple.md (load_multiple): Disallow
3887 (store_multiple): Ditto.
3889 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3891 * config.gcc: Add --with-cpu support for nds32 target.
3892 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
3893 * config/nds32/nds32.opt: Add -mcpu= option.
3895 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
3897 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
3898 isel=yes): Warn for these deprecated options.
3900 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
3902 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
3903 ISA_2_5_MASKS_EMBEDDED.
3905 2018-02-23 Jakub Jelinek <jakub@redhat.com>
3907 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
3908 p->max as pointers rather than using iterative_hash_expr.
3910 2018-02-23 Carl Love <cel@us.ibm.com>
3912 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
3913 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
3915 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
3916 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
3917 P8V_BUILTIN_VEC_VUNSIGNED2.
3919 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
3922 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
3923 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
3925 * lra-constraints.c (curr_insn_transform): Set up
3926 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
3928 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
3929 * lra-eliminations.c (spill_pseudos): Ditto.
3930 (process_insn_for_elimination): Ditto.
3931 * lra-lives.c (reg_early_clobber_p): Use the new macros.
3932 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
3933 LRA_NON_CLOBBERED_ALT.
3935 2018-02-22 Martin Sebor <msebor@redhat.com>
3937 PR tree-optimization/84480
3938 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
3939 to maybe_diag_stxncpy_trunc. Call it.
3940 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
3941 from gimple_fold_builtin_strcpy. Print inlining stack.
3942 (handle_builtin_stxncpy): Print inlining stack.
3943 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
3945 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
3948 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
3949 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
3950 and -fcheck-pointer-bounds are used together.
3951 (indirect_thunk_prefix): New enum.
3952 (indirect_thunk_need_prefix): New function.
3953 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
3954 "_nt" instead of "_bnd" for NOTRACK prefix.
3955 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
3956 (output_indirect_thunk_function): Likewise.
3958 (ix86_code_end): Update output_indirect_thunk_function calls.
3959 (ix86_output_indirect_branch_via_reg): Replace
3960 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
3961 (ix86_output_indirect_branch_via_push): Likewise.
3962 (ix86_output_function_return): Likewise.
3963 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
3964 incompatible with -fcf-protection=branch and
3965 -fcheck-pointer-bounds.
3967 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3970 * config/aarch64/aarch64.c (aarch64_print_address_internal):
3971 Change gcc_assert call to output_operand_lossage.
3973 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3975 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
3977 2018-02-22 DJ Delorie <dj@redhat.com>
3978 Sebastian Perta <sebastian.perta@renesas.com>
3979 Oleg Endo <olegendo@gcc.gnu.org>
3981 * config/rx/rx.c (rx_rtx_costs): New function.
3982 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
3984 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
3986 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
3988 2018-02-22 Martin Liska <mliska@suse.cz>
3991 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
3992 Add "native" as a possible value.
3994 2018-02-22 Martin Liska <mliska@suse.cz>
3997 * config/i386/i386.c (ix86_option_override_internal):
3998 Add "native" as a possible value for -march and -mtune.
4000 2018-02-22 Jakub Jelinek <jakub@redhat.com>
4003 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
4004 to all type variants.
4006 PR tree-optimization/84503
4007 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
4008 width as info->bitpos + info->bitsize - start.
4009 (merged_store_group::merge_overlapping): Simplify width computation.
4010 (check_no_overlap): New function.
4011 (imm_store_chain_info::try_coalesce_bswap): Compute expected
4012 start + width and last_order of the group, fail if check_no_overlap
4014 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
4015 to group if check_no_overlap fails.
4017 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4019 * config/rs6000/altivec.md: Delete contraint arguments to
4020 define_expand, define_split, and define_peephole2, and in
4021 define_insn_and_split if always unused.
4022 * config/rs6000/darwin.md: Ditto.
4023 * config/rs6000/dfp.md: Ditto.
4024 * config/rs6000/rs6000.md: Ditto.
4025 * config/rs6000/sync.md: Ditto.
4026 * config/rs6000/vector.md: Ditto.
4027 * config/rs6000/vsx.md: Ditto.
4029 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4031 * config/rs6000/altivec.md: Write output control strings as braced
4032 blocks instead of double-quoted strings.
4033 * config/rs6000/darwin.md: Ditto.
4034 * config/rs6000/rs6000.md: Ditto.
4035 * config/rs6000/vector.md: Ditto.
4036 * config/rs6000/vsx.md: Ditto.
4038 2018-02-21 Jason Merrill <jason@redhat.com>
4040 PR c++/84314 - ICE with templates and fastcall attribute.
4041 * attribs.c (build_type_attribute_qual_variant): Remove assert.
4043 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4045 * ipa-cp.c (determine_versionability): Fix comment typos.
4047 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4050 * ipa-cp.c (determine_versionability): Do not version functions caling
4053 2018-02-21 Martin Liska <mliska@suse.cz>
4056 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
4057 Add "native" as a possible value.
4058 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
4059 the macro when native cpu detection is available.
4061 2018-02-21 Martin Liska <mliska@suse.cz>
4064 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
4065 Add "native" as a possible value.
4066 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
4067 when native cpu detection is available.
4069 2018-02-21 Jakub Jelinek <jakub@redhat.com>
4070 Martin Sebor <msebor@redhat.com>
4072 PR tree-optimization/84478
4073 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
4075 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
4076 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
4077 support which is conservatively correct, for 2 only stay conservative
4078 for maxlen. Formatting and comment capitalization fixes. Add STRICT
4079 argument to the 2 argument get_range_strlen, adjust 6 arg
4080 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
4082 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
4083 (gimple_fold_builtin_strlen): Pass true as last argument to
4086 2018-02-20 Martin Sebor <msebor@redhat.com>
4089 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
4090 (builtin_memref::set_base_and_offset): Same. Handle inner references.
4091 (builtin_memref::builtin_memref): Factor out parts into
4092 set_base_and_offset and call it.
4094 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4097 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
4098 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
4099 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
4100 search at the associated MODE_INT.
4102 2018-02-20 Jeff Law <law@redhat.com>
4105 PR tree-optimization/81592
4107 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
4108 for range data rather than using global data.
4109 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
4110 range data rather than using global data.
4111 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
4112 pass it to children as needed.
4113 (struct directive::fmtresult): Similarly.
4114 (struct directive::set_width): Similarly.
4115 (struct directive::set_precision): Similarly.
4116 (format_integer, format_directive, parse_directive): Similarly.
4117 (format_none): Accept unnamed vr_values parameter.
4118 (format_percent, format_floating, format_character): Similarly.
4119 (format_string, format_plain): Similarly.
4120 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
4121 the EVRP range analyzer for range data rather than using global data.
4122 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
4123 gimple-ssa-evrp-analyze.h
4124 (class sprintf_dom_walker): Add after_dom_children member function.
4125 Add evrp_range_analyzer member.
4126 (sprintf_dom_walker::before_dom_children): Call into the EVRP
4127 range analyzer as needed.
4128 (sprintf_dom_walker::after_dom_children): New member function.
4129 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
4131 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
4132 (evrp_range_analyzer::pop_to_marker): Likewise.
4134 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4136 PR tree-optimization/84419
4137 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
4138 with the required type if its current type is compatible but
4141 2018-02-20 Jakub Jelinek <jakub@redhat.com>
4144 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
4145 after vectorization.
4147 2018-02-20 Martin Liska <mliska@suse.cz>
4150 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
4151 possible values if we don't have a hint.
4153 2018-02-20 Martin Liska <mliska@suse.cz>
4157 * final.c (shorten_branches): Build align_tab array with one
4159 * opts.c (finish_options): Add alignment option limit check.
4160 (MAX_CODE_ALIGN): Likewise.
4161 (MAX_CODE_ALIGN_VALUE): Likewise.
4162 * doc/invoke.texi: Document maximum allowed option value for
4163 all -falign-* options.
4165 2018-02-19 Jakub Jelinek <jakub@redhat.com>
4168 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
4169 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
4170 * var-tracking.c (emit_note_insn_var_location): Remove all references
4171 to NOTE_INSN_CALL_ARG_LOCATION.
4172 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
4173 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
4174 Use copy_rtx_if_shared.
4175 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
4176 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
4177 (dwarf2out_var_location): Remove handling of
4178 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
4180 * final.c (final_scan_insn): Remove all references to
4181 NOTE_INSN_CALL_ARG_LOCATION.
4182 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
4183 before dumping final insns.
4184 * except.c (emit_note_eh_region_end): Remove all references to
4185 NOTE_INSN_CALL_ARG_LOCATION.
4186 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
4187 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
4188 * config/arc/arc.c (hwloop_optimize): Likewise.
4189 * config/arm/arm.c (create_fix_barrier): Likewise.
4190 * config/s390/s390.c (s390_chunkify_start): Likewise.
4191 * config/sh/sh.c (find_barrier): Likewise.
4192 * config/i386/i386.c (rest_of_insert_endbranch,
4193 ix86_seh_fixup_eh_fallthru): Likewise.
4194 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
4195 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
4196 * config/frv/frv.c (frv_function_prologue): Likewise.
4197 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
4199 (note_outside_basic_block_p): Remove all references to
4200 NOTE_INSN_CALL_ARG_LOCATION.
4201 * gengtype.c (adjust_field_rtx_def): Likewise.
4202 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
4204 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
4205 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
4208 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
4211 PR tree-optimization/84452
4212 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
4213 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
4216 2018-02-19 Martin Liska <mliska@suse.cz>
4219 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
4221 2018-02-19 Martin Liska <mliska@suse.cz>
4222 Richard Sandiford <richard.sandiford@linaro.org>
4224 PR tree-optimization/82491
4225 * gimple-fold.c (get_base_constructor): Make earlier bail out
4228 2018-02-19 Carl Love <cel@us.ibm.com>
4230 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
4231 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
4233 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
4234 P8V_BUILTIN_VEC_NEG.
4236 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
4238 * config/rl78/rl78.md (movdf): New define expand.
4240 2018-02-19 Martin Liska <mliska@suse.cz>
4243 * doc/invoke.texi: Fix typo.
4244 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
4246 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
4248 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
4249 handle rs6000_single_float and rs6000_double_float specially for
4252 2018-02-16 Jeff Law <law@redhat.com>
4254 * config/rx/rx.c (add_pop_cfi_notes): New function.;
4257 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4260 * ipa-inline.c (inline_small_functions): Fix a typo.
4262 2018-02-16 Nathan Sidwell <nathan@acm.org>
4264 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
4266 2018-02-16 Carl Love <cel@us.ibm.com>
4268 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
4269 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
4270 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
4271 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
4272 expansion to P8V_BUILTIN_VEC_FLOAT2.
4274 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
4276 PR rtl-optimization/70023
4277 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
4278 src_regno into account.
4280 2018-02-16 Carl Love <cel@us.ibm.com>
4282 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
4283 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
4284 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
4285 * config/rs6000/rs6000.c: Remove case statements for
4286 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
4287 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
4288 and P9V_BUILTIN_VEC_VINSERT4B.
4289 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
4290 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
4291 * config/rs6000/vsx.md:
4292 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
4295 2018-02-16 Carl Love <cel@us.ibm.com>
4297 * config/rs6000/altivec.h: Add builtin names vec_extract4b
4299 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
4301 * config/rs6000/rs6000-c.c: Add the definitions for
4302 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
4303 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
4304 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
4305 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
4306 definition for insert4b and define insn *insert3b_internal.
4307 * doc/extend.texi: Add documentation for vec_extract4b.
4309 2018-02-16 Nathan Sidwell <nathan@acm.org>
4311 * doc/extend.texi (Backwards Compatibility): Mention friend
4312 injection. Note for-scope is deprecated.
4313 * doc/invoke.texi (-ffriend-injection): Deprecate.
4315 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
4317 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
4318 that moved to I2, also allow destinations that are a paradoxical
4319 subreg (instead of a normal reg).
4321 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
4324 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
4327 2018-02-16 Richard Biener <rguenther@suse.de>
4329 PR tree-optimization/84037
4330 PR tree-optimization/84016
4332 * config/i386/i386.c (ix86_builtin_vectorization_cost):
4333 Adjust vec_construct for the fact we need additional higher latency
4334 128bit inserts for AVX256 and AVX512 vector builds.
4335 (ix86_add_stmt_cost): Scale vector construction cost for
4338 2018-02-16 Richard Biener <rguenther@suse.de>
4340 PR tree-optimization/84417
4341 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
4342 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
4343 (non_rewritable_lvalue_p): Likewise, use poly-ints.
4345 2018-02-16 Martin Liska <mliska@suse.cz>
4348 * internal-fn.def (ASAN_CHECK): Set proper flags.
4349 (ASAN_MARK): Likewise.
4351 2018-02-16 Julia Koval <julia.koval@intel.com>
4353 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
4354 from PTA_CANNONLAKE.
4356 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4359 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
4360 Use ++iter rather than iter++ for std::list iterators.
4361 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
4362 defer deleting them until all nodes in the forest are processed. Do
4363 free even leaf nodes. Change to_process into auto_vec.
4366 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
4367 * vec.h (vec_default_construct): Use memset instead of placement new
4368 if BROKEN_VALUE_INITIALIZATION is defined.
4369 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
4370 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
4373 PR rtl-optimization/83723
4374 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
4375 * lra.c (lra_substitute_pseudo): Likewise. If true, use
4376 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
4378 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
4380 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
4382 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
4384 PR rtl-optimization/81443
4385 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
4386 from inner REGs to paradoxical SUBREGs.
4388 2018-02-16 Richard Biener <rguenther@suse.de>
4390 PR tree-optimization/84399
4391 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4392 For operands we can analyze at their definition make sure we can
4393 analyze them at each use as well.
4395 2018-02-16 Richard Biener <rguenther@suse.de>
4397 PR tree-optimization/84190
4398 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
4399 volatile accesses if the decl isn't volatile.
4401 2018-02-15 Jason Merrill <jason@redhat.com>
4403 PR c++/84314 - ICE with templates and fastcall attribute.
4404 * attribs.c (build_type_attribute_qual_variant): Don't clobber
4405 TYPE_CANONICAL on an existing type.
4407 2018-02-15 Jakub Jelinek <jakub@redhat.com>
4409 PR tree-optimization/84383
4410 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
4411 dstoff nor call operand_equal_p if dstbase is NULL.
4413 PR tree-optimization/84334
4414 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
4415 also a CONSTANT_CLASS_P, punt.
4417 2018-02-14 Jim Wilson <jimw@sifive.com>
4419 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
4420 first SMALL_OPERAND check. New local min_second_step. Move assert
4421 to where locals are set. Add TARGET_RVC support.
4422 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
4424 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
4426 * doc/invoke.texi: Correct -Wformat-overflow code sample.
4428 2018-02-14 Martin Sebor <msebor@redhat.com>
4430 PR tree-optimization/83698
4431 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
4432 arrays constrain the offset range to their bounds.
4433 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
4434 (builtin_access::overlap): Avoid setting the size of overlap if it's
4436 (maybe_diag_overlap): Also consider arrays when deciding what values
4437 of offsets to include in diagnostics.
4439 2018-02-14 Martin Sebor <msebor@redhat.com>
4442 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
4443 that correspond to the kind of a declaration.
4445 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
4448 * config/pa/pa.md: Load address of PIC label using the linkage table
4449 if the label is nonlocal.
4451 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
4453 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
4454 warning message if user requests -maltivec=be.
4455 * doc/invoke.texi: Document deprecation of -maltivec=be.
4457 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
4460 * config/rs6000/rs6000-c.c: Update definitions for
4461 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
4462 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
4464 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4467 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
4468 add _get_ssp intrinsics. Remove argument from
4469 __builtin_ia32_rdssp[d|q].
4470 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
4471 * config/i386/i386-builtin.def: Remove argument from
4472 __builtin_ia32_rdssp[d|q].
4473 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
4474 ix86_expand_special_args_builtin for _rdssp[d|q].
4475 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
4476 Clear register before usage.
4477 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
4478 Add documentation for new _get_ssp and _inc_ssp intrinsics.
4480 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
4482 PR tree-optimization/84357
4483 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
4484 operand 1 of an ARRAY_REF too.
4486 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
4489 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
4490 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4492 (set_of_reg): New struct.
4493 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
4494 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
4495 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4497 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
4498 Split into bitclr, bitset, bitinvert patterns if appropriate.
4499 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
4500 use rx_fuse_in_memory_bitop.
4501 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
4502 to named insn, correct maximum insn length.
4504 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
4507 * machmode.def: Define a complex mode for PARTIAL_INT.
4508 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
4510 * doc/rtl.texi: Document CSPImode.
4511 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
4513 (msp430_hard_regno_nregs_with_padding): Likewise.
4515 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
4518 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
4520 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
4522 PR rtl-optimization/84169
4523 * combine.c (try_combine): New variable split_i2i3. Set it to true if
4524 we generated a parallel as new i3 and we split that to new i2 and i3
4525 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
4526 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
4527 those to i2, not i1. Partially rewrite this scan code.
4529 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4532 * stor-layout.c (place_field): For variable length fields, adjust
4533 offset_align afterwards not just based on the field's alignment,
4534 but also on the size.
4537 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
4538 of exps and logs in the use_exp2 case.
4540 2018-02-13 Jeff Law <law@redhat.com>
4542 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
4545 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
4548 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
4552 * common.opt (gas-loc-support, gas-locview-support): New.
4553 (ginline-points, ginternal-reset-location-views): New.
4554 * doc/invoke.texi: Document them. Use @itemx where intended.
4555 (gvariable-location-views): Adjust.
4556 * target.def (reset_location_view): New.
4557 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
4558 (TARGET_RESET_LOCATION_VIEW): New.
4559 * doc/tm.texi: Rebuilt.
4560 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
4561 (dwarf2out_default_as_locview_support): New.
4562 (output_asm_line_debug_info): Use option variables.
4563 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
4564 (output_loc_list): Likewise.
4565 (add_high_low_attributes): Check option variables.
4566 Don't output entry view attribute in strict mode.
4567 (gen_inlined_subroutine_die): Check option variables.
4568 (dwarf2out_inline_entry): Likewise.
4569 (init_sections_and_labels): Likewise.
4570 (dwarf2out_early_finish): Likewise.
4571 (maybe_reset_location_view): New, from...
4572 (dwarf2out_var_location): ... here. Call it.
4573 * debug.h (dwarf2out_default_as_loc_support): Declare.
4574 (dwarf2out_default_as_locview_support): Declare.
4575 * hooks.c (hook_int_rtx_insn_0): New.
4576 * hooks.h (hook_int_rtx_insn_0): Declare.
4577 * toplev.c (process_options): Take -gas-loc-support and
4578 -gas-locview-support from dwarf2out. Enable
4579 -gvariable-location-views by default only with locview
4580 assembler support. Enable -ginternal-reset-location-views by
4581 default only if the target defines the corresponding hook.
4582 Enable -ginline-points by default if location views are
4583 enabled; force it disabled if statement frontiers are
4585 * tree-inline.c (expand_call_inline): Check option variables.
4586 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
4588 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
4590 PR tree-optimization/84321
4591 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
4592 handling. Also check whether the anti-range contains any values
4593 that satisfy the mask; switch to a VR_RANGE if not.
4595 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
4598 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
4600 2018-02-13 Martin Jambor <mjambor@suse.cz>
4603 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
4604 of call statements, also set location of a load to a temporary.
4606 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
4608 * config/rl78/rl78.c (add_vector_labels): New function.
4609 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
4610 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
4611 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
4612 which checks that no arguments are passed.
4613 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
4614 * doc/extend.texi: Documentation for the new attribute.
4616 2018-02-13 Andreas Schwab <schwab@suse.de>
4618 * config/riscv/linux.h (CPP_SPEC): Define.
4620 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4623 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
4624 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
4625 OPTION_MASK_ISA_AES as first argument to def_builtin_const
4626 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
4627 instead of OPTION_MASK_ISA_PCLMUL as first argument to
4628 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
4629 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
4630 temporarily for AES and PCLMUL builtins.
4632 PR tree-optimization/84339
4633 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
4634 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
4638 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
4639 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
4640 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
4642 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
4644 * omp-simd-clone.h: New file.
4645 * omp-simd-clone.c: Include omp-simd-clone.h.
4646 (expand_simd_clones): No longer static.
4647 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
4648 cgraph.h and omp-simd-clone.h.
4649 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
4650 (vect_recog_widen_shift_pattern): Formatting fix.
4651 (vect_pattern_recog_1): Don't check optab for calls.
4654 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
4655 operands[2] into a REG before using gen_lowpart on it.
4657 2018-02-12 Jeff Law <law@redhat.com>
4660 * config/sh/sh.c (find_barrier): Consider a sibling call
4663 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
4664 successfully back substituting a reg.
4666 2018-02-12 Richard Biener <rguenther@suse.de>
4668 PR tree-optimization/84037
4669 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
4670 parameter, move visited init to caller.
4671 (vect_slp_analyze_operations): Separate cost from validity
4672 check, initialize visited once for all instances.
4673 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
4675 * tree-vect-stmts.c (vect_model_simple_cost): Make early
4677 (vect_model_promotion_demotion_cost): Likewise.
4678 (vectorizable_bswap): Guard cost modeling with !slp_node
4679 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
4681 (vectorizable_call): Likewise.
4682 (vectorizable_conversion): Likewise.
4683 (vectorizable_assignment): Likewise.
4684 (vectorizable_shift): Likewise.
4685 (vectorizable_operation): Likewise.
4686 (vectorizable_store): Likewise.
4687 (vectorizable_load): Likewise.
4688 (vectorizable_condition): Likewise.
4689 (vectorizable_comparison): Likewise.
4691 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
4694 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
4695 (ASAN_MARK): Fix fnspec to account for return value, change pointer
4696 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
4698 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4701 * params.def (inline-min-speedup): Increase from 8 to 15.
4702 (max-inline-insns-auto): Decrease from 40 to 30.
4703 * ipa-split.c (consider_split): Add some buffer for function to
4704 be considered inlining candidate.
4705 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
4708 2018-02-12 Richard Biener <rguenther@suse.de>
4710 PR tree-optimization/84037
4711 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
4712 matched stmts if we cannot swap the non-matched ones.
4714 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4716 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
4717 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
4718 _mm_maskz_scalef_round_ss): New intrinsics.
4719 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
4720 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
4721 __builtin_ia32_scalefss_round): Remove.
4722 (__builtin_ia32_scalefsd_mask_round,
4723 __builtin_ia32_scalefss_mask_round): New intrinsics.
4724 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
4725 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
4726 ((match_operand:VF_128 2 "<round_nimm_predicate>"
4727 "<round_constraint>")): Changed to ...
4728 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
4729 "<round_scalar_constraint>")): ... this.
4730 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
4731 %0, %1, %2<round_op3>}"): Changed to ...
4732 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
4733 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
4734 %2<round_scalar_mask_op3>}"): ... this.
4735 * config/i386/subst.md (round_scalar_nimm_predicate): New.
4737 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4739 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
4740 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
4741 (_mm_maskz_sqrt_round_ss): New intrinsics.
4742 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
4743 (__builtin_ia32_sqrtsd_mask_round)
4744 (__builtin_ia32_sqrtss_mask_round): New builtins.
4745 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
4746 (__builtin_ia32_sqrtss_round): Remove.
4747 (__builtin_ia32_sqrtsd_mask_round)
4748 (__builtin_ia32_sqrtss_mask_round): New builtins.
4749 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
4750 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
4751 ((match_operand:VF_128 1 "vector_operand"
4752 "xBm,<round_constraint>")): Changed to ...
4753 ((match_operand:VF_128 1 "vector_operand"
4754 "xBm,<round_scalar_constraint>")): ... this.
4755 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
4756 %0, %2, %<iptr>1<round_op3>}): Changed to ...
4757 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
4758 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
4759 %<iptr>1<round_scalar_mask_op3>}): ... this.
4760 ((set_attr "prefix" "<round_prefix>")): Changed to ...
4761 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
4763 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
4766 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
4767 Cast vec_cmpeq result to correct type.
4768 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
4769 Cast vec_cmpgt result to correct type.
4771 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
4773 * final.c (final_scan_insn_1): Renamed from...
4774 (final_scan_insn): ... this. New wrapper, to recover
4775 seen from the outermost call in recursive ones.
4776 * config/sparc/sparc.c (output_return): Drop seen from call.
4777 (output_sibcall): Likewise.
4778 * config/visium/visium.c (output_branch): Likewise.
4780 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
4782 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
4785 2018-02-10 Alan Modra <amodra@gmail.com>
4788 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
4789 Specify LR as an input.
4791 2018-02-10 Jakub Jelinek <jakub@redhat.com>
4794 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
4795 remove_member_access_dummy_vars): New functions.
4796 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
4797 lower_omp_1, execute_lower_omp): Use them.
4799 PR rtl-optimization/84308
4800 * shrink-wrap.c (spread_components): Release todo vector.
4802 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
4804 PR rtl-optimization/57193
4805 * ira-color.c (struct allocno_color_data): Add member
4806 conflict_allocno_hard_prefs.
4807 (update_conflict_allocno_hard_prefs): New.
4808 (bucket_allocno_compare_func): Add a preference based on
4809 conflict_allocno_hard_prefs.
4810 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
4811 (color_allocnos): Remove a dead code. Initiate
4812 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
4814 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4817 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
4818 constraint from =wa to wa. Avoid a subreg on the output operand,
4819 instead use a pseudo and subreg it in a move.
4820 (p9_xxbrd_<mode>): Changed to ...
4821 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
4822 (p9_xxbrd_v2df): New expander.
4823 (p9_xxbrw_<mode>): Changed to ...
4824 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
4825 (p9_xxbrw_v4sf): New expander.
4827 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4829 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
4831 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
4834 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
4835 multiply in 32-bit mode.
4836 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
4837 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
4840 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4842 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
4843 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
4844 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
4845 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
4847 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
4850 * dwarf2out.c (is_trivial_indirect_ref): New function.
4851 (dwarf2out_late_global_decl): Do not generate a location
4852 attribute for variables that have a non-trivial DECL_VALUE_EXPR
4853 and that are not defined in the current unit.
4855 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4857 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
4858 instead of a libcall for UNORDERED.
4860 2018-02-09 Tamar Christina <tamar.christina@arm.com>
4863 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
4864 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
4866 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4869 * config/s390/s390.c (s390_set_current_function): Invoke
4870 s390_indirect_branch_settings also if fndecl didn't change.
4872 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4874 * config/rs6000/rs6000.md (blockage): Set length to zero.
4876 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4878 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
4880 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4883 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
4884 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
4888 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
4889 PARALLEL incoming that failed vt_get_decl_and_offset check.
4892 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
4893 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
4895 (get_variable_section): For decls in named .bss* sections pass true as
4896 second argument to bss_initializer_p.
4898 2018-02-09 Marek Polacek <polacek@redhat.com>
4899 Jakub Jelinek <jakub@redhat.com>
4902 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
4903 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
4904 Sync some changes from cxx_fold_indirect_ref.
4906 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4908 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
4910 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
4911 (BLOCK_INLINE_ENTRY_LABEL): New.
4912 (dwarf2out_var_location): Disregard inline entry markers.
4913 (inline_entry_data): New struct.
4914 (inline_entry_data_hasher): New hashtable type.
4915 (inline_entry_data_hasher::hash): New.
4916 (inline_entry_data_hasher::equal): New.
4917 (inline_entry_data_table): New variable.
4918 (add_high_low_attributes): Add DW_AT_entry_pc and
4919 DW_AT_GNU_entry_view attributes if a pending entry is found
4920 in inline_entry_data_table. Add old entry_pc attribute only
4921 if debug nonbinding markers are disabled.
4922 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
4923 markers are enabled.
4924 (block_within_block_p, dwarf2out_inline_entry): New.
4925 (dwarf2out_finish): Check that no entries remained in
4926 inline_entry_data_table.
4927 * final.c (reemit_insn_block_notes): Handle inline entry notes.
4928 (final_scan_insn, notice_source_line): Likewise.
4929 (rest_of_clean_state): Skip inline entry markers.
4930 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
4932 * gimple.c (gimple_build_debug_inline_entry): New.
4933 * gimple.h (enum gimple_debug_subcode): Add
4934 GIMPLE_DEBUG_INLINE_ENTRY.
4935 (gimple_build_debug_inline_entry): Declare.
4936 (gimple_debug_inline_entry_p): New.
4937 (gimple_debug_nonbind_marker_p): Adjust.
4938 * insn-notes.def (INLINE_ENTRY): New.
4939 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
4940 inline entry marker notes.
4941 (print_insn): Likewise.
4942 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
4943 (INSN_DEBUG_MARKER_KIND): Likewise.
4944 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
4945 * tree-inline.c (expand_call_inline): Build and insert
4946 debug_inline_entry stmt.
4947 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
4948 inline entry blocks early, if nonbind markers are enabled.
4949 (dump_scope_block): Dump fragment info.
4950 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
4951 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
4952 (gimple_build_debug_inline_entry): New.
