1 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
3 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
4 in __abskf2 and __powikf2.
6 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9 * config/aarch64/constraints.md (Usg, Usj): New constraints.
10 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
11 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
12 Use the above on operand 2. Reindent.
13 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
15 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
18 * common/config/i386/i386-common.c (ix86_handle_option): Don't
20 * config/i386/i386.opt (mcet): Removed.
21 * doc/install.texi: Remove -mcet documentation.
22 * doc/invoke.texi: Likewise.
24 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
27 * doc/install.texi: Remove -mcet from bootstrap-cet.
29 2018-04-24 Jakub Jelinek <jakub@redhat.com>
32 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
33 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
37 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
38 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
39 containing a CONST_VECTOR.
41 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
43 * doc/install.texi: Update newlib dependency for nvptx.
45 2018-04-24 Jakub Jelinek <jakub@redhat.com>
48 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
49 instead of INTVAL when shifting x left.
51 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
53 PR tree-optimization/85478
54 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
55 vect_grouped_store_supported for single element vectors.
57 2018-04-24 Richard Biener <rguenther@suse.de>
60 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
61 load cost increase to the case of non-constant step.
63 2018-04-24 Jakub Jelinek <jakub@redhat.com>
66 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
67 destination if any_malformed_asm.
69 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
72 * expr.c (store_field): In the bitfield case, if the value comes from
73 a function call and is returned in registers by means of a PARALLEL,
74 do not change the mode of the temporary unless BLKmode and VOIDmode.
76 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
78 PR rtl-optimization/85423
79 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
80 dependencies to debug insns when the previous insn is non-debug.
82 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
84 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
85 enums into a single definition.
86 (fls): Fix predicates and printing.
89 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
91 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
92 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
93 and short u6 immediate.
94 (check_if_valid_sleep_operand): Remove.
95 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
97 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
99 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
100 flag_always_save_lp condition.
101 * config/nds32/nds32.opt (malways-save-lp): New option.
103 2018-04-22 Shiva Chen <shiva0217@gmail.com>
105 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
106 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
107 * config/nds32/nds32.h
108 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
109 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
111 2018-04-22 Shiva Chen <shiva0217@gmail.com>
113 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
114 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
116 2018-04-22 Shiva Chen <shiva0217@gmail.com>
117 Chung-Ju Wu <jasonwucj@gmail.com>
119 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
121 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
122 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
124 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
126 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
128 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
130 * config/nds32/nds32-protos.h (nds32_data_alignment,
131 nds32_local_alignment): Declare.
132 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
133 nds32_local_alignment): New functions.
134 (TARGET_CONSTANT_ALIGNMENT): Define.
135 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
137 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
139 * config/nds32/nds32.c
140 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
141 (TARGET_MODES_TIEABLE_P): Likewise.
143 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
145 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
148 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
149 Chung-Ju Wu <jasonwucj@gmail.com>
151 * config/nds32/constants.md (unspec_volatile_element): Add enum values
152 for unaligned access.
153 * config/nds32/nds32-intrinsic.c: Implementation of expanding
155 * config/nds32/nds32-intrinsic.md: Likewise.
156 * config/nds32/nds32_intrinsic.h: Likewise.
157 * config/nds32/nds32.h (nds32_builtins): Likewise.
158 * config/nds32/nds32.opt (munaligned-access): New option.
159 * config/nds32/nds32.c (nds32_asm_file_start): Display
160 flag_unaligned_access status.
162 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
164 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
165 -mno-relax is present.
166 * config/riscv/linux.h (LINK_SPEC): Ditto.
168 2018-04-20 Martin Sebor <msebor@redhat.com>
171 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
173 (gimple_fold_builtin_stxcpy_chk): Same.
174 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
176 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
179 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
180 __powikf2 when long double is IEEE 128-bit.
182 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
184 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
185 step to make sure stack always aligned.
187 2018-04-20 Carl Love <cel@us.ibm.com>
190 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
193 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
194 Tom de Vries <tom@codesourcery.com>
197 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
198 Emit insns for calls too.
199 (nvptx_find_par): Always look for worker-level predecessor insn.
200 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
202 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
203 (nvptx_process_pars): Propagate frames for calls.
205 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
208 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
210 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
211 (ix86_handle_option): Don't handle OPT_mibt.
212 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
214 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
216 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
217 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
218 (ix86_target_macros): Define __CET__ with flag_cf_protection
220 * config/i386/i386.c (isa2_opts): Remove -mibt.
221 * config/i386/i386.h (TARGET_IBT): Removed.
222 (TARGET_IBT_P): Likewise.
223 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
224 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
225 * config/i386/i386.opt (mcet): Update help message.
228 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
229 -mcet as an alias for -mshstk.
231 2018-04-20 Richard Biener <rguenther@suse.de>
234 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
235 complexity by forcing a single use of the multiply operand.
237 2018-04-20 Martin Jambor <mjambor@suse.cz>
240 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
241 recursion dependency to only apply to non-clones.
243 2018-04-20 Martin Jambor <mjambor@suse.cz>
246 * ipa-cp.c (create_specialized_node): Check that clones of
247 self-recursive edges exist during IPA-CP.
249 2018-04-19 Toon Moene <toon@moene.org>
251 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
254 2018-04-19 Jakub Jelinek <jakub@redhat.com>
256 PR tree-optimization/85467
257 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
258 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
259 VECTOR_CST element to type.
261 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
264 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
265 * config/i386/i386.md (builtin_setjmp_setup): Removed.
266 (builtin_longjmp): Likewise.
267 (save_stack_nonlocal): New pattern.
268 (restore_stack_nonlocal): Likewise.
270 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
273 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
274 Replace ASM_OUTPUT_LABEL with fprintf.
276 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
279 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
280 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
281 * config/i386/i386-c.c (ix86_target_macros_internal): Also
282 define __IBT__ and __SHSTK__ for -fcf-protection.
283 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
285 (ix86_trampoline_init): Likewise.
286 (x86_output_mi_thunk): Likewise.
287 (ix86_notrack_prefixed_insn_p): Likewise.
288 (ix86_option_override_internal): Don't disallow -fcf-protection.
289 * config/i386/i386.md (rdssp<mode>): Also enable for
291 (incssp<mode>): Likewise.
292 (nop_endbr): Likewise.
293 * config/i386/i386.opt (mcet): Change help message to built-in
297 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
298 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
299 enable CET built-in functions.
301 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
303 * common/config/i386/i386-common.c
304 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
305 OPTION_MASK_ISA_MOVDIRI_UNSET,
306 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
307 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
308 * config.gcc (movdirintrin.h): New header.
309 * config/i386/cpuid.h (bit_MOVDIRI,
310 bit_MOVDIR64B): New bits.
311 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
313 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
314 (VOID, PVOID, PCVOID)): New function types.
315 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
316 __builtin_ia32_directstoreu_u64,
317 __builtin_ia32_movdir64b): New builtins.
318 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
319 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
321 (ix86_valid_target_attribute_inner_p): Ditto.
322 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
323 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
324 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
325 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
326 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
327 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
328 (movdiri<mode>, movdir64b_<mode>): New.
329 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
330 * config/i386/immintrin.h: Include movdirintrin.h.
331 * config/i386/movdirintrin.h: New file.
332 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
334 2018-04-19 Richard Biener <rguenther@suse.de>
337 * cfg.c (clear_bb_flags): When loop state says we have
338 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
340 2018-04-19 Richard Biener <rguenther@suse.de>
342 PR tree-optimization/84737
343 * tree-vect-data-refs.c (vect_copy_ref_info): New function
344 copying restrict info.
345 (vect_setup_realignment): Use it.
346 * tree-vectorizer.h (vect_copy_ref_info): Declare.
347 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
348 the first DR to all generated stores.
349 (vectorizable_load): Likewise for loads.
351 2018-04-19 Jakub Jelinek <jakub@redhat.com>
353 PR tree-optimization/85446
354 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
355 the integral and pointer types to have the same precision.
357 * doc/install.texi: Document --disable-cet being the default and
360 2018-04-18 Martin Liska <mliska@suse.cz>
362 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
365 2018-04-18 Martin Liska <mliska@suse.cz>
368 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
371 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
372 arguments if they are comparable.
374 2018-04-18 Martin Liska <mliska@suse.cz>
377 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
380 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
383 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
386 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
387 ENDBR after calling __morestack.
389 2018-04-18 David Malcolm <dmalcolm@redhat.com>
392 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
393 by using gcc_base_ver to generate a gcc_driver_version, and use
394 it when generating GCC_DRIVER_NAME.
395 * configure: Regenerate.
397 2018-04-18 Jakub Jelinek <jakub@redhat.com>
400 * config.gcc: Obsolete powerpc*-*-*spe*.
402 2018-04-17 Jakub Jelinek <jakub@redhat.com>
405 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
406 (stabstr_D): Change type of unum from unsigned int to
407 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
410 2018-04-17 Jim Wilson <jimw@sifive.com>
413 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
414 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
415 Set arg_pointer_offset after using pretend_args_size.
417 2018-04-17 Jakub Jelinek <jakub@redhat.com>
419 PR rtl-optimization/85431
420 * dse.c (record_store): Ignore zero width stores.
423 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
424 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
425 __builtin_stack_restore rather than after it.
426 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
427 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
428 argument instead of virtual_dynamic_stack_rtx.
430 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
432 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
434 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
435 Add note to error message to explain internal mapping of overloaded
436 built-in function name to non-overloaded built-in function name.
437 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
440 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
443 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
444 where the inputs overlap with the output.
446 2018-04-17 Jakub Jelinek <jakub@redhat.com>
448 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
449 (=v, v) alternative and explicit "memory" attribute.
450 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
451 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
453 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
454 "sselog1" type instead of "sselog".
455 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
456 "sselog". Remove explicit "memory" attribute.
457 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
458 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
460 (vec_extract_hi_v32hi): Merge all alternatives into one, use
461 "sselog1" type instead of "sselog". Remove explicit "memory"
463 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
464 use "sselog1" type instead of "sselog". Remove explicit "memory"
466 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
467 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
469 (vec_extract_hi_v64qi): Merge all alternatives into one, use
470 "sselog1" type instead of "sselog". Remove explicit "memory"
472 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
473 use "sselog1" type instead of "sselog". Remove explicit "memory"
477 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
480 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
483 2018-04-17 Martin Jambor <mjambor@suse.cz>
486 * ipa-cp.c (create_specialized_node): Call
487 expand_all_artificial_thunks if necessary.
489 2018-04-17 Martin Liska <mliska@suse.cz>
492 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
493 in message, remote space in between '_G' and '('.
495 2018-04-17 Jakub Jelinek <jakub@redhat.com>
498 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
499 avx512f_vmcmp<mode>3<round_saeonly_name>,
500 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
501 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
502 avx512f_rndscale<mode><round_saeonly_name>,
503 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
504 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
505 Use %<iptr>2 instead of %2 for -masm=intel.
506 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
507 avx512f_vcvttss2usi<round_saeonly_name>,
508 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
510 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
511 avx512f_vcvttsd2usi<round_saeonly_name>,
512 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
513 Use %q1 instead of %1 for -masm=intel.
514 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
515 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
516 of %3 for -masm=intel.
517 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
519 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
521 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
523 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
524 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
526 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
528 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
529 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
530 %g1 and one with %0 and %1.
531 (avx512er_vmrcp28<mode><round_saeonly_name>,
532 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
534 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
535 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
536 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
537 of %0 and %{%4%} for -masm=intel.
538 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
539 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
540 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
541 order of %0 and %{%5%}%{z%} for -masm=intel.
543 2018-04-17 Jan Hubicka <jh@suse.cz>
546 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
548 2018-04-17 Martin Liska <mliska@suse.cz>
551 * multiple_target.c (create_dispatcher_calls): Set apostrophes
552 for target_clone error message. Make default implementation
553 clone to be a local declaration.
554 (separate_attrs): Add new argument and check for an empty
556 (expand_target_clones): Handle it.
557 (ipa_target_clone): Make redirection just for target_clones
560 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
561 Tom de Vries <tom@codesourcery.com>
564 * omp-expand.c (expand_oacc_for): Add dummy false branch for
565 tiled basic blocks without omp continue statements.
567 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
570 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
571 vec_extract expression as having side effects to make sure it gets
574 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
577 * config/i386/i386.c (get_builtin_code_for_version): Check
580 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
583 * gcc/config.gcc: Support "skylake".
584 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
586 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
587 (processor_target_table): Add "skylake".
588 (ix86_option_override_internal): Add "skylake".
589 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
590 PROCESSOR_CANNONLAKE.
591 (get_builtin_code_for_version): Fix priority for
592 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
593 PROCESSOR_SKYLAKE-AVX512.
594 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
595 (processor_type): Add PROCESSOR_SKYLAKE.
597 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
598 Jason Merrill <jason@redhat.com>
601 * convert.c (convert_to_integer_1): Use direct recursion for
602 enumeral types and types with a precision less than the number
603 of bits in their mode.
605 2018-04-16 Julia Koval <julia.koval@intel.com>
608 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
609 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
611 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
614 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
615 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
616 and -mno-direct-move.
618 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
621 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
622 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
623 Ensure negative shifts result in {0}.
625 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
627 PR rtl-optimization/79916
628 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
629 regs (if any) to define how to gnerate SD moves when LRA is in
632 2018-04-13 Jakub Jelinek <jakub@redhat.com>
634 PR rtl-optimization/85393
635 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
636 * except.c (expand_dw2_landing_pad_for_region): Make static.
637 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
638 a label and unconditional jump to old_bb, rather than
639 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
642 PR rtl-optimization/85376
643 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
644 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
645 instead of a specific value.
647 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
648 Bin Cheng <bin.cheng@arm.com>
650 PR tree-optimization/82965
651 PR tree-optimization/83991
652 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
653 by_profile_only parameter.
654 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
655 information if the loop was predicted to iterate too many times.
656 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
658 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
661 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
664 2018-04-13 Martin Liska <mliska@suse.cz>
665 Jakub Jelinek <jakub@redhat.com>
668 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
669 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
670 * builtins.c (expand_builtin_memory_copy_args): Use
671 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
672 handle dest_addr == pc_rtx.
674 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
677 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
678 asked to not generate direct moves.
679 (fix_trunc<mode>si2_stfiwx): Similar.
680 (fix_trunc<mode>si2_internal): Similar.
682 2018-04-12 Jakub Jelinek <jakub@redhat.com>
685 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
686 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
687 lookup if dest in some wider mode is known to be const0_rtx and
688 if so, record permanent equivalence for it to be ZERO_EXTEND of
689 the narrower mode destination.
691 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
693 * lto-streamer-out.c (output_function): Revert 259346.
694 * omp-expand.c (expand_oacc_for): Likewise.
696 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
698 PR rtl-optimization/85354
699 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
700 * sel-sched.c (sel_global_init): ... here.
702 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
705 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
706 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
707 mode for PE-COFF targets.
708 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
709 (i386_pe_asm_lto_end): Likewise.
710 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
711 (TARGET_ASM_LTO_END): Likewise.
712 * config/i386/winnt.c (saved_debug_info_level): New static variable.
713 (i386_pe_asm_lto_start): New function.
714 (i386_pe_asm_lto_end): Likewise.
716 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
717 Richard Biener <rguenther@suse.de>
720 * lto-streamer-out.c (output_function): Fix CFG loop state before
722 * omp-expand.c (expand_oacc_for): Handle calls to internal
723 functions like regular functions.
725 2018-04-12 Richard Biener <rguenther@suse.de>
728 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
729 for the early LTO debug to properly generate references to it
730 during DIE emission. Do not re-use that for the skeleton for
732 (dwarf2out_early_finish): Likewise.
734 2018-04-12 Jakub Jelinek <jakub@redhat.com>
738 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
739 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
740 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
741 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
742 and output is a reg, avoid creating invalid lowpart subreg, but
743 instead split into a 512-bit move. Don't split if not AVX512VL,
744 input is xmm16+ reg and output is a mem.
745 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
746 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
747 xmm16+ reg and output is a mem.
749 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
751 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
752 also for flag_dwarf2_cfi_asm.
754 2018-04-12 Jakub Jelinek <jakub@redhat.com>
756 PR rtl-optimization/85342
757 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
758 a bool scalar var inside of the loop instead. Don't try to update
759 recog_data.operand after failed apply_change_group.
761 2018-04-12 Tom de Vries <tom@codesourcery.com>
764 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
765 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
766 array with flexible array member as array without given dimension.
767 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
768 argument for undefined param to true.
770 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
773 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
774 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
775 from PowerPC section.
776 * config/rs6000/sysv4.opt (mcall-): Improve help text.
777 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
778 help text that is too long.
779 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
780 help text that is too long.
781 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
782 help text that is too long.
784 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
786 * config/alpha/alpha.md (stack_probe_internal): Rename
787 from "probe_stack". Update all callers.
789 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
791 PR rtl-optimization/84566
792 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
793 sched_macro_fuse_insns.
795 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
798 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
799 (compute_block_dependences): ... from here.
801 2018-04-11 Jakub Jelinek <jakub@redhat.com>
803 PR tree-optimization/85331
804 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
805 from int to HOST_WIDE_INT.
807 2018-04-11 Martin Jambor <mjambor@suse.cz>
810 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
811 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
812 not the same as the source val.
813 (cgraph_edge_brings_value_p): New parameter.
814 (gather_edges_for_value): Pass destination value to
815 cgraph_edge_brings_value_p.
816 (perhaps_add_new_callers): Likewise.
817 (get_info_about_necessary_edges): Likewise and exclude values brought
818 only by self-recursive edges.
819 (create_specialized_node): Redirect only clones of self-calling edges.
820 (+self_recursive_pass_through_p): New function.
821 (find_more_scalar_values_for_callers_subset): Use it.
822 (find_aggregate_values_for_callers_subset): Likewise.
823 (known_aggs_to_agg_replacement_list): Removed.
824 (decide_whether_version_node): Re-calculate known constants for all
825 remaining context clones.
827 2018-04-11 Richard Biener <rguenther@suse.de>
830 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
831 from early DWARF output.
832 (dwarf2out_early_finish): Output line info unconditionally into
833 early DWARF and add reference to it.
835 2018-04-11 Jakub Jelinek <jakub@redhat.com>
838 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
839 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
840 other than V2DFmode using iptr mode attribute.
841 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
843 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
845 PR rtl-optimization/84659
846 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
848 2018-04-11 Jakub Jelinek <jakub@redhat.com>
851 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
853 (output_loc_list): Pass address of a dummy size variable even in the
854 locview handling loop.
855 (index_location_lists): Add comment on why skip_loc_list_entry can't
858 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
861 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
864 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
867 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
868 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
869 and -mstring-compare-inline-limit.
871 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
874 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
875 for stack clash protection in a register whenever we need it to be in
878 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
880 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
881 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
883 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
886 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
888 (mlong-double-): Ditto.
889 * config/rs6000/sysv4.opt (msdata=): Ditto.
892 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
894 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
895 erroneous entries for
896 "vector int vec_ldl (int, long int *)", and
897 "vector unsigned int vec_ldl (int, unsigned long int *)".
898 Add comments and entries for
899 "vector bool char vec_ldl (int, bool char *)",
900 "vector bool short vec_ldl (int, bool short *)",
901 "vector bool int vec_ldl (int, bool int *)",
902 "vector bool long long vec_ldl (int, bool long long *)",
903 "vector pixel vec_ldl (int, pixel *)",
904 "vector long long vec_ldl (int, long long *)",
905 "vector unsigned long long vec_ldl (int, unsigned long long *)".
906 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
907 type tree bool_long_long_type_node and correct definition of
908 bool_V2DI_type_node to make reference to this new type tree.
909 (rs6000_mangle_type): Replace erroneous reference to
910 bool_long_type_node with bool_long_long_type_node.
911 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
912 comments to emphasize sign distinctions for char and int types and
913 replace RS6000_BTI_bool_long constant with
914 RS6000_BTI_bool_long_long constant. Also add comment to restrict
915 use of RS6000_BTI_pixel.
916 (bool_long_type_node): Remove this macro definition.
917 (bool_long_long_type_node): New macro definition
919 2018-04-10 Jakub Jelinek <jakub@redhat.com>
921 PR rtl-optimization/85300
922 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
923 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
924 simplify_unary_operation fails.
926 2018-04-10 Martin Liska <mliska@suse.cz>
928 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
929 cgraph_edge and ipa_ref.
931 2018-04-10 Jakub Jelinek <jakub@redhat.com>
936 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
937 computation of the VEC_MERGE selector from mask.
938 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
939 Fix decoding of the VEC_MERGE selector into mask.
941 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
943 PR tree-optimization/85286
944 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
946 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
948 * final.c (final_1): Set insn_last_address as well as
949 insn_current_address.
951 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
954 * explow.c (emit_stack_probe): Call validize_mem on memory location
955 before passing it to gen_probe_stack. Create address operand and
956 legitimize it for the probe_stack_address case.
958 2018-04-09 Jan Hubicka <jh@suse.cz>
961 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
962 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
963 * tree.c (free_lang_data_in_type): Fix handling of binfos;
965 (free_lang_data): Rebuild type inheritance graph.
967 2018-04-09 Martin Sebor <msebor@redhat.com>
969 * invoke.texi (-finline-small-functions): Mention other optimization
971 (-findirect-inlining, -fpartial-inlining): Same.
972 (-finline-functions-called-once): Same.
973 (-freorder-blocks-and-partition): Same.
975 2018-04-09 Jan Hubicka <jh@suse.cz>
978 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
979 jumps; choose last target that matches the criteria (i.e.
980 no partition changes for non-crossing jumps).
981 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
982 support for redirecting crossing jumps to non-crossing.
984 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
986 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
987 also for naked functions.
989 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
991 * config/arc/arc.md (add_shift): New pattern.
992 (add_shift2): Likewise.
993 (sub_shift): Likewise.
994 (sub_shift_cmp0_noout): Likewise.
995 (compare_si_ashiftsi): Likewise.
996 (xbfu_cmp0_noout): New combine pattern.
997 (xbfu_cmp0"): Likewise.
998 (movsi_set_cc_insn): Place the predicable variant first.
999 (commutative_binary_cmp0_noout): Remove clobber.
1000 (commutative_binary_cmp0): New pattern.
1001 (noncommutative_binary_cmp0): Likewise.
1002 (noncommutative_binary_cmp0_noout): Likewise.
1003 (noncommutative_binary_comparison_result_used): Removed.
1004 (rsub_cmp0): New pattern.
1005 (rsub_cmp0_noout): Likewise.
1006 (extzvsi): Changed, keep only meaningful variants.
1007 (SQH, SEZ): New iterators.
1008 (SQH_postfix): New mode attribute.
1009 (SEZ_prefix): New code attribute.
1010 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
1011 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
1012 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
1014 (noncommutative_operator): Check the availability of barrel
1017 2018-04-09 Richard Biener <rguenther@suse.de>
1019 PR tree-optimization/85284
1020 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
1021 Only use the niter constraining form of simple_iv when the exit
1024 2018-04-09 Tom de Vries <tom@codesourcery.com>
1027 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1028 (define_expand "*memory_barrier"): New define_expand.
1029 (define_insn "memory_barrier"): New insn.
1031 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1033 PR rtl-optimization/80463
1034 PR rtl-optimization/83972
1035 PR rtl-optimization/83480
1037 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
1038 correct producer for the insn.
1039 (tidy_control_flow): Fixup seqnos in case of debug insns.
1041 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1043 PR rtl-optimization/83913
1045 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
1046 different sched-times when merging exprs.
1048 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1050 PR rtl-optimization/83962
1052 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
1053 tidy_fallthru_edge and tidy_control_flow.
1055 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1057 PR rtl-optimization/83530
1059 * sel-sched.c (force_next_insn): New global variable.
1060 (remove_insn_for_debug): When force_next_insn is true, also leave only
1061 next insn in the ready list.
1062 (sel_sched_region): When the region wasn't scheduled, make another pass
1063 over it with force_next_insn set to 1.
1065 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1067 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
1069 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1070 for interrupt control.
1071 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1072 functions for interrupt control.
1073 * config/nds32/nds32-intrinsic.md: Likewise.
1074 * config/nds32/nds32_intrinsic.h: Likewise.
1075 * config/nds32/nds32.h (nds32_builtins): Likewise.
1077 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1079 * config/nds32/nds32.c (nds32_init_machine_status,
1080 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
1081 strict_aligned_p field.
1082 (nds32_expand_to_rtl_hook): New function.
1083 (TARGET_EXPAND_TO_RTL_HOOK): Define.
1084 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
1086 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1087 Chung-Ju Wu <jasonwucj@gmail.com>
1089 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
1090 * config/nds32/nds32-n7.md: New file.
1091 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
1092 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
1094 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
1095 * config/nds32/nds32.md (pipeline_model): Add n7.
1096 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
1097 * config/nds32/pipelines.md: Include n7 settings.
1099 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1100 Chung-Ju Wu <jasonwucj@gmail.com>
1102 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
1103 * config/nds32/nds32-e8.md: New file.
1104 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
1105 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
1107 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
1108 * config/nds32/nds32.md (pipeline_model): Add e8.
1109 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
1110 * config/nds32/pipelines.md: Include e8 settings.
1112 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1113 Chung-Ju Wu <jasonwucj@gmail.com>
1115 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
1116 * config/nds32/nds32-n8.md: New file.
1117 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
1118 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
1120 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
1121 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
1122 * config/nds32/nds32.md (pipeline_model): Add n8.
1123 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
1124 * config/nds32/pipelines.md: Include n8 settings.
1126 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1127 Chung-Ju Wu <jasonwucj@gmail.com>
1129 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
1130 * config/nds32/nds32-n9-2r1w.md: New file.
1131 * config/nds32/nds32-n9-3r2w.md: New file.
1132 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
1133 nds32_register_ports): New or modify for cpu n9.
1134 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
1136 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
1137 * config/nds32/nds32-utils.c: New file.
1138 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
1139 TARGET_MUL_SLOW): Define.
1140 * config/nds32/nds32.md (pipeline_model): New attribute.
1141 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
1142 New options that support cpu n9.
1143 * config/nds32/pipelines.md: Include n9 settings.
1144 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
1146 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1148 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
1149 information if necessary.
1150 (output_cond_branch_compare_zero): Likewise.
1151 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
1152 (nds32_target_alignment): Refine for alignment.
1153 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
1154 (FUNCTION_BOUNDARY): Modify.
1155 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
1157 * config/nds32/nds32.opt (malways-align, malign-functions): New.
1159 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1161 * config/nds32/constants.md (unspec_volatile_element): Add values for
1162 TLB operation and data prefetch.
1163 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1164 functions for TLB operation and data prefetch.
1165 * config/nds32/nds32-intrinsic.md: Likewise.
1166 * config/nds32/nds32_intrinsic.h: Likewise.
1167 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
1168 (nds32_print_operand): Likewise.
1169 * config/nds32/nds32.h (nds32_builtins): Likewise.
1171 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
1172 Andrew Pinski <pinsika@gcc.gnu.org>
1175 * match.pd: Use constant_boolean_node of correct type instead of
1176 boolean_true_node or boolean_false_node for simplifying
1177 pointer comparisons to zero.
1179 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1181 PR tree-optimization/80021
1182 * tree.c (verify_type_variant): Make error call in verify_variant_match
1183 translatable and remove final full stop.
1185 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1187 * config/nds32/constants.md (unspec_volatile_element): Add
1188 UNSPEC_VOLATILE_EH_RETURN.
1189 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
1190 nds32_output_stack_pop): Support dwarf exception handling process.
1191 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
1192 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
1193 exception handling process.
1194 (nds32_compute_stack_frame): Likewise.
1195 (nds32_return_addr_rtx): Likewise.
1196 (nds32_initial_elimination_offset): Likewise.
1197 (nds32_expand_prologue): Likewise.
1198 (nds32_expand_epilogue): Likewise.
1199 (nds32_dynamic_chain_address): New function.