4953 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
4954 Enable/disable inline entry points too.
4955 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
4956 (DEBUG_INSN): Describe inline entry markers.
4958 * common.opt (gvariable-location-views): New.
4959 (gvariable-location-views=incompat5): New.
4960 * config.in: Rebuilt.
4961 * configure: Rebuilt.
4962 * configure.ac: Test assembler for view support.
4963 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
4964 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
4965 * dwarf2out.c (var_loc_view): New typedef.
4966 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
4967 (dwarf2out_locviews_in_attribute): New.
4968 (dwarf2out_locviews_in_loclist): New.
4969 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
4970 (enum dw_line_info_opcode): Add LI_adv_address.
4971 (struct dw_line_info_table): Add view.
4972 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
4973 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
4974 (zero_view_p): New variable.
4975 (ZERO_VIEW_P): New macro.
4976 (output_asm_line_debug_info): New.
4977 (struct var_loc_node): Add view.
4978 (add_AT_view_list, AT_loc_list): New.
4979 (add_var_loc_to_decl): Add view param. Test it against last.
4980 (new_loc_list): Add view params. Record them.
4981 (AT_loc_list_ptr): Handle loc and view lists.
4982 (view_list_to_loc_list_val_node): New.
4983 (print_dw_val): Handle dw_val_class_view_list.
4984 (size_of_die): Likewise.
4985 (value_format): Likewise.
4986 (loc_list_has_views): New.
4987 (gen_llsym): Set vl_symbol too.
4988 (maybe_gen_llsym, skip_loc_list_entry): New.
4989 (dwarf2out_maybe_output_loclist_view_pair): New.
4990 (output_loc_list): Output view list or entries too.
4991 (output_view_list_offset): New.
4992 (output_die): Handle dw_val_class_view_list.
4993 (output_dwarf_version): New.
4994 (output_compilation_unit_header): Use it.
4995 (output_skeleton_debug_sections): Likewise.
4996 (output_rnglists, output_line_info): Likewise.
4997 (output_pubnames, output_aranges): Update version comments.
4998 (output_one_line_info_table): Output view numbers in asm comments.
4999 (dw_loc_list): Determine current endview, pass it to new_loc_list.
5000 Call maybe_gen_llsym.
5001 (loc_list_from_tree_1): Adjust.
5002 (add_AT_location_description): Create view list attribute if
5003 needed, check it's absent otherwise.
5004 (convert_cfa_to_fb_loc_list): Adjust.
5005 (maybe_emit_file): Call output_asm_line_debug_info for test.
5006 (dwarf2out_var_location): Reset views as needed. Precompute
5007 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
5008 attribute. Set view.
5009 (new_line_info_table): Reset next view.
5010 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
5011 (dwarf2out_source_line): Likewise. Output view resets and labels to
5012 the assembler, or select appropriate line info opcodes.
5013 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
5014 (optimize_string_length): Catch it. Adjust.
5015 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
5016 dw_val_class_view_list, and remove it if no longer needed.
5017 (hash_loc_list): Hash view numbers.
5018 (loc_list_hasher::equal): Compare them.
5019 (optimize_location_lists): Check whether a view list symbol is
5020 needed, and whether the locview attribute is present, and
5021 whether they match. Remove the locview attribute if no longer
5023 (index_location_lists): Call skip_loc_list_entry for test.
5024 (dwarf2out_finish): Call output_asm_line_debug_info for test.
5025 Use output_dwarf_version.
5026 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
5027 (struct dw_val_node): Add val_view_list.
5028 * final.c (SEEN_NEXT_VIEW): New.
5029 (set_next_view_needed): New.
5030 (clear_next_view_needed): New.
5031 (maybe_output_next_view): New.
5032 (final_start_function): Rename to...
5033 (final_start_function_1): ... this. Take pointer to FIRST,
5034 add SEEN parameter. Emit param bindings in the initial view.
5035 (final_start_function): Reintroduce SEEN-less interface.
5036 (final): Rename to...
5037 (final_1): ... this. Take SEEN parameter. Output final pending
5038 next view at the end.
5039 (final): Reintroduce seen-less interface.
5040 (final_scan_insn): Output pending next view before switching
5041 sections or ending a block. Mark the next view as needed when
5042 outputting variable locations. Notify debug backend of section
5043 changes, and of location view changes.
5044 (rest_of_handle_final): Adjust.
5045 * toplev.c (process_options): Autodetect value for debug variable
5046 location views option. Warn on incompat5 without -gdwarf-5.
5047 * doc/invoke.texi (gvariable-location-views): New.
5048 (gvariable-location-views=incompat5): New.
5049 (gno-variable-location-views): New.
5051 2018-02-08 David Malcolm <dmalcolm@redhat.com>
5053 PR tree-optimization/84136
5054 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
5055 that the result of find_edge is non-NULL.
5057 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5060 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
5061 storing integer register in SImode. Fix cost of 256 and 512
5062 byte aligned SSE register store.
5064 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5066 * config/i386/i386.c (ix86_multiplication_cost): Fix
5067 multiplication cost for TARGET_AVX512DQ.
5069 2018-02-08 Marek Polacek <polacek@redhat.com>
5071 PR tree-optimization/84238
5072 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
5075 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5077 PR tree-optimization/84265
5078 * tree-vect-stmts.c (vectorizable_store): Don't treat
5079 VMAT_CONTIGUOUS accesses as grouped.
5080 (vectorizable_load): Likewise.
5082 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5084 PR tree-optimization/81635
5085 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
5086 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
5087 (test_round_for_mask): New functions.
5088 (wide_int_cc_tests): Call test_round_for_mask.
5089 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
5090 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
5091 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
5092 range returned by get_range_info.
5094 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5097 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
5098 * symtab.c: Include builtins.h
5099 (symtab_node::output_to_lto_symbol_table_p): Move here
5100 from lto-streamer-out.c:output_symbol_p.
5101 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
5102 (output_symbol_p): Move all logic to symtab.c
5103 (produce_symtab): Update.
5105 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5107 * config/s390/s390-opts.h (enum indirect_branch): Define.
5108 * config/s390/s390-protos.h (s390_return_addr_from_memory)
5109 (s390_indirect_branch_via_thunk)
5110 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
5111 (enum s390_indirect_branch_type): Define.
5112 * config/s390/s390.c (struct s390_frame_layout, struct
5113 machine_function): Remove.
5114 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
5115 (indirect_branch_table_label_no, indirect_branch_table_name):
5117 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
5118 (enum s390_indirect_branch_option): Define.
5119 (s390_return_addr_from_memory): New function.
5120 (s390_handle_string_attribute): New function.
5121 (s390_attribute_table): Add new attribute handler.
5122 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
5123 (s390_indirect_branch_via_thunk): New function.
5124 (s390_indirect_branch_via_inline_thunk): New function.
5125 (s390_function_ok_for_sibcall): When jumping via thunk disallow
5126 sibling call optimization for non z10 compiles.
5127 (s390_emit_call): Force indirect branch target to be a single
5128 register. Add r1 clobber for non-z10 compiles.
5129 (s390_emit_epilogue): Emit return jump via return_use expander.
5130 (s390_reorg): Handle JUMP_INSNs as execute targets.
5131 (s390_option_override_internal): Perform validity checks for the
5132 new command line options.
5133 (s390_indirect_branch_attrvalue): New function.
5134 (s390_indirect_branch_settings): New function.
5135 (s390_set_current_function): Invoke s390_indirect_branch_settings.
5136 (s390_output_indirect_thunk_function): New function.
5137 (s390_code_end): Implement target hook.
5138 (s390_case_values_threshold): Implement target hook.
5139 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
5141 * config/s390/s390.h (struct s390_frame_layout)
5142 (struct machine_function): Move here from s390.c.
5143 (TARGET_INDIRECT_BRANCH_NOBP_RET)
5144 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
5145 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
5146 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
5147 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
5148 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
5149 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
5150 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
5151 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
5152 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
5153 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
5154 (mnemonic attribute): Add values which aren't recognized
5156 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
5157 pattern for branch conversion. Fix mnemonic attribute.
5158 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
5159 indirect branch via thunk if requested.
5160 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
5161 ("*indirect_jump"): Disable for branch conversion using out of
5163 ("indirect_jump_via_thunk<mode>_z10")
5164 ("indirect_jump_via_thunk<mode>")
5165 ("indirect_jump_via_inlinethunk<mode>_z10")
5166 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
5167 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
5168 ("casesi_jump_via_inlinethunk<mode>_z10")
5169 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
5170 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
5171 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
5172 ("*indirect2_jump"): Disable for branch conversion.
5173 ("casesi_jump"): Turn into expander and expand patterns for branch
5175 ("return_use"): New expander.
5176 ("*return"): Emit return via thunk and rename it to ...
5177 ("*return<mode>"): ... this one.
5178 * config/s390/s390.opt: Add new options and and enum for the
5181 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5183 * lra-constraints.c (match_reload): Unconditionally use
5184 gen_lowpart_SUBREG, rather than selecting between that
5185 and equivalent gen_rtx_SUBREG code.
5187 2018-02-08 Richard Biener <rguenther@suse.de>
5189 PR tree-optimization/84233
5190 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
5191 changed flag instead of boguously re-using phi_inserted.
5193 2018-02-08 Martin Jambor <mjambor@suse.cz>
5195 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
5196 static local variables.
5198 2018-02-08 Richard Biener <rguenther@suse.de>
5200 PR tree-optimization/84278
5201 * tree-vect-stmts.c (vectorizable_store): When looking for
5202 smaller vector types to perform grouped strided loads/stores
5203 make sure the mode is supported by the target.
5204 (vectorizable_load): Likewise.
5206 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5208 * config/aarch64/aarch64.c (aarch64_components_for_bb):
5209 Increase LDP/STP opportunities by adding adjacent callee-saves.
5211 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5213 PR rtl-optimization/84068
5214 PR rtl-optimization/83459
5215 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
5217 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
5219 PR tree-optimization/84224
5220 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
5221 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
5224 2018-02-07 Iain Sandoe <iain@codesourcery.com>
5227 * config/rs6000/altivec.md (*restore_world): Remove LR use.
5228 * config/rs6000/predicates.md (restore_world_operation): Adjust op
5229 count, remove one USE.
5231 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
5233 * doc/install.texi (Configuration): Document the
5234 --with-long-double-format={ibm,ieee} PowerPC configuration
5238 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
5239 Convert from define_expand to be define_insn_and_split. Rework
5240 float/double/_Float128 conversions to QI/HI/SImode to work with
5241 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
5242 conversions to QI/HImode types did a store and then a load to
5243 truncate the value. For conversions to VSX registers, don't split
5244 the insn, instead emit the code directly. Use the code iterator
5245 any_fix to combine signed and unsigned conversions.
5246 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
5247 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
5248 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5249 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
5250 (fix_<mode>di2_hw): Likewise.
5251 (fixuns_<mode>di2_hw): Likewise.
5252 (fix_<mode>si2_hw): Likewise.
5253 (fixuns_<mode>si2_hw): Likewise.
5254 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
5255 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5256 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
5257 fix<uns>_trunc<SFDF:mode>si2_p8.
5258 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
5260 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
5261 (fix<uns>_<mode>_mem): Likewise.
5262 (fctiw<u>z_<mode>_mem): Likewise.
5263 (fix<uns>_<mode>_mem): Likewise.
5264 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
5265 the register allocator from doing a direct move to the GPRs to do
5266 a store, and instead use the ISA 3.0 store byte/half-word from
5267 vector register instruction. For IEEE 128-bit floating point,
5268 also optimize stores of 32-bit ints.
5269 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
5271 2018-02-07 Alan Hayward <alan.hayward@arm.com>
5273 * genextract.c (push_pathstr_operand): New function to support
5275 (walk_rtx): Call push_pathstr_operand.
5276 (print_path): Support [a-zA-Z].
5278 2018-02-07 Richard Biener <rguenther@suse.de>
5280 PR tree-optimization/84037
5281 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
5282 (cse_and_gimplify_to_preheader): Declare.
5283 (vect_get_place_in_interleaving_chain): Likewise.
5284 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5286 (_loop_vec_info::~_loop_vec_info): Delete it.
5287 (cse_and_gimplify_to_preheader): New function.
5288 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
5289 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
5290 (vectorizable_load): Likewise. For grouped stores always base
5291 the IV on the first element.
5292 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
5293 condition before gimplifying.
5295 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5297 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
5298 *DIV_EXPR and *MOD_EXPR.
5300 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
5303 * config/i386/i386.c (ix86_option_override_internal): Mask out
5304 the CF_SET bit when checking -fcf-protection.
5306 2018-02-07 Tom de Vries <tom@codesourcery.com>
5309 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
5312 2018-02-07 Richard Biener <rguenther@suse.de>
5314 PR tree-optimization/84204
5315 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
5318 PR tree-optimization/84205
5319 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
5320 special-case isl_ast_op_zdiv_r.
5322 PR tree-optimization/84223
5323 * graphite-scop-detection.c (gather_bbs::before_dom_children):
5324 Only add conditions from within the region.
5325 (gather_bbs::after_dom_children): Adjust.
5327 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
5330 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
5331 * config/avr/avr.md: Only post-reload split REG-REG moves if
5332 either register is GENERAL_REG_P.
5334 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5336 PR tree-optimization/84235
5337 * tree-ssa-scopedtables.c
5338 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
5339 if the subtraction is performed in floating point type where NaNs are
5340 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
5341 build 1. Formatting fix.
5343 2018-02-06 Jakub Jelinek <jakub@redhat.com>
5346 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
5347 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
5348 and skip it regardless of bb boundaries. Use CALL_P macro,
5349 don't test INSN_P (insn) together with CALL_P or JUMP_P check
5350 unnecessarily, formatting fix.
5352 2018-02-06 Michael Collison <michael.collison@arm.com>
5354 * config/arm/thumb2.md:
5355 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
5356 (*thumb_mov_notscc): Ditto.
5358 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
5361 * config/rs6000/rs6000.md (su code attribute): Use "u" for
5362 unsigned_fix, not "s".
5364 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5366 * configure.ac (gcc_fn_eh_frame_ro): New function.
5367 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
5368 correct .eh_frame permissions.
5369 * configure: Regenerate.
5371 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
5373 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
5376 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5378 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5379 Display warning message for -mno-speculate-indirect-jumps.
5381 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
5383 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
5385 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
5387 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
5389 PR tree-optimization/84225
5390 * tree-eh.c (find_trapping_overflow): Only call
5391 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
5393 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5396 * config/i386/i386.c: Reimplement the check of possible options
5397 -mibt/-mshstk conbination. Change error messages.
5398 * doc/invoke.texi: Fix a typo: remove extra '='.
5400 2018-02-06 Marek Polacek <polacek@redhat.com>
5402 PR tree-optimization/84228
5403 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
5405 2018-02-06 Tamar Christina <tamar.christina@arm.com>
5408 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
5409 emitted arch directives.
5410 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
5411 __ARM_FEATURE_COPROC before changing architectures.
5413 2018-02-06 Richard Biener <rguenther@suse.de>
5415 * config/i386/i386.c (print_reg): Fix typo.
5416 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
5418 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
5420 * configure: Regenerate.
5422 2018-02-05 Martin Sebor <msebor@redhat.com>
5424 PR tree-optimization/83369
5425 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
5428 2018-02-05 Martin Liska <mliska@suse.cz>
5430 * doc/invoke.texi: Cherry-pick upstream r323995.
5432 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
5434 * ira.c (ira_init_register_move_cost): Adjust comment.
5436 2018-02-05 Martin Liska <mliska@suse.cz>
5438 PR gcov-profile/84137
5439 * doc/gcov.texi: Fix typo in documentation.
5441 2018-02-05 Martin Liska <mliska@suse.cz>
5443 PR gcov-profile/83879
5444 * doc/gcov.texi: Document necessity of --dynamic-list-data when
5445 using dlopen functionality.
5447 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
5449 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
5450 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
5451 _mm_maskz_range_ss, _mm_mask_range_round_ss,
5452 _mm_maskz_range_round_ss): New intrinsics.
5453 (__builtin_ia32_rangesd128_round)
5454 (__builtin_ia32_rangess128_round): Remove.
5455 (__builtin_ia32_rangesd128_mask_round,
5456 __builtin_ia32_rangess128_mask_round): New builtins.
5457 * config/i386/i386-builtin.def (__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/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
5462 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
5463 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
5464 "<round_saeonly_constraint>")): Changed to ...
5465 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
5466 "<round_saeonly_scalar_constraint>")): ... this.
5467 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
5468 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
5469 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
5470 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
5471 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
5473 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
5475 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
5477 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
5478 Remove all values except native, 8540 and 8548.
5480 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
5482 * config/i386/i386.c (ix86_output_function_return): Pass
5483 INVALID_REGNUM, instead of -1, as invalid register number to
5484 indirect_thunk_name and output_indirect_thunk.
5486 2018-02-02 Julia Koval <julia.koval@intel.com>
5488 * config.gcc: Add -march=icelake.
5489 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
5490 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
5491 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
5492 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
5493 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
5494 (processor_target_table): Add icelake.
5495 (ix86_option_override_internal): Handle new PTAs.
5496 (get_builtin_code_for_version): Handle icelake.
5497 (M_INTEL_COREI7_ICELAKE): New.
5498 (fold_builtin_cpu): Handle icelake.
5499 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
5500 * doc/invoke.texi: Add -march=icelake.
5502 2018-02-02 Julia Koval <julia.koval@intel.com>
5504 * config/i386/i386.c (ix86_option_override_internal): Change flags type
5505 to wide_int_bitmask.
5506 * wide-int-bitmask.h: New.
5508 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5511 * config/i386/i386.md: Replace Pmode with word_mode in
5512 builtin_setjmp_setup and builtin_longjmp to support x32.
5514 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
5518 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
5520 (rs6000_supported_cpu_names): New static variable.
5521 (linux_cpu_translation_table): Likewise.
5522 (elf_platform) <cpu>: Define new static variable and use it.
5523 Translate kernel AT_PLATFORM name to canonical name if needed.
5524 Error if platform name is unknown.
5526 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
5529 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
5531 2018-02-01 Jeff Law <law@redhat.com>
5534 * config/i386/i386.c (release_scratch_register_on_entry): Add new
5535 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
5536 the scratch if RELEASE_VIA_POP is false.
5537 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
5538 If we have to save a temporary register, decrement SIZE appropriately.
5539 Pass new arguments to release_scratch_register_on_entry.
5540 (ix86_adjust_stack_and_probe): Likewise.
5541 (ix86_emit_probe_stack_range): Pass new arguments to
5542 release_scratch_register_on_entry.
5544 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
5546 PR rtl-optimization/84157
5547 * combine.c (change_zero_ext): Use REG_P predicate in
5548 front of HARD_REGISTER_P predicate.
5550 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
5552 * config/avr/avr.c (avr_option_override): Move disabling of
5553 -fdelete-null-pointer-checks to...
5554 * common/config/avr/avr-common.c (avr_option_optimization_table):
5557 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5559 PR tree-optimization/81635
5560 * tree-data-ref.c (split_constant_offset_1): For types that
5561 wrap on overflow, try to use range info to prove that wrapping
5564 2018-02-01 Renlin Li <renlin.li@arm.com>
5567 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
5569 (aarch64_register_move_cost): Likewise.
5570 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
5572 (REG_CLASS_NAMES): Likewise.
5573 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
5574 TAILCALL_ADDR_REGS. Remove IP registers.
5575 * config/aarch64/aarch64.md (Ucs): Update register constraint.
5577 2018-02-01 Richard Biener <rguenther@suse.de>
5579 * domwalk.h (dom_walker::dom_walker): Add additional constructor
5580 for specifying RPO order and allow NULL for that.
5581 * domwalk.c (dom_walker::dom_walker): Likewise.
5582 (dom_walker::walk): Handle NULL RPO order.
5583 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
5585 (rewrite_update_dom_walker): Likewise.
5586 (mark_def_dom_walker): Likewise.
5588 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5590 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
5591 (aarch64_maybe_expand_sve_subreg_move): Declare.
5592 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
5593 * config/aarch64/predicates.md (aarch64_any_register_operand): New
5595 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
5596 that are semantically a reverse operation.
5597 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
5598 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
5599 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
5601 (aarch64_can_change_mode_class): For big-endian, forbid changes
5602 between two SVE modes if they have different element sizes.
5604 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5606 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
5607 the TImode handling for big-endian targets.
5609 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5611 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
5612 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
5614 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
5615 Remove BSWAP handing for big-endian targets and use the form of
5616 LD1RQ appropariate for the mode.
5618 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5620 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
5621 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
5624 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5627 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
5628 check for operands that need to go through aarch64_sve_reload_be.
5630 2018-02-01 Jakub Jelinek <jakub@redhat.com>
5632 PR tree-optimization/81661
5633 PR tree-optimization/84117
5634 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
5635 * tree-eh.c: Include gimplify.h.
5636 (find_trapping_overflow, replace_trapping_overflow,
5637 rewrite_to_non_trapping_overflow): New functions.
5638 * tree-vect-loop.c: Include tree-eh.h.
5639 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
5640 * tree-data-ref.c: Include tree-eh.h.
5641 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
5643 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
5645 PR rtl-optimization/84123
5646 * combine.c (change_zero_ext): Check if hard register satisfies
5647 can_change_dest_mode before calling gen_lowpart_SUBREG.
5649 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
5652 * ira.c (ira_init_register_move_cost): Remove assert.
5654 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5656 PR rtl-optimization/84071
5657 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
5658 * doc/tm.texi: Regenerate.
5660 2018-01-31 Richard Biener <rguenther@suse.de>
5662 PR tree-optimization/84132
5663 * tree-data-ref.c (analyze_miv_subscript): Properly
5664 check whether evolution_function_is_affine_multivariate_p
5665 before calling gcd_of_steps_may_divide_p.
5667 2018-01-31 Julia Koval <julia.koval@intel.com>
5670 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
5671 * config/i386/i386.md (rdpid_rex64) New.
5672 (rdpid): Make 32bit only.
5674 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
5677 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
5678 an IDENTIFIER_NODE for FUNCTION_TYPE's.
5680 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5683 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5685 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5687 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5689 PR rtl-optimization/84071
5690 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
5691 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
5693 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5695 * config/arc/arc.c (arc_handle_aux_attribute): New function.
5696 (arc_attribute_table): Add 'aux' attribute.
5697 (arc_in_small_data_p): Consider aux like variables.
5698 (arc_is_aux_reg_p): New function.
5699 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
5700 (arc_get_aux_arg): New function.
5701 (prepare_move_operands): Handle aux-register access.
5702 (arc_handle_aux_attribute): New function.
5703 * doc/extend.texi (ARC Variable attributes): Add subsection.
5705 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5707 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
5708 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
5709 (arc_attribute_table): Add 'uncached' attribute.
5710 (arc_print_operand): Print '.di' flag for uncached memory
5712 (arc_in_small_data_p): Do not consider for small data the uncached
5714 (arc_is_uncached_mem_p): New function.
5715 * config/arc/predicates.md (compact_store_memory_operand): Check
5716 for uncached memory accesses.
5717 (nonvol_nonimm_operand): Likewise.
5718 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
5720 2018-01-31 Jakub Jelinek <jakub@redhat.com>
5723 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
5724 falign-loops=): Add Optimization flag.
5726 2018-01-30 Jeff Law <law@redhat.com>
5729 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
5730 INT_REGISTERS_SAVED. Check it prior to calling
5731 get_scratch_register_on_entry.
5732 (ix86_adjust_stack_and_probe): Similarly.
5733 (ix86_emit_probe_stack_range): Similarly.
5734 (ix86_expand_prologue): Corresponding changes.
5736 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5739 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
5740 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
5742 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
5745 * lra-constraints.c (curr_insn_transform): Process AND in the
5748 2018-01-30 Jakub Jelinek <jakub@redhat.com>
5750 PR rtl-optimization/83986
5751 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
5752 dependence against last_pending_memory_flush in addition to
5755 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
5757 PR tree-optimization/81611
5758 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
5761 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5764 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
5767 2018-01-30 Richard Biener <rguenther@suse.de>
5768 Jakub Jelinek <jakub@redhat.com>
5770 PR tree-optimization/84111
5771 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
5772 inner loops added during recursion, as they don't have up-to-date
5775 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5778 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
5779 (can_inline_edge_by_limits_p): ... here.
5780 (can_early_inline_edge_p, check_callers,
5781 update_caller_keys, update_callee_keys, recursive_inlining,
5782 add_new_edges_to_heap, speculation_useful_p,
5783 inline_small_functions,
5784 inline_small_functions, flatten_function,
5785 inline_to_all_callers_1): Update.
5787 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5789 * profile-count.c (profile_count::combine_with_ipa_count): Handle
5792 2018-01-30 Richard Biener <rguenther@suse.de>
5794 PR tree-optimization/83008
5795 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
5796 invariant and constant vector uses in stmts when they need
5799 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5802 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
5803 * configure: Regenerate.
5805 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5807 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
5809 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
5810 Use gen_rtx_REG rather than gen_lowpart.
5812 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5814 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
5815 rather than 0 when creating partial subregs.
5817 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5819 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
5822 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
5825 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
5826 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
5827 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
5828 flags. This restores the settings used before the 2017-07-24.
5829 Turning off pre increment/decrement/modify allows IVOPTS to
5830 optimize DF/SF loops where the index is an int.
5832 2018-01-29 Richard Biener <rguenther@suse.de>
5833 Kelvin Nilsen <kelvin@gcc.gnu.org>
5836 * tree-vect-stmts.c (vectorizable_call): Don't call
5837 targetm.vectorize_builtin_md_vectorized_function if callee is
5840 2018-01-22 Carl Love <cel@us.ibm.com>
5842 * doc/extend.tex: Fix typo in second arg in
5843 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
5845 2018-01-29 Richard Biener <rguenther@suse.de>
5847 PR tree-optimization/84086
5848 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
5849 (flush_ssaname_freelist): When SSA names were released reset
5850 the SCEV hash table.
5852 2018-01-29 Richard Biener <rguenther@suse.de>
5854 PR tree-optimization/84057
5855 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
5856 removed paths when removing edges.
5858 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
5860 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
5861 -mfunction-return=@var{choice}.
5863 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
5866 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
5868 (layout::print_source_line): Likewise.
5869 (test_get_line_width_without_trailing_whitespace): Add test cases.
5871 2018-01-27 Jakub Jelinek <jakub@redhat.com>
5874 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
5877 2018-01-26 Jim Wilson <jimw@sifive.com>
5879 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
5881 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
5884 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5886 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
5887 and CMP + SUB-immediate -> SUBS.
5889 2018-01-26 Martin Sebor <msebor@redhat.com>
5891 PR tree-optimization/83896
5892 * tree-ssa-strlen.c (get_string_len): Rename...
5893 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
5894 Avoid assuming length is constant.
5895 (handle_char_store): Use HOST_WIDE_INT for string length.
5897 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
5900 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
5901 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
5903 2018-01-26 Richard Biener <rguenther@suse.de>
5905 PR rtl-optimization/84003
5906 * dse.c (record_store): Only record redundant stores when
5907 the earlier store aliases at least all accesses the later one does.
5909 2018-01-26 Jakub Jelinek <jakub@redhat.com>
5911 PR rtl-optimization/83985
5912 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
5913 REG_CFA_RESTORE insns.
5914 (delete_unmarked_insns): Don't ignore separate shrink wrapping
5915 REG_CFA_RESTORE insns here.
5918 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
5919 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
5921 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5923 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
5924 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
5925 (arc_init): Likewise.
5926 (arc_override_options): Likewise.
5927 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
5929 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
5931 * config/arc/arc.h (TARGET_DBNZ): Define.
5932 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
5933 properly set the tune attribute.
5934 (dbnz): Use TARGET_DBNZ guard.
5935 * config/arc/arc.opt (mtune): Add core3 option.
5937 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5939 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
5940 recognize new pic like addresses.
5941 (arc_delegitimize_address): Clean up.
5943 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5945 * config/arc/arc-arches.def: Option mrf16 valid for all
5947 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
5948 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
5949 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
5950 * config/arc/arc-tables.opt: Regenerate.
5951 * config/arc/arc.c (arc_conditional_register_usage): Handle
5952 reduced register file case.
5953 (arc_file_start): Set must have build attributes.
5954 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
5956 * config/arc/arc.opt (mrf16): Add new option.
5957 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
5958 * config/arc/genmultilib.awk: Handle new mrf16 option.
5959 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
5960 * config/arc/t-multilib: Regenerate.
5961 * doc/invoke.texi (ARC Options): Document mrf16 option.
5963 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5965 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
5966 * config/arc/arc.c (arc_handle_secure_attribute): New function.
5967 (arc_attribute_table): Add 'secure_call' attribute.
5968 (arc_print_operand): Print secure call operand.
5969 (arc_function_ok_for_sibcall): Don't optimize tail calls when
5971 (arc_is_secure_call_p): New function. * config/arc/arc.md
5972 (call_i): Add support for sjli instruction.
5973 (call_value_i): Likewise.
5974 * config/arc/constraints.md (Csc): New constraint.
5976 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5977 John Eric Martin <John.Martin@emmicro-us.com>
5979 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
5980 * config/arc/arc.c (_arc_jli_section): New struct.
5981 (arc_jli_section): New type.
5982 (rc_jli_sections): New static variable.
5983 (arc_handle_jli_attribute): New function.
5984 (arc_attribute_table): Add jli_always and jli_fixed attribute.
5985 (arc_file_end): New function.
5986 (TARGET_ASM_FILE_END): Define.
5987 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
5988 (arc_add_jli_section): New function.
5989 (jli_call_scan): Likewise.
5990 (arc_reorg): Call jli_call_scan.
5991 (arc_output_addsi): Remove 'S' from printing asm operand.
5992 (arc_is_jli_call_p): New function.
5993 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
5995 (movhi_insn): Likewise.
5996 (movsi_insn): Likewise.
5997 (movsi_set_cc_insn): Likewise.
5998 (loadqi_update): Likewise.
5999 (load_zeroextendqisi_update): Likewise.
6000 (load_signextendqisi_update): Likewise.
6001 (loadhi_update): Likewise.
6002 (load_zeroextendhisi_update): Likewise.
6003 (load_signextendhisi_update): Likewise.
6004 (loadsi_update): Likewise.
6005 (loadsf_update): Likewise.
6006 (movsicc_insn): Likewise.
6007 (bset_insn): Likewise.
6008 (bxor_insn): Likewise.
6009 (bclr_insn): Likewise.
6010 (bmsk_insn): Likewise.
6011 (bicsi3_insn): Likewise.
6012 (cmpsi_cc_c_insn): Likewise.
6013 (movsi_ne): Likewise.
6014 (movsi_cond_exec): Likewise.
6015 (clrsbsi2): Likewise.
6026 (call_i): Remove 'S' asm letter, add jli instruction.
6027 (call_value_i): Likewise.
6028 * config/arc/arc.op (mjli-always): New option.
6029 * config/arc/constraints.md (Cji): New constraint.
6030 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
6032 (subsf3_fpx): Likewise.
6033 (mulsf3_fpx): Likewise.
6034 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
6036 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
6038 * doc/invoke.texi (ARC): Document mjli-always option.
6040 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
6042 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
6043 avoid addition with 0 and use incw and decw where possible.
6045 2018-01-26 Richard Biener <rguenther@suse.de>
6047 PR tree-optimization/81082
6048 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
6049 association if it requires casting to unsigned.
6050 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
6051 from fold_plusminus_mult_expr to catch important cases late when
6052 range info is available.
6054 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6056 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
6057 * configure.ac (hidden_linkonce): New test.
6058 * configure: Regenerate.
6059 * config.in: Regenerate.
6061 2018-01-26 Julia Koval <julia.koval@intel.com>
6063 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
6064 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
6065 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
6066 _mm_mask_bitshuffle_epi64_mask): Fix type.
6067 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
6068 USI_FTYPE_V4DI_V4DI_USI): Remove.
6069 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
6070 __builtin_ia32_vpshufbitqmb256_mask,
6071 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
6072 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
6073 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
6075 2018-01-26 Alan Modra <amodra@gmail.com>
6078 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
6079 UNSPEC_VBPERMQ. Sort other unspecs.
6081 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
6083 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
6085 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
6088 * predict.c (drop_profile): Do not push/pop cfun; update also
6090 (handle_missing_profiles): Fix logic looking for zero profiles.
6092 2018-01-25 Jakub Jelinek <jakub@redhat.com>
6095 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
6096 on functions with #pragma omp declare simd or functions with simd
6098 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
6099 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
6100 Remove trailing \n from warning_at calls.
6102 2018-01-25 Tom de Vries <tom@codesourcery.com>
6105 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6106 for neutered workers.
6108 2018-01-24 Joseph Myers <joseph@codesourcery.com>
6111 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
6112 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
6114 2018-01-24 Jeff Law <law@redhat.com>
6117 * i386.c (get_probe_interval): Move to earlier point.
6118 (ix86_compute_frame_layout): If -fstack-clash-protection and
6119 the frame is larger than the probe interval, then use pushes
6120 to save registers rather than reg->mem moves.
6121 (ix86_expand_prologue): Remove conditional for int_registers_saved
6124 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
6127 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
6128 min/max for never referenced object.
6130 2018-01-24 Jakub Jelinek <jakub@redhat.com>
6133 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
6135 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
6136 attributes from DECL_ATTRIBUTES (decl) without affecting
6137 DECL_ATTRIBUTES (current_function_decl).
6138 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
6139 functions with non-NULL DECL_ABSTRACT_ORIGIN.
6141 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
6143 PR tree-optimization/83979
6144 * fold-const.c (fold_comparison): Use constant_boolean_node
6145 instead of boolean_{true,false}_node.
6147 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
6149 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
6152 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6154 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6155 Simplify the clause that sets the length attribute.
6156 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6157 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
6158 clause that sets the length attribute.
6159 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6161 2018-01-24 Tom de Vries <tom@codesourcery.com>
6164 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
6165 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
6166 Add strict parameter.
6167 (prevent_branch_around_nothing): Insert dummy insn between branch to
6168 label and label with no ptx insn inbetween.
6169 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
6171 2018-01-24 Tom de Vries <tom@codesourcery.com>
6174 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6175 for neutered threads in warp.
6176 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
6178 2018-01-24 Richard Biener <rguenther@suse.de>
6180 PR tree-optimization/83176
6181 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
6184 2018-01-24 Richard Biener <rguenther@suse.de>
6186 PR tree-optimization/82819
6187 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
6188 code generating pluses that are no-ops in the target precision.
6190 2018-01-24 Richard Biener <rguenther@suse.de>
6193 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
6195 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6197 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
6198 to merge probabilities.
6199 * predict.c (probably_never_executed): Also mark as cold functions
6200 with global 0 profile and guessed local profile.
6201 * profile-count.c (profile_probability::combine_with_count): New
6203 * profile-count.h (profile_probability::operator*,
6204 profile_probability::operator*=, profile_probability::operator/,
6205 profile_probability::operator/=): Reduce precision to adjusted
6206 and set value to guessed on contradictory divisions.
6207 (profile_probability::combine_with_freq): Remove.
6208 (profile_probability::combine_wiht_count): Declare.
6209 (profile_count::force_nonzero):: Set to adjusted.
6210 (profile_count::probability_in):: Set quality to adjusted.
6211 * tree-ssa-tail-merge.c (replace_block_by): Use
6214 2018-01-23 Andrew Waterman <andrew@sifive.com>
6215 Jim Wilson <jimw@sifive.com>
6217 * config/riscv/riscv.c (riscv_stack_boundary): New.
6218 (riscv_option_override): Set riscv_stack_boundary. Handle
6219 riscv_preferred_stack_boundary_arg.
6220 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
6221 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
6222 (STACK_BOUNDARY): Set to riscv_stack_boundary.
6223 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
6224 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
6225 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
6227 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
6230 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
6231 of struct ix86_frame.
6232 (ix86_expand_epilogue): Likewise. Add a local variable for
6233 the reg_save_offset field in struct ix86_frame.
6235 2018-01-23 Bin Cheng <bin.cheng@arm.com>
6237 PR tree-optimization/82604
6238 * tree-loop-distribution.c (enum partition_kind): New enum item
6239 PKIND_PARTIAL_MEMSET.
6240 (partition_builtin_p): Support above new enum item.
6241 (generate_code_for_partition): Ditto.
6242 (compute_access_range): Differentiate cases that equality can be
6243 proven at all loops, the innermost loops or no loops.
6244 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
6245 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
6246 (finalize_partitions, distribute_loop): Don't fuse partition of
6247 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
6248 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
6251 2018-01-23 Martin Liska <mliska@suse.cz>
6253 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
6254 order to ignore the predictor.
6255 (PRED_POLYMORPHIC_CALL): Likewise.
6256 (PRED_RECURSIVE_CALL): Likewise.
6258 2018-01-23 Martin Liska <mliska@suse.cz>
6260 * tree-profile.c (tree_profiling): Print function header to
6261 aware reader which function we are working on.
6262 * value-prof.c (gimple_find_values_to_profile): Do not print
6263 not interesting value histograms.
6265 2018-01-23 Martin Liska <mliska@suse.cz>
6267 * profile-count.h (enum profile_quality): Add
6268 profile_uninitialized as the first value. Do not number values
6269 as they are zero based.
6270 (profile_count::verify): Update sanity check.
6271 (profile_probability::verify): Likewise.
6273 2018-01-23 Nathan Sidwell <nathan@acm.org>
6275 * doc/invoke.texi (ffor-scope): Deprecate.
6277 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6279 PR tree-optimization/83510
6280 * domwalk.c (set_all_edges_as_executable): New function.
6281 (dom_walker::dom_walker): Convert bool param
6282 "skip_unreachable_blocks" to enum reachability. Move setup of
6283 edge flags to set_all_edges_as_executable and only do it when
6284 reachability is REACHABLE_BLOCKS.
6285 * domwalk.h (enum dom_walker::reachability): New enum.
6286 (dom_walker::dom_walker): Convert bool param
6287 "skip_unreachable_blocks" to enum reachability.
6288 (set_all_edges_as_executable): New decl.
6289 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
6290 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
6292 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
6293 but converting true to REACHABLE_BLOCKS.
6294 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
6296 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
6297 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
6298 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
6300 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
6301 if check_all_array_refs will be called.
6303 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6305 * tree.c (selftest::test_location_wrappers): Add more test
6308 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6310 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
6311 (selftest::test_bit_in_range): Likewise.
6313 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6316 * doc/sourcebuild.texi (vect_float): Say that the selector
6317 only describes the situation when -funsafe-math-optimizations is on.
6318 (vect_float_strict): Document.
6320 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6322 PR tree-optimization/83965
6323 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
6324 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
6325 instead of checking only for a reduction.
6326 (vect_recog_widen_sum_pattern): Likewise.
6328 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6330 * predict.c (probably_never_executed): Only use precise profile info.
6331 (compute_function_frequency): Skip after inlining hack since we now
6332 have quality checking.
6334 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6336 * profile-count.h (profile_probability::very_unlikely,
6337 profile_probability::unlikely, profile_probability::even): Set
6338 precision to guessed.
6340 2018-01-23 Richard Biener <rguenther@suse.de>
6342 PR tree-optimization/83963
6343 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
6344 Properly terminate dominator walk when crossing the exit edge not
6345 when visiting its source block.
6347 2018-01-23 Jakub Jelinek <jakub@redhat.com>
6350 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
6351 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
6353 2018-01-22 Jakub Jelinek <jakub@redhat.com>
6355 PR tree-optimization/83957
6356 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
6357 semicolon after for body surrounded by braces.
6359 PR tree-optimization/83081
6360 * profile-count.h (profile_probability::split): New method.
6361 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
6362 Use profile_probability::split.
6363 (do_compare_rtx_and_jump): Fix adjustment of probabilities
6364 when splitting a single conditional jump into 2.
6366 2018-01-22 David Malcolm <dmalcolm@redhat.com>
6368 PR tree-optimization/69452
6369 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
6372 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6374 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
6375 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
6376 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
6378 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6380 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
6381 * config/rl78/rl78.md (movdi): New define_expand.
6382 * config/rl78/rl78.c (rl78_split_movdi): New function.
6384 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
6387 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
6389 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
6390 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
6391 128-bit to produce an UNSPEC move to get the double word with the
6392 signbit and then a shift directly to do signbit.
6393 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
6394 implementation with a new version that just does either a direct
6395 move or a regular move. Move memory interface to separate insns.
6396 Move insns so they are next to the expander.
6397 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
6398 with signbit move. Split big and little endian case.
6399 (signbit<mode>2_dm_mem_le): Likewise.
6400 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
6401 (signbit<mode>2_dm2): Likewise.
6403 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6405 * config/rl78/rl78.md (anddi3): New define_expand.
6407 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6409 * config/rl78/rl78.md (umindi3): New define_expand.
6411 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6413 * config/rl78/rl78.md (smindi3): New define_expand.
6415 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6417 * config/rl78/rl78.md (smaxdi3): New define_expand.
6419 2018-01-22 Carl Love <cel@us.ibm.com>
6421 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
6422 LVX_V1TI): Add macro expansion.
6423 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
6424 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
6425 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
6426 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
6427 Change check to determine if the instruction is a byte reversing
6428 entry. Fix typo in comment.
6429 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
6430 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
6431 Add def_builtin calls for new builtins.
6432 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
6433 Add define_insn expansion.
6435 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6437 * config/rl78/rl78.md (umaxdi3): New define_expand.
6439 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6441 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
6442 for non-QImode registers.
6444 2018-01-22 Richard Biener <rguenther@suse.de>
6446 PR tree-optimization/83963
6447 * graphite-scop-detection.c (scop_detection::get_sese): Delay
6448 including the loop exit block.
6449 (scop_detection::merge_sese): Likewise.
6450 (scop_detection::add_scop): Do it here instead.
6452 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6454 * doc/sourcebuild.texi (arm_softfloat): Document.
6456 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
6459 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
6460 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
6461 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
6463 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6464 David Edelsohn <dje.gcc@gmail.com>
6467 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6468 Change "crset eq" to "crset 2".
6469 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6470 (*call_indirect_aix<mode>_nospec): Likewise.
6471 (*call_value_indirect_aix<mode>_nospec): Likewise.
6472 (*call_indirect_elfv2<mode>_nospec): Likewise.
6473 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
6474 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
6475 change assembly output from . to $.
6476 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6477 (indirect_jump<mode>_nospec): Change assembly output from . to $.
6478 (*tablejump<mode>_internal1_nospec): Likewise.
6480 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
6483 * config/sh/sh_optimize_sett_clrt.cc:
6484 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
6486 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
6488 PR tree-optimization/83940
6489 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
6490 offset_dt to vect_constant_def rather than vect_unknown_def_type.
6491 (vect_check_load_store_mask): Add a mask_dt_out parameter and
6492 use it to pass back the definition type.
6493 (vect_check_store_rhs): Likewise rhs_dt_out.
6494 (vect_build_gather_load_calls): Add a mask_dt argument and use
6495 it instead of a call to vect_is_simple_use.
6496 (vectorizable_store): Update calls to vect_check_load_store_mask
6497 and vect_check_store_rhs. Use the dt returned by the latter instead
6498 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
6499 instead of calls to vect_is_simple_use. Pass the scalar rather
6500 than the vector operand to vect_is_simple_use when handling
6501 second and subsequent copies of an rhs value.
6502 (vectorizable_load): Update calls to vect_check_load_store_mask
6503 and vect_build_gather_load_calls. Use the cached mask_dt and
6504 gs_info.offset_dt instead of calls to vect_is_simple_use.
6506 2018-01-20 Jakub Jelinek <jakub@redhat.com>
6509 * tree-emutls.c: Include gimplify.h.
6510 (lower_emutls_2): New function.
6511 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
6512 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
6513 it before further processing.
6516 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
6517 UINTVAL (trueop1) instead of INTVAL (op1).
6519 2018-01-19 Jakub Jelinek <jakub@redhat.com>
6523 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
6524 INCOMING_FRAME_SP_OFFSET if not defined.
6525 (scan_trace): Add ENTRY argument. If true and
6526 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
6527 emit a note to adjust the CFA offset.
6528 (create_cfi_notes): Adjust scan_trace callers.
6529 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
6530 INCOMING_FRAME_SP_OFFSET in the CIE.
6531 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
6532 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
6534 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
6535 * doc/tm.texi: Regenerated.
6537 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6539 PR rtl-optimization/83147
6540 * lra-constraints.c (remove_inheritance_pseudos): Use
6541 lra_substitute_pseudo_within_insn.
6543 2018-01-19 Tom de Vries <tom@codesourcery.com>
6544 Cesar Philippidis <cesar@codesourcery.com>
6547 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
6549 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
6552 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
6553 spaces for function labels.
6555 2018-01-19 Martin Liska <mliska@suse.cz>
6557 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
6558 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
6559 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
6560 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
6561 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
6562 (PRED_CONST_RETURN): Change from 69 to 65.
6563 (PRED_NULL_RETURN): Change from 91 to 71.
6564 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
6565 (PRED_LOOP_GUARD): Change from 66 to 73.
6567 2018-01-19 Martin Liska <mliska@suse.cz>
6569 * predict.c (predict_insn_def): Add new assert.
6570 (struct branch_predictor): Change type to signed integer.
6571 (test_prediction_value_range): Amend test to cover
6573 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
6574 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
6575 (PRED_LOOP_ITERATIONS_MAX): Likewise.
6576 (PRED_LOOP_IV_COMPARE): Likewise.
6577 * predict.h (PROB_UNINITIALIZED): Define new constant.
6579 2018-01-19 Martin Liska <mliska@suse.cz>
6581 * predict.c (dump_prediction): Add new format for
6582 analyze_brprob.py script which is enabled with -details
6584 * profile-count.h (precise_p): New function.
6586 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6588 PR tree-optimization/83922
6589 * tree-vect-loop.c (vect_verify_full_masking): Return false if
6590 there are no statements that need masking.
6591 (vect_active_double_reduction_p): New function.
6592 (vect_analyze_loop_operations): Use it when handling phis that
6593 are not in the loop header.
6595 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6597 PR tree-optimization/83914
6598 * tree-vect-loop.c (vectorizable_induction): Don't convert
6599 init_expr or apply the peeling adjustment for inductions
6600 that are nested within the vectorized loop.
6602 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6604 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
6607 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6611 * function.h (gimplify_parameters): Add gimple_seq * argument.
6612 * function.c: Include gimple.h and options.h.
6613 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
6614 for the added local temporaries if needed.
6615 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
6616 if there are any parameter cleanups, wrap whole body into a
6617 try/finally with the cleanups.
6619 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
6622 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
6623 Use GET_MODE_CLASS for scalar floating point.
6625 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6629 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
6630 Fix call of call_cgraph_insertion_hooks.
6632 2018-01-18 Martin Sebor <msebor@redhat.com>
6634 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
6636 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6639 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
6642 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
6645 * common.opt: (-ffile-prefix-map): New option.
6646 * opts.c (common_handle_option): Defer it.
6647 * opts-global.c (handle_common_deferred_options): Handle it.
6648 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
6649 * file-prefix-map.h: New file.
6650 (remap_debug_filename, add_debug_prefix_map): ...here.
6651 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
6652 * final.c (debug_prefix_map, add_debug_prefix_map
6653 remap_debug_filename): Move to...
6654 * file-prefix-map.c: New file.
6655 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
6656 generalize, get rid of alloca(), use strrchr() instead of strchr().
6657 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
6658 Implement in terms of add_prefix_map().
6659 (remap_macro_filename, remap_debug_filename): Implement in term of
6661 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
6662 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
6663 * dbxout.c: Include file-prefix-map.h.
6664 * varasm.c: Likewise.
6665 * vmsdbgout.c: Likewise.
6666 * xcoffout.c: Likewise.
6667 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
6668 * doc/cppopts.texi (-fmacro-prefix-map): Document.
6669 * doc/invoke.texi (-ffile-prefix-map): Document.
6670 (-fdebug-prefix-map): Update description.
6672 2018-01-18 Martin Liska <mliska@suse.cz>
6674 * config/i386/i386.c (indirect_thunk_name): Document that also
6676 (output_indirect_thunk): Document why both instructions
6677 (pause and lfence) are generated.
6679 2018-01-18 Richard Biener <rguenther@suse.de>
6681 PR tree-optimization/83887
6682 * graphite-scop-detection.c
6683 (scop_detection::get_nearest_dom_with_single_entry): Remove.
6684 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
6685 (scop_detection::merge_sese): Re-implement with a flood-fill
6686 algorithm that properly finds a SESE region if it exists.
6688 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6691 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
6692 pointer_diff optimizations use view_convert instead of convert.
6694 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6696 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6697 Generate different code for -mno-speculate-indirect-jumps.
6698 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6699 (*call_indirect_aix<mode>): Disable for
6700 -mno-speculate-indirect-jumps.
6701 (*call_indirect_aix<mode>_nospec): New define_insn.
6702 (*call_value_indirect_aix<mode>): Disable for
6703 -mno-speculate-indirect-jumps.
6704 (*call_value_indirect_aix<mode>_nospec): New define_insn.
6705 (*sibcall_nonlocal_sysv<mode>): Generate different code for
6706 -mno-speculate-indirect-jumps.
6707 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6709 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6711 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
6712 long double type, set the flags for noting the default long double
6713 type, even if we don't pass or return a long double type.
6715 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
6718 * ipa-inline.c (flatten_function): Do not overwrite final inlining
6721 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
6723 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
6724 support for merge[hl].
6725 (fold_mergehl_helper): New helper function.
6726 (tree-vector-builder.h): New #include for tree_vector_builder usage.
6727 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
6728 (altivec_vmrglw_direct): Add xxmrglw insn.
6730 2018-01-17 Andrew Waterman <andrew@sifive.com>
6732 * config/riscv/riscv.c (riscv_conditional_register_usage): If
6733 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
6735 2018-01-17 David Malcolm <dmalcolm@redhat.com>
6738 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
6739 call the lto_location_cache before reading the
6740 DECL_SOURCE_LOCATION of the types.
6742 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
6743 Richard Sandiford <richard.sandiford@linaro.org>
6745 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
6746 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
6747 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
6748 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
6749 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
6751 * config/aarch64/constraints.md (aarch64_movti_operand):
6753 * config/aarch64/predicates.md (Uti): Add new constraint.
6755 2018-01-17 Carl Love <cel@us.ibm.com>
6757 * config/rs6000/vsx.md (define_expand xl_len_r,
6758 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
6759 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
6761 (define_expand, define_insn): Move the shift left from the
6762 define_insn to the define_expand for lxvl and stxvl instructions.
6763 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
6764 and XL_LEN_R definitions to PURE.
6766 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
6768 * config/i386/i386.c (indirect_thunk_name): Declare regno
6769 as unsigned int. Compare regno with INVALID_REGNUM.
6770 (output_indirect_thunk): Ditto.
6771 (output_indirect_thunk_function): Ditto.
6772 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
6773 in the call to output_indirect_thunk_function.
6775 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
6778 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
6779 rather than the size of inner_type to determine the stack slot size
6780 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
6782 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
6785 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
6788 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
6790 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
6791 endian Linux systems to optionally enable multilibs for selecting
6792 the long double type if the user configured an explicit type.
6793 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
6794 have no long double multilibs if not defined.
6795 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
6796 warn if the user used -mabi={ieee,ibm}longdouble and we built
6797 multilibs for long double.
6798 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
6799 appropriate multilib option.
6800 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
6802 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
6803 for building long double multilibs.
6804 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
6806 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
6808 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
6811 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
6813 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
6816 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
6819 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
6822 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6824 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
6825 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
6827 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
6829 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
6830 different rtl trees depending on TARGET_64BIT.
6831 (rs6000_gen_lvx): Likewise.
6833 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6835 * config/visium/visium.md (nop): Tweak comment.
6836 (hazard_nop): Likewise.
6838 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6840 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
6841 -mspeculate-indirect-jumps.
6842 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
6843 for -mno-speculate-indirect-jumps.
6844 (*call_indirect_elfv2<mode>_nospec): New define_insn.
6845 (*call_value_indirect_elfv2<mode>): Disable for
6846 -mno-speculate-indirect-jumps.
6847 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
6848 (indirect_jump): Emit different RTL for
6849 -mno-speculate-indirect-jumps.
6850 (*indirect_jump<mode>): Disable for
6851 -mno-speculate-indirect-jumps.
6852 (*indirect_jump<mode>_nospec): New define_insn.
6853 (tablejump): Emit different RTL for
6854 -mno-speculate-indirect-jumps.
6855 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
6856 (tablejumpsi_nospec): New define_expand.
6857 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
6858 (tablejumpdi_nospec): New define_expand.
6859 (*tablejump<mode>_internal1): Disable for
6860 -mno-speculate-indirect-jumps.
6861 (*tablejump<mode>_internal1_nospec): New define_insn.
6862 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
6865 2018-01-16 Artyom Skrobov tyomitch@gmail.com
6867 * caller-save.c (insert_save): Drop unnecessary parameter. All
6870 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6871 Richard Biener <rguenth@suse.de>
6874 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
6875 return early, inline manually is_gimple_sizepos. Make sure if we
6876 call gimplify_expr we don't end up with a gimple constant.
6877 * tree.c (variably_modified_type_p): Don't return true for
6878 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
6879 * gimplify.h (is_gimple_sizepos): Remove.
6881 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6883 PR tree-optimization/83857
6884 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
6885 vectorizable_live_operation for pure SLP statements.
6886 (vectorizable_live_operation): Handle PHIs.
6888 2018-01-16 Richard Biener <rguenther@suse.de>
6890 PR tree-optimization/83867
6891 * tree-vect-stmts.c (vect_transform_stmt): Precompute
6892 nested_in_vect_loop_p since the scalar stmt may get invalidated.
6894 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6897 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
6898 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
6899 If off is not INTEGER_CST, issue a may not be aligned warning
6900 rather than isn't aligned. Use isn%'t rather than isn't.
6901 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
6903 <case MULT_EXPR>: Improve the case when bottom and one of the
6904 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
6905 operand, in that case check if the other operand is multiple of
6906 bottom divided by the INTEGER_CST operand.
6908 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6911 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
6912 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
6913 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
6914 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
6915 * config/pa/pa.c (pa_function_arg_advance): Likewise.
6916 (pa_function_arg, pa_arg_partial_bytes): Likewise.
6917 (pa_function_arg_size): New function.
6919 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6921 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
6922 in a separate statement.
6924 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6926 PR tree-optimization/83847
6927 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
6928 group gathers and scatters.
6930 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6932 PR rtl-optimization/86620
6933 * params.def (max-sched-ready-insns): Bump minimum value to 1.
6935 PR rtl-optimization/83213
6936 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
6937 to last if both are JUMP_INSNs.
6939 PR tree-optimization/83843
6940 * gimple-ssa-store-merging.c
6941 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
6942 store_immediate_info for bswap/nop orig_stores.
6944 2018-01-15 Andrew Waterman <andrew@sifive.com>
6946 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
6948 <UDIV>: Increase cost if !TARGET_DIV.
6950 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
6952 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
6953 (define_attr "cr_logical_3op"): New.
6954 (cceq_ior_compare): Adjust.
6955 (cceq_ior_compare_complement): Adjust.
6956 (*cceq_rev_compare): Adjust.
6957 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
6958 (is_cracked_insn): Adjust.
6959 (insn_must_be_first_in_group): Adjust.
6960 * config/rs6000/40x.md: Adjust.
6961 * config/rs6000/440.md: Adjust.
6962 * config/rs6000/476.md: Adjust.
6963 * config/rs6000/601.md: Adjust.
6964 * config/rs6000/603.md: Adjust.
6965 * config/rs6000/6xx.md: Adjust.
6966 * config/rs6000/7450.md: Adjust.
6967 * config/rs6000/7xx.md: Adjust.
6968 * config/rs6000/8540.md: Adjust.
6969 * config/rs6000/cell.md: Adjust.
6970 * config/rs6000/e300c2c3.md: Adjust.
6971 * config/rs6000/e500mc.md: Adjust.
6972 * config/rs6000/e500mc64.md: Adjust.
6973 * config/rs6000/e5500.md: Adjust.
6974 * config/rs6000/e6500.md: Adjust.
6975 * config/rs6000/mpc.md: Adjust.
6976 * config/rs6000/power4.md: Adjust.
6977 * config/rs6000/power5.md: Adjust.
6978 * config/rs6000/power6.md: Adjust.
6979 * config/rs6000/power7.md: Adjust.
6980 * config/rs6000/power8.md: Adjust.
6981 * config/rs6000/power9.md: Adjust.
6982 * config/rs6000/rs64.md: Adjust.
6983 * config/rs6000/titan.md: Adjust.
6985 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6987 * config/i386/predicates.md (indirect_branch_operand): Rewrite
6988 ix86_indirect_branch_register logic.
6990 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6992 * config/i386/constraints.md (Bs): Update
6993 ix86_indirect_branch_register check. Don't check
6994 ix86_indirect_branch_register with GOT_memory_operand.
6996 * config/i386/predicates.md (GOT_memory_operand): Don't check
6997 ix86_indirect_branch_register here.
6998 (GOT32_symbol_operand): Likewise.
7000 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7002 * config/i386/predicates.md (constant_call_address_operand):
7003 Rewrite ix86_indirect_branch_register logic.
7004 (sibcall_insn_operand): Likewise.
7006 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7008 * config/i386/constraints.md (Bs): Replace
7009 ix86_indirect_branch_thunk_register with
7010 ix86_indirect_branch_register.
7012 * config/i386/i386.md (indirect_jump): Likewise.
7013 (tablejump): Likewise.