1200 * config/nds32/nds32.h (machine_function): Add fields for dwarf
1202 (DYNAMIC_CHAIN_ADDRESS): Define.
1203 (EH_RETURN_DATA_REGNO): Define.
1204 (EH_RETURN_STACKADJ_RTX): Define.
1205 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
1206 patterns for dwarf exception handling.
1208 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1210 * config/nds32/nds32.h: Clean up obsolete macros.
1212 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1214 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1215 Add enum values for particular instructions.
1216 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1217 particular intrinsic functions.
1218 * config/nds32/nds32-intrinsic.md: Likewise.
1219 * config/nds32/nds32_intrinsic.h: Likewise.
1220 * config/nds32/nds32.h (nds32_builtins): Likewise.
1221 * config/nds32/nds32.md (type): Add pbsad and pbsada.
1222 (btst, ave): New patterns for particular instructions.
1224 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1226 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1227 Add enum values for atomic load/store and memory sync.
1228 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
1230 * config/nds32/nds32-intrinsic.md: Likewise.
1231 * config/nds32/nds32_intrinsic.h: Likewise.
1232 * config/nds32/nds32.h (nds32_builtins): Likewise.
1234 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1236 PR tree-optimization/85257
1237 * fold-const.c (native_encode_vector): If not all elts could fit
1238 and off is -1, return 0 rather than offset.
1239 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
1240 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
1241 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
1242 adjust buffer in native_interpret_expr call.
1244 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1246 * config/nds32/constants.md (unspec_volatile_element): Add cache
1247 control enum values.
1248 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
1249 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
1250 * config/nds32/nds32.c (nds32_cctl_names): New.
1251 (nds32_print_operand): Handle cache control register names.
1252 * config/nds32/nds32.h (nds32_builtins): New enum values.
1253 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
1255 * config/nds32/nds32.md (type): Add mmu.
1256 * config/nds32/pipelines.md (simple_insn): Add mmu.
1258 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1260 * config/nds32/nds32.md (type): Remove call.
1261 * config/nds32/pipelines.md (simple_insn): Likewise.
1263 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1265 * config/nds32/constants.md (unspec_volatile_element): Add
1266 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
1267 UNSPEC_VOLATILE_FMFCFG.
1268 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
1269 description for fmfcfg and fmfcsr.
1270 (bdesc_1arg): Add fmtcsr.
1271 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
1272 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
1273 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
1274 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
1275 unspec_fmfcfg): New patterns.
1276 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
1277 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
1278 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
1279 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
1280 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
1281 __nds32__fmfcfg): Define.
1283 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1285 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
1286 intrinsic register names.
1287 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
1288 intrinsic register enum values and macros.
1290 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1292 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
1293 for load/store addressing form.
1294 (nds32_print_operand_address): Likewise.
1296 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
1299 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
1300 based on LABEL_REF. Remove useless assertion.
1301 (pic_address_needs_scratch): Fix formatting.
1302 (sparc_legitimize_pic_address): Minor tweaks.
1303 (sparc_delegitimize_address): Adjust assertion accordingly.
1304 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
1305 into symbolic_operand.
1306 (movsi_high_pic_label_ref): Likewise.
1307 (movsi_lo_sum_pic_label_ref): Likewise.
1308 (movdi_pic_label_ref): Likewise.
1309 (movdi_high_pic_label_ref): Likewise.
1310 (movdi_lo_sum_pic_label_ref): Likewise.
1312 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
1314 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
1315 custom LIB_SPEC setup.
1317 2018-04-06 Ruslan Bukin <br@bsdpad.com>
1318 Kito Cheng <kito.cheng@gmail.com>
1320 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
1321 * config/riscv/freebsd.h: New.
1323 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1325 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
1326 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
1329 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1330 Kito Cheng <kito.cheng@gmail.com>
1332 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
1333 nds32_output_call, nds32_symbol_binds_local_p): New functions.
1334 * config/nds32/nds32-protos.h (nds32_output_call,
1335 nds32_output_return): Declare.
1336 * config/nds32/nds32.md: Refine all the call and return patterns.
1338 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1341 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
1342 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
1344 PR rtl-optimization/84872
1345 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
1346 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
1349 2018-04-06 Tamar Christina <tamar.christina@arm.com>
1351 * expr.c (copy_blkmode_to_reg): Revert 254862.
1352 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
1354 2018-04-06 Richard Biener <rguenther@suse.de>
1357 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
1358 after seeing a component reference with an adjacent field. Treat
1359 refs to arrays at struct end of external decls similar to
1360 refs to unconstrained commons.
1362 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1365 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
1366 look through SAVE_EXPRs with non-side-effects argument. Adjust
1368 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
1371 2018-04-06 Richard Biener <rguenther@suse.de>
1374 * alias.c (find_base_term): New wrapper around find_base_term
1375 unwinding CSELIB_VAL_PTR changes.
1376 (find_base_term): Do not restore CSELIB_VAL_PTR during the
1379 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1381 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
1383 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
1384 constant definitions.
1385 ("nop"): lr 0,0 -> nopr r0
1386 ("nop_lr0", "nop_lr1"): New insn definitions.
1388 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1390 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
1391 NDS32_V3PUSH_AVAILABLE_P macro.
1393 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
1394 Chung-Ju Wu <jasonwucj@gmail.com>
1396 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
1397 (nds32*-*-*): Add float and fpu_config into supported_defaults.
1398 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
1399 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
1400 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
1401 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
1402 * config/nds32/constraints.md: New constraints and checking for hard
1403 float configuration.
1404 * config/nds32/iterators.md: New mode iterator and attribute for hard
1405 float configuration.
1406 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
1408 * config/nds32/nds32-fpu.md: New file.
1409 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
1410 deal with hard float code generation.
1411 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
1413 (abi_type, float_reg_number): New enum type.
1414 * config/nds32/nds32-predicates.c: New predicates for hard float.
1415 * config/nds32/nds32-protos.h: Declare functions for hard float.
1416 * config/nds32/nds32.c: Implementation for hard float configuration.
1417 * config/nds32/nds32.h: Definitions for hard float configuration.
1418 * config/nds32/nds32.md: Include hard float machine description and
1419 modify patterns for hard float configuration.
1420 * config/nds32/nds32.opt: New options for hard float configuration.
1421 * config/nds32/predicates.md: New predicates for hard float
1424 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1426 * common/config/nds32/nds32-common.c
1427 (nds32_option_optimization_table): Enable -mreleax-hint by default.
1429 2018-04-05 Jakub Jelinek <jakub@redhat.com>
1432 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
1433 CONSTRUCTOR_ELT (ctor, ...)->value.
1435 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
1438 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
1440 2018-04-05 Tom de Vries <tom@codesourcery.com>
1443 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
1446 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1447 Kito Cheng <kito.cheng@gmail.com>
1449 * config/nds32/constraints.md (U33): Fine-tune checking condition.
1450 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
1451 * config/nds32/nds32.h (nds32_16bit_address_type): Add
1452 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
1454 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1455 Kito Cheng <kito.cheng@gmail.com>
1457 * config/nds32/constraints.md (Ufe): New memory constraint.
1458 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
1459 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
1460 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
1462 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
1463 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
1465 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1467 * config/nds32/nds32.md: Use optimize_size in the condition for
1468 alu-shift instructions.
1470 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1472 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
1474 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1476 * config/nds32/nds32.md (negsi2): Refine pattern.
1478 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1479 Chung-Ju Wu <jasonwucj@gmail.com>
1481 * config/nds32/iterators.md (shift_rotate): New code iterator.
1482 (shift): New code attribute.
1483 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
1484 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
1485 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
1486 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
1487 bit-wise operations.
1488 (andsi3, *andsi3): Ditto.
1489 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
1490 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
1491 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
1492 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
1493 nds32_ior_operand, nds32_xor_operand): New predicates.
1495 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1497 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
1498 (addsi3, subsi3): ... this.
1500 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1502 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
1504 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1506 * config/nds32/nds32.md: Adjust indention.
1508 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1510 * config/nds32/nds32.md (feature): New attribute.
1512 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1514 * config/nds32/nds32.md (subtype): New attribute.
1516 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
1519 * config/arm/arm-builtins.c (arm_expand_builtin): Change
1520 expansion to perform a bitwise AND of the argument followed by a
1521 boolean negation of the result.
1523 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
1525 PR rtl-optimization/84878
1526 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
1527 the basic block. Assert the use reference is not artificial and that
1528 it has an associated insn.
1530 2018-04-04 Michael Matz <matz@suse.de>
1532 * builtins.c (compute_objsize): Pass correct operand
1533 to array_at_struct_end_p.
1535 2018-04-04 Richard Biener <rguenther@suse.de>
1538 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
1539 from contexts for DINFO_LEVEL_TERSE and below.
1541 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1543 * config/nds32/nds32-doubleword.md (move_<mode>): Require
1544 resiter_operand condition.
1545 * config/nds32/nds32.md (*move<mode>): Ditto.
1547 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1548 Monk Chiang <sh.chiang04@gmail.com>
1550 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
1552 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1554 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
1556 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1557 Kito Cheng <kito.cheng@gmail.com>
1559 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
1560 nds32_cond_code_str, output_cond_branch,
1561 output_cond_branch_compare_zero, nds32_expand_cbranch,
1562 nds32_expand_cstore, nds32_expand_movcc,
1563 nds32_output_cbranchsi4_equality_zero,
1564 nds32_output_cbranchsi4_equality_reg,
1565 nds32_output_cbranchsi4_equality_reg_or_const_int,
1566 nds32_output_cbranchsi4_greater_less_zero: New functions.
1567 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
1568 nds32_expand_cstore, nds32_expand_movcc,
1569 nds32_output_cbranchsi4_equality_zero,
1570 nds32_output_cbranchsi4_equality_reg,
1571 nds32_output_cbranchsi4_equality_reg_or_const_int,
1572 nds32_output_cbranchsi4_greater_less_zero): Declare.
1573 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
1574 nds32_rimm11s_operand): New predicates.
1575 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
1576 * config/nds32/nds32.md: Rewrite all the branch and conditional move
1579 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1581 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
1582 * config/nds32/nds32.md: Ditto.
1583 * config/nds32/pipelines.md: Ditto.
1585 2018-04-04 Richard Biener <rguenther@suse.de>
1587 PR tree-optimization/85168
1588 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
1589 propagating abnormals.
1591 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1593 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
1595 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1596 Kito Cheng <kito.cheng@gmail.com>
1598 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
1599 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
1600 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
1601 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
1602 * config/nds32/nds32.md (sibcall_internal): New.
1603 (sibcall_register): Remove.
1604 (sibcall_immediate): Remove.
1605 (sibcall_value_internal): New.
1606 (sibcall_value_register): Remove.
1607 (sibcall_value_immediate): Remove.
1608 * config/nds32/predicates.md (nds32_general_register_operand): New.
1609 (nds32_call_address_operand): New.
1611 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1613 PR rtl-optimization/85167
1614 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
1615 bb_defs if *split_p, instead preinitialize it to NULL.
1617 PR tree-optimization/85156
1618 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
1619 evaluating the argument multiple times.
1621 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
1623 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
1625 (_mm_cvtpd_ps): Likewise.
1626 (_mm_cvttpd_epi32): Likewise.
1627 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
1628 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
1629 vector, pixel, and bool following altivec.h include.
1631 2018-04-03 Martin Sebor <msebor@redhat.com>
1633 * doc/extend.texi (Common Function Attributes): Clarify.
1634 (const attribute): Likewise.
1635 (pure attribute): Likewise.
1637 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1640 * config/i386/i386.c (ix86_expand_vector_set): Use
1641 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
1643 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
1645 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
1646 instructions when changing rounding bits to preserve precision bits
1647 in the x87 control word.
1649 2018-04-03 Martin Liska <mliska@suse.cz>
1651 PR tree-optimization/82491
1652 * rtl.h (strip_offset_and_add): Replace += suboffset with
1653 poly_uint64 () + suboffset.
1655 2018-03-29 Martin Liska <mliska@suse.cz>
1656 Martin Jambor <mjambor@suse.cz>
1659 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
1660 param_type is not an integral or pointer type.
1662 2018-04-03 Richard Biener <rguenther@suse.de>
1664 * sese.h (recompute_all_dominators): Remove.
1666 2018-04-02 Martin Sebor <msebor@redhat.com>
1668 * doc/invoke.texi (-Wrestrict): Fix typos.
1670 2018-04-02 Jim Wilson <jimw@sifive.com>
1672 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
1673 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
1674 (<optab>di3, <optab>si3_extend): Likewise.
1675 (<optab>si3_mask, <optab>si3_mask_1): New.
1676 (<optab>di3_mask, <optab>di3_mask_1): New.
1677 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
1678 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
1679 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
1681 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
1683 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
1686 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
1688 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
1689 (nds32_canonicalize_comparison): New function.
1691 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1692 Kito Cheng <kito.cheng@gmail.com>
1693 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1695 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
1696 * config/nds32/constants.md (unspec_volatile_element): Add
1697 UNSPEC_VOLATILE_RELAX_GROUP.
1698 * config/nds32/nds32-relax-opt.c: New file.
1699 * config/nds32/nds32-predicates.c
1700 (nds32_symbol_load_store_p): New function.
1701 * config/nds32/nds32-protos.h
1702 (nds32_symbol_load_store_p): Declare function.
1703 (make_pass_nds32_relax_opt): Declare new rtl pass function.
1704 * config/nds32/nds32.c
1705 (nds32_register_pass): New function to register pass.
1706 (nds32_register_passes): New function to register passes.
1707 * config/nds32/nds32.md (relax_group): New pattern.
1708 * config/nds32/nds32.opt (mrelax-hint): New option.
1709 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
1711 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
1713 * config/nds32/t-nds32: Modify files dependency.
1715 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1717 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
1718 (PROFILE_HOOK): Define its implementation.
1720 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1722 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
1723 type and 32-bit size.
1725 2018-04-01 Jakub Jelinek <jakub@redhat.com>
1728 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
1729 (V_128_256): New mode iterator.
1730 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
1731 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
1732 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
1734 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
1735 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
1737 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
1740 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
1741 NaN inputs correctly.
1743 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
1746 * config/rs6000/vsx.md (??r): New mode attribute.
1747 (*vsx_mov<mode>_64bit): Use it.
1748 (*vsx_mov<mode>_32bit): Likewise.
1750 2018-03-30 Martin Sebor <msebor@redhat.com>
1752 PR tree-optimization/84818
1753 * builtins.c (check_access): Use warning_n.
1755 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1758 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
1760 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
1763 2018-03-30 Julia Koval <julia.koval@intel.com>
1766 * x86-tune.def (movx, partial_reg_dependency): Enable for
1769 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
1772 * lra-constraints.c (process_alt_operands): Move setting
1773 this_alternative_matches below.
1775 2018-03-29 Martin Liska <mliska@suse.cz>
1778 * doc/invoke.texi: Document how LTO works with debug info.
1779 Describe auto-load support of binutils. Mention 'x86-64'
1780 as valid option value of -march option.
1782 2018-03-29 Jakub Jelinek <jakub@redhat.com>
1784 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
1787 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
1788 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
1789 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
1792 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1795 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
1796 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
1797 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
1798 for RS6000_BTM_POWERPC64.
1799 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
1800 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
1801 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
1806 2018-03-28 Carl Love <cel@us.ibm.com>
1809 2017-09-27 Carl Love <cel@us.ibm.com>
1811 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
1812 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
1813 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
1816 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1818 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
1819 instead of __vector bool.
1820 (_mm_max_pu8): Likewise.
1821 (_mm_min_pi16): Likewise.
1823 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1826 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
1827 (DIVWEUO): Likewise.
1829 (DIVDEUO): Likewise.
1830 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
1831 DIVWEUO and DIVDEUO.
1832 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
1833 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
1834 (div_extend): Likewise.
1835 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
1837 (__builtin_divweuo): Likewise.
1838 (__builtin_divdeo): Likewise.
1839 (__builtin_divdeuo): Likewise.
1841 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1844 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
1845 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
1847 PR tree-optimization/82004
1848 * gimple-match-head.c (optimize_pow_to_exp): New function.
1849 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
1850 Don't fold to exp if optimize_pow_to_exp is false.
1852 2018-03-28 Martin Liska <mliska@suse.cz>
1855 * calls.c (initialize_argument_information): Fix trailing space.
1856 * common.opt: Fix typo and provide better explanation for
1857 -fsanitize-coverage option.
1858 * config/i386/i386.opt: Fix typo.
1860 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1861 Martin Liska <mliska@suse.cz>
1864 * gimplify.c (asan_poison_variable): Don't do the check for
1865 gimplify_omp_ctxp here.
1866 (gimplify_decl_expr): Do it here.
1867 (gimplify_target_expr): Likewise.
1869 2018-03-28 Martin Liska <mliska@suse.cz>
1872 * config/i386/i386.c (ix86_function_arg_advance): Do not call
1873 chkp_type_bounds_count if MPX is not enabled.
1875 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
1877 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
1879 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
1882 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
1883 function to create the function decl for complex long double
1884 multiply and divide for -mabi=ieeelongdouble.
1885 (init_float128_ieee): Call it.
1887 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
1890 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
1891 -fcf-protection=branch -mibt.
1892 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
1894 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1897 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
1899 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
1902 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
1903 extern array declarations.
1905 2018-03-27 Richard Biener <rguenther@suse.de>
1908 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
1909 explicit single_use checks.
1911 2018-03-27 Richard Biener <rguenther@suse.de>
1913 PR tree-optimization/85082
1914 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
1917 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1919 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
1920 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
1921 Turn on fasynchronous-unwind-tables and funwind-tables.
1923 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
1926 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
1927 (*bmi_blsr_<mode>_ccz): Ditto.
1929 2018-03-26 Tom de Vries <tom@codesourcery.com>
1931 PR tree-optimization/85063
1932 * omp-general.c (offloading_function_p): New function. Factor out
1934 * omp-offload.c (pass_omp_target_link::gate): ... here.
1935 * omp-general.h (offloading_function_p): Declare.
1936 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
1937 with attribute omp declare target for offloading functions.
1939 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
1941 PR tree-optimization/84005
1942 * tree-data-ref.h (get_base_for_alignment): Declare.
1943 * tree-data-ref.c (get_base_for_alignment_1): New function.
1944 (get_base_for_alignment): Likewise.
1945 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
1946 get_base_for_alignment to find a suitable base object, instead
1947 of always using drb->base_address.
1949 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1952 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
1953 known size by default.
1955 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
1958 * lra-constraints.c (process_alt_operands): Don't match BLKmode
1959 and non BLKmode operands.
1961 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1964 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
1965 Clean up attributes.
1967 2018-03-23 Richard Biener <rguenther@suse.de>
1970 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
1971 we are going to emit early debug for LTO.
1973 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1976 * function.c (match_asm_constraints_1): Don't optimize if input
1977 doesn't satisfy general_operand predicate for output's mode.
1980 * alias.c (write_dependence_p): Don't require for x_canonicalized
1981 non-VOIDmode if x has VOIDmode.
1984 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
1985 just don't try to optimize it rather than assert it never happens.
1987 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1989 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
1990 macro expansions for definition of ST_INTERNAL_<mode> and
1991 LD_INTERNAL_<mode> builtins.
1992 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
1994 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
1996 (altivec_expand_st_builtin): Likewise.
1997 (altivec_expand_builtin): Remove calls to deleted functions.
1998 (rs6000_address_for_altivec): Delete this function.
1999 * config/rs6000/vector.md: Remove expands for
2000 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
2002 2018-03-22 Sudakshina Das <sudi.das@arm.com>
2005 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
2006 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
2007 re-computing once computed.
2008 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
2009 (arm_init_machine_status): Initialize
2010 machine->static_chain_stack_bytes.
2012 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2015 * doc/extend.texi: Add four new prototypes for vec_ld.
2016 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
2017 definitions for more logical presentation.
2018 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
2019 entries for V1TI variants of __builtin_altivec_ld builtin.
2020 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
2021 handling of V1TI variant of LVX icode pattern.
2022 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
2023 (rs6000_gimple_fold_builtin): Likewise.
2024 (altivec_init_builtins): Add code to define
2025 __builtin_altivec_lvx_v1ti function.
2027 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2030 * function.c (match_asm_constraints_1): Don't do the optimization
2031 if input isn't a REG, SUBREG, MEM or constant.
2033 2018-03-22 Tom de Vries <tom@codesourcery.com>
2035 PR tree-optimization/84956
2036 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
2037 bb_has_abnormal_pred.
2039 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2042 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
2043 DECL_INITIAL (decl) to decl at the end.
2044 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
2047 2018-03-21 Joseph Myers <joseph@codesourcery.com>
2049 * doc/extend.texi (__builtin_tgmath): Document when complex
2050 integer types are treated as _Complex _Float64.
2052 2018-03-21 Tom de Vries <tom@codesourcery.com>
2054 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
2056 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2058 PR tree-optimization/84960
2059 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
2060 if it is ENTRY block, move them into single succ of ENTRY in that case.
2062 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
2064 PR tree-optimization/84811
2065 * poly-int.h (poly_span_traits): Remove the T3 parameter and
2066 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
2067 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
2068 (known_subrange_p): Update accordingly. Cast each value involved
2069 in the size comparison, rather than casting the result of the
2072 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2074 PR tree-optimization/84982
2075 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
2076 by flipping the least significant bit rather than all bits from
2077 bitpos to bitpos + bitsize - 1.
2079 2018-03-21 Nathan Sidwell <nathan@acm.org>
2081 * doc/extend.texi (Deprecated Features): Remove mention of
2082 long-deleted deprecations.
2084 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2087 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
2088 * configure: Regenerate.
2090 2018-03-21 Tom de Vries <tom@codesourcery.com>
2092 PR tree-optimization/83126
2093 * tree-parloops.c (num_phis): New function.
2094 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
2096 2018-03-21 Nathan Sidwell <nathan@acm.org>
2098 * doc/extend.texi (Deprecated Features): Update deprecated flags,
2099 mention anon-struct/union members and trailing attributes.
2101 2018-03-21 Bin Cheng <bin.cheng@arm.com>
2103 PR tree-optimization/84969
2104 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
2105 builtin memset partitions if they set different rhs values.
2107 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2109 PR rtl-optimization/84989
2110 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
2111 VEC_DUPLICATE with scalar result mode.
2113 2018-03-21 Martin Liska <mliska@suse.cz>
2116 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
2117 not intended return statement.
2119 2018-03-21 Martin Liska <mliska@suse.cz>
2122 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
2123 (chkp_find_bound_slots_1): Limit number of iterations.
2125 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
2128 * Minor grammar fixes for x86 options.
2130 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2133 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
2134 holding REG_CFA_RESTORE notes, instead turn them into a USE.
2136 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
2139 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
2140 (altivec_lvx_<mode>_1op): Likewise.
2141 (altivec_stvx_<mode>_2op): Likewise.
2142 (altivec_stvx_<mode>_1op): Likewise.
2143 (altivec_lvx_<VM2:mode>): New define_expand.
2144 (altivec_stvx_<VM2:mode>): Likewise.
2145 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
2146 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2147 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
2148 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2149 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
2150 (rs6000_gen_lvx): Likewise.
2151 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
2152 (altivec_expand_stv_builtin): Likewise.
2153 (altivec_expand_builtin): Likewise.
2154 * config/rs6000/vector.md: Likewise.
2156 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2159 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
2162 2018-03-20 Richard Biener <rguenther@suse.de>
2165 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
2166 sign-conversions as zero, fall back to standard scalar_stmt
2169 2018-03-20 Martin Liska <mliska@suse.cz>
2172 * predict.c (rebuild_frequencies): Handle case when we have
2173 PROFILE_ABSENT, but flag_guess_branch_prob is false.
2175 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2178 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
2179 flag_section_anchors.
2180 * varasm.c (use_blocks_for_decl_p): Remove hack for
2181 dw2_force_const_mem.
2184 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
2186 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
2187 be created, use lowpart_subreg of operands[0] rather than operands[0]
2189 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
2190 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
2191 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
2192 and n constraint instead of aarch64_shift_imm_di and Usd.
2193 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
2194 (*aarch64_<optab>_reg_minus<mode>3): ... this.
2196 2018-03-20 Sudakshina Das <sudi.das@arm.com>
2199 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
2200 to favor GPR over NEON registers.
2201 (<shift>di3_neon): Likewise.
2203 2018-03-20 Tom de Vries <tom@codesourcery.com>
2206 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
2207 (nvptx_process_pars): Emit bar.sync asap and alap.
2209 2018-03-20 Tom de Vries <tom@codesourcery.com>
2212 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
2213 seen_label if seen_label is already set.
2215 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2218 * config/i386/i386.c (fold_builtin_cpu): For features above 31
2219 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
2220 Use 1U instead of 1. Formatting fixes.
2223 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
2224 instead of TREE_TYPE (s1) for the return value.
2226 2018-03-19 Jakub Jelinek <jakub@redhat.com>
2228 PR tree-optimization/84946
2229 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
2230 bitsize + bitsize in poly_uint64 rather than poly_int64.
2233 * dwarf2asm.c: Include fold-const.c.
2234 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
2235 of decl rather than decl itself.
2237 PR rtl-optimization/84643
2238 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
2240 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
2243 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
2244 calling assemble_variable.
2246 2018-03-19 Sudakshina Das <sudi.das@arm.com>
2249 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
2250 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
2252 2018-03-19 Jim Wilson <jimw@sifive.com>
2255 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
2256 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
2257 (riscv_first_stack_step): Likewise.
2258 (riscv_option_override): Use STACK_BOUNDARY instead of
2260 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
2262 (BIGGEST_ALIGNMENT): Set to 128.
2263 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
2264 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
2267 2018-03-19 Richard Biener <rguenther@suse.de>
2269 PR tree-optimization/84933
2270 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
2271 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
2273 2018-03-19 Richard Biener <rguenther@suse.de>
2275 PR tree-optimization/84859
2276 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
2277 (cond_if_else_store_replacement): Perform sinking operation on
2278 single-store BBs regardless of MAX_STORES_TO_SINK setting.
2279 Generalize what a BB with a single eligible store is.
2281 2018-03-19 Richard Biener <rguenther@suse.de>
2283 PR tree-optimization/84929
2284 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
2285 chrec_is_positive against non-chrec arg.
2287 2018-03-19 Tamar Christina <tamar.christina@arm.com>
2290 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
2292 2018-03-18 Martin Liska <mliska@suse.cz>
2294 PR rtl-optimization/84635
2295 * regrename.c (build_def_use): Use matches_mode only when
2298 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
2300 PR tree-optimization/84913
2301 * tree-vect-loop.c (vectorizable_reduction): Don't try to
2302 vectorize chains of COND_EXPRs.
2304 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2306 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
2308 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2310 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
2312 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2314 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
2316 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2317 Kito Cheng <kito.cheng@gmail.com>
2319 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
2320 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
2321 (nds32_adjust_reg_alloc_order): New function.
2322 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
2324 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
2326 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
2327 nds32_print_operand, nds32_print_operand_address): Use
2328 HOST_WIDE_INT_PRINT_DEC instead.
2330 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2332 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
2334 2018-03-17 Jakub Jelinek <jakub@redhat.com>
2337 * config/i386/i386.c (initial_ix86_tune_features,
2338 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
2340 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
2341 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
2342 rather than 1u << ix86_tune. Formatting fix.
2343 (ix86_option_override_internal): Change ix86_arch_mask from
2344 unsigned int to unsigned HOST_WIDE_INT, initialize to
2345 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
2346 (ix86_function_specific_restore): Likewise.
2348 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2351 * postreload.c (reload_combine_recognize_pattern): Perform
2352 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
2353 truncate_int_for_mode the result for the destination's mode.
2356 * hsa-gen.c (mem_type_for_type): Fix comment typo.
2357 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
2359 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2362 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
2365 * lra-assigns.c (lra_split_hard_reg_for): Don't use
2366 regno_allocno_class_array and sorted_pseudos.
2367 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
2368 insns where regno is used.