7014 (*sibcall_memory): Likewise.
7015 (*sibcall_value_memory): Likewise.
7016 Peepholes of indirect call and jump via memory: Likewise.
7017 * config/i386/i386.opt: Likewise.
7018 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7019 (GOT_memory_operand): Likewise.
7020 (call_insn_operand): Likewise.
7021 (sibcall_insn_operand): Likewise.
7022 (GOT32_symbol_operand): Likewise.
7024 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7027 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
7028 type rather than type addr's type points to.
7029 (expand_omp_atomic_mutex): Likewise.
7030 (expand_omp_atomic): Likewise.
7032 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7035 * config/i386/i386.c (output_indirect_thunk_function): Use
7036 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
7037 for __x86_return_thunk.
7039 2018-01-15 Richard Biener <rguenther@suse.de>
7042 * expmed.c (extract_bit_field_1): Fix typo.
7044 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7047 * config/arm/iterators.md (VF): New mode iterator.
7048 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
7049 Remove integer-related logic from pattern.
7050 (neon_vabd<mode>_3): Likewise.
7052 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7055 * common.opt (fstrict-overflow): No longer an alias.
7056 (fwrapv-pointer): New option.
7057 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
7058 also for pointer types based on flag_wrapv_pointer.
7059 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
7060 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
7061 opts->x_flag_wrapv got set.
7062 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
7063 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
7064 POINTER_TYPE_OVERFLOW_UNDEFINED.
7065 * match.pd: Likewise in address comparison pattern.
7066 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
7068 2018-01-15 Richard Biener <rguenther@suse.de>
7071 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
7072 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
7073 Reset type names to their identifier if their TYPE_DECL doesn't
7074 have linkage (and thus is used for ODR and devirt).
7075 (save_debug_info_for_decl): Remove.
7076 (save_debug_info_for_type): Likewise.
7077 (add_tree_to_fld_list): Adjust.
7078 * tree-pretty-print.c (dump_generic_node): Make dumping of
7079 type names more robust.
7081 2018-01-15 Richard Biener <rguenther@suse.de>
7083 * BASE-VER: Bump to 8.0.1.
7085 2018-01-14 Martin Sebor <msebor@redhat.com>
7088 * builtins.c (check_access): Avoid warning when the no-warning bit
7091 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
7093 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
7094 * ira-color (allocno_hard_regs_compare): Likewise.
7096 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
7099 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
7100 Use .pushsection/.popsection.
7102 2018-01-14 Martin Sebor <msebor@redhat.com>
7105 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
7107 2018-01-14 Jakub Jelinek <jakub@redhat.com>
7109 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
7110 entry from extra_headers.
7111 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
7112 extra_headers, make the list bitwise identical to the i?86-*-* one.
7114 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7116 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
7117 -mcmodel=large with -mindirect-branch=thunk,
7118 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
7119 -mfunction-return=thunk-extern.
7120 * doc/invoke.texi: Document -mcmodel=large is incompatible with
7121 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
7122 -mfunction-return=thunk and -mfunction-return=thunk-extern.
7124 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7126 * config/i386/i386.c (print_reg): Print the name of the full
7127 integer register without '%'.
7128 (ix86_print_operand): Handle 'V'.
7129 * doc/extend.texi: Document 'V' modifier.
7131 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7133 * config/i386/constraints.md (Bs): Disallow memory operand for
7134 -mindirect-branch-register.
7136 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7137 (GOT_memory_operand): Likewise.
7138 (call_insn_operand): Likewise.
7139 (sibcall_insn_operand): Likewise.
7140 (GOT32_symbol_operand): Likewise.
7141 * config/i386/i386.md (indirect_jump): Call convert_memory_address
7142 for -mindirect-branch-register.
7143 (tablejump): Likewise.
7144 (*sibcall_memory): Likewise.
7145 (*sibcall_value_memory): Likewise.
7146 Disallow peepholes of indirect call and jump via memory for
7147 -mindirect-branch-register.
7148 (*call_pop): Replace m with Bw.
7149 (*call_value_pop): Likewise.
7150 (*sibcall_pop_memory): Replace m with Bs.
7151 * config/i386/i386.opt (mindirect-branch-register): New option.
7152 * doc/invoke.texi: Document -mindirect-branch-register option.
7154 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7156 * config/i386/i386-protos.h (ix86_output_function_return): New.
7157 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
7158 set function_return_type.
7159 (indirect_thunk_name): Add ret_p to indicate thunk for function
7161 (output_indirect_thunk_function): Pass false to
7162 indirect_thunk_name.
7163 (ix86_output_indirect_branch_via_reg): Likewise.
7164 (ix86_output_indirect_branch_via_push): Likewise.
7165 (output_indirect_thunk_function): Create alias for function
7166 return thunk if regno < 0.
7167 (ix86_output_function_return): New function.
7168 (ix86_handle_fndecl_attribute): Handle function_return.
7169 (ix86_attribute_table): Add function_return.
7170 * config/i386/i386.h (machine_function): Add
7171 function_return_type.
7172 * config/i386/i386.md (simple_return_internal): Use
7173 ix86_output_function_return.
7174 (simple_return_internal_long): Likewise.
7175 * config/i386/i386.opt (mfunction-return=): New option.
7176 (indirect_branch): Mention -mfunction-return=.
7177 * doc/extend.texi: Document function_return function attribute.
7178 * doc/invoke.texi: Document -mfunction-return= option.
7180 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7182 * config/i386/i386-opts.h (indirect_branch): New.
7183 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
7184 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
7185 with local indirect jump when converting indirect call and jump.
7186 (ix86_set_indirect_branch_type): New.
7187 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
7188 (indirectlabelno): New.
7189 (indirect_thunk_needed): Likewise.
7190 (indirect_thunk_bnd_needed): Likewise.
7191 (indirect_thunks_used): Likewise.
7192 (indirect_thunks_bnd_used): Likewise.
7193 (INDIRECT_LABEL): Likewise.
7194 (indirect_thunk_name): Likewise.
7195 (output_indirect_thunk): Likewise.
7196 (output_indirect_thunk_function): Likewise.
7197 (ix86_output_indirect_branch_via_reg): Likewise.
7198 (ix86_output_indirect_branch_via_push): Likewise.
7199 (ix86_output_indirect_branch): Likewise.
7200 (ix86_output_indirect_jmp): Likewise.
7201 (ix86_code_end): Call output_indirect_thunk_function if needed.
7202 (ix86_output_call_insn): Call ix86_output_indirect_branch if
7204 (ix86_handle_fndecl_attribute): Handle indirect_branch.
7205 (ix86_attribute_table): Add indirect_branch.
7206 * config/i386/i386.h (machine_function): Add indirect_branch_type
7207 and has_local_indirect_jump.
7208 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
7210 (tablejump): Likewise.
7211 (*indirect_jump): Use ix86_output_indirect_jmp.
7212 (*tablejump_1): Likewise.
7213 (simple_return_indirect_internal): Likewise.
7214 * config/i386/i386.opt (mindirect-branch=): New option.
7215 (indirect_branch): New.
7218 (thunk-inline): Likewise.
7219 (thunk-extern): Likewise.
7220 * doc/extend.texi: Document indirect_branch function attribute.
7221 * doc/invoke.texi: Document -mindirect-branch= option.
7223 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
7226 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
7228 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
7230 * ipa-inline.c (want_inline_small_function_p): Return false if
7231 inlining has already failed with CIF_FINAL_ERROR.
7232 (update_caller_keys): Call want_inline_small_function_p before
7234 (update_callee_keys): Likewise.
7236 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7238 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
7240 (rs6000_quadword_masked_address_p): Likewise.
7241 (quad_aligned_load_p): Likewise.
7242 (quad_aligned_store_p): Likewise.
7243 (const_load_sequence_p): Add comment to describe the outer-most loop.
7244 (mimic_memory_attributes_and_flags): New function.
7245 (rs6000_gen_stvx): Likewise.
7246 (replace_swapped_aligned_store): Likewise.
7247 (rs6000_gen_lvx): Likewise.
7248 (replace_swapped_aligned_load): Likewise.
7249 (replace_swapped_load_constant): Capitalize argument name in
7250 comment describing this function.
7251 (rs6000_analyze_swaps): Add a third pass to search for vector loads
7252 and stores that access quad-word aligned addresses and replace
7253 with stvx or lvx instructions when appropriate.
7254 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
7255 New function prototype.
7256 (rs6000_quadword_masked_address_p): Likewise.
7257 (rs6000_gen_lvx): Likewise.
7258 (rs6000_gen_stvx): Likewise.
7259 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
7260 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
7261 when memory address is aligned.
7262 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
7263 this split to select lvx instruction when memory address is aligned.
7264 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
7265 instruction when memory address is aligned.
7266 (*vsx_le_perm_load_v16qi): Likewise.
7267 (four unnamed splitters): Modify to select the stvx instruction
7268 when memory is aligned.
7270 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
7272 * predict.c (determine_unlikely_bbs): Handle correctly BBs
7273 which appears in the queue multiple times.
7275 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7276 Alan Hayward <alan.hayward@arm.com>
7277 David Sherwood <david.sherwood@arm.com>
7279 * tree-vectorizer.h (vec_lower_bound): New structure.
7280 (_loop_vec_info): Add check_nonzero and lower_bounds.
7281 (LOOP_VINFO_CHECK_NONZERO): New macro.
7282 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
7283 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
7284 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
7285 fields. Make seg_len the distance travelled, not including the
7287 (dr_direction_indicator): Declare.
7288 (dr_zero_step_indicator): Likewise.
7289 (dr_known_forward_stride_p): Likewise.
7290 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
7292 (runtime_alias_check_p): Allow runtime alias checks with
7294 (operator ==): Compare access_size and align.
7295 (prune_runtime_alias_test_list): Rework for new distinction between
7296 the access_size and seg_len.
7297 (create_intersect_range_checks_index): Likewise. Cope with polynomial
7299 (get_segment_min_max): New function.
7300 (create_intersect_range_checks): Use it.
7301 (dr_step_indicator): New function.
7302 (dr_direction_indicator): Likewise.
7303 (dr_zero_step_indicator): Likewise.
7304 (dr_known_forward_stride_p): Likewise.
7305 * tree-loop-distribution.c (data_ref_segment_size): Return
7306 DR_STEP * (niters - 1).
7307 (compute_alias_check_pairs): Update call to the dr_with_seg_len
7309 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
7310 (vect_preserves_scalar_order_p): New function, split out from...
7311 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
7312 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
7313 (vect_vfa_access_size): New function.
7314 (vect_vfa_align): Likewise.
7315 (vect_compile_time_alias): Take access_size_a and access_b arguments.
7316 (dump_lower_bound): New function.
7317 (vect_check_lower_bound): Likewise.
7318 (vect_small_gap_p): Likewise.
7319 (vectorizable_with_step_bound_p): Likewise.
7320 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
7321 depencies if the vectorization factor is 1. Convert the checks
7322 for nonzero steps into checks on the bounds of DR_STEP. Try using
7323 a bunds check for variable steps if the minimum required step is
7324 relatively small. Update calls to the dr_with_seg_len
7325 constructor and to vect_compile_time_alias.
7326 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
7328 (vect_loop_versioning): Call it.
7329 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
7331 (vect_estimate_min_profitable_iters): Account for any bounds checks.
7333 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7334 Alan Hayward <alan.hayward@arm.com>
7335 David Sherwood <david.sherwood@arm.com>
7337 * doc/sourcebuild.texi (vect_scatter_store): Document.
7338 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
7340 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
7342 * genopinit.c (main): Add supports_vec_scatter_store and
7343 supports_vec_scatter_store_cached to target_optabs.
7344 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
7345 IFN_MASK_SCATTER_STORE.
7346 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
7348 * internal-fn.h (internal_store_fn_p): Declare.
7349 (internal_fn_stored_value_index): Likewise.
7350 * internal-fn.c (scatter_store_direct): New macro.
7351 (expand_scatter_store_optab_fn): New function.
7352 (direct_scatter_store_optab_supported_p): New macro.
7353 (internal_store_fn_p): New function.
7354 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
7355 IFN_MASK_SCATTER_STORE.
7356 (internal_fn_mask_index): Likewise.
7357 (internal_fn_stored_value_index): New function.
7358 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
7360 * optabs-query.h (supports_vec_scatter_store_p): Declare.
7361 * optabs-query.c (supports_vec_scatter_store_p): New function.
7362 * tree-vectorizer.h (vect_get_store_rhs): Declare.
7363 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
7364 true for scatter stores.
7365 (vect_gather_scatter_fn_p): Handle scatter stores too.
7366 (vect_check_gather_scatter): Consider using scatter stores if
7367 supports_vec_scatter_store_p.
7368 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
7370 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7371 internal_fn_stored_value_index.
7372 (check_load_store_masking): Handle scatter stores too.
7373 (vect_get_store_rhs): Make public.
7374 (vectorizable_call): Use internal_store_fn_p.
7375 (vectorizable_store): Handle scatter store internal functions.
7376 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
7377 when deciding whether the end of the group has been reached.
7378 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
7379 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
7380 (mask_scatter_store<mode>): New insns.
7382 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7383 Alan Hayward <alan.hayward@arm.com>
7384 David Sherwood <david.sherwood@arm.com>
7386 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
7387 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
7388 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
7390 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
7391 Use vect_truncate_gather_scatter_offset if we can't treat the
7392 operation as a normal gather load or scatter store.
7393 (get_group_load_store_type): Take the gather_scatter_info
7394 as argument. Try using a gather load or scatter store for
7395 single-element groups.
7396 (get_load_store_type): Update calls to get_group_load_store_type
7397 and vect_use_strided_gather_scatters_p.
7399 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7400 Alan Hayward <alan.hayward@arm.com>
7401 David Sherwood <david.sherwood@arm.com>
7403 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
7404 optional tree argument.
7405 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
7407 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
7408 but continue to use the current value as a fallback.
7409 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
7410 to compare the updates.
7411 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
7412 (get_load_store_type): Use it when handling a strided access.
7413 (vect_get_strided_load_store_ops): New function.
7414 (vect_get_data_ptr_increment): Likewise.
7415 (vectorizable_load): Handle strided gather loads. Always pass
7416 a step to vect_create_data_ref_ptr and bump_vector_ptr.
7418 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7419 Alan Hayward <alan.hayward@arm.com>
7420 David Sherwood <david.sherwood@arm.com>
7422 * doc/md.texi (gather_load@var{m}): Document.
7423 (mask_gather_load@var{m}): Likewise.
7424 * genopinit.c (main): Add supports_vec_gather_load and
7425 supports_vec_gather_load_cached to target_optabs.
7426 * optabs-tree.c (init_tree_optimization_optabs): Use
7427 ggc_cleared_alloc to allocate target_optabs.
7428 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
7429 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
7431 * internal-fn.h (internal_load_fn_p): Declare.
7432 (internal_gather_scatter_fn_p): Likewise.
7433 (internal_fn_mask_index): Likewise.
7434 (internal_gather_scatter_fn_supported_p): Likewise.
7435 * internal-fn.c (gather_load_direct): New macro.
7436 (expand_gather_load_optab_fn): New function.
7437 (direct_gather_load_optab_supported_p): New macro.
7438 (direct_internal_fn_optab): New function.
7439 (internal_load_fn_p): Likewise.
7440 (internal_gather_scatter_fn_p): Likewise.
7441 (internal_fn_mask_index): Likewise.
7442 (internal_gather_scatter_fn_supported_p): Likewise.
7443 * optabs-query.c (supports_at_least_one_mode_p): New function.
7444 (supports_vec_gather_load_p): Likewise.
7445 * optabs-query.h (supports_vec_gather_load_p): Declare.
7446 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
7447 and memory_type field.
7448 (NUM_PATTERNS): Bump to 15.
7449 * tree-vect-data-refs.c: Include internal-fn.h.
7450 (vect_gather_scatter_fn_p): New function.
7451 (vect_describe_gather_scatter_call): Likewise.
7452 (vect_check_gather_scatter): Try using internal functions for
7453 gather loads. Recognize existing calls to a gather load function.
7454 (vect_analyze_data_refs): Consider using gather loads if
7455 supports_vec_gather_load_p.
7456 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
7457 (vect_get_gather_scatter_offset_type): Likewise.
7458 (vect_convert_mask_for_vectype): Likewise.
7459 (vect_add_conversion_to_patterm): Likewise.
7460 (vect_try_gather_scatter_pattern): Likewise.
7461 (vect_recog_gather_scatter_pattern): New pattern recognizer.
7462 (vect_vect_recog_func_ptrs): Add it.
7463 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7464 internal_fn_mask_index and internal_gather_scatter_fn_p.
7465 (check_load_store_masking): Take the gather_scatter_info as an
7466 argument and handle gather loads.
7467 (vect_get_gather_scatter_ops): New function.
7468 (vectorizable_call): Check internal_load_fn_p.
7469 (vectorizable_load): Likewise. Handle gather load internal
7471 (vectorizable_store): Update call to check_load_store_masking.
7472 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
7473 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
7474 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
7475 (aarch64_gather_scale_operand_d): New predicates.
7476 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
7477 (mask_gather_load<mode>): New insns.
7479 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7480 Alan Hayward <alan.hayward@arm.com>
7481 David Sherwood <david.sherwood@arm.com>
7483 * optabs.def (fold_left_plus_optab): New optab.
7484 * doc/md.texi (fold_left_plus_@var{m}): Document.
7485 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
7486 * internal-fn.c (fold_left_direct): Define.
7487 (expand_fold_left_optab_fn): Likewise.
7488 (direct_fold_left_optab_supported_p): Likewise.
7489 * fold-const-call.c (fold_const_fold_left): New function.
7490 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
7491 * tree-parloops.c (valid_reduction_p): New function.
7492 (gather_scalar_reductions): Use it.
7493 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
7494 (vect_finish_replace_stmt): Declare.
7495 * tree-vect-loop.c (fold_left_reduction_fn): New function.
7496 (needs_fold_left_reduction_p): New function, split out from...
7497 (vect_is_simple_reduction): ...here. Accept reductions that
7498 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
7499 (vect_force_simple_reduction): Also store the reduction type in
7500 the assignment's STMT_VINFO_REDUC_TYPE.
7501 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
7502 (merge_with_identity): New function.
7503 (vect_expand_fold_left): Likewise.
7504 (vectorize_fold_left_reduction): Likewise.
7505 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
7506 scalar phi in place for it. Check for target support and reject
7507 cases that would reassociate the operation. Defer the transform
7508 phase to vectorize_fold_left_reduction.
7509 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
7510 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
7511 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
7513 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7515 * tree-if-conv.c (predicate_mem_writes): Remove redundant
7516 call to ifc_temp_var.
7518 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7519 Alan Hayward <alan.hayward@arm.com>
7520 David Sherwood <david.sherwood@arm.com>
7522 * target.def (legitimize_address_displacement): Take the original
7523 offset as a poly_int.
7524 * targhooks.h (default_legitimize_address_displacement): Update
7526 * targhooks.c (default_legitimize_address_displacement): Likewise.
7527 * doc/tm.texi: Regenerate.
7528 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
7529 as an argument, moving assert of ad->disp == ad->disp_term to...
7530 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
7531 Try calling targetm.legitimize_address_displacement before expanding
7532 the address rather than afterwards, and adjust for the new interface.
7533 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
7534 Match the new hook interface. Handle SVE addresses.
7535 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
7538 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7540 * Makefile.in (OBJS): Add early-remat.o.
7541 * target.def (select_early_remat_modes): New hook.
7542 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
7543 * doc/tm.texi: Regenerate.
7544 * targhooks.h (default_select_early_remat_modes): Declare.
7545 * targhooks.c (default_select_early_remat_modes): New function.
7546 * timevar.def (TV_EARLY_REMAT): New timevar.
7547 * passes.def (pass_early_remat): New pass.
7548 * tree-pass.h (make_pass_early_remat): Declare.
7549 * early-remat.c: New file.
7550 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
7552 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
7554 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7555 Alan Hayward <alan.hayward@arm.com>
7556 David Sherwood <david.sherwood@arm.com>
7558 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
7559 vfm1 with a bound_epilog parameter.
7560 (vect_do_peeling): Update calls accordingly, and move the prologue
7561 call earlier in the function. Treat the base bound_epilog as 0 for
7562 fully-masked loops and retain vf - 1 for other loops. Add 1 to
7563 this base when peeling for gaps.
7564 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
7565 with fully-masked loops.
7566 (vect_estimate_min_profitable_iters): Handle the single peeled
7567 iteration in that case.
7569 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7570 Alan Hayward <alan.hayward@arm.com>
7571 David Sherwood <david.sherwood@arm.com>
7573 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
7574 single-element interleaving even if the size is not a power of 2.
7575 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
7576 accesses for single-element interleaving if the group size is
7579 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7580 Alan Hayward <alan.hayward@arm.com>
7581 David Sherwood <david.sherwood@arm.com>
7583 * doc/md.texi (fold_extract_last_@var{m}): Document.
7584 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
7585 * optabs.def (fold_extract_last_optab): New optab.
7586 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
7587 * internal-fn.c (fold_extract_direct): New macro.
7588 (expand_fold_extract_optab_fn): Likewise.
7589 (direct_fold_extract_optab_supported_p): Likewise.
7590 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
7591 * tree-vect-loop.c (vect_model_reduction_cost): Handle
7592 EXTRACT_LAST_REDUCTION.
7593 (get_initial_def_for_reduction): Do not create an initial vector
7594 for EXTRACT_LAST_REDUCTION reductions.
7595 (vectorizable_reduction): Leave the scalar phi in place for
7596 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
7597 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
7598 epilogue code for EXTRACT_LAST_REDUCTION and defer the
7599 transform phase to vectorizable_condition.
7600 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
7602 (vect_finish_stmt_generation): ...here.
7603 (vect_finish_replace_stmt): New function.
7604 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
7605 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
7607 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
7609 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7610 Alan Hayward <alan.hayward@arm.com>
7611 David Sherwood <david.sherwood@arm.com>
7613 * doc/md.texi (extract_last_@var{m}): Document.
7614 * optabs.def (extract_last_optab): New optab.
7615 * internal-fn.def (EXTRACT_LAST): New internal function.
7616 * internal-fn.c (cond_unary_direct): New macro.
7617 (expand_cond_unary_optab_fn): Likewise.
7618 (direct_cond_unary_optab_supported_p): Likewise.
7619 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
7620 loops using EXTRACT_LAST.
7621 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
7622 (extract_last_<mode>): ...this optab.
7623 (vec_extract<mode><Vel>): Update accordingly.
7625 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7626 Alan Hayward <alan.hayward@arm.com>
7627 David Sherwood <david.sherwood@arm.com>
7629 * target.def (empty_mask_is_expensive): New hook.
7630 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
7631 * doc/tm.texi: Regenerate.
7632 * targhooks.h (default_empty_mask_is_expensive): Declare.
7633 * targhooks.c (default_empty_mask_is_expensive): New function.
7634 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
7635 if the target says that empty masks are expensive.
7636 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
7638 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
7640 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7641 Alan Hayward <alan.hayward@arm.com>
7642 David Sherwood <david.sherwood@arm.com>
7644 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
7645 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
7646 (vect_use_loop_mask_for_alignment_p): New function.
7647 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
7648 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
7649 niters_skip argument. Make sure that the first niters_skip elements
7650 of the first iteration are inactive.
7651 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
7652 Update call to vect_set_loop_masks_directly.
7653 (get_misalign_in_elems): New function, split out from...
7654 (vect_gen_prolog_loop_niters): ...here.
7655 (vect_update_init_of_dr): Take a code argument that specifies whether
7656 the adjustment should be added or subtracted.
7657 (vect_update_init_of_drs): Likewise.
7658 (vect_prepare_for_masked_peels): New function.
7659 (vect_do_peeling): Skip prologue peeling if we're using a mask
7660 instead. Update call to vect_update_inits_of_drs.
7661 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7663 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
7664 alignment. Do not include the number of peeled iterations in
7665 the minimum threshold in that case.
7666 (vectorizable_induction): Adjust the start value down by
7667 LOOP_VINFO_MASK_SKIP_NITERS iterations.
7668 (vect_transform_loop): Call vect_prepare_for_masked_peels.
7669 Take the number of skipped iterations into account when calculating
7671 * tree-vect-stmts.c (vect_gen_while_not): New function.
7673 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7674 Alan Hayward <alan.hayward@arm.com>
7675 David Sherwood <david.sherwood@arm.com>
7677 * doc/sourcebuild.texi (vect_fully_masked): Document.
7678 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
7680 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
7682 (vect_analyze_loop_2): ...here. Don't check the vectorization
7683 factor against the number of loop iterations if the loop is
7686 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7687 Alan Hayward <alan.hayward@arm.com>
7688 David Sherwood <david.sherwood@arm.com>
7690 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
7691 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
7692 (dump_groups): Update accordingly.
7693 (iv_use::mem_type): New member variable.
7694 (address_p): New function.
7695 (record_use): Add a mem_type argument and initialize the new
7697 (record_group_use): Add a mem_type argument. Use address_p.
7698 Remove obsolete null checks of base_object. Update call to record_use.
7699 (find_interesting_uses_op): Update call to record_group_use.
7700 (find_interesting_uses_cond): Likewise.
7701 (find_interesting_uses_address): Likewise.
7702 (get_mem_type_for_internal_fn): New function.
7703 (find_address_like_use): Likewise.
7704 (find_interesting_uses_stmt): Try find_address_like_use before
7705 calling find_interesting_uses_op.
7706 (addr_offset_valid_p): Use the iv mem_type field as the type
7707 of the addressed memory.
7708 (add_autoinc_candidates): Likewise.
7709 (get_address_cost): Likewise.
7710 (split_small_address_groups_p): Use address_p.
7711 (split_address_groups): Likewise.
7712 (add_iv_candidate_for_use): Likewise.
7713 (autoinc_possible_for_pair): Likewise.
7714 (rewrite_groups): Likewise.
7715 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
7716 (determine_group_iv_cost): Update after split of USE_ADDRESS.
7717 (get_alias_ptr_type_for_ptr_address): New function.
7718 (rewrite_use_address): Rewrite address uses in calls that were
7719 identified by find_address_like_use.
7721 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7722 Alan Hayward <alan.hayward@arm.com>
7723 David Sherwood <david.sherwood@arm.com>
7725 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
7727 * gimple-expr.h (is_gimple_addressable: Likewise.
7728 * gimple-expr.c (is_gimple_address): Likewise.
7729 * internal-fn.c (expand_call_mem_ref): New function.
7730 (expand_mask_load_optab_fn): Use it.
7731 (expand_mask_store_optab_fn): Likewise.
7733 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7734 Alan Hayward <alan.hayward@arm.com>
7735 David Sherwood <david.sherwood@arm.com>
7737 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
7738 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
7739 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
7740 (cond_umax@var{mode}): Document.
7741 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
7742 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
7743 (cond_umin_optab, cond_umax_optab): New optabs.
7744 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
7745 (COND_IOR, COND_XOR): New internal functions.
7746 * internal-fn.h (get_conditional_internal_fn): Declare.
7747 * internal-fn.c (cond_binary_direct): New macro.
7748 (expand_cond_binary_optab_fn): Likewise.
7749 (direct_cond_binary_optab_supported_p): Likewise.
7750 (get_conditional_internal_fn): New function.
7751 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
7752 Cope with reduction statements that are vectorized as calls rather
7754 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
7755 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
7756 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
7757 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
7758 (UNSPEC_COND_EOR): New unspecs.
7759 (optab): Add mappings for them.
7760 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
7761 (sve_int_op, sve_fp_op): New int attributes.
7763 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7764 Alan Hayward <alan.hayward@arm.com>
7765 David Sherwood <david.sherwood@arm.com>
7767 * optabs.def (while_ult_optab): New optab.
7768 * doc/md.texi (while_ult@var{m}@var{n}): Document.
7769 * internal-fn.def (WHILE_ULT): New internal function.
7770 * internal-fn.h (direct_internal_fn_supported_p): New override
7771 that takes two types as argument.
7772 * internal-fn.c (while_direct): New macro.
7773 (expand_while_optab_fn): New function.
7774 (convert_optab_supported_p): Likewise.
7775 (direct_while_optab_supported_p): New macro.
7776 * wide-int.h (wi::udiv_ceil): New function.
7777 * tree-vectorizer.h (rgroup_masks): New structure.
7778 (vec_loop_masks): New typedef.
7779 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
7781 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
7782 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
7783 (vect_max_vf): New function.
7784 (slpeel_make_loop_iterate_ntimes): Delete.
7785 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
7786 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
7787 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
7788 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
7789 internal-fn.h, stor-layout.h and optabs-query.h.
7790 (vect_set_loop_mask): New function.
7791 (add_preheader_seq): Likewise.
7792 (add_header_seq): Likewise.
7793 (interleave_supported_p): Likewise.
7794 (vect_maybe_permute_loop_masks): Likewise.
7795 (vect_set_loop_masks_directly): Likewise.
7796 (vect_set_loop_condition_masked): Likewise.
7797 (vect_set_loop_condition_unmasked): New function, split out from
7798 slpeel_make_loop_iterate_ntimes.
7799 (slpeel_make_loop_iterate_ntimes): Rename to..
7800 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
7801 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
7802 (vect_do_peeling): Update call accordingly.
7803 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
7805 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7806 mask_compare_type, can_fully_mask_p and fully_masked_p.
7807 (release_vec_loop_masks): New function.
7808 (_loop_vec_info): Use it to free the loop masks.
7809 (can_produce_all_loop_masks_p): New function.
7810 (vect_get_max_nscalars_per_iter): Likewise.
7811 (vect_verify_full_masking): Likewise.
7812 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
7813 retries, and free the mask rgroups before retrying. Check loop-wide
7814 reasons for disallowing fully-masked loops. Make the final decision
7815 about whether use a fully-masked loop or not.
7816 (vect_estimate_min_profitable_iters): Do not assume that peeling
7817 for the number of iterations will be needed for fully-masked loops.
7818 (vectorizable_reduction): Disable fully-masked loops.
7819 (vectorizable_live_operation): Likewise.
7820 (vect_halve_mask_nunits): New function.
7821 (vect_double_mask_nunits): Likewise.
7822 (vect_record_loop_mask): Likewise.
7823 (vect_get_loop_mask): Likewise.
7824 (vect_transform_loop): Handle the case in which the final loop
7825 iteration might handle a partial vector. Call vect_set_loop_condition
7826 instead of slpeel_make_loop_iterate_ntimes.
7827 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
7828 (check_load_store_masking): New function.
7829 (prepare_load_store_mask): Likewise.
7830 (vectorizable_store): Handle fully-masked loops.
7831 (vectorizable_load): Likewise.
7832 (supportable_widening_operation): Use vect_halve_mask_nunits for
7834 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
7835 (vect_gen_while): New function.
7836 * config/aarch64/aarch64.md (umax<mode>3): New expander.
7837 (aarch64_uqdec<mode>): New insn.
7839 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7840 Alan Hayward <alan.hayward@arm.com>
7841 David Sherwood <david.sherwood@arm.com>
7843 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
7844 (reduc_xor_scal_optab): New optabs.
7845 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
7846 (reduc_xor_scal_@var{m}): Document.
7847 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
7848 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
7850 * fold-const-call.c (fold_const_call): Handle them.
7851 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
7852 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
7853 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
7854 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
7855 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
7856 (UNSPEC_XORV): New unspecs.
7857 (optab): Add entries for them.
7858 (BITWISEV): New int iterator.
7859 (bit_reduc_op): New int attributes.
7861 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7862 Alan Hayward <alan.hayward@arm.com>
7863 David Sherwood <david.sherwood@arm.com>
7865 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
7866 * internal-fn.def (VEC_SHL_INSERT): New internal function.
7867 * optabs.def (vec_shl_insert_optab): New optab.
7868 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
7869 (duplicate_and_interleave): Likewise.
7870 * tree-vect-loop.c: Include internal-fn.h.
7871 (neutral_op_for_slp_reduction): New function, split out from
7872 get_initial_defs_for_reduction.
7873 (get_initial_def_for_reduction): Handle option 2 for variable-length
7874 vectors by loading the neutral value into a vector and then shifting
7875 the initial value into element 0.
7876 (get_initial_defs_for_reduction): Replace the code argument with
7877 the neutral value calculated by neutral_op_for_slp_reduction.
7878 Use gimple_build_vector for constant-length vectors.
7879 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
7880 but the first group_size elements have a neutral value.
7881 Use duplicate_and_interleave otherwise.
7882 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
7883 Update call to get_initial_defs_for_reduction. Handle SLP
7884 reductions for variable-length vectors by creating one vector
7885 result for each scalar result, with the elements associated
7886 with other scalar results stubbed out with the neutral value.
7887 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
7888 Require IFN_VEC_SHL_INSERT for double reductions on
7889 variable-length vectors, or SLP reductions that have
7890 a neutral value. Require can_duplicate_and_interleave_p
7891 support for variable-length unchained SLP reductions if there
7892 is no neutral value, such as for MIN/MAX reductions. Also require
7893 the number of vector elements to be a multiple of the number of
7894 SLP statements when doing variable-length unchained SLP reductions.
7895 Update call to vect_create_epilog_for_reduction.
7896 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
7897 and remove initial values.
7898 (duplicate_and_interleave): Make public.
7899 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
7900 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
7902 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7903 Alan Hayward <alan.hayward@arm.com>
7904 David Sherwood <david.sherwood@arm.com>
7906 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
7907 (can_duplicate_and_interleave_p): New function.
7908 (vect_get_and_check_slp_defs): Take the vector of statements
7909 rather than just the current one. Remove excess parentheses.
7910 Restriction rejectinon of vect_constant_def and vect_external_def
7911 for variable-length vectors to boolean types, or types for which
7912 can_duplicate_and_interleave_p is false.
7913 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
7914 (duplicate_and_interleave): New function.
7915 (vect_get_constant_vectors): Use gimple_build_vector for
7916 constant-length vectors and suitable variable-length constant
7917 vectors. Use duplicate_and_interleave for other variable-length
7918 vectors. Don't defer the update when inserting new statements.
7920 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7921 Alan Hayward <alan.hayward@arm.com>
7922 David Sherwood <david.sherwood@arm.com>
7924 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
7925 min_profitable_iters doesn't go negative.
7927 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7928 Alan Hayward <alan.hayward@arm.com>
7929 David Sherwood <david.sherwood@arm.com>
7931 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
7932 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
7933 * optabs.def (vec_mask_load_lanes_optab): New optab.
7934 (vec_mask_store_lanes_optab): Likewise.
7935 * internal-fn.def (MASK_LOAD_LANES): New internal function.
7936 (MASK_STORE_LANES): Likewise.
7937 * internal-fn.c (mask_load_lanes_direct): New macro.
7938 (mask_store_lanes_direct): Likewise.
7939 (expand_mask_load_optab_fn): Handle masked operations.
7940 (expand_mask_load_lanes_optab_fn): New macro.
7941 (expand_mask_store_optab_fn): Handle masked operations.
7942 (expand_mask_store_lanes_optab_fn): New macro.
7943 (direct_mask_load_lanes_optab_supported_p): Likewise.
7944 (direct_mask_store_lanes_optab_supported_p): Likewise.
7945 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
7947 (vect_load_lanes_supported): Likewise.
7948 * tree-vect-data-refs.c (strip_conversion): New function.
7949 (can_group_stmts_p): Likewise.
7950 (vect_analyze_data_ref_accesses): Use it instead of checking
7951 for a pair of assignments.
7952 (vect_store_lanes_supported): Take a masked_p parameter.
7953 (vect_load_lanes_supported): Likewise.
7954 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
7955 vect_store_lanes_supported and vect_load_lanes_supported.
7956 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7957 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
7958 parameter. Don't allow gaps for masked accesses.
7959 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
7960 and vect_load_lanes_supported.
7961 (get_load_store_type): Take a masked_p parameter and update
7962 call to get_group_load_store_type.
7963 (vectorizable_store): Update call to get_load_store_type.
7964 Handle IFN_MASK_STORE_LANES.
7965 (vectorizable_load): Update call to get_load_store_type.
7966 Handle IFN_MASK_LOAD_LANES.
7968 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7969 Alan Hayward <alan.hayward@arm.com>
7970 David Sherwood <david.sherwood@arm.com>
7972 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
7974 * config/aarch64/aarch64-protos.h
7975 (aarch64_sve_struct_memory_operand_p): Declare.
7976 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
7977 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
7978 (VPRED, vpred): Handle SVE structure modes.
7979 * config/aarch64/constraints.md (Utx): New constraint.
7980 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
7981 (aarch64_sve_struct_nonimmediate_operand): New predicates.
7982 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
7983 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
7984 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
7985 structure modes. Split into pieces after RA.
7986 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
7987 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
7989 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
7990 SVE structure modes.
7991 (aarch64_classify_address): Likewise.
7992 (sizetochar): Move earlier in file.
7993 (aarch64_print_operand): Handle SVE register lists.
7994 (aarch64_array_mode): New function.
7995 (aarch64_sve_struct_memory_operand_p): Likewise.
7996 (TARGET_ARRAY_MODE): Redefine.
7998 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7999 Alan Hayward <alan.hayward@arm.com>
8000 David Sherwood <david.sherwood@arm.com>
8002 * target.def (array_mode): New target hook.
8003 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
8004 * doc/tm.texi: Regenerate.
8005 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
8006 * hooks.c (hook_optmode_mode_uhwi_none): New function.
8007 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
8009 * stor-layout.c (mode_for_array): Likewise. Support polynomial
8012 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8013 Alan Hayward <alan.hayward@arm.com>
8014 David Sherwood <david.sherwood@arm.com>
8016 * fold-const.c (fold_binary_loc): Check the argument types
8017 rather than the result type when testing for a vector operation.
8019 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8021 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
8022 * doc/tm.texi: Regenerate.
8024 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8025 Alan Hayward <alan.hayward@arm.com>
8026 David Sherwood <david.sherwood@arm.com>
8028 * doc/invoke.texi (-msve-vector-bits=): Document new option.
8029 (sve): Document new AArch64 extension.
8030 * doc/md.texi (w): Extend the description of the AArch64
8031 constraint to include SVE vectors.
8032 (Upl, Upa): Document new AArch64 predicate constraints.
8033 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
8035 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
8036 (msve-vector-bits=): New option.
8037 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
8038 SVE when these are disabled.
8039 (sve): New extension.
8040 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
8041 modes. Adjust their number of units based on aarch64_sve_vg.
8042 (MAX_BITSIZE_MODE_ANY_MODE): Define.
8043 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
8044 aarch64_addr_query_type.
8045 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
8046 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
8047 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
8048 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
8049 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
8050 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
8051 (aarch64_simd_imm_zero_p): Delete.
8052 (aarch64_check_zero_based_sve_index_immediate): Declare.
8053 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8054 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8055 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8056 (aarch64_sve_float_mul_immediate_p): Likewise.
8057 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8059 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
8060 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
8061 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
8062 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
8063 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
8064 (aarch64_regmode_natural_size): Likewise.
8065 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
8066 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
8068 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
8069 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
8070 for VG and the SVE predicate registers.
8071 (V_ALIASES): Add a "z"-prefixed alias.
8072 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
8073 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
8074 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
8075 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
8076 (REG_CLASS_NAMES): Add entries for them.
8077 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
8078 and the predicate registers.
8079 (aarch64_sve_vg): Declare.
8080 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
8081 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
8082 (REGMODE_NATURAL_SIZE): Define.
8083 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
8085 * config/aarch64/aarch64.c: Include cfgrtl.h.
8086 (simd_immediate_info): Add a constructor for series vectors,
8087 and an associated step field.
8088 (aarch64_sve_vg): New variable.
8089 (aarch64_dbx_register_number): Handle VG and the predicate registers.
8090 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
8091 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
8092 (VEC_ANY_DATA, VEC_STRUCT): New constants.
8093 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
8094 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
8095 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
8096 (aarch64_get_mask_mode): New functions.
8097 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
8098 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
8099 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
8100 predicate modes and predicate registers. Explicitly restrict
8101 GPRs to modes of 16 bytes or smaller. Only allow FP registers
8102 to store a vector mode if it is recognized by
8103 aarch64_classify_vector_mode.
8104 (aarch64_regmode_natural_size): New function.
8105 (aarch64_hard_regno_caller_save_mode): Return the original mode
8107 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
8108 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
8109 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
8110 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
8112 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
8113 does not overlap dest if the function is frame-related. Handle
8115 (aarch64_split_add_offset): New function.
8116 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
8117 them aarch64_add_offset.
8118 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
8119 and update call to aarch64_sub_sp.
8120 (aarch64_add_cfa_expression): New function.
8121 (aarch64_expand_prologue): Pass extra temporary registers to the
8122 functions above. Handle the case in which we need to emit new
8123 DW_CFA_expressions for registers that were originally saved
8124 relative to the stack pointer, but now have to be expressed
8125 relative to the frame pointer.
8126 (aarch64_output_mi_thunk): Pass extra temporary registers to the
8128 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
8129 IP0 and IP1 values for SVE frames.
8130 (aarch64_expand_vec_series): New function.
8131 (aarch64_expand_sve_widened_duplicate): Likewise.
8132 (aarch64_expand_sve_const_vector): Likewise.
8133 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
8134 Handle SVE constants. Use emit_move_insn to move a force_const_mem
8135 into the register, rather than emitting a SET directly.
8136 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
8137 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
8138 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
8139 (offset_9bit_signed_scaled_p): New functions.
8140 (aarch64_replicate_bitmask_imm): New function.
8141 (aarch64_bitmask_imm): Use it.
8142 (aarch64_cannot_force_const_mem): Reject expressions involving
8143 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
8144 (aarch64_classify_index): Handle SVE indices, by requiring
8145 a plain register index with a scale that matches the element size.
8146 (aarch64_classify_address): Handle SVE addresses. Assert that
8147 the mode of the address is VOIDmode or an integer mode.
8148 Update call to aarch64_classify_symbol.
8149 (aarch64_classify_symbolic_expression): Update call to
8150 aarch64_classify_symbol.
8151 (aarch64_const_vec_all_in_range_p): New function.
8152 (aarch64_print_vector_float_operand): Likewise.
8153 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
8154 "vN" for FP registers with SVE modes. Handle (const ...) vectors
8155 and the FP immediates 1.0 and 0.5.
8156 (aarch64_print_address_internal): Handle SVE addresses.
8157 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
8158 (aarch64_regno_regclass): Handle predicate registers.
8159 (aarch64_secondary_reload): Handle big-endian reloads of SVE
8161 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
8162 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
8163 (aarch64_convert_sve_vector_bits): New function.
8164 (aarch64_override_options): Use it to handle -msve-vector-bits=.
8165 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8167 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
8168 Handle SVE vector and predicate modes. Accept VL-based constants
8169 that need only one temporary register, and VL offsets that require
8170 no temporary registers.
8171 (aarch64_conditional_register_usage): Mark the predicate registers
8172 as fixed if SVE isn't available.
8173 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
8174 Return true for SVE vector and predicate modes.
8175 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
8176 rather than an unsigned int. Handle SVE modes.
8177 (aarch64_preferred_simd_mode): Update call accordingly. Handle
8179 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
8181 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8182 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8183 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8184 (aarch64_sve_float_mul_immediate_p): New functions.
8185 (aarch64_sve_valid_immediate): New function.
8186 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
8187 Explicitly reject structure modes. Check for INDEX constants.
8188 Handle PTRUE and PFALSE constants.
8189 (aarch64_check_zero_based_sve_index_immediate): New function.
8190 (aarch64_simd_imm_zero_p): Delete.
8191 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
8192 vector modes. Accept constants in the range of CNT[BHWD].
8193 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
8194 ask for an Advanced SIMD mode.
8195 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
8196 (aarch64_simd_vector_alignment): Handle SVE predicates.
8197 (aarch64_vectorize_preferred_vector_alignment): New function.
8198 (aarch64_simd_vector_alignment_reachable): Use it instead of
8200 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
8201 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
8203 (MAX_VECT_LEN): Delete.
8204 (expand_vec_perm_d): Add a vec_flags field.
8205 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
8206 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
8207 (aarch64_evpc_ext): Don't apply a big-endian lane correction
8209 (aarch64_evpc_rev): Rename to...
8210 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
8211 (aarch64_evpc_rev_global): New function.
8212 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
8213 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
8215 (aarch64_evpc_sve_tbl): New function.
8216 (aarch64_expand_vec_perm_const_1): Update after rename of
8217 aarch64_evpc_rev. Handle SVE permutes too, trying
8218 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
8219 than aarch64_evpc_tbl.
8220 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
8221 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
8222 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
8223 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
8224 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
8225 (aarch64_expand_sve_vcond): New functions.
8226 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
8227 of aarch64_vector_mode_p.
8228 (aarch64_dwarf_poly_indeterminate_value): New function.
8229 (aarch64_compute_pressure_classes): Likewise.
8230 (aarch64_can_change_mode_class): Likewise.
8231 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
8232 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
8233 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
8234 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
8235 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
8236 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
8237 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
8238 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
8240 (Dn, Dl, Dr): Accept const as well as const_vector.
8241 (Dz): Likewise. Compare against CONST0_RTX.
8242 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
8243 of "vector" where appropriate.
8244 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
8245 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
8246 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
8247 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
8248 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
8249 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
8250 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
8251 (v_int_equiv): Extend to SVE modes.
8252 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
8254 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
8255 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
8256 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
8257 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
8258 (SVE_COND_FP_CMP): New int iterators.
8259 (perm_hilo): Handle the new unpack unspecs.
8260 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
8262 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
8263 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
8264 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
8265 (aarch64_equality_operator, aarch64_constant_vector_operand)
8266 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
8267 (aarch64_sve_nonimmediate_operand): Likewise.
8268 (aarch64_sve_general_operand): Likewise.
8269 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
8270 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
8271 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
8272 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
8273 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
8274 (aarch64_sve_float_arith_immediate): Likewise.
8275 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
8276 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
8277 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
8278 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
8279 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
8280 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
8281 (aarch64_sve_float_arith_operand): Likewise.
8282 (aarch64_sve_float_arith_with_sub_operand): Likewise.
8283 (aarch64_sve_float_mul_operand): Likewise.
8284 (aarch64_sve_vec_perm_operand): Likewise.
8285 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
8286 (aarch64_mov_operand): Accept const_poly_int and const_vector.
8287 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
8288 as well as const_vector.
8289 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
8290 in file. Use CONST0_RTX and CONSTM1_RTX.
8291 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
8292 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
8293 Use aarch64_simd_imm_zero.
8294 * config/aarch64/aarch64-sve.md: New file.
8295 * config/aarch64/aarch64.md: Include it.
8296 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
8297 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
8298 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
8299 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
8300 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
8301 (sve): New attribute.
8302 (enabled): Disable instructions with the sve attribute unless
8304 (movqi, movhi): Pass CONST_POLY_INT operaneds through
8305 aarch64_expand_mov_immediate.
8306 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
8307 CNT[BHSD] immediates.
8308 (movti): Split CONST_POLY_INT moves into two halves.
8309 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
8310 Split additions that need a temporary here if the destination
8311 is the stack pointer.
8312 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
8313 (*add<mode>3_poly_1): New instruction.
8314 (set_clobber_cc): New expander.
8316 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8318 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
8319 parameter and use it instead of GET_MODE_SIZE (innermode). Use
8320 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
8321 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
8322 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
8323 Change innermode from fixed_mode_size to machine_mode.
8324 (simplify_subreg): Update call accordingly. Handle a constant-sized
8325 subreg of a variable-length CONST_VECTOR.
8327 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8328 Alan Hayward <alan.hayward@arm.com>
8329 David Sherwood <david.sherwood@arm.com>
8331 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
8332 (add_offset_to_base): New function, split out from...
8333 (create_mem_ref): ...here. When handling a scale other than 1,
8334 check first whether the address is valid without the offset.
8335 Add it into the base if so, leaving the index and scale as-is.
8337 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8340 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
8341 fold_for_warn before checking if arg2 is INTEGER_CST.
8343 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
8345 * config/rs6000/predicates.md (load_multiple_operation): Delete.
8346 (store_multiple_operation): Delete.
8347 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
8348 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
8349 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
8350 guarded by TARGET_STRING.
8351 (rs6000_output_load_multiple): Delete.
8352 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
8353 OPTION_MASK_STRING / TARGET_STRING handling.
8354 (print_operand) <'N', 'O'>: Add comment that these are unused now.
8355 (const rs6000_opt_masks) <"string">: Change mask to 0.
8356 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
8357 (MASK_STRING): Delete.
8358 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
8360 (load_multiple): Delete.
8367 (store_multiple): Delete.
8374 (movmemsi_8reg): Delete.
8375 (corresponding unnamed define_insn): Delete.
8376 (movmemsi_6reg): Delete.
8377 (corresponding unnamed define_insn): Delete.
8378 (movmemsi_4reg): Delete.
8379 (corresponding unnamed define_insn): Delete.
8380 (movmemsi_2reg): Delete.
8381 (corresponding unnamed define_insn): Delete.
8382 (movmemsi_1reg): Delete.
8383 (corresponding unnamed define_insn): Delete.
8384 * config/rs6000/rs6000.opt (mno-string): New.
8385 (mstring): Replace by deprecation warning stub.
8386 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
8388 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8390 * regrename.c (regrename_do_replace): If replacing the same
8391 reg multiple times, try to reuse last created gen_raw_REG.
8394 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
8395 main to workaround a bug in GDB.
8397 2018-01-12 Tom de Vries <tom@codesourcery.com>
8400 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
8402 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
8404 PR rtl-optimization/80481
8405 * ira-color.c (get_cap_member): New function.
8406 (allocnos_conflict_by_live_ranges_p): Use it.
8407 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
8408 (setup_slot_coalesced_allocno_live_ranges): Ditto.
8410 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
8413 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
8414 (*saddl_se_1): Ditto.
8416 (*ssubl_se_1): Ditto.
8418 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8420 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
8421 rather than wi::to_widest for DR_INITs.
8422 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
8423 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
8424 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
8426 (vect_analyze_group_access_1): Note that here.
8428 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8430 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
8431 polynomial type sizes.
8433 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8435 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
8436 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
8437 (gimple_add_tmp_var): Likewise.
8439 2018-01-12 Martin Liska <mliska@suse.cz>
8441 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
8442 (gimple_alloc_sizes): Likewise.
8443 (dump_gimple_statistics): Use PRIu64 in printf format.
8444 * gimple.h: Change uint64_t to int.
8446 2018-01-12 Martin Liska <mliska@suse.cz>
8448 * tree-core.h: Use uint64_t instead of int.
8449 * tree.c (tree_node_counts): Likewise.
8450 (tree_node_sizes): Likewise.
8451 (dump_tree_statistics): Use PRIu64 in printf format.
8453 2018-01-12 Martin Liska <mliska@suse.cz>
8455 * Makefile.in: As qsort_chk is implemented in vec.c, add
8456 vec.o to linkage of gencfn-macros.
8457 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
8458 passing the info to record_node_allocation_statistics.
8459 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
8461 * ggc-common.c (struct ggc_usage): Add operator== and use
8462 it in operator< and compare function.
8463 * mem-stats.h (struct mem_usage): Likewise.
8464 * vec.c (struct vec_usage): Remove operator< and compare
8465 function. Can be simply inherited.
8467 2018-01-12 Martin Jambor <mjambor@suse.cz>
8470 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
8471 * tree-ssa-math-opts.c: Include domwalk.h.
8472 (convert_mult_to_fma_1): New function.
8473 (fma_transformation_info): New type.
8474 (fma_deferring_state): Likewise.
8475 (cancel_fma_deferring): New function.
8476 (result_of_phi): Likewise.
8477 (last_fma_candidate_feeds_initial_phi): Likewise.
8478 (convert_mult_to_fma): Added deferring logic, split actual
8479 transformation to convert_mult_to_fma_1.
8480 (math_opts_dom_walker): New type.
8481 (math_opts_dom_walker::after_dom_children): New method, body moved
8482 here from pass_optimize_widening_mul::execute, added deferring logic
8484 (pass_optimize_widening_mul::execute): Moved most of code to
8485 math_opts_dom_walker::after_dom_children.
8486 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
8487 * config/i386/i386.c (ix86_option_override_internal): Added
8488 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
8490 2018-01-12 Richard Biener <rguenther@suse.de>
8493 * dwarf2out.c (gen_variable_die): Do not reset old_die for
8494 inline instance vars.
8496 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
8499 * config/rx/rx.c (rx_is_restricted_memory_address):
8502 2018-01-12 Richard Biener <rguenther@suse.de>
8504 PR tree-optimization/80846
8505 * target.def (split_reduction): New target hook.
8506 * targhooks.c (default_split_reduction): New function.
8507 * targhooks.h (default_split_reduction): Declare.
8508 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
8509 target requests first reduce vectors by combining low and high
8511 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
8512 (get_vectype_for_scalar_type_and_size): Export.
8513 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
8514 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
8515 * doc/tm.texi: Regenerate.
8516 * config/i386/i386.c (ix86_split_reduction): Implement
8517 TARGET_VECTORIZE_SPLIT_REDUCTION.
8519 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8522 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
8523 in PIC mode except for TARGET_VXWORKS_RTP.
8524 * config/sparc/sparc.c: Include cfgrtl.h.
8525 (TARGET_INIT_PIC_REG): Define.
8526 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
8527 (sparc_pic_register_p): New predicate.
8528 (sparc_legitimate_address_p): Use it.
8529 (sparc_legitimize_pic_address): Likewise.
8530 (sparc_delegitimize_address): Likewise.
8531 (sparc_mode_dependent_address_p): Likewise.
8532 (gen_load_pcrel_sym): Remove 4th parameter.
8533 (load_got_register): Adjust call to above. Remove obsolete stuff.
8534 (sparc_expand_prologue): Do not call load_got_register here.
8535 (sparc_flat_expand_prologue): Likewise.
8536 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
8537 (sparc_use_pseudo_pic_reg): New function.
8538 (sparc_init_pic_reg): Likewise.
8539 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8540 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
8542 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
8544 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
8545 Add item for branch_cost.
8547 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8549 PR rtl-optimization/83565
8550 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
8551 not extend the result to a larger mode for rotate operations.
8552 (num_sign_bit_copies1): Likewise.
8554 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8557 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
8559 Use values-Xc.o for -pedantic.
8560 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
8562 2018-01-12 Martin Liska <mliska@suse.cz>
8565 * ipa-devirt.c (final_warning_record::grow_type_warnings):
8567 (possible_polymorphic_call_targets): Use it.
8568 (ipa_devirt): Likewise.
8570 2018-01-12 Martin Liska <mliska@suse.cz>
8572 * profile-count.h (enum profile_quality): Use 0 as invalid
8573 enum value of profile_quality.
8575 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
8577 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
8578 -mext-string options.
8580 2018-01-12 Richard Biener <rguenther@suse.de>
8582 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
8583 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
8584 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
8586 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
8588 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
8590 * configure.ac (--with-long-double-format): Add support for the
8591 configuration option to change the default long double format on
8593 * config.gcc (powerpc*-linux*-*): Likewise.
8594 * configure: Regenerate.
8595 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
8596 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
8597 used without modification.
8599 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8601 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
8602 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
8603 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
8604 MISC_BUILTIN_SPEC_BARRIER.
8605 (rs6000_init_builtins): Likewise.
8606 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
8608 (speculation_barrier): New define_insn.
8609 * doc/extend.texi: Document __builtin_speculation_barrier.
8611 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8614 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
8615 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
8616 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
8618 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
8619 integral modes instead of "ss" and "sd".
8620 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
8621 vectors with 32-bit and 64-bit elements.
8622 (vecdupssescalarmodesuffix): New mode attribute.
8623 (vec_dup<mode>): Use it.
8625 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
8628 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
8629 frame if argument is passed on stack.
8631 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8634 * ree.c (combine_reaching_defs): Optimize also
8635 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
8636 reg2=any_extend(exp); reg1=reg2;, formatting fix.
8638 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8641 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
8643 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8646 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
8647 after they are computed.
8649 2018-01-11 Bin Cheng <bin.cheng@arm.com>
8651 PR tree-optimization/83695
8652 * gimple-loop-linterchange.cc
8653 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
8654 reset cached scev information after interchange.
8655 (pass_linterchange::execute): Remove call to scev_reset_htab.
8657 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8659 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
8660 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
8661 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
8662 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
8663 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
8664 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
8665 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
8666 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
8667 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
8668 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
8669 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
8670 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
8671 (V_lane_reg): Likewise.
8672 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
8674 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
8675 (vfmal_lane_low<mode>_intrinsic,
8676 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
8677 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
8678 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
8679 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
8680 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
8681 vfmsl_lane_high<mode>_intrinsic): New define_insns.
8683 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8685 * config/arm/arm-cpus.in (fp16fml): New feature.
8686 (ALL_SIMD): Add fp16fml.
8687 (armv8.2-a): Add fp16fml as an option.
8688 (armv8.3-a): Likewise.
8689 (armv8.4-a): Add fp16fml as part of fp16.
8690 * config/arm/arm.h (TARGET_FP16FML): Define.
8691 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
8693 * config/arm/arm-modes.def (V2HF): Define.
8694 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
8695 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
8696 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
8697 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
8698 vfmsl_low, vfmsl_high): New set of builtins.
8699 * config/arm/iterators.md (PLUSMINUS): New code iterator.
8700 (vfml_op): New code attribute.
8701 (VFMLHALVES): New int iterator.
8702 (VFML, VFMLSEL): New mode attributes.
8703 (V_reg): Define mapping for V2HF.
8704 (V_hi, V_lo): New mode attributes.
8705 (VF_constraint): Likewise.