2370 2018-03-16 Martin Liska <mliska@suse.cz>
2373 * multiple_target.c (create_dispatcher_calls): Redirect
2374 reference in the symbol table.
2376 2018-03-16 Martin Liska <mliska@suse.cz>
2379 * multiple_target.c (create_dispatcher_calls): Redirect also
2382 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2386 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
2387 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
2388 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
2390 2018-03-16 Julia Koval <julia.koval@intel.com>
2392 * doc/invoke.texi (Skylake Server): Add CLWB.
2393 Cannonlake): Remove CLWB.
2395 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2397 PR tree-optimization/84841
2398 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
2400 (FLOAT_ONE_CONST_TYPE): Define.
2401 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
2402 (sort_by_operand_rank): Put entries with higher constant_type last
2403 rather than first to match comments.
2405 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
2407 * config/nios2/nios2.md (movsi_internal): Fix thinko in
2410 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2413 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
2414 check and use address of target always.
2416 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
2419 * config/i386/i386.c (indirect_thunk_needed): Update comments.
2420 (indirect_thunk_bnd_needed): Likewise.
2421 (indirect_thunks_used): Likewise.
2422 (indirect_thunks_bnd_used): Likewise.
2423 (indirect_return_needed): New.
2424 (indirect_return_bnd_needed): Likewise.
2425 (output_indirect_thunk_function): Add a bool argument for
2427 (output_indirect_thunk_function): Don't generate alias for
2428 function return thunk.
2429 (ix86_code_end): Call output_indirect_thunk_function to generate
2430 function return thunks.
2431 (ix86_output_function_return): Set indirect_return_bnd_needed
2432 and indirect_return_needed instead of indirect_thunk_bnd_needed
2433 and indirect_thunk_needed.
2435 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
2437 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
2438 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
2439 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
2441 2018-03-15 David Malcolm <dmalcolm@redhat.com>
2442 Paul Hua <paul.hua.gm@gmail.com>
2445 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
2447 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
2449 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
2450 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
2453 2018-03-15 Tamar Christina <tamar.christina@arm.com>
2456 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
2457 instead of GET_MODE_SIZE when comparing Units.
2459 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2462 * varasm.c (hash_section): Return an unchangeble hash value
2463 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
2464 Return !aarch64_can_use_per_function_literal_pools_p ().
2466 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2469 * optabs.c (emit_conditional_move): Pass address of cmode's copy
2470 rather than address of cmode as last argument to prepare_cmp_insn.
2472 2018-03-15 Julia Koval <julia.koval@intel.com>
2474 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
2475 F_AVX512VNNI, F_AVX512BITALG): New.
2477 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
2480 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
2481 insn for floating-point loads and stores.
2483 2018-03-14 Carl Love <cel@us.ibm.com>
2485 * config/rs6000/rs6000-c.c: Add macro definitions for
2486 ALTIVEC_BUILTIN_VEC_PERMXOR.
2487 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
2488 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
2489 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
2490 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
2492 * config/doc/extend.texi: Add prototypes for vec_permxor.
2494 2018-03-14 David Malcolm <dmalcolm@redhat.com>
2497 * diagnostic-show-locus.c (class layout_point): Convert m_line
2498 from int to linenum_type.
2499 (line_span::comparator): Use linenum "compare" function when
2500 comparing line numbers.
2501 (test_line_span): New function.
2502 (layout_range::contains_point): Convert param "row" from int to
2504 (layout_range::intersects_line_p): Likewise.
2505 (layout::will_show_line_p): Likewise.
2506 (layout::print_source_line): Likewise.
2507 (layout::should_print_annotation_line_p): Likewise.
2508 (layout::print_annotation_line): Likewise.
2509 (layout::print_leading_fixits): Likewise.
2510 (layout::annotation_line_showed_range_p): Likewise.
2511 (struct line_corrections): Likewise for field m_row.
2512 (line_corrections::line_corrections): Likewise for param "row".
2513 (layout::print_trailing_fixits): Likewise.
2514 (layout::get_state_at_point): Likewise.
2515 (layout::get_x_bound_for_row): Likewise.
2516 (layout::print_line): Likewise.
2517 (diagnostic_show_locus): Likewise for locals "last_line" and
2519 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
2520 * input.c (selftest::test_linenum_comparisons): New function.
2521 (selftest::input_c_tests): Call it.
2522 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
2523 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
2524 * selftest.h (ASSERT_GT): New macro.
2525 (ASSERT_GT_AT): New macro.
2526 (ASSERT_LT): New macro.
2527 (ASSERT_LT_AT): New macro.
2529 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
2531 PR rtl-optimization/84780
2532 * combine.c (distribute_links): Don't make a link based on pc_rtx.
2534 2018-03-14 Martin Liska <mliska@suse.cz>
2536 * tree.c (record_node_allocation_statistics): Use
2537 get_stats_node_kind.
2538 (get_stats_node_kind): New function extracted from
2539 record_node_allocation_statistics.
2540 (free_node): Use get_stats_node_kind.
2542 2018-03-14 Richard Biener <rguenther@suse.de>
2544 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
2545 that the value-set of ANTIC_IN doesn't grow.
2548 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2550 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2551 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2553 2018-03-14 Julia Koval <julia.koval@intel.com>
2555 * config.gcc (icelake-client, icelake-server): New.
2557 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
2558 (initial_ix86_arch_features): Ditto.
2559 (PTA_SKYLAKE): Add SGX.
2560 (PTA_ICELAKE): Remove.
2561 (PTA_ICELAKE_CLIENT): New.
2562 (PTA_ICELAKE_SERVER): New.
2563 (ix86_option_override_internal): Split up icelake on icelake client and
2565 (get_builtin_code_for_version): Ditto.
2566 (fold_builtin_cpu): Ditto.
2567 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
2568 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
2569 * config/i386/i386.h (processor_type): Ditto.
2570 * doc/invoke.texi: Ditto.
2572 2018-03-14 Jakub Jelinek <jakub@redhat.com>
2575 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
2576 INTEGER_CST offset, add it together with bitpos / 8 and
2577 sign extend based on POINTER_SIZE.
2581 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
2584 * config/i386/constraints.md (Yc): New register constraint.
2585 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
2586 Use Yc constraint for alternative 2 of operand 0. Remove
2587 preferred_for_speed attribute.
2589 2018-03-14 Richard Biener <rguenther@suse.de>
2591 PR tree-optimization/84830
2592 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
2593 with the old one to avoid oscillations.
2595 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
2598 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
2600 (assign_by_spills): Return a flag of reload assignment failure.
2601 Do not process the reload assignment failures. Do not spill other
2602 reload pseudos if they has the same reg class. Update n if
2604 (lra_assign): Add a return arg. Set up from the result of
2605 assign_by_spills call.
2606 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2607 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2608 usage_insns if it is not NULL.
2609 (spill_hard_reg_in_range): New function.
2610 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2611 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2612 function prototypes.
2613 (lra_assign): Change prototype.
2614 * lra.c (lra): Add code to deal with fails by splitting hard reg
2617 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
2619 * config/riscv/riscv.opt (mrelax): New option.
2620 * config/riscv/riscv.c (riscv_file_start): Emit ".option
2621 "norelax" when riscv_mrelax is disabled.
2622 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
2624 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2627 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
2628 reassociation for int modes.
2630 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2632 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
2633 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
2635 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
2636 * config/aarch64/aarch64-sve.md
2637 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
2638 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
2639 (*extend<mode><Vwide>2): Rename to...
2640 (aarch64_sve_extend<mode><Vwide>2): ...this.
2641 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
2642 renaming the old pattern to...
2643 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
2645 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
2646 define_expand, renaming the old pattern to...
2647 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
2648 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
2649 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
2650 account when deciding which SVE instruction the optab should use.
2651 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
2653 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2655 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
2656 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
2657 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
2658 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
2659 (tlsdesc_small_<mode>): Turn a define_expand and use
2660 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
2661 (tlsdesc_small_advsimd_<mode>): ...this.
2662 (tlsdesc_small_sve_<mode>): New pattern.
2664 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2666 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
2667 (UNSPEC_UMUL_HIGHPART): New constants.
2668 (MUL_HIGHPART): New int iteraor.
2669 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
2670 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
2672 (*<su>mul<mode>3_highpart): New define_insn.
2674 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2677 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2680 2018-03-13 Martin Liska <mliska@suse.cz>
2683 * (sem_item_optimizer::sem_item_optimizer): Initialize new
2685 (sem_item_optimizer::~sem_item_optimizer): Release it.
2686 (sem_item_optimizer::merge_classes): Register variable aliases.
2687 (sem_item_optimizer::fixup_pt_set): New function.
2688 (sem_item_optimizer::fixup_points_to_sets): Likewise.
2689 * ipa-icf.h: Declare new variables and functions.
2691 2018-03-13 Jakub Jelinek <jakub@redhat.com>
2694 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
2695 integer_pow2p@2 and test integer_pow2p in condition.
2696 (A < 0 ? C : 0): Similarly for @1.
2699 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
2700 characters starting at p contain '\0' character, don't look beyond
2704 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
2705 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
2708 * reg-stack.c (change_stack): Change update_end var from int to
2709 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
2710 also call set_block_for_insn on the newly added insns and rescan.
2713 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
2714 on the last operand.
2717 * tree.c (stabilize_reference_1): Return save_expr (e) for
2718 STATEMENT_LIST even if it doesn't have side-effects.
2720 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
2722 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
2724 2018-03-12 Renlin Li <renlin.li@arm.com>
2726 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
2727 aarch64_output_scalar_simd_mov_immediate.
2729 2018-03-12 Martin Sebor <msebor@redhat.com>
2731 PR tree-optimization/83456
2732 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
2733 for perfectly overlapping calls to memcpy.
2734 (gimple_fold_builtin_memory_chk): Same.
2735 (gimple_fold_builtin_strcpy): Handle no-warning.
2736 (gimple_fold_builtin_stxcpy_chk): Same.
2737 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
2739 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2741 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
2742 parameter. Use it for SFmode.
2743 (rs6000_function_arg_advance_1): Adjust.
2744 (rs6000_function_arg): Adjust.
2745 (rs6000_gimplify_va_arg): Pass false for that new parameter.
2747 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2749 PR rtl-optimization/84169
2750 PR rtl-optimization/84780
2751 * combine.c (can_combine_p): Check for a 2-insn combination whether
2752 the destination register is used between the two insns, too.
2754 2018-03-12 Richard Biener <rguenther@suse.de>
2756 PR tree-optimization/84803
2757 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
2758 for refs DR analysis didn't process.
2760 2018-03-12 Richard Biener <rguenther@suse.de>
2762 PR tree-optimization/84777
2763 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
2764 force-vectorize loops ignore whether we are optimizing for size.
2766 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
2768 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
2769 (TARGET_MD_ASM_ADJUST): Define.
2771 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
2772 Kito Cheng <kito.cheng@gmail.com>
2773 Chung-Ju Wu <jasonwucj@gmail.com>
2775 * config/nds32/nds32.c (nds32_compute_stack_frame,
2776 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
2777 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
2778 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
2779 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
2780 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
2781 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
2782 * config/nds32/nds32.md (prologue, epilogue): Use macro
2783 NDS32_V3PUSH_AVAILABLE_P to do checking.
2785 2018-03-11 Jakub Jelinek <jakub@redhat.com>
2788 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
2789 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
2790 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
2791 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
2792 addition of most attributes on !orig_type_die or the attribute not
2793 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
2795 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2796 Chung-Ju Wu <jasonwucj@gmail.com>
2798 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
2800 * config/nds32/nds32.opt (mvh): New option.
2802 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2803 Chung-Ju Wu <jasonwucj@gmail.com>
2805 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
2807 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
2808 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
2811 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2812 Chung-Ju Wu <jasonwucj@gmail.com>
2814 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
2816 * config/nds32/nds32-multiple.md (strlensi): New pattern.
2817 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
2819 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
2820 Kito Cheng <kito.cheng@gmail.com>
2821 Chung-Ju Wu <jasonwucj@gmail.com>
2823 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
2824 UNSPEC_FFMISM and UNSPEC_FLMISM.
2825 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
2826 for ffb, ffmism and flmism.
2827 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
2828 (unspec_ffmism): Ditto.
2829 (unspec_flmism): Ditto.
2830 (nds32_expand_builtin_impl): Check if string extension is available.
2831 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
2832 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
2834 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
2837 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2840 * lra-assigns.c (assign_by_spills): Return a flag of reload
2841 assignment failure. Do not process the reload assignment
2842 failures. Do not spill other reload pseudos if they has the same
2844 (lra_assign): Add a return arg. Set up from the result of
2845 assign_by_spills call.
2846 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2847 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2848 usage_insns if it is not NULL.
2849 (spill_hard_reg_in_range): New function.
2850 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2851 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2852 function prototypes.
2853 (lra_assign): Change prototype.
2854 * lra.c (lra): Add code to deal with fails by splitting hard reg
2857 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
2860 * config/i386/i386.opt: Replace Enforcment with Enforcement.
2862 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
2865 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
2866 (dw_val_node): Add val_symbolic_view.
2867 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
2868 (symview_upper_bound): New.
2869 (new_line_info_table): Initialize symviews_since_reset.
2870 (dwarf2out_source_line): Count symviews_since_reset and set
2871 symview_upper_bound.
2872 (dw_val_equal_p): Handle symview.
2873 (add_AT_symview): New.
2874 (print_dw_val): Handle symview.
2875 (attr_checksum, attr_checksum_ordered): Likewise.
2876 (same_dw_val_p, size_of_die): Likewise.
2877 (value_format, output_die): Likewise.
2878 (add_high_low_attributes): Use add_AT_symview for entry_view.
2879 (dwarf2out_finish): Reset symview_upper_bound, clear
2882 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
2885 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
2886 Add strict argument and use it.
2887 (rs6000_split_multireg_move): Update for new strict argument.
2888 (mem_operand_gpr): Disallow all non-offsettable addresses.
2889 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
2891 2018-03-09 Jakub Jelinek <jakub@redhat.com>
2894 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
2895 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
2896 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
2899 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
2900 decl, use remap_type if we want to use the type.
2902 2018-03-09 Martin Sebor <msebor@redhat.com>
2904 PR tree-optimization/84526
2905 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2907 (builtin_access::generic_overlap): Be prepared to handle non-array
2910 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
2912 PR rtl-optimization/84682
2913 * lra-constraints.c (process_address_1): Check is_address flag
2914 for address constraints.
2915 (process_alt_operands): Likewise.
2916 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
2917 preprocess_constraints.
2918 * recog.h (preprocess_constraints): Add oploc parameter.
2920 * recog.c (preprocess_constraints): Test address_operand for
2921 CT_ADDRESS constraints.
2923 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2926 * lra-assigns.c (assign_by_spills): Return a flag of reload
2927 assignment failure. Do not process the reload assignment
2928 failures. Do not spill other reload pseudos if they has the same
2930 (lra_assign): Add a return arg. Set up from the result of
2931 assign_by_spills call.
2932 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2933 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2934 usage_insns if it is not NULL.
2935 (spill_hard_reg_in_range): New function.
2936 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2937 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2938 function prototypes.
2939 (lra_assign): Change prototype.
2940 * lra.c (lra): Add code to deal with fails by splitting hard reg
2943 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2946 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
2947 Accept complain bool parameter. Only emit errors if it is true.
2948 (arm_parse_cpu_option_name): Likewise.
2949 (arm_target_thumb_only): Adjust callers of the above.
2950 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
2951 prototype to take a default true bool parameter.
2952 (arm_parse_arch_option_name): Likewise.
2954 2018-03-09 David Malcolm <dmalcolm@redhat.com>
2955 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
2959 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
2960 * configure: Regenerate.
2961 * configure.ac ("linker --version-script option"): New.
2962 ("linker soname option"): New.
2964 2018-03-09 Richard Biener <rguenther@suse.de>
2966 PR tree-optimization/84775
2967 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
2968 immediate uses of predicate stmts and mark them modified.
2971 PR tree-optimization/84178
2972 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2974 (version_loop_for_if_conversion): Delay update_ssa call.
2975 (tree_if_conversion): Delay update_ssa until after predicate
2978 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
2981 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
2982 when the function accesses prior frames.
2984 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2987 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
2988 gen_llsym, otherwise call maybe_gen_llsym.
2991 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
2992 has ',' character inside of it.
2994 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2997 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
2998 as clobbering CC_REGNUM.
3000 2018-03-08 Richard Biener <rguenther@suse.de>
3003 * tree-scalar-evolution.c: Include tree-into-ssa.h.
3004 (follow_copies_to_constant): Do not follow SSA names registered
3007 2018-03-08 Richard Biener <rguenther@suse.de>
3009 PR tree-optimization/84178
3010 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3012 (version_loop_for_if_conversion): Delay update_ssa call.
3013 (tree_if_conversion): Delay update_ssa until after predicate
3016 2018-03-08 David Malcolm <dmalcolm@redhat.com>
3018 PR tree-optimization/84178
3019 * tree-if-conv.c (release_bb_predicate): Remove the
3020 the assertion that the stmts have NULL use_ops.
3021 Discard the statements, asserting that they haven't
3022 yet been added to a BB.
3024 2018-03-08 Richard Biener <rguenther@suse.de>
3026 PR tree-optimization/84746
3027 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
3028 (phi_translate): Pass in destination ANTIC_OUT set.
3029 (phi_translate_1): Likewise. For a simplified result lookup
3030 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
3031 (phi_translate_set): Adjust.
3032 (do_pre_regular_insertion): Likewise.
3033 (do_pre_partial_partial_insertion): Likewise.
3035 2018-03-08 Martin Liska <mliska@suse.cz>
3037 PR gcov-profile/84735
3038 * doc/gcov.texi: Document usage of profile files.
3039 * gcov-io.h: Document changes in the format.
3041 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
3045 * dwarf2out.c (struct dw_line_info_table): Update comments for
3047 (FORCE_RESET_NEXT_VIEW): New.
3048 (FORCE_RESETTING_VIEW_P): New.
3049 (RESETTING_VIEW_P): Check for -1 too.
3050 (ZERO_VIEW_P): Likewise.
3051 (new_line_info_table): Force-reset next view.
3052 (dwarf2out_begin_function): Likewise.
3053 (dwarf2out_source_line): Simplify zero_view_p initialization.
3054 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
3055 view directly. Omit view when omitting .loc at line 0.
3057 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3059 PR tree-optimization/84740
3060 * tree-switch-conversion.c (process_switch): Call build_constructors
3061 only if info.phi_count is non-zero.
3063 PR tree-optimization/84739
3064 * tree-tailcall.c (find_tail_calls): Check call arguments against
3065 DECL_ARGUMENTS (current_function_decl) rather than
3066 DECL_ARGUMENTS (func) when checking for tail recursion.
3068 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3070 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
3071 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
3072 Volker Reichelt's entry and add entries for people that perform
3073 GCC fuzzy testing and report numerous bugs.
3075 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
3078 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
3079 readonly data in sdata, if that is disabled.
3080 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
3081 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3082 -mreadonly-in-sdata option.
3084 2018-03-07 Martin Sebor <msebor@redhat.com>
3086 PR tree-optimization/84468
3087 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
3088 basic block when looking for nul assignment.
3090 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
3093 * except.h (output_function_exception_table): Adjust prototype.
3094 * except.c (output_function_exception_table): Remove FNNAME parameter
3095 and add SECTION parameter. Ouput one part of the table at a time.
3096 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
3097 the first part of the exception table and emit unwind directives.
3098 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
3099 (i386_pe_seh_cold_init): Likewise.
3100 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
3101 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
3102 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
3103 (ix86_output_call_insn): Emit a nop in one more case for SEH.
3104 * config/i386/winnt.c: Include except.h.
3105 (struct seh_frame_state): Add reg_offset, after_prologue and
3106 in_cold_section fields.
3107 (i386_pe_seh_end_prologue): Set seh->after_prologue.
3108 (i386_pe_seh_cold_init): New function.
3109 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
3110 to seh->in_cold_section.
3111 (seh_emit_push): Record the offset of the push.
3112 (seh_emit_save): Record the offet of the save.
3113 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
3114 Test seh->after_prologue to disregard the epilogue.
3115 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
3116 (i386_pe_end_cold_function): New function.
3118 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3121 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
3122 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
3125 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
3127 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
3128 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
3131 * multiple_target.c: Include tree-inline.h and intl.h.
3132 (expand_target_clones): Diagnose and fail if node->definition and
3133 !tree_versionable_function_p (node->decl).
3135 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
3137 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
3139 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
3140 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
3141 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
3143 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3146 * combine.c (try_combine): Use reg_or_subregno instead of handling
3147 just paradoxical SUBREGs and REGs.
3149 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
3151 * config/arc/arc.c (arc_finalize_pic): Remove function.
3152 (arc_must_save_register): We use single base PIC register, remove
3153 checks to save/restore the PIC register.
3154 (arc_expand_prologue): Likewise.
3155 * config/arc/arc-protos.h (arc_set_default_type_attributes):
3157 (arc_verify_short): Likewise.
3158 (arc_attr_type): Likewise.
3159 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
3160 (walk_stores): Likewise.
3161 (arc_address_cost): Make it static.
3162 (arc_verify_short): Likewise.
3163 (branch_dest): Likewise.
3164 (arc_attr_type): Likewise.
3165 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
3166 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
3167 (arc_final_prescan_insn): Remove inserting the nops due to
3168 hardware hazards. It is done in reorg step.
3169 (insn_length_variant_t): Remove.
3170 (insn_length_parameters_t): Likewise.
3171 (arc_insn_length_parameters): Likewise.
3172 (arc_get_insn_variants): Likewise.
3173 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
3175 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3178 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
3181 PR tree-optimization/84687
3182 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
3184 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
3186 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3188 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
3189 Rename to ppc_speculation_barrier.
3190 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
3191 __builtin_ppc_speculation_barrier.
3193 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3196 * combine.c (combine_simplify_rtx): Don't try to simplify if
3197 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
3200 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
3202 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
3203 to 32 bytes when compiling for POWER9.
3205 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3208 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
3209 regparm >= 3 with no arg reg available also for calls with
3210 flag_force_indirect_call. Pass decl to ix86_function_regparm.
3213 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
3215 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
3217 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
3220 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
3222 2018-03-05 Richard Biener <rguenther@suse.de>
3224 PR tree-optimization/84486
3225 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
3226 When inserting a __builtin_assume_aligned call set the LHS
3227 SSA name alignment info accordingly.
3229 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
3231 PR tree-optimization/84114
3232 * config/aarch64/aarch64.c (aarch64_reassociation_width)
3233 Avoid reassociation of FLOAT_MODE addition.
3235 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
3237 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
3238 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
3239 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
3240 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
3241 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
3242 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
3243 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
3245 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
3246 __builtin_ia32_wbinvd): New builtins.
3247 (SPECIAL_ARGS2): New.
3248 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
3249 (SPECIAL_ARGS2): New.
3250 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
3251 (ix86_valid_target_attribute_inner_p): Ditto.
3252 (ix86_init_mmx_sse_builtins): Add special_args2.
3253 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
3254 TARGET_WBNOINVD_P): New.
3255 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
3256 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
3257 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
3258 * config/i386/immintrin.h (_wbinvd): New intrinsic.
3259 * config/i386/pconfigintrin.h: New file.
3260 * config/i386/wbnoinvdintrin.h: Ditto.
3261 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
3262 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
3264 2018-03-05 Richard Biener <rguenther@suse.de>
3266 PR tree-optimization/84670
3267 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3269 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3270 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3271 (compute_antic_aux): Only assert the number of values in ANTIC_IN
3272 doesn't grow if all successors (recursively) were visited at least
3275 2018-03-05 Richard Biener <rguenther@suse.de>
3277 PR tree-optimization/84650
3278 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
3279 if executed in the loop pipeline.
3281 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
3283 * doc/configfiles.texi (Configuration Files): Move info about
3284 conditionalizing $target-protos.h to...
3285 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
3286 differs from $target-protos.h.
3288 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
3289 Chung-Ju Wu <jasonwucj@gmail.com>
3291 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
3292 * config/nds32/nds32-multiple.md (setmemsi): Define.
3293 * config/nds32/nds32-memory-manipulation.c
3294 (nds32_gen_dup_4_byte_to_word_value): New.
3295 (emit_setmem_word_loop): New.
3296 (emit_setmem_byte_loop): New.
3297 (nds32_expand_setmem_loop): New.
3298 (nds32_expand_setmem_loop_v3m): New.
3299 (nds32_expand_setmem_unroll): New.
3300 (nds32_expand_setmem): New.
3302 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3303 Chung-Ju Wu <jasonwucj@gmail.com>
3305 * config/nds32/nds32-memory-manipulation.c
3306 (nds32_emit_load_store): New.
3307 (nds32_emit_post_inc_load_store): New.
3308 (nds32_emit_mem_move): New.
3309 (nds32_emit_mem_move_block): New.
3310 (nds32_expand_movmemsi_loop_unknown_size): New.
3311 (nds32_expand_movmemsi_loop_known_size): New.
3312 (nds32_expand_movmemsi_loop): New.
3313 (nds32_expand_movmemsi_unroll): New.
3314 (nds32_expand_movmemqi): Rename ...
3315 (nds32_expand_movmemsi): ... to this.
3316 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
3317 (movmemsi): ... to this.
3318 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
3319 (nds32_expand_movmemsi): ... to this.
3321 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3322 Monk Chiang <sh.chiang04@gmail.com>
3323 Chung-Ju Wu <jasonwucj@gmail.com>
3325 * config/nds32/nds32-protos.h
3326 (nds32_expand_load_multiple): New arguments.
3327 (nds32_expand_store_multiple): Ditto.
3328 (nds32_valid_multiple_load_store): Rename ...
3329 (nds32_valid_multiple_load_store_p): ... to this.
3330 * config/nds32/nds32-memory-manipulation.c
3331 (nds32_expand_load_multiple): Refine implementation.
3332 (nds32_expand_store_multiple): Ditto.
3333 * config/nds32/nds32-multiple.md
3334 (load_multiple): Update nds32_expand_load_multiple interface.
3335 (store_multiple): Update nds32_expand_store_multiple interface.
3336 * config/nds32/nds32-predicates.c
3337 (nds32_valid_multiple_load_store): Rename ...
3338 (nds32_valid_multiple_load_store_p): ... to this and refine
3340 * config/nds32/predicates.md
3341 (nds32_load_multiple_and_update_address_operation): New predicate.
3342 (nds32_store_multiple_and_update_address_operation): New predicate.
3344 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3345 Chung-Ju Wu <jasonwucj@gmail.com>
3347 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
3348 (combo): New attribute.
3349 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
3351 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
3353 * config/nds32/nds32.opt: Change -mcmodel= default value.
3355 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
3356 Monk Chiang <sh.chiang04@gmail.com>
3357 Chung-Ju Wu <jasonwucj@gmail.com>
3359 * config/nds32/constants.md (unspec_element): New enum.
3360 * config/nds32/constraints.md (Umw): New constraint.
3361 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
3362 * config/nds32/nds32-intrinsic.md: Likewise.
3363 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
3364 (nds32_valid_smw_lwm_base_p): New.
3365 (nds32_output_smw_single_word): New.
3366 (nds32_output_lmw_single_word): New.
3367 (nds32_expand_unaligned_load): New.
3368 (nds32_expand_unaligned_store): New.
3369 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
3370 (nds32_output_smw_single_word): Declare.
3371 (nds32_output_lmw_single_word): Declare.
3372 (nds32_expand_unaligned_load): Declare.
3373 (nds32_expand_unaligned_store): Declare.
3374 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
3375 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
3376 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
3377 NDS32_BUILTIN_UASTORE_DW.
3378 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
3381 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
3382 Kito Cheng <kito.cheng@gmail.com>
3383 Chung-Ju Wu <jasonwucj@gmail.com>
3385 * config/nds32/nds32-intrinsic.c
3386 (nds32_expand_builtin_null_ftype_reg): Delete.
3387 (nds32_expand_builtin_reg_ftype_imm): Ditto.