8706 (vfml_half, vfml_half_selector): New int attributes.
8707 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
8709 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
8710 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
8712 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
8713 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
8714 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
8715 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
8717 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
8718 Document new effective target and option set.
8720 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8722 * config/arm/arm-cpus.in (armv8_4): New feature.
8723 (ARMv8_4a): New fgroup.
8724 (armv8.4-a): New arch.
8725 * config/arm/arm-tables.opt: Regenerate.
8726 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
8727 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
8728 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
8729 Add matching rules for -march=armv8.4-a and extensions.
8730 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
8732 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
8735 * config/rx/rx.md (BW): New mode attribute.
8736 (sync_lock_test_and_setsi): Add mode suffix to insn output.
8738 2018-01-11 Richard Biener <rguenther@suse.de>
8740 PR tree-optimization/83435
8741 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
8742 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
8743 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
8745 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8746 Alan Hayward <alan.hayward@arm.com>
8747 David Sherwood <david.sherwood@arm.com>
8749 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
8751 (aarch64_classify_address): Initialize it. Track polynomial offsets.
8752 (aarch64_print_address_internal): Use it to check for a zero offset.
8754 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8755 Alan Hayward <alan.hayward@arm.com>
8756 David Sherwood <david.sherwood@arm.com>
8758 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
8759 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
8760 Return a poly_int64 rather than a HOST_WIDE_INT.
8761 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
8762 rather than a HOST_WIDE_INT.
8763 * config/aarch64/aarch64.h (aarch64_frame): Protect with
8764 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
8765 hard_fp_offset, frame_size, initial_adjust, callee_offset and
8766 final_offset from HOST_WIDE_INT to poly_int64.
8767 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
8768 to_constant when getting the number of units in an Advanced SIMD
8770 (aarch64_builtin_vectorized_function): Check for a constant number
8772 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
8774 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
8775 attribute instead of GET_MODE_NUNITS.
8776 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
8777 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
8778 GET_MODE_SIZE for fixed-size registers.
8779 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
8780 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
8781 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
8782 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
8783 (aarch64_print_operand, aarch64_print_address_internal)
8784 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
8785 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
8786 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
8787 Handle polynomial GET_MODE_SIZE.
8788 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
8789 wider than SImode without modification.
8790 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
8791 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
8792 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
8793 passing and returning SVE modes.
8794 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
8795 rather than GEN_INT.
8796 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
8797 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
8798 (aarch64_allocate_and_probe_stack_space): Likewise.
8799 (aarch64_layout_frame): Cope with polynomial offsets.
8800 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
8801 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
8803 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
8804 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
8805 poly_int64 rather than a HOST_WIDE_INT.
8806 (aarch64_get_separate_components, aarch64_process_components)
8807 (aarch64_expand_prologue, aarch64_expand_epilogue)
8808 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
8809 (aarch64_anchor_offset): New function, split out from...
8810 (aarch64_legitimize_address): ...here.
8811 (aarch64_builtin_vectorization_cost): Handle polynomial
8812 TYPE_VECTOR_SUBPARTS.
8813 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
8815 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
8816 number of elements from the PARALLEL rather than the mode.
8817 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
8818 rather than GET_MODE_BITSIZE.
8819 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
8820 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
8821 (aarch64_expand_vec_perm_const_1): Handle polynomial
8822 d->perm.length () and d->perm elements.
8823 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
8824 Apply to_constant to d->perm elements.
8825 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
8826 polynomial CONST_VECTOR_NUNITS.
8827 (aarch64_move_pointer): Take amount as a poly_int64 rather
8829 (aarch64_progress_pointer): Avoid temporary variable.
8830 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
8831 the mode attribute instead of GET_MODE.
8833 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8834 Alan Hayward <alan.hayward@arm.com>
8835 David Sherwood <david.sherwood@arm.com>
8837 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
8838 x exists before using it.
8839 (aarch64_add_constant_internal): Rename to...
8840 (aarch64_add_offset_1): ...this. Replace regnum with separate
8841 src and dest rtxes. Handle the case in which they're different,
8842 including when the offset is zero. Replace scratchreg with an rtx.
8843 Use 2 additions if there is no spare register into which we can
8844 move a 16-bit constant.
8845 (aarch64_add_constant): Delete.
8846 (aarch64_add_offset): Replace reg with separate src and dest
8847 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
8848 Use aarch64_add_offset_1.
8849 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
8850 an rtx rather than an int. Take the delta as a poly_int64
8851 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
8852 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
8853 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
8854 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
8855 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
8857 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
8858 aarch64_add_constant.
8860 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8862 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
8863 Use scalar_float_mode.
8865 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8867 * config/aarch64/aarch64-simd.md
8868 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
8869 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
8870 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
8871 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
8872 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
8873 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
8874 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
8875 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
8876 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
8877 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
8879 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8882 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
8883 targ_options->x_arm_arch_string is non NULL.
8885 2018-01-11 Tamar Christina <tamar.christina@arm.com>
8887 * config/aarch64/aarch64.h
8888 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
8890 2018-01-11 Sudakshina Das <sudi.das@arm.com>
8893 * expmed.c (emit_store_flag_force): Swap if const op0
8894 and change VOIDmode to mode of op0.
8896 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8898 PR rtl-optimization/83761
8899 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
8900 than bytes to mode_for_size.
8902 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8905 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
8906 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
8909 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8912 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
8913 when in layout mode.
8914 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
8915 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
8918 2018-01-10 Michael Collison <michael.collison@arm.com>
8920 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
8921 * config/aarch64/aarch64-option-extension.def: Add
8922 AARCH64_OPT_EXTENSION of 'fp16fml'.
8923 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8924 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
8925 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
8926 * config/aarch64/constraints.md (Ui7): New constraint.
8927 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
8928 (VFMLA_SEL_W): Ditto.
8931 (VFMLA16_LOW): New int iterator.
8932 (VFMLA16_HIGH): Ditto.
8933 (UNSPEC_FMLAL): New unspec.
8934 (UNSPEC_FMLSL): Ditto.
8935 (UNSPEC_FMLAL2): Ditto.
8936 (UNSPEC_FMLSL2): Ditto.
8937 (f16mac): New code attribute.
8938 * config/aarch64/aarch64-simd-builtins.def
8939 (aarch64_fmlal_lowv2sf): Ditto.
8940 (aarch64_fmlsl_lowv2sf): Ditto.
8941 (aarch64_fmlalq_lowv4sf): Ditto.
8942 (aarch64_fmlslq_lowv4sf): Ditto.
8943 (aarch64_fmlal_highv2sf): Ditto.
8944 (aarch64_fmlsl_highv2sf): Ditto.
8945 (aarch64_fmlalq_highv4sf): Ditto.
8946 (aarch64_fmlslq_highv4sf): Ditto.
8947 (aarch64_fmlal_lane_lowv2sf): Ditto.
8948 (aarch64_fmlsl_lane_lowv2sf): Ditto.
8949 (aarch64_fmlal_laneq_lowv2sf): Ditto.
8950 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
8951 (aarch64_fmlalq_lane_lowv4sf): Ditto.
8952 (aarch64_fmlsl_lane_lowv4sf): Ditto.
8953 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
8954 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
8955 (aarch64_fmlal_lane_highv2sf): Ditto.
8956 (aarch64_fmlsl_lane_highv2sf): Ditto.
8957 (aarch64_fmlal_laneq_highv2sf): Ditto.
8958 (aarch64_fmlsl_laneq_highv2sf): Ditto.
8959 (aarch64_fmlalq_lane_highv4sf): Ditto.
8960 (aarch64_fmlsl_lane_highv4sf): Ditto.
8961 (aarch64_fmlalq_laneq_highv4sf): Ditto.
8962 (aarch64_fmlsl_laneq_highv4sf): Ditto.
8963 * config/aarch64/aarch64-simd.md:
8964 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
8965 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8966 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
8967 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8968 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
8969 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
8970 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
8971 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
8972 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
8973 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
8974 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
8975 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
8976 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
8977 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
8978 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
8979 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
8980 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
8981 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
8982 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
8983 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
8984 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
8985 (vfmlsl_low_u32): Ditto.
8986 (vfmlalq_low_u32): Ditto.
8987 (vfmlslq_low_u32): Ditto.
8988 (vfmlal_high_u32): Ditto.
8989 (vfmlsl_high_u32): Ditto.
8990 (vfmlalq_high_u32): Ditto.
8991 (vfmlslq_high_u32): Ditto.
8992 (vfmlal_lane_low_u32): Ditto.
8993 (vfmlsl_lane_low_u32): Ditto.
8994 (vfmlal_laneq_low_u32): Ditto.
8995 (vfmlsl_laneq_low_u32): Ditto.
8996 (vfmlalq_lane_low_u32): Ditto.
8997 (vfmlslq_lane_low_u32): Ditto.
8998 (vfmlalq_laneq_low_u32): Ditto.
8999 (vfmlslq_laneq_low_u32): Ditto.
9000 (vfmlal_lane_high_u32): Ditto.
9001 (vfmlsl_lane_high_u32): Ditto.
9002 (vfmlal_laneq_high_u32): Ditto.
9003 (vfmlsl_laneq_high_u32): Ditto.
9004 (vfmlalq_lane_high_u32): Ditto.
9005 (vfmlslq_lane_high_u32): Ditto.
9006 (vfmlalq_laneq_high_u32): Ditto.
9007 (vfmlslq_laneq_high_u32): Ditto.
9008 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
9009 (AARCH64_FL_FOR_ARCH8_4): New.
9010 (AARCH64_ISA_F16FML): New ISA flag.
9011 (TARGET_F16FML): New feature flag for fp16fml.
9012 (doc/invoke.texi): Document new fp16fml option.
9014 2018-01-10 Michael Collison <michael.collison@arm.com>
9016 * config/aarch64/aarch64-builtins.c:
9017 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
9018 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9019 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
9020 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
9021 (AARCH64_ISA_SHA3): New ISA flag.
9022 (TARGET_SHA3): New feature flag for sha3.
9023 * config/aarch64/iterators.md (sha512_op): New int attribute.
9024 (CRYPTO_SHA512): New int iterator.
9025 (UNSPEC_SHA512H): New unspec.
9026 (UNSPEC_SHA512H2): Ditto.
9027 (UNSPEC_SHA512SU0): Ditto.
9028 (UNSPEC_SHA512SU1): Ditto.
9029 * config/aarch64/aarch64-simd-builtins.def
9030 (aarch64_crypto_sha512hqv2di): New builtin.
9031 (aarch64_crypto_sha512h2qv2di): Ditto.
9032 (aarch64_crypto_sha512su0qv2di): Ditto.
9033 (aarch64_crypto_sha512su1qv2di): Ditto.
9034 (aarch64_eor3qv8hi): Ditto.
9035 (aarch64_rax1qv2di): Ditto.
9036 (aarch64_xarqv2di): Ditto.
9037 (aarch64_bcaxqv8hi): Ditto.
9038 * config/aarch64/aarch64-simd.md:
9039 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
9040 (aarch64_crypto_sha512su0qv2di): Ditto.
9041 (aarch64_crypto_sha512su1qv2di): Ditto.
9042 (aarch64_eor3qv8hi): Ditto.
9043 (aarch64_rax1qv2di): Ditto.
9044 (aarch64_xarqv2di): Ditto.
9045 (aarch64_bcaxqv8hi): Ditto.
9046 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
9047 (vsha512h2q_u64): Ditto.
9048 (vsha512su0q_u64): Ditto.
9049 (vsha512su1q_u64): Ditto.
9050 (veor3q_u16): Ditto.
9051 (vrax1q_u64): Ditto.
9053 (vbcaxq_u16): Ditto.
9054 * config/arm/types.md (crypto_sha512): New type attribute.
9055 (crypto_sha3): Ditto.
9056 (doc/invoke.texi): Document new sha3 option.
9058 2018-01-10 Michael Collison <michael.collison@arm.com>
9060 * config/aarch64/aarch64-builtins.c:
9061 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
9062 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9063 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
9064 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
9065 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
9066 (AARCH64_ISA_SM4): New ISA flag.
9067 (TARGET_SM4): New feature flag for sm4.
9068 * config/aarch64/aarch64-simd-builtins.def
9069 (aarch64_sm3ss1qv4si): Ditto.
9070 (aarch64_sm3tt1aq4si): Ditto.
9071 (aarch64_sm3tt1bq4si): Ditto.
9072 (aarch64_sm3tt2aq4si): Ditto.
9073 (aarch64_sm3tt2bq4si): Ditto.
9074 (aarch64_sm3partw1qv4si): Ditto.
9075 (aarch64_sm3partw2qv4si): Ditto.
9076 (aarch64_sm4eqv4si): Ditto.
9077 (aarch64_sm4ekeyqv4si): Ditto.
9078 * config/aarch64/aarch64-simd.md:
9079 (aarch64_sm3ss1qv4si): Ditto.
9080 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
9081 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
9082 (aarch64_sm4eqv4si): Ditto.
9083 (aarch64_sm4ekeyqv4si): Ditto.
9084 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
9085 (sm3part_op): Ditto.
9086 (CRYPTO_SM3TT): Ditto.
9087 (CRYPTO_SM3PART): Ditto.
9088 (UNSPEC_SM3SS1): New unspec.
9089 (UNSPEC_SM3TT1A): Ditto.
9090 (UNSPEC_SM3TT1B): Ditto.
9091 (UNSPEC_SM3TT2A): Ditto.
9092 (UNSPEC_SM3TT2B): Ditto.
9093 (UNSPEC_SM3PARTW1): Ditto.
9094 (UNSPEC_SM3PARTW2): Ditto.
9095 (UNSPEC_SM4E): Ditto.
9096 (UNSPEC_SM4EKEY): Ditto.
9097 * config/aarch64/constraints.md (Ui2): New constraint.
9098 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
9099 * config/arm/types.md (crypto_sm3): New type attribute.
9100 (crypto_sm4): Ditto.
9101 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
9102 (vsm3tt1aq_u32): Ditto.
9103 (vsm3tt1bq_u32): Ditto.
9104 (vsm3tt2aq_u32): Ditto.
9105 (vsm3tt2bq_u32): Ditto.
9106 (vsm3partw1q_u32): Ditto.
9107 (vsm3partw2q_u32): Ditto.
9108 (vsm4eq_u32): Ditto.
9109 (vsm4ekeyq_u32): Ditto.
9110 (doc/invoke.texi): Document new sm4 option.
9112 2018-01-10 Michael Collison <michael.collison@arm.com>
9114 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
9115 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
9116 (AARCH64_FL_FOR_ARCH8_4): New.
9117 (AARCH64_FL_V8_4): New flag.
9118 (doc/invoke.texi): Document new armv8.4-a option.
9120 2018-01-10 Michael Collison <michael.collison@arm.com>
9122 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9123 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
9124 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
9125 * config/aarch64/aarch64-option-extension.def: Add
9126 AARCH64_OPT_EXTENSION of 'sha2'.
9127 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
9128 (crypto): Disable sha2 and aes if crypto disabled.
9129 (crypto): Enable aes and sha2 if enabled.
9130 (simd): Disable sha2 and aes if simd disabled.
9131 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
9133 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
9134 (TARGET_SHA2): New feature flag for sha2.
9135 (TARGET_AES): New feature flag for aes.
9136 * config/aarch64/aarch64-simd.md:
9137 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
9138 conditional on TARGET_AES.
9139 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
9140 (aarch64_crypto_sha1hsi): Make pattern conditional
9142 (aarch64_crypto_sha1hv4si): Ditto.
9143 (aarch64_be_crypto_sha1hv4si): Ditto.
9144 (aarch64_crypto_sha1su1v4si): Ditto.
9145 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
9146 (aarch64_crypto_sha1su0v4si): Ditto.
9147 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
9148 (aarch64_crypto_sha256su0v4si): Ditto.
9149 (aarch64_crypto_sha256su1v4si): Ditto.
9150 (doc/invoke.texi): Document new aes and sha2 options.
9152 2018-01-10 Martin Sebor <msebor@redhat.com>
9154 PR tree-optimization/83781
9155 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
9158 2018-01-11 Martin Sebor <msebor@gmail.com>
9159 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9161 PR tree-optimization/83501
9162 PR tree-optimization/81703
9164 * tree-ssa-strlen.c (get_string_cst): Rename...
9165 (get_string_len): ...to this. Handle global constants.
9166 (handle_char_store): Adjust.
9168 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
9169 Jim Wilson <jimw@sifive.com>
9171 * config/riscv/riscv-protos.h (riscv_output_return): New.
9172 * config/riscv/riscv.c (struct machine_function): New naked_p field.
9173 (riscv_attribute_table, riscv_output_return),
9174 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
9175 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
9176 (riscv_compute_frame_info): Only compute frame->mask if not a naked
9178 (riscv_expand_prologue): Add early return for naked function.
9179 (riscv_expand_epilogue): Likewise.
9180 (riscv_function_ok_for_sibcall): Return false for naked function.
9181 (riscv_set_current_function): New.
9182 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
9183 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
9184 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
9185 * doc/extend.texi (RISC-V Function Attributes): New.
9187 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
9189 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
9190 check for 128-bit long double before checking TCmode.
9191 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
9192 128-bit long doubles before checking TFmode or TCmode.
9193 (FLOAT128_IBM_P): Likewise.
9195 2018-01-10 Martin Sebor <msebor@redhat.com>
9197 PR tree-optimization/83671
9198 * builtins.c (c_strlen): Unconditionally return zero for the empty
9200 Use -Warray-bounds for warnings.
9201 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
9202 for non-constant array indices with COMPONENT_REF, arrays of
9203 arrays, and pointers to arrays.
9204 (gimple_fold_builtin_strlen): Determine and set length range for
9205 non-constant character arrays.
9207 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
9210 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
9213 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
9215 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
9217 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9220 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
9221 VECTOR_MEM_ALTIVEC_OR_VSX_P.
9222 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
9223 indexed_or_indirect_operand predicate.
9224 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
9225 (*vsx_le_perm_load_v8hi): Likewise.
9226 (*vsx_le_perm_load_v16qi): Likewise.
9227 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
9228 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
9229 (*vsx_le_perm_store_v8hi): Likewise.
9230 (*vsx_le_perm_store_v16qi): Likewise.
9231 (eight unnamed splitters): Likewise.
9233 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9235 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
9236 * config/rs6000/emmintrin.h: Likewise.
9237 * config/rs6000/mmintrin.h: Likewise.
9238 * config/rs6000/xmmintrin.h: Likewise.
9240 2018-01-10 David Malcolm <dmalcolm@redhat.com>
9243 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
9245 * tree.c (tree_nop_conversion): Return true for location wrapper
9247 (maybe_wrap_with_location): New function.
9248 (selftest::check_strip_nops): New function.
9249 (selftest::test_location_wrappers): New function.
9250 (selftest::tree_c_tests): Call it.
9251 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
9252 (maybe_wrap_with_location): New decl.
9253 (EXPR_LOCATION_WRAPPER_P): New macro.
9254 (location_wrapper_p): New inline function.
9255 (tree_strip_any_location_wrapper): New inline function.
9257 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
9260 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
9261 stack_realign_offset for the largest alignment of stack slot
9263 (ix86_find_max_used_stack_alignment): New function.
9264 (ix86_finalize_stack_frame_flags): Use it. Set
9265 max_used_stack_alignment if we don't realign stack.
9266 * config/i386/i386.h (machine_function): Add
9267 max_used_stack_alignment.
9269 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
9271 * config/arm/arm.opt (-mbranch-cost): New option.
9272 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
9275 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
9278 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
9279 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
9281 2018-01-10 Richard Biener <rguenther@suse.de>
9284 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
9285 early out so it also covers the case where we have a non-NULL
9288 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9290 PR tree-optimization/83753
9291 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
9292 for non-strided grouped accesses if the number of elements is 1.
9294 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9297 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
9298 * i386.h (TARGET_USE_GATHER): Define.
9299 * x86-tune.def (X86_TUNE_USE_GATHER): New.
9301 2018-01-10 Martin Liska <mliska@suse.cz>
9304 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
9305 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
9307 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
9308 CLEANUP_NO_PARTITIONING is not set.
9310 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9312 * doc/rtl.texi: Remove documentation of (const ...) wrappers
9313 for vectors, as a partial revert of r254296.
9314 * rtl.h (const_vec_p): Delete.
9315 (const_vec_duplicate_p): Don't test for vector CONSTs.
9316 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
9317 * expmed.c (make_tree): Likewise.
9320 * common.md (E, F): Use CONSTANT_P instead of checking for
9322 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9323 checking for CONST_VECTOR.
9325 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9328 * predict.c (force_edge_cold): Handle in more sane way edges
9331 2018-01-09 Carl Love <cel@us.ibm.com>
9333 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
9335 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
9336 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
9337 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
9338 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
9339 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
9340 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
9341 * config/rs6000/rs6000-protos.h: Add extern defition for
9342 rs6000_generate_float2_double_code.
9343 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
9345 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
9346 (float2_v2df): Add define_expand.
9348 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
9351 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
9352 op_mode in the force_to_mode call.
9354 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9356 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
9357 instead of checking each element individually.
9358 (aarch64_evpc_uzp): Likewise.
9359 (aarch64_evpc_zip): Likewise.
9360 (aarch64_evpc_ext): Likewise.
9361 (aarch64_evpc_rev): Likewise.
9362 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
9363 instead of checking each element individually. Return true without
9365 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
9366 whether all selected elements come from the same input, instead of
9367 checking each element individually. Remove calls to gen_rtx_REG,
9368 start_sequence and end_sequence and instead assert that no rtl is
9371 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9373 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
9374 order of HIGH and CONST checks.
9376 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9378 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
9379 if the destination isn't an SSA_NAME.
9381 2018-01-09 Richard Biener <rguenther@suse.de>
9383 PR tree-optimization/83668
9384 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
9386 (canonicalize_loop_form): ... here, renamed from ...
9387 (canonicalize_loop_closed_ssa_form): ... this and amended to
9388 swap successor edges for loop exit blocks to make us use
9389 the RPO order we need for initial schedule generation.
9391 2018-01-09 Joseph Myers <joseph@codesourcery.com>
9393 PR tree-optimization/64811
9394 * match.pd: When optimizing comparisons with Inf, avoid
9395 introducing or losing exceptions from comparisons with NaN.
9397 2018-01-09 Martin Liska <mliska@suse.cz>
9400 * asan.c (shadow_mem_size): Add gcc_assert.
9402 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
9404 Don't save registers in main().
9407 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
9408 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
9409 * config/avr/avr.c (avr_set_current_function): Don't error if
9410 naked, OS_task or OS_main are specified at the same time.
9411 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
9413 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
9415 * common/config/avr/avr-common.c (avr_option_optimization_table):
9416 Switch on -mmain-is-OS_task for optimizing compilations.
9418 2018-01-09 Richard Biener <rguenther@suse.de>
9420 PR tree-optimization/83572
9421 * graphite.c: Include cfganal.h.
9422 (graphite_transform_loops): Connect infinite loops to exit
9423 and remove fake edges at the end.
9425 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9427 * ipa-inline.c (edge_badness): Revert accidental checkin.
9429 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9432 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
9433 symbols; not inline clones.
9435 2018-01-09 Jakub Jelinek <jakub@redhat.com>
9438 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
9439 hard registers. Formatting fixes.
9441 PR preprocessor/83722
9442 * gcc.c (try_generate_repro): Pass
9443 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
9444 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
9447 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
9448 Kito Cheng <kito.cheng@gmail.com>
9450 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
9451 (riscv_leaf_function_p): Delete.
9452 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
9454 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9456 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
9458 (do_ifelse): New function.
9459 (do_isel): New function.
9460 (do_sub3): New function.
9461 (do_add3): New function.
9462 (do_load_mask_compare): New function.
9463 (do_overlap_load_compare): New function.
9464 (expand_compare_loop): New function.
9465 (expand_block_compare): Call expand_compare_loop() when appropriate.
9466 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
9468 (-mblock-compare-inline-loop-limit): New option.
9470 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9473 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
9474 Reverse order of second and third operands in first alternative.
9475 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
9476 of first and second elements in UNSPEC_VPERMR vector.
9477 (altivec_expand_vec_perm_le): Likewise.
9479 2018-01-08 Jeff Law <law@redhat.com>
9481 PR rtl-optimizatin/81308
9482 * tree-switch-conversion.c (cfg_altered): New file scoped static.
9483 (process_switch): If group_case_labels makes a change, then set
9485 (pass_convert_switch::execute): If a switch is converted, then
9486 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
9488 PR rtl-optimization/81308
9489 * recog.c (split_all_insns): Conditionally cleanup the CFG after
9492 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
9494 PR target/83663 - Revert r255946
9495 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
9496 generation for cases where splatting a value is not useful.
9497 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
9498 across a vec_duplicate and a paradoxical subreg forming a vector
9499 mode to a vec_concat.
9501 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9503 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
9504 -march=armv8.3-a variants.
9505 * config/arm/t-multilib: Likewise.
9506 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
9508 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9510 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
9512 (cceq_ior_compare_complement): Give it a name so I can use it, and
9513 change boolean_or_operator predicate to boolean_operator so it can
9514 be used to generate a crand.
9515 (eqne): New code iterator.
9516 (bd/bd_neg): New code_attrs.
9517 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
9518 a single define_insn.
9519 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
9520 decrement (bdnzt/bdnzf/bdzt/bdzf).
9521 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
9522 with the new names of the branch decrement patterns, and added the
9523 names of the branch decrement conditional patterns.
9525 2018-01-08 Richard Biener <rguenther@suse.de>
9527 PR tree-optimization/83563
9528 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
9531 2018-01-08 Richard Biener <rguenther@suse.de>
9534 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
9536 2018-01-08 Richard Biener <rguenther@suse.de>
9538 PR tree-optimization/83685
9539 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
9540 references to abnormals.
9542 2018-01-08 Richard Biener <rguenther@suse.de>
9545 * dwarf2out.c (output_indirect_strings): Handle empty
9546 skeleton_debug_str_hash.
9547 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
9549 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9551 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
9552 (emit_store_direct): Likewise.
9553 (arc_trampoline_adjust_address): Likewise.
9554 (arc_asm_trampoline_template): New function.
9555 (arc_initialize_trampoline): Use asm_trampoline_template.
9556 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
9557 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
9558 * config/arc/arc.md (flush_icache): Delete pattern.
9560 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9562 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
9563 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
9566 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9569 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
9570 by not USED_FOR_TARGET.
9571 (make_pass_resolve_sw_modes): Likewise.
9573 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9575 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
9578 2018-01-08 Richard Biener <rguenther@suse.de>
9581 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
9583 2018-01-08 Richard Biener <rguenther@suse.de>
9586 * match.pd ((t * 2) / 2) -> t): Add missing :c.
9588 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
9591 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
9592 basic blocks with a small number of successors.
9593 (convert_control_dep_chain_into_preds): Improve handling of
9595 (dump_predicates): Split apart into...
9596 (dump_pred_chain): ...here...
9597 (dump_pred_info): ...and here.
9598 (can_one_predicate_be_invalidated_p): Add debugging printfs.
9599 (can_chain_union_be_invalidated_p): Improve check for invalidation
9601 (uninit_uses_cannot_happen): Avoid unnecessary if
9602 convert_control_dep_chain_into_preds yielded nothing.
9604 2018-01-06 Martin Sebor <msebor@redhat.com>
9606 PR tree-optimization/83640
9607 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
9608 subtracting negative offset from size.
9609 (builtin_access::overlap): Adjust offset bounds of the access to fall
9610 within the size of the object if possible.
9612 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
9614 PR rtl-optimization/83699
9615 * expmed.c (extract_bit_field_1): Restrict the vector usage of
9616 extract_bit_field_as_subreg to cases in which the extracted
9617 value is also a vector.
9619 * lra-constraints.c (process_alt_operands): Test for the equivalence
9620 substitutions when detecting a possible reload cycle.
9622 2018-01-06 Jakub Jelinek <jakub@redhat.com>
9625 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
9626 by default if flag_selective_schedling{,2}. Formatting fixes.
9628 PR rtl-optimization/83682
9629 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
9630 if it has non-VECTOR_MODE element mode.
9631 (vec_duplicate_p): Likewise.
9634 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
9635 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
9637 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9640 * config/i386/i386-builtin.def
9641 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
9642 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
9643 Require also OPTION_MASK_ISA_AVX512F in addition to
9644 OPTION_MASK_ISA_GFNI.
9645 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
9646 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
9647 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
9648 to OPTION_MASK_ISA_GFNI.
9649 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
9650 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
9651 OPTION_MASK_ISA_AVX512BW.
9652 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
9653 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
9654 addition to OPTION_MASK_ISA_GFNI.
9655 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
9656 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
9657 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
9658 to OPTION_MASK_ISA_GFNI.
9659 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
9660 a requirement for all ISAs rather than any of them with a few
9662 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
9664 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
9665 bitmasks to be enabled with 3 exceptions, instead of requiring any
9666 enabled ISA with lots of exceptions.
9667 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
9668 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
9669 Change avx512bw in isa attribute to avx512f.
9670 * config/i386/sgxintrin.h: Add license boilerplate.
9671 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
9672 to __AVX512F__ and __AVX512VL to __AVX512VL__.
9673 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
9674 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
9676 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
9677 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
9678 temporarily sse2 rather than sse if not enabled already.
9681 * config/i386/sse.md (VI248_VLBW): Rename to ...
9682 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
9683 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
9684 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
9685 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
9686 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
9687 mode iterator instead of VI248_VLBW.
9689 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
9691 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
9692 (record_modified): Skip clobbers; add debug output.
9693 (param_change_prob): Use sreal frequencies.
9695 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9697 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
9698 punt for user-aligned variables.
9700 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9702 * tree-chrec.c (chrec_contains_symbols): Return true for
9705 2018-01-05 Sudakshina Das <sudi.das@arm.com>
9708 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
9709 of (x|y) == x for BICS pattern.
9711 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9713 PR tree-optimization/83605
9714 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
9715 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
9718 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
9720 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
9721 * config/epiphany/rtems.h: New file.
9723 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9724 Uros Bizjak <ubizjak@gmail.com>
9727 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
9728 QIreg_operand instead of register_operand predicate.
9729 * config/i386/i386.c (ix86_rop_should_change_byte_p,
9730 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
9731 comments instead of -fmitigate[-_]rop.
9733 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9736 * cgraphunit.c (symbol_table::compile): Switch to text_section
9737 before calling assembly_start debug hook.
9738 * run-rtl-passes.c (run_rtl_passes): Likewise.
9741 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9743 * tree-vrp.c (extract_range_from_binary_expr_1): Check
9744 range_int_cst_p rather than !symbolic_range_p before calling
9745 extract_range_from_multiplicative_op_1.
9747 2018-01-04 Jeff Law <law@redhat.com>
9749 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
9750 redundant test in assertion.
9752 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9754 * doc/rtl.texi: Document machine_mode wrapper classes.
9756 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9758 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
9761 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9763 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
9764 the VEC_PERM_EXPR fold to fail.
9766 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9769 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
9770 to switched_sections.
9772 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9775 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
9778 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
9781 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
9782 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
9784 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9787 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
9788 is BLKmode and bitpos not zero or mode change is needed.
9790 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9793 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
9796 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
9799 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
9800 instead of MULT rtx. Update all corresponding splitters.
9802 (*ssub<modesuffix>): Ditto.
9804 (*cmp_sadd_di): Update split patterns.
9805 (*cmp_sadd_si): Ditto.
9806 (*cmp_sadd_sidi): Ditto.
9807 (*cmp_ssub_di): Ditto.
9808 (*cmp_ssub_si): Ditto.
9809 (*cmp_ssub_sidi): Ditto.
9810 * config/alpha/predicates.md (const23_operand): New predicate.
9811 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
9812 Look for ASHIFT, not MULT inner operand.
9813 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
9815 2018-01-04 Martin Liska <mliska@suse.cz>
9817 PR gcov-profile/83669
9818 * gcov.c (output_intermediate_file): Add version to intermediate
9820 * doc/gcov.texi: Document new field 'version' in intermediate
9821 file format. Fix location of '-k' option of gcov command.
9823 2018-01-04 Martin Liska <mliska@suse.cz>
9826 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
9828 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9830 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
9832 2018-01-03 Martin Sebor <msebor@redhat.com>
9834 PR tree-optimization/83655
9835 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
9836 checking calls with invalid arguments.
9838 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9840 * tree-vect-stmts.c (vect_get_store_rhs): New function.
9841 (vectorizable_mask_load_store): Delete.
9842 (vectorizable_call): Return false for masked loads and stores.
9843 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
9844 instead of gimple_assign_rhs1.
9845 (vectorizable_load): Handle IFN_MASK_LOAD.
9846 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
9848 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9850 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
9852 (vectorizable_mask_load_store): ...here.
9853 (vectorizable_load): ...and here.
9855 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9857 * tree-vect-stmts.c (vect_build_all_ones_mask)
9858 (vect_build_zero_merge_argument): New functions, split out from...
9859 (vectorizable_load): ...here.
9861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9863 * tree-vect-stmts.c (vect_check_store_rhs): New function,
9865 (vectorizable_mask_load_store): ...here.
9866 (vectorizable_store): ...and here.
9868 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9870 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
9872 (vectorizable_mask_load_store): ...here.
9874 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9876 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
9877 (vect_model_store_cost): Take a vec_load_store_type instead of a
9879 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
9880 (vect_model_store_cost): Take a vec_load_store_type instead of a
9882 (vectorizable_mask_load_store): Update accordingly.
9883 (vectorizable_store): Likewise.
9884 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
9886 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9888 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
9889 IFN_MASK_LOAD calls here rather than...
9890 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
9892 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9893 Alan Hayward <alan.hayward@arm.com>
9894 David Sherwood <david.sherwood@arm.com>
9896 * expmed.c (extract_bit_field_1): For vector extracts,
9897 fall back to extract_bit_field_as_subreg if vec_extract
9900 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9901 Alan Hayward <alan.hayward@arm.com>
9902 David Sherwood <david.sherwood@arm.com>
9904 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
9905 they are variable or constant sized.
9906 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
9907 slots for constant-sized data.
9909 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9910 Alan Hayward <alan.hayward@arm.com>
9911 David Sherwood <david.sherwood@arm.com>
9913 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
9914 handling COND_EXPRs with boolean comparisons, try to find a better
9915 basis for the mask type than the boolean itself.
9917 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9919 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
9920 is calculated and how it can be overridden.
9921 * genmodes.c (max_bitsize_mode_any_mode): New variable.
9922 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
9924 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
9927 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9928 Alan Hayward <alan.hayward@arm.com>
9929 David Sherwood <david.sherwood@arm.com>
9931 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
9932 Remove the mode argument.
9933 (aarch64_simd_valid_immediate): Remove the mode and inverse
9935 * config/aarch64/iterators.md (bitsize): New iterator.
9936 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
9937 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
9938 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
9939 aarch64_simd_valid_immediate.
9940 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
9941 (aarch64_reg_or_bic_imm): Likewise.
9942 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
9943 with an insn_type enum and msl with a modifier_type enum.
9944 Replace element_width with a scalar_mode. Change the shift
9945 to unsigned int. Add constructors for scalar_float_mode and
9946 scalar_int_mode elements.
9947 (aarch64_vect_float_const_representable_p): Delete.
9948 (aarch64_can_const_movi_rtx_p)
9949 (aarch64_simd_scalar_immediate_valid_for_move)
9950 (aarch64_simd_make_constant): Update call to
9951 aarch64_simd_valid_immediate.
9952 (aarch64_advsimd_valid_immediate_hs): New function.
9953 (aarch64_advsimd_valid_immediate): Likewise.
9954 (aarch64_simd_valid_immediate): Remove mode and inverse
9955 arguments. Rewrite to use the above. Use const_vec_duplicate_p
9956 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
9957 and aarch64_float_const_representable_p on the result.
9958 (aarch64_output_simd_mov_immediate): Remove mode argument.
9959 Update call to aarch64_simd_valid_immediate and use of
9960 simd_immediate_info.
9961 (aarch64_output_scalar_simd_mov_immediate): Update call
9964 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9965 Alan Hayward <alan.hayward@arm.com>
9966 David Sherwood <david.sherwood@arm.com>
9968 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
9969 (mode_nunits): Likewise CONST_MODE_NUNITS.
9970 * machmode.def (ADJUST_NUNITS): Document.
9971 * genmodes.c (mode_data::need_nunits_adj): New field.
9972 (blank_mode): Update accordingly.
9973 (adj_nunits): New variable.
9974 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
9976 (emit_mode_size_inline): Set need_bytesize_adj for all modes
9977 listed in adj_nunits.
9978 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
9979 listed in adj_nunits. Don't emit case statements for such modes.
9980 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
9981 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
9982 nothing if adj_nunits is nonnull.
9983 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
9984 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
9985 (emit_mode_fbit): Update use of print_maybe_const_decl.
9986 (emit_move_size): Likewise. Treat the array as non-const
9988 (emit_mode_adjustments): Handle adj_nunits.
9990 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9992 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
9993 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
9994 (VECTOR_MODES): Use it.
9995 (make_vector_modes): Take the prefix as an argument.
9997 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9998 Alan Hayward <alan.hayward@arm.com>
9999 David Sherwood <david.sherwood@arm.com>
10001 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
10002 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
10003 for MODE_VECTOR_BOOL.
10004 * machmode.def (VECTOR_BOOL_MODE): Document.
10005 * genmodes.c (VECTOR_BOOL_MODE): New macro.
10006 (make_vector_bool_mode): New function.
10007 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
10009 * lto-streamer-in.c (lto_input_mode_table): Likewise.
10010 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
10012 * stor-layout.c (int_mode_for_mode): Likewise.
10013 * tree.c (build_vector_type_for_mode): Likewise.
10014 * varasm.c (output_constant_pool_2): Likewise.
10015 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
10016 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
10017 for MODE_VECTOR_BOOL.
10018 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
10019 of mode class checks.
10020 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
10021 instead of a list of mode class checks.
10022 (expand_vector_scalar_condition): Likewise.
10023 (type_for_widest_vector_mode): Handle BImode as an inner mode.
10025 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10026 Alan Hayward <alan.hayward@arm.com>
10027 David Sherwood <david.sherwood@arm.com>
10029 * machmode.h (mode_size): Change from unsigned short to
10031 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
10032 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10033 or if measurement_type is not polynomial.
10034 (fixed_size_mode::includes_p): Check for constant-sized modes.
10035 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
10036 return a poly_uint16 rather than an unsigned short.
10037 (emit_mode_size): Change the type of mode_size from unsigned short
10038 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
10039 (emit_mode_adjustments): Cope with polynomial vector sizes.
10040 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10042 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10044 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
10045 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
10046 * caller-save.c (setup_save_areas): Likewise.
10047 (replace_reg_with_saved_mem): Likewise.
10048 * calls.c (emit_library_call_value_1): Likewise.
10049 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
10050 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
10051 (gen_lowpart_for_combine): Likewise.
10052 * convert.c (convert_to_integer_1): Likewise.
10053 * cse.c (equiv_constant, cse_insn): Likewise.
10054 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
10055 (cselib_subst_to_values): Likewise.
10056 * dce.c (word_dce_process_block): Likewise.
10057 * df-problems.c (df_word_lr_mark_ref): Likewise.
10058 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
10059 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
10060 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
10061 (rtl_for_decl_location): Likewise.
10062 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
10063 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
10064 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
10065 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
10066 (expand_expr_real_1): Likewise.
10067 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
10068 (pad_below): Likewise.
10069 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10070 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
10071 * ira.c (get_subreg_tracking_sizes): Likewise.
10072 * ira-build.c (ira_create_allocno_objects): Likewise.
10073 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
10074 (ira_sort_regnos_for_alter_reg): Likewise.
10075 * ira-costs.c (record_operand_costs): Likewise.
10076 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
10077 (resolve_simple_move): Likewise.
10078 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
10079 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
10080 (lra_constraints): Likewise.
10081 (CONST_POOL_OK_P): Reject variable-sized modes.
10082 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
10083 (add_pseudo_to_slot, lra_spill): Likewise.
10084 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10085 * optabs-query.c (get_best_extraction_insn): Likewise.
10086 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10087 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
10088 (expand_mult_highpart, valid_multiword_target_p): Likewise.
10089 * recog.c (offsettable_address_addr_space_p): Likewise.
10090 * regcprop.c (maybe_mode_change): Likewise.
10091 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
10092 * regrename.c (build_def_use): Likewise.
10093 * regstat.c (dump_reg_info): Likewise.
10094 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
10095 (find_reloads, find_reloads_subreg_address): Likewise.
10096 * reload1.c (eliminate_regs_1): Likewise.
10097 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
10098 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
10099 (simplify_binary_operation_1, simplify_subreg): Likewise.
10100 * targhooks.c (default_function_arg_padding): Likewise.
10101 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
10102 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
10103 (verify_gimple_assign_ternary): Likewise.
10104 * tree-inline.c (estimate_move_cost): Likewise.
10105 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10106 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
10107 (get_address_cost_ainc): Likewise.
10108 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
10109 (vect_supportable_dr_alignment): Likewise.
10110 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10111 (vectorizable_reduction): Likewise.
10112 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
10113 (vectorizable_operation, vectorizable_load): Likewise.
10114 * tree.c (build_same_sized_truth_vector_type): Likewise.
10115 * valtrack.c (cleanup_auto_inc_dec): Likewise.
10116 * var-tracking.c (emit_note_insn_var_location): Likewise.
10117 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
10118 (ADDR_VEC_ALIGN): Likewise.
10120 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10121 Alan Hayward <alan.hayward@arm.com>
10122 David Sherwood <david.sherwood@arm.com>
10124 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
10126 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10127 or if measurement_type is polynomial.
10128 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
10129 * combine.c (make_extraction): Likewise.
10130 * dse.c (find_shift_sequence): Likewise.
10131 * dwarf2out.c (mem_loc_descriptor): Likewise.
10132 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
10133 (extract_bit_field, extract_low_bits): Likewise.
10134 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
10135 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
10136 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
10137 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
10138 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10139 * reload.c (find_reloads): Likewise.
10140 * reload1.c (alter_reg): Likewise.
10141 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
10142 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
10143 * tree-if-conv.c (predicate_mem_writes): Likewise.
10144 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
10145 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
10146 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
10147 * valtrack.c (dead_debug_insert_temp): Likewise.
10148 * varasm.c (mergeable_constant_section): Likewise.
10149 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
10151 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10152 Alan Hayward <alan.hayward@arm.com>
10153 David Sherwood <david.sherwood@arm.com>
10155 * expr.c (expand_assignment): Cope with polynomial mode sizes
10156 when assigning to a CONCAT.
10158 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10159 Alan Hayward <alan.hayward@arm.com>
10160 David Sherwood <david.sherwood@arm.com>
10162 * machmode.h (mode_precision): Change from unsigned short to
10164 (mode_to_precision): Return a poly_uint16 rather than an unsigned
10166 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
10167 or if measurement_type is not polynomial.
10168 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
10169 in which the mode is already known to be a scalar_int_mode.
10170 * genmodes.c (emit_mode_precision): Change the type of mode_precision
10171 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
10173 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10174 for GET_MODE_PRECISION.
10175 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10176 for GET_MODE_PRECISION.
10177 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
10179 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
10180 (expand_field_assignment, make_extraction): Likewise.
10181 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
10182 (get_last_value): Likewise.
10183 * convert.c (convert_to_integer_1): Likewise.
10184 * cse.c (cse_insn): Likewise.
10185 * expr.c (expand_expr_real_1): Likewise.
10186 * lra-constraints.c (simplify_operand_subreg): Likewise.
10187 * optabs-query.c (can_atomic_load_p): Likewise.
10188 * optabs.c (expand_atomic_load): Likewise.
10189 (expand_atomic_store): Likewise.
10190 * ree.c (combine_reaching_defs): Likewise.
10191 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
10192 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
10193 * tree.h (type_has_mode_precision_p): Likewise.
10194 * ubsan.c (instrument_si_overflow): Likewise.
10196 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10197 Alan Hayward <alan.hayward@arm.com>
10198 David Sherwood <david.sherwood@arm.com>
10200 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
10201 polynomial numbers of units.
10202 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
10203 (valid_vector_subparts_p): New function.
10204 (build_vector_type): Remove temporary shim and take the number
10205 of units as a poly_uint64 rather than an int.
10206 (build_opaque_vector_type): Take the number of units as a
10207 poly_uint64 rather than an int.
10208 * tree.c (build_vector_from_ctor): Handle polynomial
10209 TYPE_VECTOR_SUBPARTS.
10210 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
10211 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
10212 (build_vector_from_val): If the number of units is variable,
10213 use build_vec_duplicate_cst for constant operands and
10214 VEC_DUPLICATE_EXPR otherwise.
10215 (make_vector_type): Remove temporary is_constant ().
10216 (build_vector_type, build_opaque_vector_type): Take the number of
10217 units as a poly_uint64 rather than an int.
10218 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
10220 * cfgexpand.c (expand_debug_expr): Likewise.
10221 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
10222 (store_constructor, expand_expr_real_1): Likewise.
10223 (const_scalar_mask_from_tree): Likewise.
10224 * fold-const-call.c (fold_const_reduction): Likewise.
10225 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
10226 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
10227 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
10228 (fold_relational_const): Likewise.
10229 (native_interpret_vector): Likewise. Change the size from an
10230 int to an unsigned int.
10231 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
10232 TYPE_VECTOR_SUBPARTS.
10233 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
10234 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
10235 duplicating a non-constant operand into a variable-length vector.
10236 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
10237 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
10238 * ipa-icf.c (sem_variable::equals): Likewise.
10239 * match.pd: Likewise.
10240 * omp-simd-clone.c (simd_clone_subparts): Likewise.
10241 * print-tree.c (print_node): Likewise.
10242 * stor-layout.c (layout_type): Likewise.
10243 * targhooks.c (default_builtin_vectorization_cost): Likewise.
10244 * tree-cfg.c (verify_gimple_comparison): Likewise.
10245 (verify_gimple_assign_binary): Likewise.
10246 (verify_gimple_assign_ternary): Likewise.
10247 (verify_gimple_assign_single): Likewise.
10248 * tree-pretty-print.c (dump_generic_node): Likewise.
10249 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10250 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
10251 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
10252 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
10253 (vect_shift_permute_load_chain): Likewise.
10254 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
10255 (expand_vector_condition, optimize_vector_constructor): Likewise.
10256 (lower_vec_perm, get_compute_type): Likewise.
10257 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10258 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
10259 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
10260 (vect_recog_mask_conversion_pattern): Likewise.
10261 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
10262 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
10263 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10264 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
10265 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
10266 (vectorizable_shift, vectorizable_operation, vectorizable_store)
10267 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
10268 (supportable_widening_operation): Likewise.
10269 (supportable_narrowing_operation): Likewise.
10270 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
10272 * varasm.c (output_constant): Likewise.
10274 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10275 Alan Hayward <alan.hayward@arm.com>
10276 David Sherwood <david.sherwood@arm.com>
10278 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
10279 so that both the length == 3 and length != 3 cases set up their
10280 own permute vectors. Add comments explaining why we know the
10281 number of elements is constant.
10282 (vect_permute_load_chain): Likewise.
10284 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10285 Alan Hayward <alan.hayward@arm.com>
10286 David Sherwood <david.sherwood@arm.com>
10288 * machmode.h (mode_nunits): Change from unsigned char to
10290 (ONLY_FIXED_SIZE_MODES): New macro.
10291 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
10292 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
10293 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
10295 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
10296 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
10297 or if measurement_type is not polynomial.
10298 * genmodes.c (ZERO_COEFFS): New macro.
10299 (emit_mode_nunits_inline): Make mode_nunits_inline return a
10301 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
10302 Use ZERO_COEFFS when emitting initializers.
10303 * data-streamer.h (bp_pack_poly_value): New function.
10304 (bp_unpack_poly_value): Likewise.
10305 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10306 for GET_MODE_NUNITS.
10307 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10308 for GET_MODE_NUNITS.
10309 * tree.c (make_vector_type): Remove temporary shim and make
10310 the real function take the number of units as a poly_uint64
10311 rather than an int.
10312 (build_vector_type_for_mode): Handle polynomial nunits.
10313 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
10314 * emit-rtl.c (const_vec_series_p_1): Likewise.
10315 (gen_rtx_CONST_VECTOR): Likewise.
10316 * fold-const.c (test_vec_duplicate_folding): Likewise.
10317 * genrecog.c (validate_pattern): Likewise.
10318 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
10319 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10320 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
10321 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
10322 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
10323 * rtlanal.c (subreg_get_info): Likewise.
10324 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10325 (vect_grouped_load_supported): Likewise.
10326 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
10327 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
10328 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
10329 (simplify_const_unary_operation, simplify_binary_operation_1)
10330 (simplify_const_binary_operation, simplify_ternary_operation)
10331 (test_vector_ops_duplicate, test_vector_ops): Likewise.
10332 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
10333 instead of CONST_VECTOR_NUNITS.
10334 * varasm.c (output_constant_pool_2): Likewise.
10335 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
10336 explicit-encoded elements in the XVEC for variable-length vectors.
10338 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10340 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
10342 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10343 Alan Hayward <alan.hayward@arm.com>
10344 David Sherwood <david.sherwood@arm.com>
10346 * coretypes.h (fixed_size_mode): Declare.
10347 (fixed_size_mode_pod): New typedef.
10348 * builtins.h (target_builtins::x_apply_args_mode)
10349 (target_builtins::x_apply_result_mode): Change type to
10350 fixed_size_mode_pod.
10351 * builtins.c (apply_args_size, apply_result_size, result_vector)
10352 (expand_builtin_apply_args_1, expand_builtin_apply)
10353 (expand_builtin_return): Update accordingly.
10355 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10357 * cse.c (hash_rtx_cb): Hash only the encoded elements.
10358 * cselib.c (cselib_hash_rtx): Likewise.
10359 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
10360 CONST_VECTOR encoding.
10362 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10363 Jeff Law <law@redhat.com>
10366 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
10367 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
10368 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
10369 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
10372 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
10373 explicitly probe *sp in a noreturn function if there were any callee
10374 register saves or frame pointer is needed.
10376 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10379 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
10380 BLKmode for ternary, binary or unary expressions.
10383 * var-tracking.c (delete_vta_debug_insn): New inline function.
10384 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
10385 insns from get_insns () to NULL instead of each bb separately.
10386 Use delete_vta_debug_insn. No longer static.
10387 (vt_debug_insns_local, variable_tracking_main_1): Adjust
10388 delete_vta_debug_insns callers.
10389 * rtl.h (delete_vta_debug_insns): Declare.
10390 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
10391 instead of variable_tracking_main.
10393 2018-01-03 Martin Sebor <msebor@redhat.com>
10395 PR tree-optimization/83603
10396 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
10397 arguments past the endof the argument list in functions declared
10398 without a prototype.
10399 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
10400 Avoid checking when arguments are null.
10402 2018-01-03 Martin Sebor <msebor@redhat.com>
10405 * doc/extend.texi (attribute const): Fix a typo.
10406 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
10407 issuing -Wsuggest-attribute for void functions.
10409 2018-01-03 Martin Sebor <msebor@redhat.com>
10411 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
10412 offset_int::from instead of wide_int::to_shwi.
10413 (maybe_diag_overlap): Remove assertion.
10414 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
10415 * gimple-ssa-sprintf.c (format_directive): Same.
10416 (parse_directive): Same.
10417 (sprintf_dom_walker::compute_format_length): Same.
10418 (try_substitute_return_value): Same.
10420 2018-01-03 Jeff Law <law@redhat.com>
10422 PR middle-end/83654
10423 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
10424 non-constant residual for zero at runtime and avoid probing in
10425 that case. Reorganize code for trailing problem to mirror handling
10428 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10430 PR tree-optimization/83501
10431 * tree-ssa-strlen.c (get_string_cst): New.
10432 (handle_char_store): Call get_string_cst.
10434 2018-01-03 Martin Liska <mliska@suse.cz>
10436 PR tree-optimization/83593
10437 * tree-ssa-strlen.c: Include tree-cfg.h.
10438 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
10439 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
10440 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
10442 (strlen_dom_walker::before_dom_children): Call
10443 gimple_purge_dead_eh_edges. Dump tranformation with details
10445 (strlen_dom_walker::before_dom_children): Update call by adding
10446 new argument cleanup_eh.
10447 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
10449 2018-01-03 Martin Liska <mliska@suse.cz>
10452 * cif-code.def (VARIADIC_THUNK): New enum value.
10453 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
10456 2018-01-03 Jan Beulich <jbeulich@suse.com>
10458 * sse.md (mov<mode>_internal): Tighten condition for when to use
10459 vmovdqu<ssescalarsize> for TI and OI modes.
10461 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10463 Update copyright years.
10465 2018-01-03 Martin Liska <mliska@suse.cz>
10468 * ipa-visibility.c (function_and_variable_visibility): Skip
10469 functions with noipa attribure.
10471 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10473 * gcc.c (process_command): Update copyright notice dates.
10474 * gcov-dump.c (print_version): Ditto.
10475 * gcov.c (print_version): Ditto.
10476 * gcov-tool.c (print_version): Ditto.
10477 * gengtype.c (create_file): Ditto.
10478 * doc/cpp.texi: Bump @copying's copyright year.
10479 * doc/cppinternals.texi: Ditto.
10480 * doc/gcc.texi: Ditto.
10481 * doc/gccint.texi: Ditto.
10482 * doc/gcov.texi: Ditto.
10483 * doc/install.texi: Ditto.
10484 * doc/invoke.texi: Ditto.
10486 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10488 * vector-builder.h (vector_builder::m_full_nelts): Change from
10489 unsigned int to poly_uint64.
10490 (vector_builder::full_nelts): Update prototype accordingly.
10491 (vector_builder::new_vector): Likewise.
10492 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
10493 (vector_builder::operator ==): Likewise.
10494 (vector_builder::finalize): Likewise.
10495 * int-vector-builder.h (int_vector_builder::int_vector_builder):
10496 Take the number of elements as a poly_uint64 rather than an
10498 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
10499 from unsigned int to poly_uint64.
10500 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
10501 (vec_perm_indices::new_vector): Likewise.
10502 (vec_perm_indices::length): Likewise.
10503 (vec_perm_indices::nelts_per_input): Likewise.
10504 (vec_perm_indices::input_nelts): Likewise.
10505 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
10506 number of elements per input as a poly_uint64 rather than an
10507 unsigned int. Use the original encoding for variable-length
10508 vectors, rather than clamping each individual element.
10509 For the second and subsequent elements in each pattern,
10510 clamp the step and base before clamping their sum.
10511 (vec_perm_indices::series_p): Handle polynomial element counts.
10512 (vec_perm_indices::all_in_range_p): Likewise.
10513 (vec_perm_indices_to_tree): Likewise.
10514 (vec_perm_indices_to_rtx): Likewise.
10515 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
10516 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
10517 (tree_vector_builder::new_binary_operation): Handle polynomial
10518 element counts. Return false if we need to know the number
10519 of elements at compile time.
10520 * fold-const.c (fold_vec_perm): Punt if the number of elements
10521 isn't known at compile time.
10523 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10525 * vec-perm-indices.h (vec_perm_builder): Change element type
10526 from HOST_WIDE_INT to poly_int64.
10527 (vec_perm_indices::element_type): Update accordingly.
10528 (vec_perm_indices::clamp): Handle polynomial element_types.
10529 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10530 (vec_perm_indices::all_in_range_p): Likewise.
10531 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
10533 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
10534 polynomial vec_perm_indices element types.
10535 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
10536 * fold-const.c (fold_vec_perm): Likewise.
10537 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
10538 * tree-vect-generic.c (lower_vec_perm): Likewise.
10539 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10540 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
10541 element type to HOST_WIDE_INT.
10543 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10544 Alan Hayward <alan.hayward@arm.com>
10545 David Sherwood <david.sherwood@arm.com>
10547 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
10548 rather than an int. Use plus_constant.
10549 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
10550 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
10552 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10553 Alan Hayward <alan.hayward@arm.com>
10554 David Sherwood <david.sherwood@arm.com>
10556 * calls.c (emit_call_1, expand_call): Change struct_value_size from
10557 a HOST_WIDE_INT to a poly_int64.
10559 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10560 Alan Hayward <alan.hayward@arm.com>
10561 David Sherwood <david.sherwood@arm.com>
10563 * calls.c (load_register_parameters): Cope with polynomial
10564 mode sizes. Require a constant size for BLKmode parameters
10565 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
10566 forces a parameter to be padded at the lsb end in order to
10567 fill a complete number of words, require the parameter size
10568 to be ordered wrt UNITS_PER_WORD.
10570 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10571 Alan Hayward <alan.hayward@arm.com>
10572 David Sherwood <david.sherwood@arm.com>
10574 * reload1.c (spill_stack_slot_width): Change element type
10575 from unsigned int to poly_uint64_pod.
10576 (alter_reg): Treat mode sizes as polynomial.
10578 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10579 Alan Hayward <alan.hayward@arm.com>
10580 David Sherwood <david.sherwood@arm.com>
10582 * reload.c (complex_word_subreg_p): New function.
10583 (reload_inner_reg_of_subreg, push_reload): Use it.
10585 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10586 Alan Hayward <alan.hayward@arm.com>
10587 David Sherwood <david.sherwood@arm.com>
10589 * lra-constraints.c (process_alt_operands): Reject matched
10590 operands whose sizes aren't ordered.
10591 (match_reload): Refer to this check here.
10593 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10594 Alan Hayward <alan.hayward@arm.com>
10595 David Sherwood <david.sherwood@arm.com>
10597 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
10598 that the mode size is in the set {1, 2, 4, 8, 16}.