3388 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
3389 (nds32_read_argument): New.
3390 (nds32_legitimize_target): Ditto.
3391 (nds32_legitimize_argument): Ditto.
3392 (nds32_check_constant_argument): Ditto.
3393 (nds32_expand_unop_builtin): Ditto.
3394 (nds32_expand_unopimm_builtin): Ditto.
3395 (nds32_expand_binop_builtin): Ditto.
3396 (nds32_builtin_decl_impl): Ditto.
3397 (builtin_description): Ditto.
3398 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
3399 (nds32_init_builtins_impl): Ditto.
3400 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
3401 (nds32_builtin_decl): New.
3402 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
3403 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
3405 2018-03-02 Jeff Law <law@redhat.com>
3407 * reorg.c (stop_search_p): Handle DEBUG_INSN.
3408 (redundant_insn, fill_simple_delay_slots): Likewise.
3409 (fill_slots_from_thread): Likewise.
3410 * resource.c (mark_referenced_resources): Likewise.
3411 (mark_set_resources, find_dead_or_set_registers): Likewise.
3413 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3415 * substring-locations.h (format_warning_va): Formatting fix for
3417 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
3419 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
3420 * substring-locations.c: Include intl.h.
3421 (format_warning_va): Turned into small wrapper around
3422 format_warning_n_va, renamed to ...
3423 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
3424 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
3426 (format_warning_at_substring_n): New function.
3427 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
3428 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
3429 format_warning_at_substring with just a shorter name instead of
3430 const function pointer.
3431 (fmtwarn_n): New function.
3432 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
3433 appropriate, get rid of all the fmtstr temporaries, move conditionals
3434 with G_() wrapped string literals directly into fmtwarn arguments,
3435 cast dir.len to (int), formatting fixes.
3437 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
3439 * doc/invoke.texi: Remove "Cilk Plus" references.
3441 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3442 Richard Biener <rguenther@suse.de>
3445 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
3446 for error or warning attributes if CALL_FROM_THUNK_P is set.
3449 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3452 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
3453 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
3456 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
3457 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
3459 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
3460 instead of -1U in last predictors element's probability member.
3462 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3465 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3466 arguments if they are comparable.
3468 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3470 PR tree-optimization/84634
3471 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
3472 masks and masked_loop_p with a single loop_masks, making sure it's
3473 null for bb vectorization.
3475 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3477 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3478 (vect_analyze_data_ref_access): Use loop->safe_len rather than
3479 loop->force_vectorize to check whether there is no alias.
3481 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3484 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
3486 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
3488 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
3489 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
3490 instead of a loop around prev_real_insn.
3491 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
3495 * config/i386/i386.c (ix86_print_operand): Use conditional
3496 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
3499 2018-03-02 Richard Biener <rguenther@suse.de>
3501 PR tree-optimization/84427
3502 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
3503 (bitmap_set_subtract_values): Rewrite to handle multiple
3506 (prune_clobbered_mems): Likewise.
3507 (phi_translate): Take edge instead of pred/phiblock.
3508 (phi_translate_1): Likewise.
3509 (phi_translate_set): Likewise. Insert all translated
3510 exprs for a value into the set, keeping possibly multiple
3511 expressions per value.
3512 (compute_antic_aux): Adjust for phi_translate changes.
3513 When intersecting union the expressions and prune those
3514 not in the final value set, keeping possibly multiple
3515 expressions per value. Do not use value-insertion
3516 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
3517 all expressions. Add verification that the value-sets
3518 only shrink during iteration.
3519 (compute_partial_antic_aux): Adjust for the phi_translate changes.
3520 (do_pre_regular_insertion): Likewise.
3521 (do_pre_partial_partial_insertion): Likewise.
3523 2018-03-02 Richard Biener <rguenther@suse.de>
3526 * config/darwin.c (saved_debug_info_level): New static global.
3527 (darwin_asm_lto_start): Disable debug info generation for LTO out.
3528 (darwin_asm_lto_end): Restore debug info generation settings.
3530 2018-03-01 Martin Liska <mliska@suse.cz>
3533 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
3536 2018-03-01 Richard Biener <rguenther@suse.de>
3539 * dwarf2out.c (gen_variable_die): Properly handle late VLA
3540 type annotation with LTO when debug was disabled at compile-time.
3542 2018-03-01 Matthew Fortune <mfortune@gmail.com>
3544 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
3546 (mips_final_postscan_insn): Likewise.
3548 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
3550 PR rtl-optimization/84528
3551 * alias.c (init_alias_target): Add commentary.
3552 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
3553 a unique base value if the frame pointer is not eliminated
3554 to the stack pointer.
3556 2018-03-01 Tom de Vries <tom@codesourcery.com>
3558 PR rtl-optimization/83327
3559 * lra-int.h (hard_regs_spilled_into): Declare.
3560 * lra.c (hard_regs_spilled_into): Define.
3561 (init_reg_info): Init hard_regs_spilled_into.
3562 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
3563 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
3564 (process_bb_lives): Handle hard_regs_spilled_into.
3565 (lra_create_live_ranges_1): Before doing liveness propagation, clear
3566 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
3568 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
3570 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
3571 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
3572 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
3573 * config/rs6000/aix72.h: New file.
3575 2018-02-28 Jakub Jelinek <jakub@redhat.com>
3577 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
3578 instead of warning_at with conditional singular and plural messages
3582 * stor-layout.c (update_alignment_for_field): For
3583 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
3584 && !DECL_PACKED (field), do the alignment update, just use
3585 only desired_align instead of MAX (type_align, desired_align)
3587 (place_field): Don't do known_align < desired_align handling
3588 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
3589 is non-NULL, instead do it after rli->prev_field handling and
3590 only if not within a bitfield word. For DECL_PACKED (field)
3591 use type_align of BITS_PER_UNIT.
3593 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
3595 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
3596 superfluous parentheses and trailing spaces.
3598 2018-02-28 Richard Biener <rguenther@suse.de>
3600 PR tree-optimization/84584
3601 * graphite-scop-detection.c (scop_detection::add_scop): Discard
3602 SCoPs with fake exit edge.
3604 2018-02-28 Martin Liska <mliska@suse.cz>
3607 * timevar.c (timer::print): Fix format to properly print 100%
3610 2018-02-28 Richard Biener <rguenther@suse.de>
3613 * genmatch.c (capture_info::walk_match): Do not mark
3614 captured expressions without operands as expr_p given
3615 they act more like predicates and should be subject to
3616 "lost tail" side-effect preserving.
3618 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
3620 PR rtl-optimization/81611
3621 * auto-inc-dec.c (attempt_change): Move dead note from
3622 mem_insn if it's the next use of regno
3623 (find_address): Take address use of reg holding
3624 non-incremented value. Add parm to limit search to the named
3626 (merge_in_block): Attempt to use a mem insn that is the next
3627 use of the original regno.
3629 2018-02-27 Martin Sebor <msebor@redhat.com>
3632 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
3633 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
3635 2018-02-27 Martin Sebor <msebor@redhat.com>
3637 PR translation/84207
3638 * diagnostic-core.h (warning_n, error_n, inform_n): Change
3639 n argument to unsigned HOST_WIDE_INT.
3640 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
3641 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
3642 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
3643 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
3645 2018-02-27 Richard Biener <rguenther@suse.de>
3647 PR tree-optimization/84512
3648 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
3649 Do not use the estimate returned from record_stmt_cost for
3650 the scalar iteration cost but sum properly using add_stmt_cost.
3652 2018-02-27 Richard Biener <rguenther@suse.de>
3654 PR tree-optimization/84466
3655 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3656 Adjust last change to less strictly validate use operands.
3658 2018-02-27 Martin Liska <mliska@suse.cz>
3660 PR gcov-profile/84548
3661 * gcov.c (process_file): Allow partial overlap and consider it
3662 also as group functions.
3663 (output_lines): Properly calculate range of lines for a group.
3665 2018-02-27 Martin Liska <mliska@suse.cz>
3667 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
3668 'ggc' suffixes. Change first column width.
3669 (timer::print): Fix formatting of the column.
3671 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
3673 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
3674 preserve inline entry blocks for the sake of debug inline
3675 entry point markers alone.
3676 (remove_unused_locals): Suggest in comments a better place to
3677 force the preservation of inline entry blocks that are
3678 otherwise unused, but do not preserve them.
3680 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3682 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
3684 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3687 * config/i386/constraints.md (Bs): Replace
3688 ix86_indirect_branch_register with
3689 TARGET_INDIRECT_BRANCH_REGISTER.
3691 * config/i386/i386.md (indirect_jump): Likewise.
3692 (tablejump): Likewise.
3693 (*sibcall_memory): Likewise.
3694 (*sibcall_value_memory): Likewise.
3695 Peepholes of indirect call and jump via memory: Likewise.
3696 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
3697 (*sibcall_value_GOT_32): Likewise.
3698 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3699 (GOT_memory_operand): Likewise.
3700 (call_insn_operand): Likewise.
3701 (sibcall_insn_operand): Likewise.
3702 (GOT32_symbol_operand): Likewise.
3703 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
3705 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3707 PR rtl-optimization/83496
3708 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
3709 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
3710 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
3711 redundant insn, if any.
3712 (relax_delay_slots): Likewise.
3713 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
3715 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
3717 PR tree-optimization/83965
3718 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
3719 that grouped statements are part of a reduction chain. Return
3720 true if the statement is not marked as a reduction itself but
3722 (vect_recog_dot_prod_pattern): Don't check whether the statement
3723 is part of a group here.
3724 (vect_recog_sad_pattern): Likewise.
3725 (vect_recog_widen_sum_pattern): Likewise.
3727 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3730 * final.c (rest_of_clean_state): Also look for calls inside sequences.
3732 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3735 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
3737 (ix86_output_indirect_function_return): New prototype.
3738 (ix86_split_simple_return_pop_internal): Likewise.
3739 * config/i386/i386.c (indirect_return_via_cx): New.
3740 (indirect_return_via_cx_bnd): Likewise.
3741 (indirect_thunk_name): Handle return va CX_REG.
3742 (output_indirect_thunk_function): Create alias for
3743 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
3744 (ix86_output_indirect_jmp): Remove the bool argument.
3745 (ix86_output_indirect_function_return): New function.
3746 (ix86_split_simple_return_pop_internal): Likewise.
3747 * config/i386/i386.md (*indirect_jump): Don't pass false
3748 to ix86_output_indirect_jmp.
3749 (*tablejump_1): Likewise.
3750 (simple_return_pop_internal): Change it to define_insn_and_split.
3751 Call ix86_split_simple_return_pop_internal to split it for
3753 (simple_return_indirect_internal): Call
3754 ix86_output_indirect_function_return instead of
3755 ix86_output_indirect_jmp.
3757 2018-02-26 Jakub Jelinek <jakub@redhat.com>
3760 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
3761 memset and value initialization afterwards.
3763 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
3765 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
3767 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3770 * common/config/aarch64/aarch64-common.c
3771 (aarch_option_optimization_table[]): Switch
3772 off fomit-frame-pointer
3774 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3775 Chung-Ju Wu <jasonwucj@gmail.com>
3777 * config/nds32/nds32-multiple.md (load_multiple): Disallow
3779 (store_multiple): Ditto.
3781 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3783 * config.gcc: Add --with-cpu support for nds32 target.
3784 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
3785 * config/nds32/nds32.opt: Add -mcpu= option.
3787 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
3789 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
3790 isel=yes): Warn for these deprecated options.
3792 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
3794 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
3795 ISA_2_5_MASKS_EMBEDDED.
3797 2018-02-23 Jakub Jelinek <jakub@redhat.com>
3799 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
3800 p->max as pointers rather than using iterative_hash_expr.
3802 2018-02-23 Carl Love <cel@us.ibm.com>
3804 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
3805 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
3807 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
3808 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
3809 P8V_BUILTIN_VEC_VUNSIGNED2.
3811 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
3814 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
3815 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
3817 * lra-constraints.c (curr_insn_transform): Set up
3818 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
3820 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
3821 * lra-eliminations.c (spill_pseudos): Ditto.
3822 (process_insn_for_elimination): Ditto.
3823 * lra-lives.c (reg_early_clobber_p): Use the new macros.
3824 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
3825 LRA_NON_CLOBBERED_ALT.
3827 2018-02-22 Martin Sebor <msebor@redhat.com>
3829 PR tree-optimization/84480
3830 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
3831 to maybe_diag_stxncpy_trunc. Call it.
3832 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
3833 from gimple_fold_builtin_strcpy. Print inlining stack.
3834 (handle_builtin_stxncpy): Print inlining stack.
3835 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
3837 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
3840 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
3841 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
3842 and -fcheck-pointer-bounds are used together.
3843 (indirect_thunk_prefix): New enum.
3844 (indirect_thunk_need_prefix): New function.
3845 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
3846 "_nt" instead of "_bnd" for NOTRACK prefix.
3847 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
3848 (output_indirect_thunk_function): Likewise.
3850 (ix86_code_end): Update output_indirect_thunk_function calls.
3851 (ix86_output_indirect_branch_via_reg): Replace
3852 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
3853 (ix86_output_indirect_branch_via_push): Likewise.
3854 (ix86_output_function_return): Likewise.
3855 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
3856 incompatible with -fcf-protection=branch and
3857 -fcheck-pointer-bounds.
3859 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3862 * config/aarch64/aarch64.c (aarch64_print_address_internal):
3863 Change gcc_assert call to output_operand_lossage.
3865 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3867 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
3869 2018-02-22 DJ Delorie <dj@redhat.com>
3870 Sebastian Perta <sebastian.perta@renesas.com>
3871 Oleg Endo <olegendo@gcc.gnu.org>
3873 * config/rx/rx.c (rx_rtx_costs): New function.
3874 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
3876 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
3878 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
3880 2018-02-22 Martin Liska <mliska@suse.cz>
3883 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
3884 Add "native" as a possible value.
3886 2018-02-22 Martin Liska <mliska@suse.cz>
3889 * config/i386/i386.c (ix86_option_override_internal):
3890 Add "native" as a possible value for -march and -mtune.
3892 2018-02-22 Jakub Jelinek <jakub@redhat.com>
3895 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
3896 to all type variants.
3898 PR tree-optimization/84503
3899 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
3900 width as info->bitpos + info->bitsize - start.
3901 (merged_store_group::merge_overlapping): Simplify width computation.
3902 (check_no_overlap): New function.
3903 (imm_store_chain_info::try_coalesce_bswap): Compute expected
3904 start + width and last_order of the group, fail if check_no_overlap
3906 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
3907 to group if check_no_overlap fails.
3909 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3911 * config/rs6000/altivec.md: Delete contraint arguments to
3912 define_expand, define_split, and define_peephole2, and in
3913 define_insn_and_split if always unused.
3914 * config/rs6000/darwin.md: Ditto.
3915 * config/rs6000/dfp.md: Ditto.
3916 * config/rs6000/rs6000.md: Ditto.
3917 * config/rs6000/sync.md: Ditto.
3918 * config/rs6000/vector.md: Ditto.
3919 * config/rs6000/vsx.md: Ditto.
3921 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3923 * config/rs6000/altivec.md: Write output control strings as braced
3924 blocks instead of double-quoted strings.
3925 * config/rs6000/darwin.md: Ditto.
3926 * config/rs6000/rs6000.md: Ditto.
3927 * config/rs6000/vector.md: Ditto.
3928 * config/rs6000/vsx.md: Ditto.
3930 2018-02-21 Jason Merrill <jason@redhat.com>
3932 PR c++/84314 - ICE with templates and fastcall attribute.
3933 * attribs.c (build_type_attribute_qual_variant): Remove assert.
3935 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3937 * ipa-cp.c (determine_versionability): Fix comment typos.
3939 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3942 * ipa-cp.c (determine_versionability): Do not version functions caling
3945 2018-02-21 Martin Liska <mliska@suse.cz>
3948 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
3949 Add "native" as a possible value.
3950 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
3951 the macro when native cpu detection is available.
3953 2018-02-21 Martin Liska <mliska@suse.cz>
3956 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
3957 Add "native" as a possible value.
3958 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
3959 when native cpu detection is available.
3961 2018-02-21 Jakub Jelinek <jakub@redhat.com>
3962 Martin Sebor <msebor@redhat.com>
3964 PR tree-optimization/84478
3965 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
3967 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
3968 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
3969 support which is conservatively correct, for 2 only stay conservative
3970 for maxlen. Formatting and comment capitalization fixes. Add STRICT
3971 argument to the 2 argument get_range_strlen, adjust 6 arg
3972 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
3974 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
3975 (gimple_fold_builtin_strlen): Pass true as last argument to
3978 2018-02-20 Martin Sebor <msebor@redhat.com>
3981 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
3982 (builtin_memref::set_base_and_offset): Same. Handle inner references.
3983 (builtin_memref::builtin_memref): Factor out parts into
3984 set_base_and_offset and call it.
3986 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3989 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
3990 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
3991 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
3992 search at the associated MODE_INT.
3994 2018-02-20 Jeff Law <law@redhat.com>
3997 PR tree-optimization/81592
3999 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
4000 for range data rather than using global data.
4001 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
4002 range data rather than using global data.
4003 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
4004 pass it to children as needed.
4005 (struct directive::fmtresult): Similarly.
4006 (struct directive::set_width): Similarly.
4007 (struct directive::set_precision): Similarly.
4008 (format_integer, format_directive, parse_directive): Similarly.
4009 (format_none): Accept unnamed vr_values parameter.
4010 (format_percent, format_floating, format_character): Similarly.
4011 (format_string, format_plain): Similarly.
4012 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
4013 the EVRP range analyzer for range data rather than using global data.
4014 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
4015 gimple-ssa-evrp-analyze.h
4016 (class sprintf_dom_walker): Add after_dom_children member function.
4017 Add evrp_range_analyzer member.
4018 (sprintf_dom_walker::before_dom_children): Call into the EVRP
4019 range analyzer as needed.
4020 (sprintf_dom_walker::after_dom_children): New member function.
4021 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
4023 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
4024 (evrp_range_analyzer::pop_to_marker): Likewise.
4026 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4028 PR tree-optimization/84419
4029 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
4030 with the required type if its current type is compatible but
4033 2018-02-20 Jakub Jelinek <jakub@redhat.com>
4036 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
4037 after vectorization.
4039 2018-02-20 Martin Liska <mliska@suse.cz>
4042 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
4043 possible values if we don't have a hint.
4045 2018-02-20 Martin Liska <mliska@suse.cz>
4049 * final.c (shorten_branches): Build align_tab array with one
4051 * opts.c (finish_options): Add alignment option limit check.
4052 (MAX_CODE_ALIGN): Likewise.
4053 (MAX_CODE_ALIGN_VALUE): Likewise.
4054 * doc/invoke.texi: Document maximum allowed option value for
4055 all -falign-* options.
4057 2018-02-19 Jakub Jelinek <jakub@redhat.com>
4060 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
4061 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
4062 * var-tracking.c (emit_note_insn_var_location): Remove all references
4063 to NOTE_INSN_CALL_ARG_LOCATION.
4064 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
4065 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
4066 Use copy_rtx_if_shared.
4067 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
4068 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
4069 (dwarf2out_var_location): Remove handling of
4070 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
4072 * final.c (final_scan_insn): Remove all references to
4073 NOTE_INSN_CALL_ARG_LOCATION.
4074 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
4075 before dumping final insns.
4076 * except.c (emit_note_eh_region_end): Remove all references to
4077 NOTE_INSN_CALL_ARG_LOCATION.
4078 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
4079 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
4080 * config/arc/arc.c (hwloop_optimize): Likewise.
4081 * config/arm/arm.c (create_fix_barrier): Likewise.
4082 * config/s390/s390.c (s390_chunkify_start): Likewise.
4083 * config/sh/sh.c (find_barrier): Likewise.
4084 * config/i386/i386.c (rest_of_insert_endbranch,
4085 ix86_seh_fixup_eh_fallthru): Likewise.
4086 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
4087 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
4088 * config/frv/frv.c (frv_function_prologue): Likewise.
4089 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
4091 (note_outside_basic_block_p): Remove all references to
4092 NOTE_INSN_CALL_ARG_LOCATION.
4093 * gengtype.c (adjust_field_rtx_def): Likewise.
4094 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
4096 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
4097 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
4100 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
4103 PR tree-optimization/84452
4104 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
4105 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
4108 2018-02-19 Martin Liska <mliska@suse.cz>
4111 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
4113 2018-02-19 Martin Liska <mliska@suse.cz>
4114 Richard Sandiford <richard.sandiford@linaro.org>
4116 PR tree-optimization/82491
4117 * gimple-fold.c (get_base_constructor): Make earlier bail out
4120 2018-02-19 Carl Love <cel@us.ibm.com>
4122 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
4123 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
4125 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
4126 P8V_BUILTIN_VEC_NEG.
4128 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
4130 * config/rl78/rl78.md (movdf): New define expand.
4132 2018-02-19 Martin Liska <mliska@suse.cz>
4135 * doc/invoke.texi: Fix typo.
4136 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
4138 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
4140 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
4141 handle rs6000_single_float and rs6000_double_float specially for
4144 2018-02-16 Jeff Law <law@redhat.com>
4146 * config/rx/rx.c (add_pop_cfi_notes): New function.;
4149 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4152 * ipa-inline.c (inline_small_functions): Fix a typo.
4154 2018-02-16 Nathan Sidwell <nathan@acm.org>
4156 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
4158 2018-02-16 Carl Love <cel@us.ibm.com>
4160 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
4161 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
4162 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
4163 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
4164 expansion to P8V_BUILTIN_VEC_FLOAT2.
4166 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
4168 PR rtl-optimization/70023
4169 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
4170 src_regno into account.
4172 2018-02-16 Carl Love <cel@us.ibm.com>
4174 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
4175 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
4176 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
4177 * config/rs6000/rs6000.c: Remove case statements for
4178 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
4179 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
4180 and P9V_BUILTIN_VEC_VINSERT4B.
4181 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
4182 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
4183 * config/rs6000/vsx.md:
4184 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
4187 2018-02-16 Carl Love <cel@us.ibm.com>
4189 * config/rs6000/altivec.h: Add builtin names vec_extract4b
4191 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
4193 * config/rs6000/rs6000-c.c: Add the definitions for
4194 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
4195 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
4196 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
4197 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
4198 definition for insert4b and define insn *insert3b_internal.
4199 * doc/extend.texi: Add documentation for vec_extract4b.
4201 2018-02-16 Nathan Sidwell <nathan@acm.org>
4203 * doc/extend.texi (Backwards Compatibility): Mention friend
4204 injection. Note for-scope is deprecated.
4205 * doc/invoke.texi (-ffriend-injection): Deprecate.
4207 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
4209 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
4210 that moved to I2, also allow destinations that are a paradoxical
4211 subreg (instead of a normal reg).
4213 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
4216 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
4219 2018-02-16 Richard Biener <rguenther@suse.de>
4221 PR tree-optimization/84037
4222 PR tree-optimization/84016
4224 * config/i386/i386.c (ix86_builtin_vectorization_cost):
4225 Adjust vec_construct for the fact we need additional higher latency
4226 128bit inserts for AVX256 and AVX512 vector builds.
4227 (ix86_add_stmt_cost): Scale vector construction cost for
4230 2018-02-16 Richard Biener <rguenther@suse.de>
4232 PR tree-optimization/84417
4233 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
4234 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
4235 (non_rewritable_lvalue_p): Likewise, use poly-ints.
4237 2018-02-16 Martin Liska <mliska@suse.cz>
4240 * internal-fn.def (ASAN_CHECK): Set proper flags.
4241 (ASAN_MARK): Likewise.
4243 2018-02-16 Julia Koval <julia.koval@intel.com>
4245 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
4246 from PTA_CANNONLAKE.
4248 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4251 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
4252 Use ++iter rather than iter++ for std::list iterators.
4253 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
4254 defer deleting them until all nodes in the forest are processed. Do
4255 free even leaf nodes. Change to_process into auto_vec.
4258 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
4259 * vec.h (vec_default_construct): Use memset instead of placement new
4260 if BROKEN_VALUE_INITIALIZATION is defined.
4261 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
4262 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
4265 PR rtl-optimization/83723
4266 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
4267 * lra.c (lra_substitute_pseudo): Likewise. If true, use
4268 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
4270 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
4272 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
4274 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
4276 PR rtl-optimization/81443
4277 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
4278 from inner REGs to paradoxical SUBREGs.
4280 2018-02-16 Richard Biener <rguenther@suse.de>
4282 PR tree-optimization/84399
4283 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4284 For operands we can analyze at their definition make sure we can
4285 analyze them at each use as well.
4287 2018-02-16 Richard Biener <rguenther@suse.de>
4289 PR tree-optimization/84190
4290 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
4291 volatile accesses if the decl isn't volatile.
4293 2018-02-15 Jason Merrill <jason@redhat.com>
4295 PR c++/84314 - ICE with templates and fastcall attribute.
4296 * attribs.c (build_type_attribute_qual_variant): Don't clobber
4297 TYPE_CANONICAL on an existing type.
4299 2018-02-15 Jakub Jelinek <jakub@redhat.com>
4301 PR tree-optimization/84383
4302 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
4303 dstoff nor call operand_equal_p if dstbase is NULL.
4305 PR tree-optimization/84334
4306 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
4307 also a CONSTANT_CLASS_P, punt.
4309 2018-02-14 Jim Wilson <jimw@sifive.com>
4311 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
4312 first SMALL_OPERAND check. New local min_second_step. Move assert
4313 to where locals are set. Add TARGET_RVC support.
4314 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
4316 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
4318 * doc/invoke.texi: Correct -Wformat-overflow code sample.
4320 2018-02-14 Martin Sebor <msebor@redhat.com>
4322 PR tree-optimization/83698
4323 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
4324 arrays constrain the offset range to their bounds.
4325 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
4326 (builtin_access::overlap): Avoid setting the size of overlap if it's
4328 (maybe_diag_overlap): Also consider arrays when deciding what values
4329 of offsets to include in diagnostics.
4331 2018-02-14 Martin Sebor <msebor@redhat.com>
4334 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
4335 that correspond to the kind of a declaration.
4337 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
4340 * config/pa/pa.md: Load address of PIC label using the linkage table
4341 if the label is nonlocal.
4343 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
4345 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
4346 warning message if user requests -maltivec=be.
4347 * doc/invoke.texi: Document deprecation of -maltivec=be.
4349 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
4352 * config/rs6000/rs6000-c.c: Update definitions for
4353 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
4354 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
4356 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4359 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
4360 add _get_ssp intrinsics. Remove argument from
4361 __builtin_ia32_rdssp[d|q].
4362 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
4363 * config/i386/i386-builtin.def: Remove argument from
4364 __builtin_ia32_rdssp[d|q].
4365 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
4366 ix86_expand_special_args_builtin for _rdssp[d|q].
4367 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
4368 Clear register before usage.
4369 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
4370 Add documentation for new _get_ssp and _inc_ssp intrinsics.
4372 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
4374 PR tree-optimization/84357
4375 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
4376 operand 1 of an ARRAY_REF too.
4378 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
4381 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
4382 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4384 (set_of_reg): New struct.
4385 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
4386 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
4387 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4389 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
4390 Split into bitclr, bitset, bitinvert patterns if appropriate.
4391 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
4392 use rx_fuse_in_memory_bitop.
4393 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
4394 to named insn, correct maximum insn length.
4396 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
4399 * machmode.def: Define a complex mode for PARTIAL_INT.
4400 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
4402 * doc/rtl.texi: Document CSPImode.
4403 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
4405 (msp430_hard_regno_nregs_with_padding): Likewise.
4407 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
4410 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
4412 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
4414 PR rtl-optimization/84169
4415 * combine.c (try_combine): New variable split_i2i3. Set it to true if
4416 we generated a parallel as new i3 and we split that to new i2 and i3
4417 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
4418 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
4419 those to i2, not i1. Partially rewrite this scan code.