10600 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10601 Alan Hayward <alan.hayward@arm.com>
10602 David Sherwood <david.sherwood@arm.com>
10604 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
10605 Use plus_constant instead of gen_rtx_PLUS.
10607 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10608 Alan Hayward <alan.hayward@arm.com>
10609 David Sherwood <david.sherwood@arm.com>
10611 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
10612 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
10613 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
10614 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
10615 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
10616 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
10617 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
10618 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
10619 * config/i386/i386.c (ix86_push_rounding): ...this new function.
10620 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
10622 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
10623 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
10624 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
10625 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
10626 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
10627 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
10628 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
10629 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
10630 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
10631 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
10633 * expr.c (emit_move_resolve_push): Treat the input and result
10634 of PUSH_ROUNDING as a poly_int64.
10635 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
10636 (emit_push_insn): Likewise.
10637 * lra-eliminations.c (mark_not_eliminable): Likewise.
10638 * recog.c (push_operand): Likewise.
10639 * reload1.c (elimination_effects): Likewise.
10640 * rtlanal.c (nonzero_bits1): Likewise.
10641 * calls.c (store_one_arg): Likewise. Require the padding to be
10642 known at compile time.
10644 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10645 Alan Hayward <alan.hayward@arm.com>
10646 David Sherwood <david.sherwood@arm.com>
10648 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
10649 Use plus_constant instead of gen_rtx_PLUS.
10651 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10652 Alan Hayward <alan.hayward@arm.com>
10653 David Sherwood <david.sherwood@arm.com>
10655 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
10656 rather than an int.
10658 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10659 Alan Hayward <alan.hayward@arm.com>
10660 David Sherwood <david.sherwood@arm.com>
10662 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
10663 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
10664 via stack temporaries. Treat the mode size as polynomial too.
10666 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10667 Alan Hayward <alan.hayward@arm.com>
10668 David Sherwood <david.sherwood@arm.com>
10670 * expr.c (expand_expr_real_2): When handling conversions involving
10671 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
10672 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
10673 as a poly_uint64 too.
10675 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10676 Alan Hayward <alan.hayward@arm.com>
10677 David Sherwood <david.sherwood@arm.com>
10679 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
10681 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10682 Alan Hayward <alan.hayward@arm.com>
10683 David Sherwood <david.sherwood@arm.com>
10685 * combine.c (can_change_dest_mode): Handle polynomial
10686 REGMODE_NATURAL_SIZE.
10687 * expmed.c (store_bit_field_1): Likewise.
10688 * expr.c (store_constructor): Likewise.
10689 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
10690 and polynomial REGMODE_NATURAL_SIZE.
10691 (gen_lowpart_common): Likewise.
10692 * reginfo.c (record_subregs_of_mode): Likewise.
10693 * rtlanal.c (read_modify_subreg_p): Likewise.
10695 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10696 Alan Hayward <alan.hayward@arm.com>
10697 David Sherwood <david.sherwood@arm.com>
10699 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
10700 numbers of elements.
10702 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10703 Alan Hayward <alan.hayward@arm.com>
10704 David Sherwood <david.sherwood@arm.com>
10706 * match.pd: Cope with polynomial numbers of vector elements.
10708 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10709 Alan Hayward <alan.hayward@arm.com>
10710 David Sherwood <david.sherwood@arm.com>
10712 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
10713 in a POINTER_PLUS_EXPR.
10715 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10716 Alan Hayward <alan.hayward@arm.com>
10717 David Sherwood <david.sherwood@arm.com>
10719 * omp-simd-clone.c (simd_clone_subparts): New function.
10720 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
10721 (ipa_simd_modify_function_body): Likewise.
10723 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10724 Alan Hayward <alan.hayward@arm.com>
10725 David Sherwood <david.sherwood@arm.com>
10727 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
10728 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
10729 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
10730 (expand_vector_condition, vector_element): Likewise.
10731 (subparts_gt): New function.
10732 (get_compute_type): Use subparts_gt.
10733 (count_type_subparts): Delete.
10734 (expand_vector_operations_1): Use subparts_gt instead of
10735 count_type_subparts.
10737 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10738 Alan Hayward <alan.hayward@arm.com>
10739 David Sherwood <david.sherwood@arm.com>
10741 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
10742 (vect_compile_time_alias): ...this new function. Do the calculation
10743 on poly_ints rather than trees.
10744 (vect_prune_runtime_alias_test_list): Update call accordingly.
10746 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10747 Alan Hayward <alan.hayward@arm.com>
10748 David Sherwood <david.sherwood@arm.com>
10750 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
10752 (vect_schedule_slp_instance): Likewise.
10754 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10755 Alan Hayward <alan.hayward@arm.com>
10756 David Sherwood <david.sherwood@arm.com>
10758 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
10759 constant and extern definitions for variable-length vectors.
10760 (vect_get_constant_vectors): Note that the number of units
10761 is known to be constant.
10763 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10764 Alan Hayward <alan.hayward@arm.com>
10765 David Sherwood <david.sherwood@arm.com>
10767 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
10768 of units as polynomial. Choose between WIDE and NARROW based
10771 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10772 Alan Hayward <alan.hayward@arm.com>
10773 David Sherwood <david.sherwood@arm.com>
10775 * tree-vect-stmts.c (simd_clone_subparts): New function.
10776 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
10778 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10779 Alan Hayward <alan.hayward@arm.com>
10780 David Sherwood <david.sherwood@arm.com>
10782 * tree-vect-stmts.c (vectorizable_call): Treat the number of
10783 vectors as polynomial. Use build_index_vector for
10784 IFN_GOMP_SIMD_LANE.
10786 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10787 Alan Hayward <alan.hayward@arm.com>
10788 David Sherwood <david.sherwood@arm.com>
10790 * tree-vect-stmts.c (get_load_store_type): Treat the number of
10791 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
10792 for variable-length vectors.
10793 (vectorizable_mask_load_store): Treat the number of units as
10794 polynomial, asserting that it is constant if the condition has
10795 already been enforced.
10796 (vectorizable_store, vectorizable_load): Likewise.
10798 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10799 Alan Hayward <alan.hayward@arm.com>
10800 David Sherwood <david.sherwood@arm.com>
10802 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
10803 of units as polynomial. Punt if we can't tell at compile time
10804 which vector contains the final result.
10806 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10807 Alan Hayward <alan.hayward@arm.com>
10808 David Sherwood <david.sherwood@arm.com>
10810 * tree-vect-loop.c (vectorizable_induction): Treat the number
10811 of units as polynomial. Punt on SLP inductions. Use an integer
10812 VEC_SERIES_EXPR for variable-length integer reductions. Use a
10813 cast of such a series for variable-length floating-point
10816 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10817 Alan Hayward <alan.hayward@arm.com>
10818 David Sherwood <david.sherwood@arm.com>
10820 * tree.h (build_index_vector): Declare.
10821 * tree.c (build_index_vector): New function.
10822 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
10823 of units as polynomial, forcibly converting it to a constant if
10824 vectorizable_reduction has already enforced the condition.
10825 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
10826 to create a {1,2,3,...} vector.
10827 (vectorizable_reduction): Treat the number of units as polynomial.
10828 Choose vectype_in based on the largest scalar element size rather
10829 than the smallest number of units. Enforce the restrictions
10832 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10833 Alan Hayward <alan.hayward@arm.com>
10834 David Sherwood <david.sherwood@arm.com>
10836 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
10837 number of units as polynomial.
10839 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10840 Alan Hayward <alan.hayward@arm.com>
10841 David Sherwood <david.sherwood@arm.com>
10843 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
10844 * target.def (autovectorize_vector_sizes): Return the vector sizes
10845 by pointer, using vector_sizes rather than a bitmask.
10846 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
10847 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
10848 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
10850 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
10851 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
10852 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
10853 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
10854 * omp-general.c (omp_max_vf): Likewise.
10855 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10856 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
10857 * tree-vect-loop.c (vect_analyze_loop): Likewise.
10858 * tree-vect-slp.c (vect_slp_bb): Likewise.
10859 * doc/tm.texi: Regenerate.
10860 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
10862 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
10863 the vector size as a poly_uint64 rather than an unsigned int.
10864 (current_vector_size): Change from an unsigned int to a poly_uint64.
10865 (get_vectype_for_scalar_type): Update accordingly.
10866 * tree.h (build_truth_vector_type): Take the size and number of
10867 units as a poly_uint64 rather than an unsigned int.
10868 (build_vector_type): Add a temporary overload that takes
10869 the number of units as a poly_uint64 rather than an unsigned int.
10870 * tree.c (make_vector_type): Likewise.
10871 (build_truth_vector_type): Take the number of units as a poly_uint64
10872 rather than an unsigned int.
10874 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10875 Alan Hayward <alan.hayward@arm.com>
10876 David Sherwood <david.sherwood@arm.com>
10878 * target.def (get_mask_mode): Take the number of units and length
10879 as poly_uint64s rather than unsigned ints.
10880 * targhooks.h (default_get_mask_mode): Update accordingly.
10881 * targhooks.c (default_get_mask_mode): Likewise.
10882 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
10883 * doc/tm.texi: Regenerate.
10885 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10886 Alan Hayward <alan.hayward@arm.com>
10887 David Sherwood <david.sherwood@arm.com>
10889 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
10890 * omp-general.c (omp_max_vf): Likewise.
10891 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
10892 (expand_omp_simd): Handle polynomial safelen.
10893 * omp-low.c (omplow_simd_context): Add a default constructor.
10894 (omplow_simd_context::max_vf): Change from int to poly_uint64.
10895 (lower_rec_simd_input_clauses): Update accordingly.
10896 (lower_rec_input_clauses): Likewise.
10898 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10899 Alan Hayward <alan.hayward@arm.com>
10900 David Sherwood <david.sherwood@arm.com>
10902 * tree-vectorizer.h (vect_nunits_for_cost): New function.
10903 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
10904 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
10905 (vect_analyze_slp_cost): Likewise.
10906 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
10907 (vect_model_load_cost): Likewise.
10909 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10910 Alan Hayward <alan.hayward@arm.com>
10911 David Sherwood <david.sherwood@arm.com>
10913 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
10914 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
10915 from an unsigned int * to a poly_uint64_pod *.
10916 (calculate_unrolling_factor): New function.
10917 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
10919 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10920 Alan Hayward <alan.hayward@arm.com>
10921 David Sherwood <david.sherwood@arm.com>
10923 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
10924 from an unsigned int to a poly_uint64.
10925 (_loop_vec_info::slp_unrolling_factor): Likewise.
10926 (_loop_vec_info::vectorization_factor): Change from an int
10928 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
10929 (vect_get_num_vectors): New function.
10930 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
10931 (vect_get_num_copies): Use vect_get_num_vectors.
10932 (vect_analyze_data_ref_dependences): Change max_vf from an int *
10933 to an unsigned int *.
10934 (vect_analyze_data_refs): Change min_vf from an int * to a
10936 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10937 than an unsigned HOST_WIDE_INT.
10938 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
10939 (vect_analyze_data_ref_dependence): Change max_vf from an int *
10940 to an unsigned int *.
10941 (vect_analyze_data_ref_dependences): Likewise.
10942 (vect_compute_data_ref_alignment): Handle polynomial vf.
10943 (vect_enhance_data_refs_alignment): Likewise.
10944 (vect_prune_runtime_alias_test_list): Likewise.
10945 (vect_shift_permute_load_chain): Likewise.
10946 (vect_supportable_dr_alignment): Likewise.
10947 (dependence_distance_ge_vf): Take the vectorization factor as a
10948 poly_uint64 rather than an unsigned HOST_WIDE_INT.
10949 (vect_analyze_data_refs): Change min_vf from an int * to a
10951 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
10952 vfm1 as a poly_uint64 rather than an int. Make the same change
10953 for the returned bound_scalar.
10954 (vect_gen_vector_loop_niters): Handle polynomial vf.
10955 (vect_do_peeling): Likewise. Update call to
10956 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
10957 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
10959 * tree-vect-loop.c (vect_determine_vectorization_factor)
10960 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
10961 (vect_get_known_peeling_cost): Likewise.
10962 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
10963 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
10964 (vect_transform_loop): Likewise. Use the lowest possible VF when
10965 updating the upper bounds of the loop.
10966 (vect_min_worthwhile_factor): Make static. Return an unsigned int
10967 rather than an int.
10968 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
10969 polynomial unroll factors.
10970 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
10971 (vect_make_slp_decision): Likewise.
10972 (vect_supported_load_permutation_p): Likewise, and polynomial
10974 (vect_analyze_slp_cost): Handle polynomial vf.
10975 (vect_slp_analyze_node_operations): Likewise.
10976 (vect_slp_analyze_bb_1): Likewise.
10977 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10978 than an unsigned HOST_WIDE_INT.
10979 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
10980 (vectorizable_load): Handle polynomial vf.
10981 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
10983 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
10985 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10986 Alan Hayward <alan.hayward@arm.com>
10987 David Sherwood <david.sherwood@arm.com>
10989 * match.pd: Handle bit operations involving three constants
10990 and try to fold one pair.
10992 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10994 * tree-vect-loop-manip.c: Include gimple-fold.h.
10995 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
10996 niters_maybe_zero parameters. Handle other cases besides a step of 1.
10997 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
10998 Add a path that uses a step of VF instead of 1, but disable it
11000 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
11001 and niters_no_overflow parameters. Update calls to
11002 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
11003 Create a new SSA name if the latter choses to use a ste other
11004 than zero, and return it via niters_vector_mult_vf_var.
11005 * tree-vect-loop.c (vect_transform_loop): Update calls to
11006 vect_do_peeling, vect_gen_vector_loop_niters and
11007 slpeel_make_loop_iterate_ntimes.
11008 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
11009 (vect_gen_vector_loop_niters): Update declarations after above changes.
11011 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
11013 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
11014 128-bit round to integer instructions.
11015 (ceil<mode>2): Likewise.
11016 (btrunc<mode>2): Likewise.
11017 (round<mode>2): Likewise.
11019 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11021 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
11022 unaligned VSX load/store on P8/P9.
11023 (expand_block_clear): Allow the use of unaligned VSX
11024 load/store on P8/P9.
11026 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11028 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
11030 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
11031 swap associated with both a load and a store.
11033 2018-01-02 Andrew Waterman <andrew@sifive.com>
11035 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
11036 * config/riscv/riscv.md (clear_cache): Use it.
11038 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
11040 * web.c: Remove out-of-date comment.
11042 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11044 * expr.c (fixup_args_size_notes): Check that any existing
11045 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
11046 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
11047 (emit_single_push_insn): ...here.
11049 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11051 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
11052 (const_vector_encoded_nelts): New function.
11053 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
11054 (const_vector_int_elt, const_vector_elt): Declare.
11055 * emit-rtl.c (const_vector_int_elt_1): New function.
11056 (const_vector_elt): Likewise.
11057 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
11058 of CONST_VECTOR_ELT.
11060 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11062 * expr.c: Include rtx-vector-builder.h.
11063 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
11064 directly on the tree encoding.
11065 (const_vector_from_tree): Likewise.
11066 * optabs.c: Include rtx-vector-builder.h.
11067 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
11068 sequence of "u" values.
11069 * vec-perm-indices.c: Include rtx-vector-builder.h.
11070 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
11071 directly on the vec_perm_indices encoding.
11073 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11075 * doc/rtl.texi (const_vector): Describe new encoding scheme.
11076 * Makefile.in (OBJS): Add rtx-vector-builder.o.
11077 * rtx-vector-builder.h: New file.
11078 * rtx-vector-builder.c: Likewise.
11079 * rtl.h (rtx_def::u2): Add a const_vector field.
11080 (CONST_VECTOR_NPATTERNS): New macro.
11081 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
11082 (CONST_VECTOR_DUPLICATE_P): Likewise.
11083 (CONST_VECTOR_STEPPED_P): Likewise.
11084 (CONST_VECTOR_ENCODED_ELT): Likewise.
11085 (const_vec_duplicate_p): Check for a duplicated vector encoding.
11086 (unwrap_const_vec_duplicate): Likewise.
11087 (const_vec_series_p): Check for a non-duplicated vector encoding.
11088 Say that the function only returns true for integer vectors.
11089 * emit-rtl.c: Include rtx-vector-builder.h.
11090 (gen_const_vec_duplicate_1): Delete.
11091 (gen_const_vector): Call gen_const_vec_duplicate instead of
11092 gen_const_vec_duplicate_1.
11093 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
11094 (gen_const_vec_duplicate): Use rtx_vector_builder.
11095 (gen_const_vec_series): Likewise.
11096 (gen_rtx_CONST_VECTOR): Likewise.
11097 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
11098 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11099 Build a new vector rather than modifying a CONST_VECTOR in-place.
11100 (handle_special_swappables): Update call accordingly.
11101 * config/rs6000/rs6000-p8swap.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.
11106 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11108 * simplify-rtx.c (simplify_const_binary_operation): Use
11109 CONST_VECTOR_ELT instead of XVECEXP.
11111 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11113 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
11114 the selector elements to be different from the data elements
11115 if the selector is a VECTOR_CST.
11116 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
11117 ssizetype for the selector.
11119 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11121 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
11122 before testing each element individually.
11123 * tree-vect-generic.c (lower_vec_perm): Likewise.
11125 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11127 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
11128 * selftest-run-tests.c (selftest::run_tests): Call it.
11129 * vector-builder.h (vector_builder::operator ==): New function.
11130 (vector_builder::operator !=): Likewise.
11131 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
11132 (vec_perm_indices::all_from_input_p): New function.
11133 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11134 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
11135 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
11136 instead of reading the VECTOR_CST directly. Detect whether both
11137 vector inputs are the same before constructing the vec_perm_indices,
11138 and update the number of inputs argument accordingly. Use the
11139 utility functions added above. Only construct sel2 if we need to.
11141 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11143 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
11144 the broadcast of the low byte.
11145 (expand_mult_highpart): Use an explicit encoding for the permutes.
11146 * optabs-query.c (can_mult_highpart_p): Likewise.
11147 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
11148 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11149 (vectorizable_bswap): Likewise.
11150 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
11151 explicit encoding for the power-of-2 permutes.
11152 (vect_permute_store_chain): Likewise.
11153 (vect_grouped_load_supported): Likewise.
11154 (vect_permute_load_chain): Likewise.
11156 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11158 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
11159 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
11160 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
11161 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11162 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
11163 (vect_gen_perm_mask_any): Likewise.
11165 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11167 * int-vector-builder.h: New file.
11168 * vec-perm-indices.h: Include int-vector-builder.h.
11169 (vec_perm_indices): Redefine as an int_vector_builder.
11170 (auto_vec_perm_indices): Delete.
11171 (vec_perm_builder): Redefine as a stand-alone class.
11172 (vec_perm_indices::vec_perm_indices): New function.
11173 (vec_perm_indices::clamp): Likewise.
11174 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
11175 (vec_perm_indices::new_vector): New function.
11176 (vec_perm_indices::new_expanded_vector): Update for new
11177 vec_perm_indices class.
11178 (vec_perm_indices::rotate_inputs): New function.
11179 (vec_perm_indices::all_in_range_p): Operate directly on the
11180 encoded form, without computing elided elements.
11181 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
11182 encoding. Update for new vec_perm_indices class.
11183 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
11184 the given vec_perm_builder.
11185 (expand_vec_perm_var): Update vec_perm_builder constructor.
11186 (expand_mult_highpart): Use vec_perm_builder instead of
11187 auto_vec_perm_indices.
11188 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
11189 vec_perm_indices instead of auto_vec_perm_indices. Use a single
11190 or double series encoding as appropriate.
11191 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
11192 vec_perm_indices instead of auto_vec_perm_indices.
11193 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11194 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11195 (vect_permute_store_chain): Likewise.
11196 (vect_grouped_load_supported): Likewise.
11197 (vect_permute_load_chain): Likewise.
11198 (vect_shift_permute_load_chain): Likewise.
11199 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11200 (vect_transform_slp_perm_load): Likewise.
11201 (vect_schedule_slp_instance): Likewise.
11202 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11203 (vectorizable_mask_load_store): Likewise.
11204 (vectorizable_bswap): Likewise.
11205 (vectorizable_store): Likewise.
11206 (vectorizable_load): Likewise.
11207 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
11208 vec_perm_indices instead of auto_vec_perm_indices. Use
11209 tree_to_vec_perm_builder to read the vector from a tree.
11210 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
11211 vec_perm_builder instead of a vec_perm_indices.
11212 (have_whole_vector_shift): Use vec_perm_builder and
11213 vec_perm_indices instead of auto_vec_perm_indices. Leave the
11214 truncation to calc_vec_perm_mask_for_shift.
11215 (vect_create_epilog_for_reduction): Likewise.
11216 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
11217 from auto_vec_perm_indices to vec_perm_indices.
11218 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11219 instead of changing individual elements.
11220 (aarch64_vectorize_vec_perm_const): Use new_vector to install
11221 the vector in d.perm.
11222 * config/arm/arm.c (expand_vec_perm_d::perm): Change
11223 from auto_vec_perm_indices to vec_perm_indices.
11224 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11225 instead of changing individual elements.
11226 (arm_vectorize_vec_perm_const): Use new_vector to install
11227 the vector in d.perm.
11228 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
11229 Update vec_perm_builder constructor.
11230 (rs6000_expand_interleave): Likewise.
11231 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
11232 (rs6000_expand_interleave): Likewise.
11234 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11236 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
11237 to qimode could truncate the indices.
11238 * optabs.c (expand_vec_perm_var): Likewise.
11240 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11242 * Makefile.in (OBJS): Add vec-perm-indices.o.
11243 * vec-perm-indices.h: New file.
11244 * vec-perm-indices.c: Likewise.
11245 * target.h (vec_perm_indices): Replace with a forward class
11247 (auto_vec_perm_indices): Move to vec-perm-indices.h.
11248 * optabs.h: Include vec-perm-indices.h.
11249 (expand_vec_perm): Delete.
11250 (selector_fits_mode_p, expand_vec_perm_var): Declare.
11251 (expand_vec_perm_const): Declare.
11252 * target.def (vec_perm_const_ok): Replace with...
11253 (vec_perm_const): ...this new hook.
11254 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
11255 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
11256 * doc/tm.texi: Regenerate.
11257 * optabs.def (vec_perm_const): Delete.
11258 * doc/md.texi (vec_perm_const): Likewise.
11259 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
11260 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
11261 expand_vec_perm for constant permutation vectors. Assert that
11262 the mode of variable permutation vectors is the integer equivalent
11263 of the mode that is being permuted.
11264 * optabs-query.h (selector_fits_mode_p): Declare.
11265 * optabs-query.c: Include vec-perm-indices.h.
11266 (selector_fits_mode_p): New function.
11267 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
11268 is defined, instead of checking whether the vec_perm_const_optab
11269 exists. Use targetm.vectorize.vec_perm_const instead of
11270 targetm.vectorize.vec_perm_const_ok. Check whether the indices
11271 fit in the vector mode before using a variable permute.
11272 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
11273 vec_perm_indices instead of an rtx.
11274 (expand_vec_perm): Replace with...
11275 (expand_vec_perm_const): ...this new function. Take the selector
11276 as a vec_perm_indices rather than an rtx. Also take the mode of
11277 the selector. Update call to shift_amt_for_vec_perm_mask.
11278 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
11279 Use vec_perm_indices::new_expanded_vector to expand the original
11280 selector into bytes. Check whether the indices fit in the vector
11281 mode before using a variable permute.
11282 (expand_vec_perm_var): Make global.
11283 (expand_mult_highpart): Use expand_vec_perm_const.
11284 * fold-const.c: Includes vec-perm-indices.h.
11285 * tree-ssa-forwprop.c: Likewise.
11286 * tree-vect-data-refs.c: Likewise.
11287 * tree-vect-generic.c: Likewise.
11288 * tree-vect-loop.c: Likewise.
11289 * tree-vect-slp.c: Likewise.
11290 * tree-vect-stmts.c: Likewise.
11291 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
11293 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
11294 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
11295 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
11296 (aarch64_vectorize_vec_perm_const): ...this new function.
11297 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11298 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11299 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
11300 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
11301 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11302 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11303 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
11305 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
11306 check for NEON modes.
11307 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
11308 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
11309 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
11310 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
11312 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
11313 the old VEC_PERM_CONST conditions.
11314 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
11315 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
11316 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
11317 (ia64_vectorize_vec_perm_const_ok): Merge into...
11318 (ia64_vectorize_vec_perm_const): ...this new function.
11319 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
11320 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
11321 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
11322 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
11323 * config/mips/mips.c (mips_expand_vec_perm_const)
11324 (mips_vectorize_vec_perm_const_ok): Merge into...
11325 (mips_vectorize_vec_perm_const): ...this new function.
11326 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
11327 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
11328 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
11329 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
11330 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
11331 (rs6000_expand_vec_perm_const): Delete.
11332 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
11334 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11335 (altivec_expand_vec_perm_const_le): Take each operand individually.
11336 Operate on constant selectors rather than rtxes.
11337 (altivec_expand_vec_perm_const): Likewise. Update call to
11338 altivec_expand_vec_perm_const_le.
11339 (rs6000_expand_vec_perm_const): Delete.
11340 (rs6000_vectorize_vec_perm_const_ok): Delete.
11341 (rs6000_vectorize_vec_perm_const): New function.
11342 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11343 an element count and rtx array.
11344 (rs6000_expand_extract_even): Update call accordingly.
11345 (rs6000_expand_interleave): Likewise.
11346 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
11347 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
11348 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
11349 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
11350 (rs6000_expand_vec_perm_const): Delete.
11351 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11352 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11353 (altivec_expand_vec_perm_const_le): Take each operand individually.
11354 Operate on constant selectors rather than rtxes.
11355 (altivec_expand_vec_perm_const): Likewise. Update call to
11356 altivec_expand_vec_perm_const_le.
11357 (rs6000_expand_vec_perm_const): Delete.
11358 (rs6000_vectorize_vec_perm_const_ok): Delete.
11359 (rs6000_vectorize_vec_perm_const): New function. Remove stray
11360 reference to the SPE evmerge intructions.
11361 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11362 an element count and rtx array.
11363 (rs6000_expand_extract_even): Update call accordingly.
11364 (rs6000_expand_interleave): Likewise.
11365 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
11366 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
11368 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11370 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11372 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
11373 vector mode and that that mode matches the mode of the data
11375 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
11376 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
11377 directly using expand_vec_perm_1 when forcing selectors into
11379 (expand_vec_perm_var): New function, split out from expand_vec_perm.
11381 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11383 * optabs-query.h (can_vec_perm_p): Delete.
11384 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
11385 * optabs-query.c (can_vec_perm_p): Split into...
11386 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
11387 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
11388 particular selector is valid.
11389 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11390 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11391 (vect_grouped_load_supported): Likewise.
11392 (vect_shift_permute_load_chain): Likewise.
11393 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11394 (vect_transform_slp_perm_load): Likewise.
11395 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11396 (vectorizable_bswap): Likewise.
11397 (vect_gen_perm_mask_checked): Likewise.
11398 * fold-const.c (fold_ternary_loc): Likewise. Don't take
11399 implementations of variable permutation vectors into account
11400 when deciding which selector to use.
11401 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
11402 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
11403 with a false third argument.
11404 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
11405 to test whether the constant selector is valid and can_vec_perm_var_p
11406 to test whether a variable selector is valid.
11408 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11410 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
11411 * optabs-query.c (can_vec_perm_p): Likewise.
11412 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
11413 instead of vec_perm_indices.
11414 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
11415 (vect_gen_perm_mask_checked): Likewise,
11416 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
11417 (vect_gen_perm_mask_checked): Likewise,
11419 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11421 * optabs-query.h (qimode_for_vec_perm): Declare.
11422 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
11423 (qimode_for_vec_perm): ...this new function.
11424 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
11426 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11428 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
11429 does not have a conditional at the top.
11431 2018-01-02 Richard Biener <rguenther@suse.de>
11433 * ipa-inline.c (big_speedup_p): Fix expression.
11435 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11438 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
11441 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11445 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
11446 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
11447 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
11448 cond_taken_branch_cost 3->4.
11450 2018-01-01 Jakub Jelinek <jakub@redhat.com>
11452 PR tree-optimization/83581
11453 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
11454 TODO_cleanup_cfg if any changes have been made.
11456 PR middle-end/83608
11457 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
11458 convert_modes if target mode has the right side, but different mode
11461 PR middle-end/83609
11462 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
11463 last argument when extracting from CONCAT. If either from_real or
11464 from_imag is NULL, use expansion through memory. If result is not
11465 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
11466 the parts directly to inner mode, if even that fails, use expansion
11469 PR middle-end/83623
11470 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
11471 check for bswap in mode rather than HImode and use that in expand_unop
11474 Copyright (C) 2018 Free Software Foundation, Inc.
11476 Copying and distribution of this file, with or without modification,
11477 are permitted in any medium without royalty provided the copyright
11478 notice and this notice are preserved.