4421 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4424 * stor-layout.c (place_field): For variable length fields, adjust
4425 offset_align afterwards not just based on the field's alignment,
4426 but also on the size.
4429 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
4430 of exps and logs in the use_exp2 case.
4432 2018-02-13 Jeff Law <law@redhat.com>
4434 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
4437 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
4440 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
4444 * common.opt (gas-loc-support, gas-locview-support): New.
4445 (ginline-points, ginternal-reset-location-views): New.
4446 * doc/invoke.texi: Document them. Use @itemx where intended.
4447 (gvariable-location-views): Adjust.
4448 * target.def (reset_location_view): New.
4449 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
4450 (TARGET_RESET_LOCATION_VIEW): New.
4451 * doc/tm.texi: Rebuilt.
4452 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
4453 (dwarf2out_default_as_locview_support): New.
4454 (output_asm_line_debug_info): Use option variables.
4455 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
4456 (output_loc_list): Likewise.
4457 (add_high_low_attributes): Check option variables.
4458 Don't output entry view attribute in strict mode.
4459 (gen_inlined_subroutine_die): Check option variables.
4460 (dwarf2out_inline_entry): Likewise.
4461 (init_sections_and_labels): Likewise.
4462 (dwarf2out_early_finish): Likewise.
4463 (maybe_reset_location_view): New, from...
4464 (dwarf2out_var_location): ... here. Call it.
4465 * debug.h (dwarf2out_default_as_loc_support): Declare.
4466 (dwarf2out_default_as_locview_support): Declare.
4467 * hooks.c (hook_int_rtx_insn_0): New.
4468 * hooks.h (hook_int_rtx_insn_0): Declare.
4469 * toplev.c (process_options): Take -gas-loc-support and
4470 -gas-locview-support from dwarf2out. Enable
4471 -gvariable-location-views by default only with locview
4472 assembler support. Enable -ginternal-reset-location-views by
4473 default only if the target defines the corresponding hook.
4474 Enable -ginline-points by default if location views are
4475 enabled; force it disabled if statement frontiers are
4477 * tree-inline.c (expand_call_inline): Check option variables.
4478 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
4480 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
4482 PR tree-optimization/84321
4483 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
4484 handling. Also check whether the anti-range contains any values
4485 that satisfy the mask; switch to a VR_RANGE if not.
4487 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
4490 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
4492 2018-02-13 Martin Jambor <mjambor@suse.cz>
4495 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
4496 of call statements, also set location of a load to a temporary.
4498 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
4500 * config/rl78/rl78.c (add_vector_labels): New function.
4501 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
4502 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
4503 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
4504 which checks that no arguments are passed.
4505 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
4506 * doc/extend.texi: Documentation for the new attribute.
4508 2018-02-13 Andreas Schwab <schwab@suse.de>
4510 * config/riscv/linux.h (CPP_SPEC): Define.
4512 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4515 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
4516 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
4517 OPTION_MASK_ISA_AES as first argument to def_builtin_const
4518 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
4519 instead of OPTION_MASK_ISA_PCLMUL as first argument to
4520 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
4521 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
4522 temporarily for AES and PCLMUL builtins.
4524 PR tree-optimization/84339
4525 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
4526 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
4530 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
4531 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
4532 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
4534 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
4536 * omp-simd-clone.h: New file.
4537 * omp-simd-clone.c: Include omp-simd-clone.h.
4538 (expand_simd_clones): No longer static.
4539 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
4540 cgraph.h and omp-simd-clone.h.
4541 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
4542 (vect_recog_widen_shift_pattern): Formatting fix.
4543 (vect_pattern_recog_1): Don't check optab for calls.
4546 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
4547 operands[2] into a REG before using gen_lowpart on it.
4549 2018-02-12 Jeff Law <law@redhat.com>
4552 * config/sh/sh.c (find_barrier): Consider a sibling call
4555 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
4556 successfully back substituting a reg.
4558 2018-02-12 Richard Biener <rguenther@suse.de>
4560 PR tree-optimization/84037
4561 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
4562 parameter, move visited init to caller.
4563 (vect_slp_analyze_operations): Separate cost from validity
4564 check, initialize visited once for all instances.
4565 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
4567 * tree-vect-stmts.c (vect_model_simple_cost): Make early
4569 (vect_model_promotion_demotion_cost): Likewise.
4570 (vectorizable_bswap): Guard cost modeling with !slp_node
4571 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
4573 (vectorizable_call): Likewise.
4574 (vectorizable_conversion): Likewise.
4575 (vectorizable_assignment): Likewise.
4576 (vectorizable_shift): Likewise.
4577 (vectorizable_operation): Likewise.
4578 (vectorizable_store): Likewise.
4579 (vectorizable_load): Likewise.
4580 (vectorizable_condition): Likewise.
4581 (vectorizable_comparison): Likewise.
4583 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
4586 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
4587 (ASAN_MARK): Fix fnspec to account for return value, change pointer
4588 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
4590 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4593 * params.def (inline-min-speedup): Increase from 8 to 15.
4594 (max-inline-insns-auto): Decrease from 40 to 30.
4595 * ipa-split.c (consider_split): Add some buffer for function to
4596 be considered inlining candidate.
4597 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
4600 2018-02-12 Richard Biener <rguenther@suse.de>
4602 PR tree-optimization/84037
4603 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
4604 matched stmts if we cannot swap the non-matched ones.
4606 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4608 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
4609 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
4610 _mm_maskz_scalef_round_ss): New intrinsics.
4611 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
4612 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
4613 __builtin_ia32_scalefss_round): Remove.
4614 (__builtin_ia32_scalefsd_mask_round,
4615 __builtin_ia32_scalefss_mask_round): New intrinsics.
4616 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
4617 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
4618 ((match_operand:VF_128 2 "<round_nimm_predicate>"
4619 "<round_constraint>")): Changed to ...
4620 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
4621 "<round_scalar_constraint>")): ... this.
4622 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
4623 %0, %1, %2<round_op3>}"): Changed to ...
4624 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
4625 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
4626 %2<round_scalar_mask_op3>}"): ... this.
4627 * config/i386/subst.md (round_scalar_nimm_predicate): New.
4629 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4631 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
4632 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
4633 (_mm_maskz_sqrt_round_ss): New intrinsics.
4634 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
4635 (__builtin_ia32_sqrtsd_mask_round)
4636 (__builtin_ia32_sqrtss_mask_round): New builtins.
4637 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
4638 (__builtin_ia32_sqrtss_round): Remove.
4639 (__builtin_ia32_sqrtsd_mask_round)
4640 (__builtin_ia32_sqrtss_mask_round): New builtins.
4641 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
4642 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
4643 ((match_operand:VF_128 1 "vector_operand"
4644 "xBm,<round_constraint>")): Changed to ...
4645 ((match_operand:VF_128 1 "vector_operand"
4646 "xBm,<round_scalar_constraint>")): ... this.
4647 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
4648 %0, %2, %<iptr>1<round_op3>}): Changed to ...
4649 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
4650 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
4651 %<iptr>1<round_scalar_mask_op3>}): ... this.
4652 ((set_attr "prefix" "<round_prefix>")): Changed to ...
4653 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
4655 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
4658 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
4659 Cast vec_cmpeq result to correct type.
4660 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
4661 Cast vec_cmpgt result to correct type.
4663 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
4665 * final.c (final_scan_insn_1): Renamed from...
4666 (final_scan_insn): ... this. New wrapper, to recover
4667 seen from the outermost call in recursive ones.
4668 * config/sparc/sparc.c (output_return): Drop seen from call.
4669 (output_sibcall): Likewise.
4670 * config/visium/visium.c (output_branch): Likewise.
4672 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
4674 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
4677 2018-02-10 Alan Modra <amodra@gmail.com>
4680 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
4681 Specify LR as an input.
4683 2018-02-10 Jakub Jelinek <jakub@redhat.com>
4686 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
4687 remove_member_access_dummy_vars): New functions.
4688 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
4689 lower_omp_1, execute_lower_omp): Use them.
4691 PR rtl-optimization/84308
4692 * shrink-wrap.c (spread_components): Release todo vector.
4694 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
4696 PR rtl-optimization/57193
4697 * ira-color.c (struct allocno_color_data): Add member
4698 conflict_allocno_hard_prefs.
4699 (update_conflict_allocno_hard_prefs): New.
4700 (bucket_allocno_compare_func): Add a preference based on
4701 conflict_allocno_hard_prefs.
4702 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
4703 (color_allocnos): Remove a dead code. Initiate
4704 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
4706 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4709 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
4710 constraint from =wa to wa. Avoid a subreg on the output operand,
4711 instead use a pseudo and subreg it in a move.
4712 (p9_xxbrd_<mode>): Changed to ...
4713 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
4714 (p9_xxbrd_v2df): New expander.
4715 (p9_xxbrw_<mode>): Changed to ...
4716 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
4717 (p9_xxbrw_v4sf): New expander.
4719 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4721 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
4723 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
4726 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
4727 multiply in 32-bit mode.
4728 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
4729 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
4732 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4734 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
4735 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
4736 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
4737 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
4739 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
4742 * dwarf2out.c (is_trivial_indirect_ref): New function.
4743 (dwarf2out_late_global_decl): Do not generate a location
4744 attribute for variables that have a non-trivial DECL_VALUE_EXPR
4745 and that are not defined in the current unit.
4747 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4749 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
4750 instead of a libcall for UNORDERED.
4752 2018-02-09 Tamar Christina <tamar.christina@arm.com>
4755 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
4756 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
4758 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4761 * config/s390/s390.c (s390_set_current_function): Invoke
4762 s390_indirect_branch_settings also if fndecl didn't change.
4764 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4766 * config/rs6000/rs6000.md (blockage): Set length to zero.
4768 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4770 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
4772 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4775 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
4776 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
4780 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
4781 PARALLEL incoming that failed vt_get_decl_and_offset check.
4784 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
4785 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
4787 (get_variable_section): For decls in named .bss* sections pass true as
4788 second argument to bss_initializer_p.
4790 2018-02-09 Marek Polacek <polacek@redhat.com>
4791 Jakub Jelinek <jakub@redhat.com>
4794 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
4795 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
4796 Sync some changes from cxx_fold_indirect_ref.
4798 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4800 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
4802 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
4803 (BLOCK_INLINE_ENTRY_LABEL): New.
4804 (dwarf2out_var_location): Disregard inline entry markers.
4805 (inline_entry_data): New struct.
4806 (inline_entry_data_hasher): New hashtable type.
4807 (inline_entry_data_hasher::hash): New.
4808 (inline_entry_data_hasher::equal): New.
4809 (inline_entry_data_table): New variable.
4810 (add_high_low_attributes): Add DW_AT_entry_pc and
4811 DW_AT_GNU_entry_view attributes if a pending entry is found
4812 in inline_entry_data_table. Add old entry_pc attribute only
4813 if debug nonbinding markers are disabled.
4814 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
4815 markers are enabled.
4816 (block_within_block_p, dwarf2out_inline_entry): New.
4817 (dwarf2out_finish): Check that no entries remained in
4818 inline_entry_data_table.
4819 * final.c (reemit_insn_block_notes): Handle inline entry notes.
4820 (final_scan_insn, notice_source_line): Likewise.
4821 (rest_of_clean_state): Skip inline entry markers.
4822 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
4824 * gimple.c (gimple_build_debug_inline_entry): New.
4825 * gimple.h (enum gimple_debug_subcode): Add
4826 GIMPLE_DEBUG_INLINE_ENTRY.
4827 (gimple_build_debug_inline_entry): Declare.
4828 (gimple_debug_inline_entry_p): New.
4829 (gimple_debug_nonbind_marker_p): Adjust.
4830 * insn-notes.def (INLINE_ENTRY): New.
4831 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
4832 inline entry marker notes.
4833 (print_insn): Likewise.
4834 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
4835 (INSN_DEBUG_MARKER_KIND): Likewise.
4836 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
4837 * tree-inline.c (expand_call_inline): Build and insert
4838 debug_inline_entry stmt.
4839 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
4840 inline entry blocks early, if nonbind markers are enabled.
4841 (dump_scope_block): Dump fragment info.
4842 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
4843 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
4844 (gimple_build_debug_inline_entry): New.
4845 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
4846 Enable/disable inline entry points too.
4847 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
4848 (DEBUG_INSN): Describe inline entry markers.
4850 * common.opt (gvariable-location-views): New.
4851 (gvariable-location-views=incompat5): New.
4852 * config.in: Rebuilt.
4853 * configure: Rebuilt.
4854 * configure.ac: Test assembler for view support.
4855 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
4856 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
4857 * dwarf2out.c (var_loc_view): New typedef.
4858 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
4859 (dwarf2out_locviews_in_attribute): New.
4860 (dwarf2out_locviews_in_loclist): New.
4861 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
4862 (enum dw_line_info_opcode): Add LI_adv_address.
4863 (struct dw_line_info_table): Add view.
4864 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
4865 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
4866 (zero_view_p): New variable.
4867 (ZERO_VIEW_P): New macro.
4868 (output_asm_line_debug_info): New.
4869 (struct var_loc_node): Add view.
4870 (add_AT_view_list, AT_loc_list): New.
4871 (add_var_loc_to_decl): Add view param. Test it against last.
4872 (new_loc_list): Add view params. Record them.
4873 (AT_loc_list_ptr): Handle loc and view lists.
4874 (view_list_to_loc_list_val_node): New.
4875 (print_dw_val): Handle dw_val_class_view_list.
4876 (size_of_die): Likewise.
4877 (value_format): Likewise.
4878 (loc_list_has_views): New.
4879 (gen_llsym): Set vl_symbol too.
4880 (maybe_gen_llsym, skip_loc_list_entry): New.
4881 (dwarf2out_maybe_output_loclist_view_pair): New.
4882 (output_loc_list): Output view list or entries too.
4883 (output_view_list_offset): New.
4884 (output_die): Handle dw_val_class_view_list.
4885 (output_dwarf_version): New.
4886 (output_compilation_unit_header): Use it.
4887 (output_skeleton_debug_sections): Likewise.
4888 (output_rnglists, output_line_info): Likewise.
4889 (output_pubnames, output_aranges): Update version comments.
4890 (output_one_line_info_table): Output view numbers in asm comments.
4891 (dw_loc_list): Determine current endview, pass it to new_loc_list.
4892 Call maybe_gen_llsym.
4893 (loc_list_from_tree_1): Adjust.
4894 (add_AT_location_description): Create view list attribute if
4895 needed, check it's absent otherwise.
4896 (convert_cfa_to_fb_loc_list): Adjust.
4897 (maybe_emit_file): Call output_asm_line_debug_info for test.
4898 (dwarf2out_var_location): Reset views as needed. Precompute
4899 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
4900 attribute. Set view.
4901 (new_line_info_table): Reset next view.
4902 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
4903 (dwarf2out_source_line): Likewise. Output view resets and labels to
4904 the assembler, or select appropriate line info opcodes.
4905 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
4906 (optimize_string_length): Catch it. Adjust.
4907 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
4908 dw_val_class_view_list, and remove it if no longer needed.
4909 (hash_loc_list): Hash view numbers.
4910 (loc_list_hasher::equal): Compare them.
4911 (optimize_location_lists): Check whether a view list symbol is
4912 needed, and whether the locview attribute is present, and
4913 whether they match. Remove the locview attribute if no longer
4915 (index_location_lists): Call skip_loc_list_entry for test.
4916 (dwarf2out_finish): Call output_asm_line_debug_info for test.
4917 Use output_dwarf_version.
4918 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
4919 (struct dw_val_node): Add val_view_list.
4920 * final.c (SEEN_NEXT_VIEW): New.
4921 (set_next_view_needed): New.
4922 (clear_next_view_needed): New.
4923 (maybe_output_next_view): New.
4924 (final_start_function): Rename to...
4925 (final_start_function_1): ... this. Take pointer to FIRST,
4926 add SEEN parameter. Emit param bindings in the initial view.
4927 (final_start_function): Reintroduce SEEN-less interface.
4928 (final): Rename to...
4929 (final_1): ... this. Take SEEN parameter. Output final pending
4930 next view at the end.
4931 (final): Reintroduce seen-less interface.
4932 (final_scan_insn): Output pending next view before switching
4933 sections or ending a block. Mark the next view as needed when
4934 outputting variable locations. Notify debug backend of section
4935 changes, and of location view changes.
4936 (rest_of_handle_final): Adjust.
4937 * toplev.c (process_options): Autodetect value for debug variable
4938 location views option. Warn on incompat5 without -gdwarf-5.
4939 * doc/invoke.texi (gvariable-location-views): New.
4940 (gvariable-location-views=incompat5): New.
4941 (gno-variable-location-views): New.
4943 2018-02-08 David Malcolm <dmalcolm@redhat.com>
4945 PR tree-optimization/84136
4946 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
4947 that the result of find_edge is non-NULL.
4949 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4952 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
4953 storing integer register in SImode. Fix cost of 256 and 512
4954 byte aligned SSE register store.
4956 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4958 * config/i386/i386.c (ix86_multiplication_cost): Fix
4959 multiplication cost for TARGET_AVX512DQ.
4961 2018-02-08 Marek Polacek <polacek@redhat.com>
4963 PR tree-optimization/84238
4964 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
4967 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4969 PR tree-optimization/84265
4970 * tree-vect-stmts.c (vectorizable_store): Don't treat
4971 VMAT_CONTIGUOUS accesses as grouped.
4972 (vectorizable_load): Likewise.
4974 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4976 PR tree-optimization/81635
4977 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
4978 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
4979 (test_round_for_mask): New functions.
4980 (wide_int_cc_tests): Call test_round_for_mask.
4981 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
4982 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
4983 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
4984 range returned by get_range_info.
4986 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4989 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
4990 * symtab.c: Include builtins.h
4991 (symtab_node::output_to_lto_symbol_table_p): Move here
4992 from lto-streamer-out.c:output_symbol_p.
4993 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
4994 (output_symbol_p): Move all logic to symtab.c
4995 (produce_symtab): Update.
4997 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4999 * config/s390/s390-opts.h (enum indirect_branch): Define.
5000 * config/s390/s390-protos.h (s390_return_addr_from_memory)
5001 (s390_indirect_branch_via_thunk)
5002 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
5003 (enum s390_indirect_branch_type): Define.
5004 * config/s390/s390.c (struct s390_frame_layout, struct
5005 machine_function): Remove.
5006 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
5007 (indirect_branch_table_label_no, indirect_branch_table_name):
5009 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
5010 (enum s390_indirect_branch_option): Define.
5011 (s390_return_addr_from_memory): New function.
5012 (s390_handle_string_attribute): New function.
5013 (s390_attribute_table): Add new attribute handler.
5014 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
5015 (s390_indirect_branch_via_thunk): New function.
5016 (s390_indirect_branch_via_inline_thunk): New function.
5017 (s390_function_ok_for_sibcall): When jumping via thunk disallow
5018 sibling call optimization for non z10 compiles.
5019 (s390_emit_call): Force indirect branch target to be a single
5020 register. Add r1 clobber for non-z10 compiles.
5021 (s390_emit_epilogue): Emit return jump via return_use expander.
5022 (s390_reorg): Handle JUMP_INSNs as execute targets.
5023 (s390_option_override_internal): Perform validity checks for the
5024 new command line options.
5025 (s390_indirect_branch_attrvalue): New function.
5026 (s390_indirect_branch_settings): New function.
5027 (s390_set_current_function): Invoke s390_indirect_branch_settings.
5028 (s390_output_indirect_thunk_function): New function.
5029 (s390_code_end): Implement target hook.
5030 (s390_case_values_threshold): Implement target hook.
5031 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
5033 * config/s390/s390.h (struct s390_frame_layout)
5034 (struct machine_function): Move here from s390.c.
5035 (TARGET_INDIRECT_BRANCH_NOBP_RET)
5036 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
5037 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
5038 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
5039 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
5040 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
5041 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
5042 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
5043 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
5044 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
5045 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
5046 (mnemonic attribute): Add values which aren't recognized
5048 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
5049 pattern for branch conversion. Fix mnemonic attribute.
5050 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
5051 indirect branch via thunk if requested.
5052 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
5053 ("*indirect_jump"): Disable for branch conversion using out of
5055 ("indirect_jump_via_thunk<mode>_z10")
5056 ("indirect_jump_via_thunk<mode>")
5057 ("indirect_jump_via_inlinethunk<mode>_z10")
5058 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
5059 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
5060 ("casesi_jump_via_inlinethunk<mode>_z10")
5061 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
5062 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
5063 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
5064 ("*indirect2_jump"): Disable for branch conversion.
5065 ("casesi_jump"): Turn into expander and expand patterns for branch
5067 ("return_use"): New expander.
5068 ("*return"): Emit return via thunk and rename it to ...
5069 ("*return<mode>"): ... this one.
5070 * config/s390/s390.opt: Add new options and and enum for the
5073 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5075 * lra-constraints.c (match_reload): Unconditionally use
5076 gen_lowpart_SUBREG, rather than selecting between that
5077 and equivalent gen_rtx_SUBREG code.
5079 2018-02-08 Richard Biener <rguenther@suse.de>
5081 PR tree-optimization/84233
5082 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
5083 changed flag instead of boguously re-using phi_inserted.
5085 2018-02-08 Martin Jambor <mjambor@suse.cz>
5087 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
5088 static local variables.
5090 2018-02-08 Richard Biener <rguenther@suse.de>
5092 PR tree-optimization/84278
5093 * tree-vect-stmts.c (vectorizable_store): When looking for
5094 smaller vector types to perform grouped strided loads/stores
5095 make sure the mode is supported by the target.
5096 (vectorizable_load): Likewise.
5098 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5100 * config/aarch64/aarch64.c (aarch64_components_for_bb):
5101 Increase LDP/STP opportunities by adding adjacent callee-saves.
5103 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5105 PR rtl-optimization/84068
5106 PR rtl-optimization/83459
5107 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
5109 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
5111 PR tree-optimization/84224
5112 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
5113 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
5116 2018-02-07 Iain Sandoe <iain@codesourcery.com>
5119 * config/rs6000/altivec.md (*restore_world): Remove LR use.
5120 * config/rs6000/predicates.md (restore_world_operation): Adjust op
5121 count, remove one USE.
5123 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
5125 * doc/install.texi (Configuration): Document the
5126 --with-long-double-format={ibm,ieee} PowerPC configuration
5130 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
5131 Convert from define_expand to be define_insn_and_split. Rework
5132 float/double/_Float128 conversions to QI/HI/SImode to work with
5133 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
5134 conversions to QI/HImode types did a store and then a load to
5135 truncate the value. For conversions to VSX registers, don't split
5136 the insn, instead emit the code directly. Use the code iterator
5137 any_fix to combine signed and unsigned conversions.
5138 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
5139 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
5140 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5141 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
5142 (fix_<mode>di2_hw): Likewise.
5143 (fixuns_<mode>di2_hw): Likewise.
5144 (fix_<mode>si2_hw): Likewise.
5145 (fixuns_<mode>si2_hw): Likewise.
5146 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
5147 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5148 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
5149 fix<uns>_trunc<SFDF:mode>si2_p8.
5150 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
5152 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
5153 (fix<uns>_<mode>_mem): Likewise.
5154 (fctiw<u>z_<mode>_mem): Likewise.
5155 (fix<uns>_<mode>_mem): Likewise.
5156 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
5157 the register allocator from doing a direct move to the GPRs to do
5158 a store, and instead use the ISA 3.0 store byte/half-word from
5159 vector register instruction. For IEEE 128-bit floating point,
5160 also optimize stores of 32-bit ints.
5161 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
5163 2018-02-07 Alan Hayward <alan.hayward@arm.com>
5165 * genextract.c (push_pathstr_operand): New function to support
5167 (walk_rtx): Call push_pathstr_operand.
5168 (print_path): Support [a-zA-Z].
5170 2018-02-07 Richard Biener <rguenther@suse.de>
5172 PR tree-optimization/84037
5173 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
5174 (cse_and_gimplify_to_preheader): Declare.
5175 (vect_get_place_in_interleaving_chain): Likewise.
5176 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5178 (_loop_vec_info::~_loop_vec_info): Delete it.
5179 (cse_and_gimplify_to_preheader): New function.
5180 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
5181 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
5182 (vectorizable_load): Likewise. For grouped stores always base
5183 the IV on the first element.
5184 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
5185 condition before gimplifying.
5187 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5189 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
5190 *DIV_EXPR and *MOD_EXPR.
5192 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
5195 * config/i386/i386.c (ix86_option_override_internal): Mask out
5196 the CF_SET bit when checking -fcf-protection.
5198 2018-02-07 Tom de Vries <tom@codesourcery.com>
5201 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
5204 2018-02-07 Richard Biener <rguenther@suse.de>
5206 PR tree-optimization/84204
5207 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
5210 PR tree-optimization/84205
5211 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
5212 special-case isl_ast_op_zdiv_r.
5214 PR tree-optimization/84223
5215 * graphite-scop-detection.c (gather_bbs::before_dom_children):
5216 Only add conditions from within the region.
5217 (gather_bbs::after_dom_children): Adjust.
5219 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
5222 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
5223 * config/avr/avr.md: Only post-reload split REG-REG moves if
5224 either register is GENERAL_REG_P.
5226 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5228 PR tree-optimization/84235
5229 * tree-ssa-scopedtables.c
5230 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
5231 if the subtraction is performed in floating point type where NaNs are
5232 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
5233 build 1. Formatting fix.
5235 2018-02-06 Jakub Jelinek <jakub@redhat.com>
5238 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
5239 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
5240 and skip it regardless of bb boundaries. Use CALL_P macro,
5241 don't test INSN_P (insn) together with CALL_P or JUMP_P check
5242 unnecessarily, formatting fix.
5244 2018-02-06 Michael Collison <michael.collison@arm.com>
5246 * config/arm/thumb2.md:
5247 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
5248 (*thumb_mov_notscc): Ditto.
5250 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
5253 * config/rs6000/rs6000.md (su code attribute): Use "u" for
5254 unsigned_fix, not "s".
5256 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5258 * configure.ac (gcc_fn_eh_frame_ro): New function.
5259 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
5260 correct .eh_frame permissions.
5261 * configure: Regenerate.
5263 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
5265 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
5268 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5270 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5271 Display warning message for -mno-speculate-indirect-jumps.
5273 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
5275 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
5277 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
5279 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
5281 PR tree-optimization/84225
5282 * tree-eh.c (find_trapping_overflow): Only call
5283 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
5285 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5288 * config/i386/i386.c: Reimplement the check of possible options
5289 -mibt/-mshstk conbination. Change error messages.
5290 * doc/invoke.texi: Fix a typo: remove extra '='.
5292 2018-02-06 Marek Polacek <polacek@redhat.com>
5294 PR tree-optimization/84228
5295 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
5297 2018-02-06 Tamar Christina <tamar.christina@arm.com>
5300 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
5301 emitted arch directives.
5302 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
5303 __ARM_FEATURE_COPROC before changing architectures.
5305 2018-02-06 Richard Biener <rguenther@suse.de>
5307 * config/i386/i386.c (print_reg): Fix typo.
5308 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
5310 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
5312 * configure: Regenerate.
5314 2018-02-05 Martin Sebor <msebor@redhat.com>
5316 PR tree-optimization/83369
5317 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
5320 2018-02-05 Martin Liska <mliska@suse.cz>
5322 * doc/invoke.texi: Cherry-pick upstream r323995.
5324 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
5326 * ira.c (ira_init_register_move_cost): Adjust comment.
5328 2018-02-05 Martin Liska <mliska@suse.cz>
5330 PR gcov-profile/84137
5331 * doc/gcov.texi: Fix typo in documentation.
5333 2018-02-05 Martin Liska <mliska@suse.cz>
5335 PR gcov-profile/83879
5336 * doc/gcov.texi: Document necessity of --dynamic-list-data when
5337 using dlopen functionality.
5339 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
5341 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
5342 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
5343 _mm_maskz_range_ss, _mm_mask_range_round_ss,
5344 _mm_maskz_range_round_ss): New intrinsics.
5345 (__builtin_ia32_rangesd128_round)
5346 (__builtin_ia32_rangess128_round): Remove.
5347 (__builtin_ia32_rangesd128_mask_round,
5348 __builtin_ia32_rangess128_mask_round): New builtins.
5349 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
5350 __builtin_ia32_rangess128_round): Remove.
5351 (__builtin_ia32_rangesd128_mask_round,
5352 __builtin_ia32_rangess128_mask_round): New builtins.
5353 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
5354 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
5355 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
5356 "<round_saeonly_constraint>")): Changed to ...
5357 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
5358 "<round_saeonly_scalar_constraint>")): ... this.
5359 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
5360 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
5361 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
5362 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
5363 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
5365 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
5367 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
5369 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
5370 Remove all values except native, 8540 and 8548.
5372 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
5374 * config/i386/i386.c (ix86_output_function_return): Pass
5375 INVALID_REGNUM, instead of -1, as invalid register number to
5376 indirect_thunk_name and output_indirect_thunk.
5378 2018-02-02 Julia Koval <julia.koval@intel.com>
5380 * config.gcc: Add -march=icelake.
5381 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
5382 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
5383 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
5384 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
5385 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
5386 (processor_target_table): Add icelake.
5387 (ix86_option_override_internal): Handle new PTAs.
5388 (get_builtin_code_for_version): Handle icelake.
5389 (M_INTEL_COREI7_ICELAKE): New.
5390 (fold_builtin_cpu): Handle icelake.
5391 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
5392 * doc/invoke.texi: Add -march=icelake.
5394 2018-02-02 Julia Koval <julia.koval@intel.com>
5396 * config/i386/i386.c (ix86_option_override_internal): Change flags type
5397 to wide_int_bitmask.
5398 * wide-int-bitmask.h: New.
5400 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5403 * config/i386/i386.md: Replace Pmode with word_mode in
5404 builtin_setjmp_setup and builtin_longjmp to support x32.
5406 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
5410 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
5412 (rs6000_supported_cpu_names): New static variable.
5413 (linux_cpu_translation_table): Likewise.
5414 (elf_platform) <cpu>: Define new static variable and use it.
5415 Translate kernel AT_PLATFORM name to canonical name if needed.
5416 Error if platform name is unknown.
5418 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
5421 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
5423 2018-02-01 Jeff Law <law@redhat.com>
5426 * config/i386/i386.c (release_scratch_register_on_entry): Add new
5427 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
5428 the scratch if RELEASE_VIA_POP is false.
5429 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
5430 If we have to save a temporary register, decrement SIZE appropriately.
5431 Pass new arguments to release_scratch_register_on_entry.
5432 (ix86_adjust_stack_and_probe): Likewise.
5433 (ix86_emit_probe_stack_range): Pass new arguments to
5434 release_scratch_register_on_entry.
5436 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
5438 PR rtl-optimization/84157
5439 * combine.c (change_zero_ext): Use REG_P predicate in
5440 front of HARD_REGISTER_P predicate.
5442 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
5444 * config/avr/avr.c (avr_option_override): Move disabling of
5445 -fdelete-null-pointer-checks to...
5446 * common/config/avr/avr-common.c (avr_option_optimization_table):
5449 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5451 PR tree-optimization/81635
5452 * tree-data-ref.c (split_constant_offset_1): For types that
5453 wrap on overflow, try to use range info to prove that wrapping
5456 2018-02-01 Renlin Li <renlin.li@arm.com>
5459 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
5461 (aarch64_register_move_cost): Likewise.
5462 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
5464 (REG_CLASS_NAMES): Likewise.
5465 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
5466 TAILCALL_ADDR_REGS. Remove IP registers.
5467 * config/aarch64/aarch64.md (Ucs): Update register constraint.
5469 2018-02-01 Richard Biener <rguenther@suse.de>
5471 * domwalk.h (dom_walker::dom_walker): Add additional constructor
5472 for specifying RPO order and allow NULL for that.
5473 * domwalk.c (dom_walker::dom_walker): Likewise.
5474 (dom_walker::walk): Handle NULL RPO order.
5475 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
5477 (rewrite_update_dom_walker): Likewise.
5478 (mark_def_dom_walker): Likewise.
5480 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5482 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
5483 (aarch64_maybe_expand_sve_subreg_move): Declare.
5484 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
5485 * config/aarch64/predicates.md (aarch64_any_register_operand): New
5487 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
5488 that are semantically a reverse operation.
5489 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
5490 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
5491 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
5493 (aarch64_can_change_mode_class): For big-endian, forbid changes
5494 between two SVE modes if they have different element sizes.
5496 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5498 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
5499 the TImode handling for big-endian targets.
5501 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5503 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
5504 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
5506 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
5507 Remove BSWAP handing for big-endian targets and use the form of
5508 LD1RQ appropariate for the mode.
5510 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5512 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
5513 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
5516 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5519 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
5520 check for operands that need to go through aarch64_sve_reload_be.
5522 2018-02-01 Jakub Jelinek <jakub@redhat.com>
5524 PR tree-optimization/81661
5525 PR tree-optimization/84117
5526 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
5527 * tree-eh.c: Include gimplify.h.
5528 (find_trapping_overflow, replace_trapping_overflow,
5529 rewrite_to_non_trapping_overflow): New functions.
5530 * tree-vect-loop.c: Include tree-eh.h.
5531 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
5532 * tree-data-ref.c: Include tree-eh.h.
5533 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
5535 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
5537 PR rtl-optimization/84123
5538 * combine.c (change_zero_ext): Check if hard register satisfies
5539 can_change_dest_mode before calling gen_lowpart_SUBREG.
5541 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
5544 * ira.c (ira_init_register_move_cost): Remove assert.
5546 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5548 PR rtl-optimization/84071
5549 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
5550 * doc/tm.texi: Regenerate.
5552 2018-01-31 Richard Biener <rguenther@suse.de>
5554 PR tree-optimization/84132
5555 * tree-data-ref.c (analyze_miv_subscript): Properly
5556 check whether evolution_function_is_affine_multivariate_p
5557 before calling gcd_of_steps_may_divide_p.
5559 2018-01-31 Julia Koval <julia.koval@intel.com>
5562 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
5563 * config/i386/i386.md (rdpid_rex64) New.
5564 (rdpid): Make 32bit only.
5566 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
5569 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
5570 an IDENTIFIER_NODE for FUNCTION_TYPE's.
5572 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5575 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5577 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5579 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5581 PR rtl-optimization/84071
5582 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
5583 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
5585 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5587 * config/arc/arc.c (arc_handle_aux_attribute): New function.
5588 (arc_attribute_table): Add 'aux' attribute.
5589 (arc_in_small_data_p): Consider aux like variables.
5590 (arc_is_aux_reg_p): New function.
5591 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
5592 (arc_get_aux_arg): New function.
5593 (prepare_move_operands): Handle aux-register access.
5594 (arc_handle_aux_attribute): New function.
5595 * doc/extend.texi (ARC Variable attributes): Add subsection.
5597 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5599 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
5600 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
5601 (arc_attribute_table): Add 'uncached' attribute.
5602 (arc_print_operand): Print '.di' flag for uncached memory
5604 (arc_in_small_data_p): Do not consider for small data the uncached
5606 (arc_is_uncached_mem_p): New function.
5607 * config/arc/predicates.md (compact_store_memory_operand): Check
5608 for uncached memory accesses.
5609 (nonvol_nonimm_operand): Likewise.
5610 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
5612 2018-01-31 Jakub Jelinek <jakub@redhat.com>
5615 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
5616 falign-loops=): Add Optimization flag.
5618 2018-01-30 Jeff Law <law@redhat.com>
5621 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
5622 INT_REGISTERS_SAVED. Check it prior to calling
5623 get_scratch_register_on_entry.
5624 (ix86_adjust_stack_and_probe): Similarly.
5625 (ix86_emit_probe_stack_range): Similarly.
5626 (ix86_expand_prologue): Corresponding changes.
5628 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5631 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
5632 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
5634 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
5637 * lra-constraints.c (curr_insn_transform): Process AND in the
5640 2018-01-30 Jakub Jelinek <jakub@redhat.com>
5642 PR rtl-optimization/83986
5643 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
5644 dependence against last_pending_memory_flush in addition to
5647 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
5649 PR tree-optimization/81611
5650 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
5653 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5656 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
5659 2018-01-30 Richard Biener <rguenther@suse.de>
5660 Jakub Jelinek <jakub@redhat.com>
5662 PR tree-optimization/84111
5663 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
5664 inner loops added during recursion, as they don't have up-to-date
5667 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5670 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
5671 (can_inline_edge_by_limits_p): ... here.
5672 (can_early_inline_edge_p, check_callers,
5673 update_caller_keys, update_callee_keys, recursive_inlining,
5674 add_new_edges_to_heap, speculation_useful_p,
5675 inline_small_functions,
5676 inline_small_functions, flatten_function,
5677 inline_to_all_callers_1): Update.
5679 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5681 * profile-count.c (profile_count::combine_with_ipa_count): Handle
5684 2018-01-30 Richard Biener <rguenther@suse.de>
5686 PR tree-optimization/83008
5687 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
5688 invariant and constant vector uses in stmts when they need
5691 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5694 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
5695 * configure: Regenerate.
5697 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5699 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
5701 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
5702 Use gen_rtx_REG rather than gen_lowpart.
5704 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5706 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
5707 rather than 0 when creating partial subregs.
5709 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5711 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
5714 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
5717 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
5718 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
5719 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
5720 flags. This restores the settings used before the 2017-07-24.
5721 Turning off pre increment/decrement/modify allows IVOPTS to
5722 optimize DF/SF loops where the index is an int.
5724 2018-01-29 Richard Biener <rguenther@suse.de>
5725 Kelvin Nilsen <kelvin@gcc.gnu.org>
5728 * tree-vect-stmts.c (vectorizable_call): Don't call
5729 targetm.vectorize_builtin_md_vectorized_function if callee is
5732 2018-01-22 Carl Love <cel@us.ibm.com>
5734 * doc/extend.tex: Fix typo in second arg in
5735 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
5737 2018-01-29 Richard Biener <rguenther@suse.de>
5739 PR tree-optimization/84086
5740 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
5741 (flush_ssaname_freelist): When SSA names were released reset
5742 the SCEV hash table.
5744 2018-01-29 Richard Biener <rguenther@suse.de>
5746 PR tree-optimization/84057
5747 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
5748 removed paths when removing edges.
5750 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
5752 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
5753 -mfunction-return=@var{choice}.
5755 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
5758 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
5760 (layout::print_source_line): Likewise.
5761 (test_get_line_width_without_trailing_whitespace): Add test cases.
5763 2018-01-27 Jakub Jelinek <jakub@redhat.com>
5766 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
5769 2018-01-26 Jim Wilson <jimw@sifive.com>
5771 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
5773 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
5776 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5778 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
5779 and CMP + SUB-immediate -> SUBS.
5781 2018-01-26 Martin Sebor <msebor@redhat.com>
5783 PR tree-optimization/83896
5784 * tree-ssa-strlen.c (get_string_len): Rename...
5785 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
5786 Avoid assuming length is constant.
5787 (handle_char_store): Use HOST_WIDE_INT for string length.
5789 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
5792 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
5793 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
5795 2018-01-26 Richard Biener <rguenther@suse.de>
5797 PR rtl-optimization/84003
5798 * dse.c (record_store): Only record redundant stores when
5799 the earlier store aliases at least all accesses the later one does.
5801 2018-01-26 Jakub Jelinek <jakub@redhat.com>
5803 PR rtl-optimization/83985
5804 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
5805 REG_CFA_RESTORE insns.
5806 (delete_unmarked_insns): Don't ignore separate shrink wrapping
5807 REG_CFA_RESTORE insns here.
5810 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
5811 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
5813 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5815 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
5816 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
5817 (arc_init): Likewise.
5818 (arc_override_options): Likewise.
5819 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
5821 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
5823 * config/arc/arc.h (TARGET_DBNZ): Define.
5824 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
5825 properly set the tune attribute.
5826 (dbnz): Use TARGET_DBNZ guard.
5827 * config/arc/arc.opt (mtune): Add core3 option.
5829 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5831 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
5832 recognize new pic like addresses.
5833 (arc_delegitimize_address): Clean up.
5835 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5837 * config/arc/arc-arches.def: Option mrf16 valid for all
5839 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
5840 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
5841 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
5842 * config/arc/arc-tables.opt: Regenerate.
5843 * config/arc/arc.c (arc_conditional_register_usage): Handle
5844 reduced register file case.
5845 (arc_file_start): Set must have build attributes.
5846 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
5848 * config/arc/arc.opt (mrf16): Add new option.
5849 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
5850 * config/arc/genmultilib.awk: Handle new mrf16 option.
5851 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
5852 * config/arc/t-multilib: Regenerate.
5853 * doc/invoke.texi (ARC Options): Document mrf16 option.
5855 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5857 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
5858 * config/arc/arc.c (arc_handle_secure_attribute): New function.
5859 (arc_attribute_table): Add 'secure_call' attribute.
5860 (arc_print_operand): Print secure call operand.
5861 (arc_function_ok_for_sibcall): Don't optimize tail calls when
5863 (arc_is_secure_call_p): New function. * config/arc/arc.md
5864 (call_i): Add support for sjli instruction.
5865 (call_value_i): Likewise.
5866 * config/arc/constraints.md (Csc): New constraint.
5868 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5869 John Eric Martin <John.Martin@emmicro-us.com>
5871 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
5872 * config/arc/arc.c (_arc_jli_section): New struct.
5873 (arc_jli_section): New type.
5874 (rc_jli_sections): New static variable.
5875 (arc_handle_jli_attribute): New function.
5876 (arc_attribute_table): Add jli_always and jli_fixed attribute.
5877 (arc_file_end): New function.
5878 (TARGET_ASM_FILE_END): Define.
5879 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
5880 (arc_add_jli_section): New function.
5881 (jli_call_scan): Likewise.
5882 (arc_reorg): Call jli_call_scan.
5883 (arc_output_addsi): Remove 'S' from printing asm operand.
5884 (arc_is_jli_call_p): New function.
5885 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
5887 (movhi_insn): Likewise.
5888 (movsi_insn): Likewise.
5889 (movsi_set_cc_insn): Likewise.
5890 (loadqi_update): Likewise.
5891 (load_zeroextendqisi_update): Likewise.
5892 (load_signextendqisi_update): Likewise.
5893 (loadhi_update): Likewise.
5894 (load_zeroextendhisi_update): Likewise.
5895 (load_signextendhisi_update): Likewise.
5896 (loadsi_update): Likewise.
5897 (loadsf_update): Likewise.
5898 (movsicc_insn): Likewise.
5899 (bset_insn): Likewise.
5900 (bxor_insn): Likewise.
5901 (bclr_insn): Likewise.
5902 (bmsk_insn): Likewise.
5903 (bicsi3_insn): Likewise.
5904 (cmpsi_cc_c_insn): Likewise.
5905 (movsi_ne): Likewise.
5906 (movsi_cond_exec): Likewise.
5907 (clrsbsi2): Likewise.
5918 (call_i): Remove 'S' asm letter, add jli instruction.
5919 (call_value_i): Likewise.
5920 * config/arc/arc.op (mjli-always): New option.
5921 * config/arc/constraints.md (Cji): New constraint.
5922 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
5924 (subsf3_fpx): Likewise.
5925 (mulsf3_fpx): Likewise.
5926 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
5928 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
5930 * doc/invoke.texi (ARC): Document mjli-always option.
5932 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
5934 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
5935 avoid addition with 0 and use incw and decw where possible.
5937 2018-01-26 Richard Biener <rguenther@suse.de>
5939 PR tree-optimization/81082
5940 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
5941 association if it requires casting to unsigned.
5942 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
5943 from fold_plusminus_mult_expr to catch important cases late when
5944 range info is available.
5946 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5948 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
5949 * configure.ac (hidden_linkonce): New test.
5950 * configure: Regenerate.
5951 * config.in: Regenerate.
5953 2018-01-26 Julia Koval <julia.koval@intel.com>
5955 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
5956 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
5957 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
5958 _mm_mask_bitshuffle_epi64_mask): Fix type.
5959 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
5960 USI_FTYPE_V4DI_V4DI_USI): Remove.
5961 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
5962 __builtin_ia32_vpshufbitqmb256_mask,
5963 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
5964 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
5965 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
5967 2018-01-26 Alan Modra <amodra@gmail.com>
5970 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
5971 UNSPEC_VBPERMQ. Sort other unspecs.
5973 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
5975 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
5977 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
5980 * predict.c (drop_profile): Do not push/pop cfun; update also
5982 (handle_missing_profiles): Fix logic looking for zero profiles.
5984 2018-01-25 Jakub Jelinek <jakub@redhat.com>
5987 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
5988 on functions with #pragma omp declare simd or functions with simd
5990 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
5991 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5992 Remove trailing \n from warning_at calls.
5994 2018-01-25 Tom de Vries <tom@codesourcery.com>
5997 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5998 for neutered workers.
6000 2018-01-24 Joseph Myers <joseph@codesourcery.com>
6003 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
6004 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
6006 2018-01-24 Jeff Law <law@redhat.com>
6009 * i386.c (get_probe_interval): Move to earlier point.
6010 (ix86_compute_frame_layout): If -fstack-clash-protection and
6011 the frame is larger than the probe interval, then use pushes
6012 to save registers rather than reg->mem moves.
6013 (ix86_expand_prologue): Remove conditional for int_registers_saved
6016 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
6019 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
6020 min/max for never referenced object.
6022 2018-01-24 Jakub Jelinek <jakub@redhat.com>
6025 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
6027 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
6028 attributes from DECL_ATTRIBUTES (decl) without affecting
6029 DECL_ATTRIBUTES (current_function_decl).
6030 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
6031 functions with non-NULL DECL_ABSTRACT_ORIGIN.
6033 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
6035 PR tree-optimization/83979
6036 * fold-const.c (fold_comparison): Use constant_boolean_node
6037 instead of boolean_{true,false}_node.
6039 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
6041 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
6044 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6046 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6047 Simplify the clause that sets the length attribute.
6048 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6049 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
6050 clause that sets the length attribute.
6051 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6053 2018-01-24 Tom de Vries <tom@codesourcery.com>
6056 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
6057 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
6058 Add strict parameter.
6059 (prevent_branch_around_nothing): Insert dummy insn between branch to
6060 label and label with no ptx insn inbetween.
6061 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
6063 2018-01-24 Tom de Vries <tom@codesourcery.com>
6066 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6067 for neutered threads in warp.
6068 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
6070 2018-01-24 Richard Biener <rguenther@suse.de>
6072 PR tree-optimization/83176
6073 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
6076 2018-01-24 Richard Biener <rguenther@suse.de>
6078 PR tree-optimization/82819
6079 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
6080 code generating pluses that are no-ops in the target precision.
6082 2018-01-24 Richard Biener <rguenther@suse.de>
6085 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
6087 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6089 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
6090 to merge probabilities.
6091 * predict.c (probably_never_executed): Also mark as cold functions
6092 with global 0 profile and guessed local profile.
6093 * profile-count.c (profile_probability::combine_with_count): New
6095 * profile-count.h (profile_probability::operator*,
6096 profile_probability::operator*=, profile_probability::operator/,
6097 profile_probability::operator/=): Reduce precision to adjusted
6098 and set value to guessed on contradictory divisions.
6099 (profile_probability::combine_with_freq): Remove.
6100 (profile_probability::combine_wiht_count): Declare.
6101 (profile_count::force_nonzero):: Set to adjusted.
6102 (profile_count::probability_in):: Set quality to adjusted.
6103 * tree-ssa-tail-merge.c (replace_block_by): Use
6106 2018-01-23 Andrew Waterman <andrew@sifive.com>
6107 Jim Wilson <jimw@sifive.com>
6109 * config/riscv/riscv.c (riscv_stack_boundary): New.
6110 (riscv_option_override): Set riscv_stack_boundary. Handle
6111 riscv_preferred_stack_boundary_arg.
6112 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
6113 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
6114 (STACK_BOUNDARY): Set to riscv_stack_boundary.
6115 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
6116 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
6117 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
6119 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
6122 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
6123 of struct ix86_frame.
6124 (ix86_expand_epilogue): Likewise. Add a local variable for
6125 the reg_save_offset field in struct ix86_frame.
6127 2018-01-23 Bin Cheng <bin.cheng@arm.com>
6129 PR tree-optimization/82604
6130 * tree-loop-distribution.c (enum partition_kind): New enum item
6131 PKIND_PARTIAL_MEMSET.
6132 (partition_builtin_p): Support above new enum item.
6133 (generate_code_for_partition): Ditto.
6134 (compute_access_range): Differentiate cases that equality can be
6135 proven at all loops, the innermost loops or no loops.
6136 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
6137 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
6138 (finalize_partitions, distribute_loop): Don't fuse partition of
6139 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
6140 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
6143 2018-01-23 Martin Liska <mliska@suse.cz>
6145 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
6146 order to ignore the predictor.
6147 (PRED_POLYMORPHIC_CALL): Likewise.
6148 (PRED_RECURSIVE_CALL): Likewise.
6150 2018-01-23 Martin Liska <mliska@suse.cz>
6152 * tree-profile.c (tree_profiling): Print function header to
6153 aware reader which function we are working on.
6154 * value-prof.c (gimple_find_values_to_profile): Do not print
6155 not interesting value histograms.
6157 2018-01-23 Martin Liska <mliska@suse.cz>
6159 * profile-count.h (enum profile_quality): Add
6160 profile_uninitialized as the first value. Do not number values
6161 as they are zero based.
6162 (profile_count::verify): Update sanity check.
6163 (profile_probability::verify): Likewise.
6165 2018-01-23 Nathan Sidwell <nathan@acm.org>
6167 * doc/invoke.texi (ffor-scope): Deprecate.
6169 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6171 PR tree-optimization/83510
6172 * domwalk.c (set_all_edges_as_executable): New function.
6173 (dom_walker::dom_walker): Convert bool param
6174 "skip_unreachable_blocks" to enum reachability. Move setup of
6175 edge flags to set_all_edges_as_executable and only do it when
6176 reachability is REACHABLE_BLOCKS.
6177 * domwalk.h (enum dom_walker::reachability): New enum.
6178 (dom_walker::dom_walker): Convert bool param
6179 "skip_unreachable_blocks" to enum reachability.
6180 (set_all_edges_as_executable): New decl.
6181 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
6182 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
6184 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
6185 but converting true to REACHABLE_BLOCKS.
6186 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
6188 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
6189 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
6190 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
6192 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
6193 if check_all_array_refs will be called.
6195 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6197 * tree.c (selftest::test_location_wrappers): Add more test
6200 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6202 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
6203 (selftest::test_bit_in_range): Likewise.
6205 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6208 * doc/sourcebuild.texi (vect_float): Say that the selector
6209 only describes the situation when -funsafe-math-optimizations is on.
6210 (vect_float_strict): Document.
6212 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6214 PR tree-optimization/83965
6215 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
6216 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
6217 instead of checking only for a reduction.
6218 (vect_recog_widen_sum_pattern): Likewise.
6220 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6222 * predict.c (probably_never_executed): Only use precise profile info.
6223 (compute_function_frequency): Skip after inlining hack since we now
6224 have quality checking.
6226 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6228 * profile-count.h (profile_probability::very_unlikely,
6229 profile_probability::unlikely, profile_probability::even): Set
6230 precision to guessed.
6232 2018-01-23 Richard Biener <rguenther@suse.de>
6234 PR tree-optimization/83963
6235 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
6236 Properly terminate dominator walk when crossing the exit edge not
6237 when visiting its source block.
6239 2018-01-23 Jakub Jelinek <jakub@redhat.com>
6242 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
6243 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
6245 2018-01-22 Jakub Jelinek <jakub@redhat.com>
6247 PR tree-optimization/83957
6248 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
6249 semicolon after for body surrounded by braces.
6251 PR tree-optimization/83081
6252 * profile-count.h (profile_probability::split): New method.
6253 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
6254 Use profile_probability::split.
6255 (do_compare_rtx_and_jump): Fix adjustment of probabilities
6256 when splitting a single conditional jump into 2.
6258 2018-01-22 David Malcolm <dmalcolm@redhat.com>
6260 PR tree-optimization/69452
6261 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
6264 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6266 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
6267 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
6268 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
6270 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6272 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
6273 * config/rl78/rl78.md (movdi): New define_expand.
6274 * config/rl78/rl78.c (rl78_split_movdi): New function.
6276 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
6279 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
6281 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
6282 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
6283 128-bit to produce an UNSPEC move to get the double word with the
6284 signbit and then a shift directly to do signbit.
6285 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
6286 implementation with a new version that just does either a direct
6287 move or a regular move. Move memory interface to separate insns.
6288 Move insns so they are next to the expander.
6289 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
6290 with signbit move. Split big and little endian case.
6291 (signbit<mode>2_dm_mem_le): Likewise.
6292 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
6293 (signbit<mode>2_dm2): Likewise.
6295 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6297 * config/rl78/rl78.md (anddi3): New define_expand.
6299 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6301 * config/rl78/rl78.md (umindi3): New define_expand.
6303 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6305 * config/rl78/rl78.md (smindi3): New define_expand.
6307 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6309 * config/rl78/rl78.md (smaxdi3): New define_expand.
6311 2018-01-22 Carl Love <cel@us.ibm.com>
6313 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
6314 LVX_V1TI): Add macro expansion.
6315 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
6316 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
6317 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
6318 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
6319 Change check to determine if the instruction is a byte reversing
6320 entry. Fix typo in comment.
6321 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
6322 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
6323 Add def_builtin calls for new builtins.
6324 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
6325 Add define_insn expansion.
6327 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6329 * config/rl78/rl78.md (umaxdi3): New define_expand.
6331 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6333 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
6334 for non-QImode registers.
6336 2018-01-22 Richard Biener <rguenther@suse.de>
6338 PR tree-optimization/83963
6339 * graphite-scop-detection.c (scop_detection::get_sese): Delay
6340 including the loop exit block.
6341 (scop_detection::merge_sese): Likewise.
6342 (scop_detection::add_scop): Do it here instead.
6344 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6346 * doc/sourcebuild.texi (arm_softfloat): Document.
6348 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
6351 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
6352 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
6353 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
6355 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6356 David Edelsohn <dje.gcc@gmail.com>
6359 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6360 Change "crset eq" to "crset 2".
6361 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6362 (*call_indirect_aix<mode>_nospec): Likewise.
6363 (*call_value_indirect_aix<mode>_nospec): Likewise.
6364 (*call_indirect_elfv2<mode>_nospec): Likewise.
6365 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
6366 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
6367 change assembly output from . to $.
6368 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6369 (indirect_jump<mode>_nospec): Change assembly output from . to $.
6370 (*tablejump<mode>_internal1_nospec): Likewise.
6372 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
6375 * config/sh/sh_optimize_sett_clrt.cc:
6376 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
6378 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
6380 PR tree-optimization/83940
6381 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
6382 offset_dt to vect_constant_def rather than vect_unknown_def_type.
6383 (vect_check_load_store_mask): Add a mask_dt_out parameter and
6384 use it to pass back the definition type.
6385 (vect_check_store_rhs): Likewise rhs_dt_out.
6386 (vect_build_gather_load_calls): Add a mask_dt argument and use
6387 it instead of a call to vect_is_simple_use.
6388 (vectorizable_store): Update calls to vect_check_load_store_mask
6389 and vect_check_store_rhs. Use the dt returned by the latter instead
6390 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
6391 instead of calls to vect_is_simple_use. Pass the scalar rather
6392 than the vector operand to vect_is_simple_use when handling
6393 second and subsequent copies of an rhs value.
6394 (vectorizable_load): Update calls to vect_check_load_store_mask
6395 and vect_build_gather_load_calls. Use the cached mask_dt and
6396 gs_info.offset_dt instead of calls to vect_is_simple_use.
6398 2018-01-20 Jakub Jelinek <jakub@redhat.com>
6401 * tree-emutls.c: Include gimplify.h.
6402 (lower_emutls_2): New function.
6403 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
6404 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
6405 it before further processing.
6408 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
6409 UINTVAL (trueop1) instead of INTVAL (op1).
6411 2018-01-19 Jakub Jelinek <jakub@redhat.com>
6415 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
6416 INCOMING_FRAME_SP_OFFSET if not defined.
6417 (scan_trace): Add ENTRY argument. If true and
6418 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
6419 emit a note to adjust the CFA offset.
6420 (create_cfi_notes): Adjust scan_trace callers.
6421 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
6422 INCOMING_FRAME_SP_OFFSET in the CIE.
6423 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
6424 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
6426 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
6427 * doc/tm.texi: Regenerated.
6429 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6431 PR rtl-optimization/83147
6432 * lra-constraints.c (remove_inheritance_pseudos): Use
6433 lra_substitute_pseudo_within_insn.
6435 2018-01-19 Tom de Vries <tom@codesourcery.com>
6436 Cesar Philippidis <cesar@codesourcery.com>
6439 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
6441 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
6444 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
6445 spaces for function labels.
6447 2018-01-19 Martin Liska <mliska@suse.cz>
6449 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
6450 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
6451 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
6452 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
6453 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
6454 (PRED_CONST_RETURN): Change from 69 to 65.
6455 (PRED_NULL_RETURN): Change from 91 to 71.
6456 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
6457 (PRED_LOOP_GUARD): Change from 66 to 73.
6459 2018-01-19 Martin Liska <mliska@suse.cz>
6461 * predict.c (predict_insn_def): Add new assert.
6462 (struct branch_predictor): Change type to signed integer.
6463 (test_prediction_value_range): Amend test to cover
6465 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
6466 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
6467 (PRED_LOOP_ITERATIONS_MAX): Likewise.
6468 (PRED_LOOP_IV_COMPARE): Likewise.
6469 * predict.h (PROB_UNINITIALIZED): Define new constant.
6471 2018-01-19 Martin Liska <mliska@suse.cz>
6473 * predict.c (dump_prediction): Add new format for
6474 analyze_brprob.py script which is enabled with -details
6476 * profile-count.h (precise_p): New function.
6478 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6480 PR tree-optimization/83922
6481 * tree-vect-loop.c (vect_verify_full_masking): Return false if
6482 there are no statements that need masking.
6483 (vect_active_double_reduction_p): New function.
6484 (vect_analyze_loop_operations): Use it when handling phis that
6485 are not in the loop header.
6487 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6489 PR tree-optimization/83914
6490 * tree-vect-loop.c (vectorizable_induction): Don't convert
6491 init_expr or apply the peeling adjustment for inductions
6492 that are nested within the vectorized loop.
6494 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6496 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
6499 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6503 * function.h (gimplify_parameters): Add gimple_seq * argument.
6504 * function.c: Include gimple.h and options.h.
6505 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
6506 for the added local temporaries if needed.
6507 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
6508 if there are any parameter cleanups, wrap whole body into a
6509 try/finally with the cleanups.
6511 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
6514 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
6515 Use GET_MODE_CLASS for scalar floating point.
6517 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6521 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
6522 Fix call of call_cgraph_insertion_hooks.
6524 2018-01-18 Martin Sebor <msebor@redhat.com>
6526 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
6528 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6531 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
6534 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
6537 * common.opt: (-ffile-prefix-map): New option.
6538 * opts.c (common_handle_option): Defer it.
6539 * opts-global.c (handle_common_deferred_options): Handle it.
6540 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
6541 * file-prefix-map.h: New file.
6542 (remap_debug_filename, add_debug_prefix_map): ...here.
6543 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
6544 * final.c (debug_prefix_map, add_debug_prefix_map
6545 remap_debug_filename): Move to...
6546 * file-prefix-map.c: New file.
6547 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
6548 generalize, get rid of alloca(), use strrchr() instead of strchr().
6549 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
6550 Implement in terms of add_prefix_map().
6551 (remap_macro_filename, remap_debug_filename): Implement in term of
6553 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
6554 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
6555 * dbxout.c: Include file-prefix-map.h.
6556 * varasm.c: Likewise.
6557 * vmsdbgout.c: Likewise.
6558 * xcoffout.c: Likewise.
6559 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
6560 * doc/cppopts.texi (-fmacro-prefix-map): Document.
6561 * doc/invoke.texi (-ffile-prefix-map): Document.
6562 (-fdebug-prefix-map): Update description.
6564 2018-01-18 Martin Liska <mliska@suse.cz>
6566 * config/i386/i386.c (indirect_thunk_name): Document that also
6568 (output_indirect_thunk): Document why both instructions
6569 (pause and lfence) are generated.
6571 2018-01-18 Richard Biener <rguenther@suse.de>
6573 PR tree-optimization/83887
6574 * graphite-scop-detection.c
6575 (scop_detection::get_nearest_dom_with_single_entry): Remove.
6576 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
6577 (scop_detection::merge_sese): Re-implement with a flood-fill
6578 algorithm that properly finds a SESE region if it exists.
6580 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6583 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
6584 pointer_diff optimizations use view_convert instead of convert.
6586 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6588 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6589 Generate different code for -mno-speculate-indirect-jumps.
6590 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6591 (*call_indirect_aix<mode>): Disable for
6592 -mno-speculate-indirect-jumps.
6593 (*call_indirect_aix<mode>_nospec): New define_insn.
6594 (*call_value_indirect_aix<mode>): Disable for
6595 -mno-speculate-indirect-jumps.
6596 (*call_value_indirect_aix<mode>_nospec): New define_insn.
6597 (*sibcall_nonlocal_sysv<mode>): Generate different code for
6598 -mno-speculate-indirect-jumps.
6599 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6601 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6603 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
6604 long double type, set the flags for noting the default long double
6605 type, even if we don't pass or return a long double type.
6607 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
6610 * ipa-inline.c (flatten_function): Do not overwrite final inlining
6613 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
6615 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
6616 support for merge[hl].
6617 (fold_mergehl_helper): New helper function.
6618 (tree-vector-builder.h): New #include for tree_vector_builder usage.
6619 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
6620 (altivec_vmrglw_direct): Add xxmrglw insn.
6622 2018-01-17 Andrew Waterman <andrew@sifive.com>
6624 * config/riscv/riscv.c (riscv_conditional_register_usage): If
6625 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
6627 2018-01-17 David Malcolm <dmalcolm@redhat.com>
6630 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
6631 call the lto_location_cache before reading the
6632 DECL_SOURCE_LOCATION of the types.
6634 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
6635 Richard Sandiford <richard.sandiford@linaro.org>
6637 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
6638 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
6639 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
6640 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
6641 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
6643 * config/aarch64/constraints.md (aarch64_movti_operand):
6645 * config/aarch64/predicates.md (Uti): Add new constraint.
6647 2018-01-17 Carl Love <cel@us.ibm.com>
6649 * config/rs6000/vsx.md (define_expand xl_len_r,
6650 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
6651 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
6653 (define_expand, define_insn): Move the shift left from the
6654 define_insn to the define_expand for lxvl and stxvl instructions.
6655 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
6656 and XL_LEN_R definitions to PURE.
6658 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
6660 * config/i386/i386.c (indirect_thunk_name): Declare regno
6661 as unsigned int. Compare regno with INVALID_REGNUM.
6662 (output_indirect_thunk): Ditto.
6663 (output_indirect_thunk_function): Ditto.
6664 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
6665 in the call to output_indirect_thunk_function.
6667 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
6670 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
6671 rather than the size of inner_type to determine the stack slot size
6672 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
6674 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
6677 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
6680 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
6682 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
6683 endian Linux systems to optionally enable multilibs for selecting
6684 the long double type if the user configured an explicit type.
6685 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
6686 have no long double multilibs if not defined.
6687 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
6688 warn if the user used -mabi={ieee,ibm}longdouble and we built
6689 multilibs for long double.
6690 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
6691 appropriate multilib option.
6692 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
6694 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
6695 for building long double multilibs.
6696 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
6698 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
6700 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
6703 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
6705 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
6708 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
6711 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
6714 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6716 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
6717 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
6719 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
6721 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
6722 different rtl trees depending on TARGET_64BIT.
6723 (rs6000_gen_lvx): Likewise.
6725 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6727 * config/visium/visium.md (nop): Tweak comment.
6728 (hazard_nop): Likewise.
6730 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6732 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
6733 -mspeculate-indirect-jumps.
6734 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
6735 for -mno-speculate-indirect-jumps.
6736 (*call_indirect_elfv2<mode>_nospec): New define_insn.
6737 (*call_value_indirect_elfv2<mode>): Disable for
6738 -mno-speculate-indirect-jumps.
6739 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
6740 (indirect_jump): Emit different RTL for
6741 -mno-speculate-indirect-jumps.
6742 (*indirect_jump<mode>): Disable for
6743 -mno-speculate-indirect-jumps.
6744 (*indirect_jump<mode>_nospec): New define_insn.
6745 (tablejump): Emit different RTL for
6746 -mno-speculate-indirect-jumps.
6747 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
6748 (tablejumpsi_nospec): New define_expand.
6749 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
6750 (tablejumpdi_nospec): New define_expand.
6751 (*tablejump<mode>_internal1): Disable for
6752 -mno-speculate-indirect-jumps.
6753 (*tablejump<mode>_internal1_nospec): New define_insn.
6754 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
6757 2018-01-16 Artyom Skrobov tyomitch@gmail.com
6759 * caller-save.c (insert_save): Drop unnecessary parameter. All
6762 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6763 Richard Biener <rguenth@suse.de>
6766 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
6767 return early, inline manually is_gimple_sizepos. Make sure if we
6768 call gimplify_expr we don't end up with a gimple constant.
6769 * tree.c (variably_modified_type_p): Don't return true for
6770 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
6771 * gimplify.h (is_gimple_sizepos): Remove.
6773 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6775 PR tree-optimization/83857
6776 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
6777 vectorizable_live_operation for pure SLP statements.
6778 (vectorizable_live_operation): Handle PHIs.
6780 2018-01-16 Richard Biener <rguenther@suse.de>
6782 PR tree-optimization/83867
6783 * tree-vect-stmts.c (vect_transform_stmt): Precompute
6784 nested_in_vect_loop_p since the scalar stmt may get invalidated.
6786 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6789 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
6790 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
6791 If off is not INTEGER_CST, issue a may not be aligned warning
6792 rather than isn't aligned. Use isn%'t rather than isn't.
6793 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
6795 <case MULT_EXPR>: Improve the case when bottom and one of the
6796 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
6797 operand, in that case check if the other operand is multiple of
6798 bottom divided by the INTEGER_CST operand.
6800 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6803 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
6804 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
6805 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
6806 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
6807 * config/pa/pa.c (pa_function_arg_advance): Likewise.
6808 (pa_function_arg, pa_arg_partial_bytes): Likewise.
6809 (pa_function_arg_size): New function.
6811 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6813 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
6814 in a separate statement.
6816 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6818 PR tree-optimization/83847
6819 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
6820 group gathers and scatters.
6822 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6824 PR rtl-optimization/86620
6825 * params.def (max-sched-ready-insns): Bump minimum value to 1.
6827 PR rtl-optimization/83213
6828 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
6829 to last if both are JUMP_INSNs.
6831 PR tree-optimization/83843
6832 * gimple-ssa-store-merging.c
6833 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
6834 store_immediate_info for bswap/nop orig_stores.
6836 2018-01-15 Andrew Waterman <andrew@sifive.com>
6838 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
6840 <UDIV>: Increase cost if !TARGET_DIV.
6842 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
6844 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
6845 (define_attr "cr_logical_3op"): New.
6846 (cceq_ior_compare): Adjust.
6847 (cceq_ior_compare_complement): Adjust.
6848 (*cceq_rev_compare): Adjust.
6849 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
6850 (is_cracked_insn): Adjust.
6851 (insn_must_be_first_in_group): Adjust.
6852 * config/rs6000/40x.md: Adjust.
6853 * config/rs6000/440.md: Adjust.
6854 * config/rs6000/476.md: Adjust.
6855 * config/rs6000/601.md: Adjust.
6856 * config/rs6000/603.md: Adjust.
6857 * config/rs6000/6xx.md: Adjust.
6858 * config/rs6000/7450.md: Adjust.
6859 * config/rs6000/7xx.md: Adjust.
6860 * config/rs6000/8540.md: Adjust.
6861 * config/rs6000/cell.md: Adjust.
6862 * config/rs6000/e300c2c3.md: Adjust.
6863 * config/rs6000/e500mc.md: Adjust.
6864 * config/rs6000/e500mc64.md: Adjust.
6865 * config/rs6000/e5500.md: Adjust.
6866 * config/rs6000/e6500.md: Adjust.
6867 * config/rs6000/mpc.md: Adjust.
6868 * config/rs6000/power4.md: Adjust.
6869 * config/rs6000/power5.md: Adjust.
6870 * config/rs6000/power6.md: Adjust.
6871 * config/rs6000/power7.md: Adjust.
6872 * config/rs6000/power8.md: Adjust.
6873 * config/rs6000/power9.md: Adjust.
6874 * config/rs6000/rs64.md: Adjust.
6875 * config/rs6000/titan.md: Adjust.
6877 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6879 * config/i386/predicates.md (indirect_branch_operand): Rewrite
6880 ix86_indirect_branch_register logic.
6882 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6884 * config/i386/constraints.md (Bs): Update
6885 ix86_indirect_branch_register check. Don't check
6886 ix86_indirect_branch_register with GOT_memory_operand.
6888 * config/i386/predicates.md (GOT_memory_operand): Don't check
6889 ix86_indirect_branch_register here.
6890 (GOT32_symbol_operand): Likewise.
6892 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6894 * config/i386/predicates.md (constant_call_address_operand):
6895 Rewrite ix86_indirect_branch_register logic.
6896 (sibcall_insn_operand): Likewise.
6898 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6900 * config/i386/constraints.md (Bs): Replace
6901 ix86_indirect_branch_thunk_register with
6902 ix86_indirect_branch_register.
6904 * config/i386/i386.md (indirect_jump): Likewise.
6905 (tablejump): Likewise.
6906 (*sibcall_memory): Likewise.
6907 (*sibcall_value_memory): Likewise.
6908 Peepholes of indirect call and jump via memory: Likewise.
6909 * config/i386/i386.opt: Likewise.
6910 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6911 (GOT_memory_operand): Likewise.
6912 (call_insn_operand): Likewise.
6913 (sibcall_insn_operand): Likewise.
6914 (GOT32_symbol_operand): Likewise.
6916 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6919 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
6920 type rather than type addr's type points to.
6921 (expand_omp_atomic_mutex): Likewise.
6922 (expand_omp_atomic): Likewise.
6924 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6927 * config/i386/i386.c (output_indirect_thunk_function): Use
6928 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
6929 for __x86_return_thunk.
6931 2018-01-15 Richard Biener <rguenther@suse.de>
6934 * expmed.c (extract_bit_field_1): Fix typo.
6936 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6939 * config/arm/iterators.md (VF): New mode iterator.
6940 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
6941 Remove integer-related logic from pattern.
6942 (neon_vabd<mode>_3): Likewise.
6944 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6947 * common.opt (fstrict-overflow): No longer an alias.
6948 (fwrapv-pointer): New option.
6949 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
6950 also for pointer types based on flag_wrapv_pointer.
6951 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
6952 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
6953 opts->x_flag_wrapv got set.
6954 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
6955 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
6956 POINTER_TYPE_OVERFLOW_UNDEFINED.
6957 * match.pd: Likewise in address comparison pattern.
6958 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
6960 2018-01-15 Richard Biener <rguenther@suse.de>
6963 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
6964 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
6965 Reset type names to their identifier if their TYPE_DECL doesn't
6966 have linkage (and thus is used for ODR and devirt).
6967 (save_debug_info_for_decl): Remove.
6968 (save_debug_info_for_type): Likewise.
6969 (add_tree_to_fld_list): Adjust.
6970 * tree-pretty-print.c (dump_generic_node): Make dumping of
6971 type names more robust.
6973 2018-01-15 Richard Biener <rguenther@suse.de>
6975 * BASE-VER: Bump to 8.0.1.
6977 2018-01-14 Martin Sebor <msebor@redhat.com>
6980 * builtins.c (check_access): Avoid warning when the no-warning bit
6983 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
6985 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
6986 * ira-color (allocno_hard_regs_compare): Likewise.
6988 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
6991 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
6992 Use .pushsection/.popsection.
6994 2018-01-14 Martin Sebor <msebor@redhat.com>
6997 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
6999 2018-01-14 Jakub Jelinek <jakub@redhat.com>
7001 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
7002 entry from extra_headers.
7003 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
7004 extra_headers, make the list bitwise identical to the i?86-*-* one.
7006 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7008 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
7009 -mcmodel=large with -mindirect-branch=thunk,
7010 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
7011 -mfunction-return=thunk-extern.
7012 * doc/invoke.texi: Document -mcmodel=large is incompatible with
7013 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
7014 -mfunction-return=thunk and -mfunction-return=thunk-extern.
7016 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7018 * config/i386/i386.c (print_reg): Print the name of the full
7019 integer register without '%'.
7020 (ix86_print_operand): Handle 'V'.
7021 * doc/extend.texi: Document 'V' modifier.
7023 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7025 * config/i386/constraints.md (Bs): Disallow memory operand for
7026 -mindirect-branch-register.
7028 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7029 (GOT_memory_operand): Likewise.
7030 (call_insn_operand): Likewise.
7031 (sibcall_insn_operand): Likewise.
7032 (GOT32_symbol_operand): Likewise.
7033 * config/i386/i386.md (indirect_jump): Call convert_memory_address
7034 for -mindirect-branch-register.
7035 (tablejump): Likewise.
7036 (*sibcall_memory): Likewise.
7037 (*sibcall_value_memory): Likewise.
7038 Disallow peepholes of indirect call and jump via memory for
7039 -mindirect-branch-register.
7040 (*call_pop): Replace m with Bw.
7041 (*call_value_pop): Likewise.
7042 (*sibcall_pop_memory): Replace m with Bs.
7043 * config/i386/i386.opt (mindirect-branch-register): New option.
7044 * doc/invoke.texi: Document -mindirect-branch-register option.
7046 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7048 * config/i386/i386-protos.h (ix86_output_function_return): New.
7049 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
7050 set function_return_type.
7051 (indirect_thunk_name): Add ret_p to indicate thunk for function
7053 (output_indirect_thunk_function): Pass false to
7054 indirect_thunk_name.
7055 (ix86_output_indirect_branch_via_reg): Likewise.
7056 (ix86_output_indirect_branch_via_push): Likewise.
7057 (output_indirect_thunk_function): Create alias for function
7058 return thunk if regno < 0.
7059 (ix86_output_function_return): New function.
7060 (ix86_handle_fndecl_attribute): Handle function_return.
7061 (ix86_attribute_table): Add function_return.
7062 * config/i386/i386.h (machine_function): Add
7063 function_return_type.
7064 * config/i386/i386.md (simple_return_internal): Use
7065 ix86_output_function_return.
7066 (simple_return_internal_long): Likewise.
7067 * config/i386/i386.opt (mfunction-return=): New option.
7068 (indirect_branch): Mention -mfunction-return=.
7069 * doc/extend.texi: Document function_return function attribute.
7070 * doc/invoke.texi: Document -mfunction-return= option.
7072 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7074 * config/i386/i386-opts.h (indirect_branch): New.
7075 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
7076 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
7077 with local indirect jump when converting indirect call and jump.
7078 (ix86_set_indirect_branch_type): New.
7079 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
7080 (indirectlabelno): New.
7081 (indirect_thunk_needed): Likewise.
7082 (indirect_thunk_bnd_needed): Likewise.
7083 (indirect_thunks_used): Likewise.
7084 (indirect_thunks_bnd_used): Likewise.
7085 (INDIRECT_LABEL): Likewise.
7086 (indirect_thunk_name): Likewise.
7087 (output_indirect_thunk): Likewise.
7088 (output_indirect_thunk_function): Likewise.
7089 (ix86_output_indirect_branch_via_reg): Likewise.
7090 (ix86_output_indirect_branch_via_push): Likewise.
7091 (ix86_output_indirect_branch): Likewise.
7092 (ix86_output_indirect_jmp): Likewise.
7093 (ix86_code_end): Call output_indirect_thunk_function if needed.
7094 (ix86_output_call_insn): Call ix86_output_indirect_branch if
7096 (ix86_handle_fndecl_attribute): Handle indirect_branch.
7097 (ix86_attribute_table): Add indirect_branch.
7098 * config/i386/i386.h (machine_function): Add indirect_branch_type
7099 and has_local_indirect_jump.
7100 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
7102 (tablejump): Likewise.
7103 (*indirect_jump): Use ix86_output_indirect_jmp.
7104 (*tablejump_1): Likewise.
7105 (simple_return_indirect_internal): Likewise.
7106 * config/i386/i386.opt (mindirect-branch=): New option.
7107 (indirect_branch): New.
7110 (thunk-inline): Likewise.
7111 (thunk-extern): Likewise.
7112 * doc/extend.texi: Document indirect_branch function attribute.
7113 * doc/invoke.texi: Document -mindirect-branch= option.
7115 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
7118 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
7120 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
7122 * ipa-inline.c (want_inline_small_function_p): Return false if
7123 inlining has already failed with CIF_FINAL_ERROR.
7124 (update_caller_keys): Call want_inline_small_function_p before
7126 (update_callee_keys): Likewise.
7128 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7130 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
7132 (rs6000_quadword_masked_address_p): Likewise.
7133 (quad_aligned_load_p): Likewise.
7134 (quad_aligned_store_p): Likewise.
7135 (const_load_sequence_p): Add comment to describe the outer-most loop.
7136 (mimic_memory_attributes_and_flags): New function.
7137 (rs6000_gen_stvx): Likewise.
7138 (replace_swapped_aligned_store): Likewise.
7139 (rs6000_gen_lvx): Likewise.
7140 (replace_swapped_aligned_load): Likewise.
7141 (replace_swapped_load_constant): Capitalize argument name in
7142 comment describing this function.
7143 (rs6000_analyze_swaps): Add a third pass to search for vector loads
7144 and stores that access quad-word aligned addresses and replace
7145 with stvx or lvx instructions when appropriate.
7146 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
7147 New function prototype.
7148 (rs6000_quadword_masked_address_p): Likewise.
7149 (rs6000_gen_lvx): Likewise.
7150 (rs6000_gen_stvx): Likewise.
7151 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
7152 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
7153 when memory address is aligned.
7154 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
7155 this split to select lvx instruction when memory address is aligned.
7156 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
7157 instruction when memory address is aligned.
7158 (*vsx_le_perm_load_v16qi): Likewise.
7159 (four unnamed splitters): Modify to select the stvx instruction
7160 when memory is aligned.
7162 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
7164 * predict.c (determine_unlikely_bbs): Handle correctly BBs
7165 which appears in the queue multiple times.
7167 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7168 Alan Hayward <alan.hayward@arm.com>
7169 David Sherwood <david.sherwood@arm.com>
7171 * tree-vectorizer.h (vec_lower_bound): New structure.
7172 (_loop_vec_info): Add check_nonzero and lower_bounds.
7173 (LOOP_VINFO_CHECK_NONZERO): New macro.
7174 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
7175 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
7176 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
7177 fields. Make seg_len the distance travelled, not including the
7179 (dr_direction_indicator): Declare.
7180 (dr_zero_step_indicator): Likewise.
7181 (dr_known_forward_stride_p): Likewise.
7182 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
7184 (runtime_alias_check_p): Allow runtime alias checks with
7186 (operator ==): Compare access_size and align.
7187 (prune_runtime_alias_test_list): Rework for new distinction between
7188 the access_size and seg_len.
7189 (create_intersect_range_checks_index): Likewise. Cope with polynomial
7191 (get_segment_min_max): New function.
7192 (create_intersect_range_checks): Use it.
7193 (dr_step_indicator): New function.
7194 (dr_direction_indicator): Likewise.
7195 (dr_zero_step_indicator): Likewise.
7196 (dr_known_forward_stride_p): Likewise.
7197 * tree-loop-distribution.c (data_ref_segment_size): Return
7198 DR_STEP * (niters - 1).
7199 (compute_alias_check_pairs): Update call to the dr_with_seg_len
7201 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
7202 (vect_preserves_scalar_order_p): New function, split out from...
7203 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
7204 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
7205 (vect_vfa_access_size): New function.
7206 (vect_vfa_align): Likewise.
7207 (vect_compile_time_alias): Take access_size_a and access_b arguments.
7208 (dump_lower_bound): New function.
7209 (vect_check_lower_bound): Likewise.
7210 (vect_small_gap_p): Likewise.
7211 (vectorizable_with_step_bound_p): Likewise.
7212 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
7213 depencies if the vectorization factor is 1. Convert the checks
7214 for nonzero steps into checks on the bounds of DR_STEP. Try using
7215 a bunds check for variable steps if the minimum required step is
7216 relatively small. Update calls to the dr_with_seg_len
7217 constructor and to vect_compile_time_alias.
7218 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
7220 (vect_loop_versioning): Call it.
7221 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
7223 (vect_estimate_min_profitable_iters): Account for any bounds checks.
7225 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7226 Alan Hayward <alan.hayward@arm.com>
7227 David Sherwood <david.sherwood@arm.com>
7229 * doc/sourcebuild.texi (vect_scatter_store): Document.
7230 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
7232 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
7234 * genopinit.c (main): Add supports_vec_scatter_store and
7235 supports_vec_scatter_store_cached to target_optabs.
7236 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
7237 IFN_MASK_SCATTER_STORE.
7238 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
7240 * internal-fn.h (internal_store_fn_p): Declare.
7241 (internal_fn_stored_value_index): Likewise.
7242 * internal-fn.c (scatter_store_direct): New macro.
7243 (expand_scatter_store_optab_fn): New function.
7244 (direct_scatter_store_optab_supported_p): New macro.
7245 (internal_store_fn_p): New function.
7246 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
7247 IFN_MASK_SCATTER_STORE.
7248 (internal_fn_mask_index): Likewise.
7249 (internal_fn_stored_value_index): New function.
7250 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
7252 * optabs-query.h (supports_vec_scatter_store_p): Declare.
7253 * optabs-query.c (supports_vec_scatter_store_p): New function.
7254 * tree-vectorizer.h (vect_get_store_rhs): Declare.
7255 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
7256 true for scatter stores.
7257 (vect_gather_scatter_fn_p): Handle scatter stores too.
7258 (vect_check_gather_scatter): Consider using scatter stores if
7259 supports_vec_scatter_store_p.
7260 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
7262 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7263 internal_fn_stored_value_index.
7264 (check_load_store_masking): Handle scatter stores too.
7265 (vect_get_store_rhs): Make public.
7266 (vectorizable_call): Use internal_store_fn_p.
7267 (vectorizable_store): Handle scatter store internal functions.
7268 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
7269 when deciding whether the end of the group has been reached.
7270 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
7271 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
7272 (mask_scatter_store<mode>): New insns.
7274 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7275 Alan Hayward <alan.hayward@arm.com>
7276 David Sherwood <david.sherwood@arm.com>
7278 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
7279 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
7280 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
7282 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
7283 Use vect_truncate_gather_scatter_offset if we can't treat the
7284 operation as a normal gather load or scatter store.
7285 (get_group_load_store_type): Take the gather_scatter_info
7286 as argument. Try using a gather load or scatter store for
7287 single-element groups.
7288 (get_load_store_type): Update calls to get_group_load_store_type
7289 and vect_use_strided_gather_scatters_p.
7291 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7292 Alan Hayward <alan.hayward@arm.com>
7293 David Sherwood <david.sherwood@arm.com>
7295 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
7296 optional tree argument.
7297 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
7299 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
7300 but continue to use the current value as a fallback.
7301 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
7302 to compare the updates.
7303 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
7304 (get_load_store_type): Use it when handling a strided access.
7305 (vect_get_strided_load_store_ops): New function.
7306 (vect_get_data_ptr_increment): Likewise.
7307 (vectorizable_load): Handle strided gather loads. Always pass
7308 a step to vect_create_data_ref_ptr and bump_vector_ptr.
7310 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7311 Alan Hayward <alan.hayward@arm.com>
7312 David Sherwood <david.sherwood@arm.com>
7314 * doc/md.texi (gather_load@var{m}): Document.
7315 (mask_gather_load@var{m}): Likewise.
7316 * genopinit.c (main): Add supports_vec_gather_load and
7317 supports_vec_gather_load_cached to target_optabs.
7318 * optabs-tree.c (init_tree_optimization_optabs): Use
7319 ggc_cleared_alloc to allocate target_optabs.
7320 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
7321 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
7323 * internal-fn.h (internal_load_fn_p): Declare.
7324 (internal_gather_scatter_fn_p): Likewise.
7325 (internal_fn_mask_index): Likewise.
7326 (internal_gather_scatter_fn_supported_p): Likewise.
7327 * internal-fn.c (gather_load_direct): New macro.
7328 (expand_gather_load_optab_fn): New function.
7329 (direct_gather_load_optab_supported_p): New macro.
7330 (direct_internal_fn_optab): New function.
7331 (internal_load_fn_p): Likewise.
7332 (internal_gather_scatter_fn_p): Likewise.
7333 (internal_fn_mask_index): Likewise.
7334 (internal_gather_scatter_fn_supported_p): Likewise.
7335 * optabs-query.c (supports_at_least_one_mode_p): New function.
7336 (supports_vec_gather_load_p): Likewise.
7337 * optabs-query.h (supports_vec_gather_load_p): Declare.
7338 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
7339 and memory_type field.
7340 (NUM_PATTERNS): Bump to 15.
7341 * tree-vect-data-refs.c: Include internal-fn.h.
7342 (vect_gather_scatter_fn_p): New function.
7343 (vect_describe_gather_scatter_call): Likewise.
7344 (vect_check_gather_scatter): Try using internal functions for
7345 gather loads. Recognize existing calls to a gather load function.
7346 (vect_analyze_data_refs): Consider using gather loads if
7347 supports_vec_gather_load_p.
7348 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
7349 (vect_get_gather_scatter_offset_type): Likewise.
7350 (vect_convert_mask_for_vectype): Likewise.
7351 (vect_add_conversion_to_patterm): Likewise.
7352 (vect_try_gather_scatter_pattern): Likewise.
7353 (vect_recog_gather_scatter_pattern): New pattern recognizer.
7354 (vect_vect_recog_func_ptrs): Add it.
7355 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7356 internal_fn_mask_index and internal_gather_scatter_fn_p.
7357 (check_load_store_masking): Take the gather_scatter_info as an
7358 argument and handle gather loads.
7359 (vect_get_gather_scatter_ops): New function.
7360 (vectorizable_call): Check internal_load_fn_p.
7361 (vectorizable_load): Likewise. Handle gather load internal
7363 (vectorizable_store): Update call to check_load_store_masking.
7364 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
7365 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
7366 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
7367 (aarch64_gather_scale_operand_d): New predicates.
7368 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
7369 (mask_gather_load<mode>): New insns.
7371 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7372 Alan Hayward <alan.hayward@arm.com>
7373 David Sherwood <david.sherwood@arm.com>
7375 * optabs.def (fold_left_plus_optab): New optab.
7376 * doc/md.texi (fold_left_plus_@var{m}): Document.
7377 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
7378 * internal-fn.c (fold_left_direct): Define.
7379 (expand_fold_left_optab_fn): Likewise.
7380 (direct_fold_left_optab_supported_p): Likewise.
7381 * fold-const-call.c (fold_const_fold_left): New function.
7382 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
7383 * tree-parloops.c (valid_reduction_p): New function.
7384 (gather_scalar_reductions): Use it.
7385 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
7386 (vect_finish_replace_stmt): Declare.
7387 * tree-vect-loop.c (fold_left_reduction_fn): New function.
7388 (needs_fold_left_reduction_p): New function, split out from...
7389 (vect_is_simple_reduction): ...here. Accept reductions that
7390 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
7391 (vect_force_simple_reduction): Also store the reduction type in
7392 the assignment's STMT_VINFO_REDUC_TYPE.
7393 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
7394 (merge_with_identity): New function.
7395 (vect_expand_fold_left): Likewise.
7396 (vectorize_fold_left_reduction): Likewise.
7397 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
7398 scalar phi in place for it. Check for target support and reject
7399 cases that would reassociate the operation. Defer the transform
7400 phase to vectorize_fold_left_reduction.
7401 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
7402 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
7403 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
7405 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7407 * tree-if-conv.c (predicate_mem_writes): Remove redundant
7408 call to ifc_temp_var.
7410 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7411 Alan Hayward <alan.hayward@arm.com>
7412 David Sherwood <david.sherwood@arm.com>
7414 * target.def (legitimize_address_displacement): Take the original
7415 offset as a poly_int.
7416 * targhooks.h (default_legitimize_address_displacement): Update
7418 * targhooks.c (default_legitimize_address_displacement): Likewise.
7419 * doc/tm.texi: Regenerate.
7420 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
7421 as an argument, moving assert of ad->disp == ad->disp_term to...
7422 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
7423 Try calling targetm.legitimize_address_displacement before expanding
7424 the address rather than afterwards, and adjust for the new interface.
7425 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
7426 Match the new hook interface. Handle SVE addresses.
7427 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
7430 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7432 * Makefile.in (OBJS): Add early-remat.o.
7433 * target.def (select_early_remat_modes): New hook.
7434 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
7435 * doc/tm.texi: Regenerate.
7436 * targhooks.h (default_select_early_remat_modes): Declare.
7437 * targhooks.c (default_select_early_remat_modes): New function.
7438 * timevar.def (TV_EARLY_REMAT): New timevar.
7439 * passes.def (pass_early_remat): New pass.
7440 * tree-pass.h (make_pass_early_remat): Declare.
7441 * early-remat.c: New file.
7442 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
7444 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
7446 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7447 Alan Hayward <alan.hayward@arm.com>
7448 David Sherwood <david.sherwood@arm.com>
7450 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
7451 vfm1 with a bound_epilog parameter.
7452 (vect_do_peeling): Update calls accordingly, and move the prologue
7453 call earlier in the function. Treat the base bound_epilog as 0 for
7454 fully-masked loops and retain vf - 1 for other loops. Add 1 to
7455 this base when peeling for gaps.
7456 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
7457 with fully-masked loops.
7458 (vect_estimate_min_profitable_iters): Handle the single peeled
7459 iteration in that case.
7461 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7462 Alan Hayward <alan.hayward@arm.com>
7463 David Sherwood <david.sherwood@arm.com>
7465 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
7466 single-element interleaving even if the size is not a power of 2.
7467 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
7468 accesses for single-element interleaving if the group size is
7471 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7472 Alan Hayward <alan.hayward@arm.com>
7473 David Sherwood <david.sherwood@arm.com>
7475 * doc/md.texi (fold_extract_last_@var{m}): Document.
7476 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
7477 * optabs.def (fold_extract_last_optab): New optab.
7478 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
7479 * internal-fn.c (fold_extract_direct): New macro.
7480 (expand_fold_extract_optab_fn): Likewise.
7481 (direct_fold_extract_optab_supported_p): Likewise.
7482 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
7483 * tree-vect-loop.c (vect_model_reduction_cost): Handle
7484 EXTRACT_LAST_REDUCTION.
7485 (get_initial_def_for_reduction): Do not create an initial vector
7486 for EXTRACT_LAST_REDUCTION reductions.
7487 (vectorizable_reduction): Leave the scalar phi in place for
7488 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
7489 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
7490 epilogue code for EXTRACT_LAST_REDUCTION and defer the
7491 transform phase to vectorizable_condition.
7492 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
7494 (vect_finish_stmt_generation): ...here.
7495 (vect_finish_replace_stmt): New function.
7496 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
7497 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
7499 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
7501 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7502 Alan Hayward <alan.hayward@arm.com>
7503 David Sherwood <david.sherwood@arm.com>
7505 * doc/md.texi (extract_last_@var{m}): Document.
7506 * optabs.def (extract_last_optab): New optab.
7507 * internal-fn.def (EXTRACT_LAST): New internal function.
7508 * internal-fn.c (cond_unary_direct): New macro.
7509 (expand_cond_unary_optab_fn): Likewise.
7510 (direct_cond_unary_optab_supported_p): Likewise.
7511 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
7512 loops using EXTRACT_LAST.
7513 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
7514 (extract_last_<mode>): ...this optab.
7515 (vec_extract<mode><Vel>): Update accordingly.
7517 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7518 Alan Hayward <alan.hayward@arm.com>
7519 David Sherwood <david.sherwood@arm.com>
7521 * target.def (empty_mask_is_expensive): New hook.
7522 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
7523 * doc/tm.texi: Regenerate.
7524 * targhooks.h (default_empty_mask_is_expensive): Declare.
7525 * targhooks.c (default_empty_mask_is_expensive): New function.
7526 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
7527 if the target says that empty masks are expensive.
7528 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
7530 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
7532 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7533 Alan Hayward <alan.hayward@arm.com>
7534 David Sherwood <david.sherwood@arm.com>
7536 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
7537 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
7538 (vect_use_loop_mask_for_alignment_p): New function.
7539 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
7540 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
7541 niters_skip argument. Make sure that the first niters_skip elements
7542 of the first iteration are inactive.
7543 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
7544 Update call to vect_set_loop_masks_directly.
7545 (get_misalign_in_elems): New function, split out from...
7546 (vect_gen_prolog_loop_niters): ...here.
7547 (vect_update_init_of_dr): Take a code argument that specifies whether
7548 the adjustment should be added or subtracted.
7549 (vect_update_init_of_drs): Likewise.
7550 (vect_prepare_for_masked_peels): New function.
7551 (vect_do_peeling): Skip prologue peeling if we're using a mask
7552 instead. Update call to vect_update_inits_of_drs.
7553 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7555 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
7556 alignment. Do not include the number of peeled iterations in
7557 the minimum threshold in that case.
7558 (vectorizable_induction): Adjust the start value down by
7559 LOOP_VINFO_MASK_SKIP_NITERS iterations.
7560 (vect_transform_loop): Call vect_prepare_for_masked_peels.
7561 Take the number of skipped iterations into account when calculating
7563 * tree-vect-stmts.c (vect_gen_while_not): New function.
7565 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7566 Alan Hayward <alan.hayward@arm.com>
7567 David Sherwood <david.sherwood@arm.com>
7569 * doc/sourcebuild.texi (vect_fully_masked): Document.
7570 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
7572 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
7574 (vect_analyze_loop_2): ...here. Don't check the vectorization
7575 factor against the number of loop iterations if the loop is
7578 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7579 Alan Hayward <alan.hayward@arm.com>
7580 David Sherwood <david.sherwood@arm.com>
7582 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
7583 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
7584 (dump_groups): Update accordingly.
7585 (iv_use::mem_type): New member variable.
7586 (address_p): New function.
7587 (record_use): Add a mem_type argument and initialize the new
7589 (record_group_use): Add a mem_type argument. Use address_p.
7590 Remove obsolete null checks of base_object. Update call to record_use.
7591 (find_interesting_uses_op): Update call to record_group_use.
7592 (find_interesting_uses_cond): Likewise.
7593 (find_interesting_uses_address): Likewise.
7594 (get_mem_type_for_internal_fn): New function.
7595 (find_address_like_use): Likewise.
7596 (find_interesting_uses_stmt): Try find_address_like_use before
7597 calling find_interesting_uses_op.
7598 (addr_offset_valid_p): Use the iv mem_type field as the type
7599 of the addressed memory.
7600 (add_autoinc_candidates): Likewise.
7601 (get_address_cost): Likewise.
7602 (split_small_address_groups_p): Use address_p.
7603 (split_address_groups): Likewise.
7604 (add_iv_candidate_for_use): Likewise.
7605 (autoinc_possible_for_pair): Likewise.
7606 (rewrite_groups): Likewise.
7607 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
7608 (determine_group_iv_cost): Update after split of USE_ADDRESS.
7609 (get_alias_ptr_type_for_ptr_address): New function.
7610 (rewrite_use_address): Rewrite address uses in calls that were
7611 identified by find_address_like_use.
7613 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7614 Alan Hayward <alan.hayward@arm.com>
7615 David Sherwood <david.sherwood@arm.com>
7617 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
7619 * gimple-expr.h (is_gimple_addressable: Likewise.
7620 * gimple-expr.c (is_gimple_address): Likewise.
7621 * internal-fn.c (expand_call_mem_ref): New function.
7622 (expand_mask_load_optab_fn): Use it.
7623 (expand_mask_store_optab_fn): Likewise.
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 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
7630 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
7631 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
7632 (cond_umax@var{mode}): Document.
7633 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
7634 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
7635 (cond_umin_optab, cond_umax_optab): New optabs.
7636 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
7637 (COND_IOR, COND_XOR): New internal functions.
7638 * internal-fn.h (get_conditional_internal_fn): Declare.
7639 * internal-fn.c (cond_binary_direct): New macro.
7640 (expand_cond_binary_optab_fn): Likewise.
7641 (direct_cond_binary_optab_supported_p): Likewise.
7642 (get_conditional_internal_fn): New function.
7643 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
7644 Cope with reduction statements that are vectorized as calls rather
7646 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
7647 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
7648 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
7649 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
7650 (UNSPEC_COND_EOR): New unspecs.
7651 (optab): Add mappings for them.
7652 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
7653 (sve_int_op, sve_fp_op): New int attributes.
7655 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7656 Alan Hayward <alan.hayward@arm.com>
7657 David Sherwood <david.sherwood@arm.com>
7659 * optabs.def (while_ult_optab): New optab.
7660 * doc/md.texi (while_ult@var{m}@var{n}): Document.
7661 * internal-fn.def (WHILE_ULT): New internal function.
7662 * internal-fn.h (direct_internal_fn_supported_p): New override
7663 that takes two types as argument.
7664 * internal-fn.c (while_direct): New macro.
7665 (expand_while_optab_fn): New function.
7666 (convert_optab_supported_p): Likewise.
7667 (direct_while_optab_supported_p): New macro.
7668 * wide-int.h (wi::udiv_ceil): New function.
7669 * tree-vectorizer.h (rgroup_masks): New structure.
7670 (vec_loop_masks): New typedef.
7671 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
7673 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
7674 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
7675 (vect_max_vf): New function.
7676 (slpeel_make_loop_iterate_ntimes): Delete.
7677 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
7678 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
7679 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
7680 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
7681 internal-fn.h, stor-layout.h and optabs-query.h.
7682 (vect_set_loop_mask): New function.
7683 (add_preheader_seq): Likewise.
7684 (add_header_seq): Likewise.
7685 (interleave_supported_p): Likewise.
7686 (vect_maybe_permute_loop_masks): Likewise.
7687 (vect_set_loop_masks_directly): Likewise.
7688 (vect_set_loop_condition_masked): Likewise.
7689 (vect_set_loop_condition_unmasked): New function, split out from
7690 slpeel_make_loop_iterate_ntimes.
7691 (slpeel_make_loop_iterate_ntimes): Rename to..
7692 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
7693 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
7694 (vect_do_peeling): Update call accordingly.
7695 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
7697 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7698 mask_compare_type, can_fully_mask_p and fully_masked_p.
7699 (release_vec_loop_masks): New function.
7700 (_loop_vec_info): Use it to free the loop masks.
7701 (can_produce_all_loop_masks_p): New function.
7702 (vect_get_max_nscalars_per_iter): Likewise.
7703 (vect_verify_full_masking): Likewise.
7704 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
7705 retries, and free the mask rgroups before retrying. Check loop-wide
7706 reasons for disallowing fully-masked loops. Make the final decision
7707 about whether use a fully-masked loop or not.
7708 (vect_estimate_min_profitable_iters): Do not assume that peeling
7709 for the number of iterations will be needed for fully-masked loops.
7710 (vectorizable_reduction): Disable fully-masked loops.
7711 (vectorizable_live_operation): Likewise.
7712 (vect_halve_mask_nunits): New function.
7713 (vect_double_mask_nunits): Likewise.
7714 (vect_record_loop_mask): Likewise.
7715 (vect_get_loop_mask): Likewise.
7716 (vect_transform_loop): Handle the case in which the final loop
7717 iteration might handle a partial vector. Call vect_set_loop_condition
7718 instead of slpeel_make_loop_iterate_ntimes.
7719 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
7720 (check_load_store_masking): New function.
7721 (prepare_load_store_mask): Likewise.
7722 (vectorizable_store): Handle fully-masked loops.
7723 (vectorizable_load): Likewise.
7724 (supportable_widening_operation): Use vect_halve_mask_nunits for
7726 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
7727 (vect_gen_while): New function.
7728 * config/aarch64/aarch64.md (umax<mode>3): New expander.
7729 (aarch64_uqdec<mode>): New insn.
7731 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7732 Alan Hayward <alan.hayward@arm.com>
7733 David Sherwood <david.sherwood@arm.com>
7735 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
7736 (reduc_xor_scal_optab): New optabs.
7737 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
7738 (reduc_xor_scal_@var{m}): Document.
7739 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
7740 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
7742 * fold-const-call.c (fold_const_call): Handle them.
7743 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
7744 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
7745 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
7746 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
7747 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
7748 (UNSPEC_XORV): New unspecs.
7749 (optab): Add entries for them.
7750 (BITWISEV): New int iterator.
7751 (bit_reduc_op): New int attributes.
7753 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7754 Alan Hayward <alan.hayward@arm.com>
7755 David Sherwood <david.sherwood@arm.com>
7757 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
7758 * internal-fn.def (VEC_SHL_INSERT): New internal function.
7759 * optabs.def (vec_shl_insert_optab): New optab.
7760 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
7761 (duplicate_and_interleave): Likewise.
7762 * tree-vect-loop.c: Include internal-fn.h.
7763 (neutral_op_for_slp_reduction): New function, split out from
7764 get_initial_defs_for_reduction.
7765 (get_initial_def_for_reduction): Handle option 2 for variable-length
7766 vectors by loading the neutral value into a vector and then shifting
7767 the initial value into element 0.
7768 (get_initial_defs_for_reduction): Replace the code argument with
7769 the neutral value calculated by neutral_op_for_slp_reduction.
7770 Use gimple_build_vector for constant-length vectors.
7771 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
7772 but the first group_size elements have a neutral value.
7773 Use duplicate_and_interleave otherwise.
7774 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
7775 Update call to get_initial_defs_for_reduction. Handle SLP
7776 reductions for variable-length vectors by creating one vector
7777 result for each scalar result, with the elements associated
7778 with other scalar results stubbed out with the neutral value.
7779 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
7780 Require IFN_VEC_SHL_INSERT for double reductions on
7781 variable-length vectors, or SLP reductions that have
7782 a neutral value. Require can_duplicate_and_interleave_p
7783 support for variable-length unchained SLP reductions if there
7784 is no neutral value, such as for MIN/MAX reductions. Also require
7785 the number of vector elements to be a multiple of the number of
7786 SLP statements when doing variable-length unchained SLP reductions.
7787 Update call to vect_create_epilog_for_reduction.
7788 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
7789 and remove initial values.
7790 (duplicate_and_interleave): Make public.
7791 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
7792 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
7794 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7795 Alan Hayward <alan.hayward@arm.com>
7796 David Sherwood <david.sherwood@arm.com>
7798 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
7799 (can_duplicate_and_interleave_p): New function.
7800 (vect_get_and_check_slp_defs): Take the vector of statements
7801 rather than just the current one. Remove excess parentheses.
7802 Restriction rejectinon of vect_constant_def and vect_external_def
7803 for variable-length vectors to boolean types, or types for which
7804 can_duplicate_and_interleave_p is false.
7805 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
7806 (duplicate_and_interleave): New function.
7807 (vect_get_constant_vectors): Use gimple_build_vector for
7808 constant-length vectors and suitable variable-length constant
7809 vectors. Use duplicate_and_interleave for other variable-length
7810 vectors. Don't defer the update when inserting new statements.
7812 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7813 Alan Hayward <alan.hayward@arm.com>
7814 David Sherwood <david.sherwood@arm.com>
7816 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
7817 min_profitable_iters doesn't go negative.
7819 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7820 Alan Hayward <alan.hayward@arm.com>
7821 David Sherwood <david.sherwood@arm.com>
7823 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
7824 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
7825 * optabs.def (vec_mask_load_lanes_optab): New optab.
7826 (vec_mask_store_lanes_optab): Likewise.
7827 * internal-fn.def (MASK_LOAD_LANES): New internal function.
7828 (MASK_STORE_LANES): Likewise.
7829 * internal-fn.c (mask_load_lanes_direct): New macro.
7830 (mask_store_lanes_direct): Likewise.
7831 (expand_mask_load_optab_fn): Handle masked operations.
7832 (expand_mask_load_lanes_optab_fn): New macro.
7833 (expand_mask_store_optab_fn): Handle masked operations.
7834 (expand_mask_store_lanes_optab_fn): New macro.
7835 (direct_mask_load_lanes_optab_supported_p): Likewise.
7836 (direct_mask_store_lanes_optab_supported_p): Likewise.
7837 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
7839 (vect_load_lanes_supported): Likewise.
7840 * tree-vect-data-refs.c (strip_conversion): New function.
7841 (can_group_stmts_p): Likewise.
7842 (vect_analyze_data_ref_accesses): Use it instead of checking
7843 for a pair of assignments.
7844 (vect_store_lanes_supported): Take a masked_p parameter.
7845 (vect_load_lanes_supported): Likewise.
7846 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
7847 vect_store_lanes_supported and vect_load_lanes_supported.
7848 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7849 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
7850 parameter. Don't allow gaps for masked accesses.
7851 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
7852 and vect_load_lanes_supported.
7853 (get_load_store_type): Take a masked_p parameter and update
7854 call to get_group_load_store_type.
7855 (vectorizable_store): Update call to get_load_store_type.
7856 Handle IFN_MASK_STORE_LANES.
7857 (vectorizable_load): Update call to get_load_store_type.
7858 Handle IFN_MASK_LOAD_LANES.
7860 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7861 Alan Hayward <alan.hayward@arm.com>
7862 David Sherwood <david.sherwood@arm.com>
7864 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
7866 * config/aarch64/aarch64-protos.h
7867 (aarch64_sve_struct_memory_operand_p): Declare.
7868 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
7869 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
7870 (VPRED, vpred): Handle SVE structure modes.
7871 * config/aarch64/constraints.md (Utx): New constraint.
7872 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
7873 (aarch64_sve_struct_nonimmediate_operand): New predicates.
7874 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
7875 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
7876 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
7877 structure modes. Split into pieces after RA.
7878 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
7879 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
7881 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
7882 SVE structure modes.
7883 (aarch64_classify_address): Likewise.
7884 (sizetochar): Move earlier in file.
7885 (aarch64_print_operand): Handle SVE register lists.
7886 (aarch64_array_mode): New function.
7887 (aarch64_sve_struct_memory_operand_p): Likewise.
7888 (TARGET_ARRAY_MODE): Redefine.
7890 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7891 Alan Hayward <alan.hayward@arm.com>
7892 David Sherwood <david.sherwood@arm.com>
7894 * target.def (array_mode): New target hook.
7895 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
7896 * doc/tm.texi: Regenerate.
7897 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
7898 * hooks.c (hook_optmode_mode_uhwi_none): New function.
7899 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
7901 * stor-layout.c (mode_for_array): Likewise. Support polynomial
7904 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7905 Alan Hayward <alan.hayward@arm.com>
7906 David Sherwood <david.sherwood@arm.com>
7908 * fold-const.c (fold_binary_loc): Check the argument types
7909 rather than the result type when testing for a vector operation.
7911 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7913 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
7914 * doc/tm.texi: Regenerate.
7916 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7917 Alan Hayward <alan.hayward@arm.com>
7918 David Sherwood <david.sherwood@arm.com>
7920 * doc/invoke.texi (-msve-vector-bits=): Document new option.
7921 (sve): Document new AArch64 extension.
7922 * doc/md.texi (w): Extend the description of the AArch64
7923 constraint to include SVE vectors.
7924 (Upl, Upa): Document new AArch64 predicate constraints.
7925 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
7927 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
7928 (msve-vector-bits=): New option.
7929 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
7930 SVE when these are disabled.
7931 (sve): New extension.
7932 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
7933 modes. Adjust their number of units based on aarch64_sve_vg.
7934 (MAX_BITSIZE_MODE_ANY_MODE): Define.
7935 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
7936 aarch64_addr_query_type.
7937 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
7938 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
7939 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
7940 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
7941 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
7942 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
7943 (aarch64_simd_imm_zero_p): Delete.
7944 (aarch64_check_zero_based_sve_index_immediate): Declare.
7945 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7946 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7947 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7948 (aarch64_sve_float_mul_immediate_p): Likewise.
7949 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7951 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
7952 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
7953 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
7954 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
7955 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
7956 (aarch64_regmode_natural_size): Likewise.
7957 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
7958 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
7960 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
7961 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
7962 for VG and the SVE predicate registers.
7963 (V_ALIASES): Add a "z"-prefixed alias.
7964 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
7965 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
7966 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
7967 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
7968 (REG_CLASS_NAMES): Add entries for them.
7969 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
7970 and the predicate registers.
7971 (aarch64_sve_vg): Declare.
7972 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
7973 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
7974 (REGMODE_NATURAL_SIZE): Define.
7975 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
7977 * config/aarch64/aarch64.c: Include cfgrtl.h.
7978 (simd_immediate_info): Add a constructor for series vectors,
7979 and an associated step field.
7980 (aarch64_sve_vg): New variable.
7981 (aarch64_dbx_register_number): Handle VG and the predicate registers.
7982 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
7983 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
7984 (VEC_ANY_DATA, VEC_STRUCT): New constants.
7985 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
7986 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
7987 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
7988 (aarch64_get_mask_mode): New functions.
7989 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
7990 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
7991 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
7992 predicate modes and predicate registers. Explicitly restrict
7993 GPRs to modes of 16 bytes or smaller. Only allow FP registers
7994 to store a vector mode if it is recognized by
7995 aarch64_classify_vector_mode.
7996 (aarch64_regmode_natural_size): New function.
7997 (aarch64_hard_regno_caller_save_mode): Return the original mode
7999 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
8000 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
8001 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
8002 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
8004 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
8005 does not overlap dest if the function is frame-related. Handle
8007 (aarch64_split_add_offset): New function.
8008 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
8009 them aarch64_add_offset.
8010 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
8011 and update call to aarch64_sub_sp.
8012 (aarch64_add_cfa_expression): New function.
8013 (aarch64_expand_prologue): Pass extra temporary registers to the
8014 functions above. Handle the case in which we need to emit new
8015 DW_CFA_expressions for registers that were originally saved
8016 relative to the stack pointer, but now have to be expressed
8017 relative to the frame pointer.
8018 (aarch64_output_mi_thunk): Pass extra temporary registers to the
8020 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
8021 IP0 and IP1 values for SVE frames.
8022 (aarch64_expand_vec_series): New function.
8023 (aarch64_expand_sve_widened_duplicate): Likewise.
8024 (aarch64_expand_sve_const_vector): Likewise.
8025 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
8026 Handle SVE constants. Use emit_move_insn to move a force_const_mem
8027 into the register, rather than emitting a SET directly.
8028 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
8029 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
8030 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
8031 (offset_9bit_signed_scaled_p): New functions.
8032 (aarch64_replicate_bitmask_imm): New function.
8033 (aarch64_bitmask_imm): Use it.
8034 (aarch64_cannot_force_const_mem): Reject expressions involving
8035 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
8036 (aarch64_classify_index): Handle SVE indices, by requiring
8037 a plain register index with a scale that matches the element size.
8038 (aarch64_classify_address): Handle SVE addresses. Assert that
8039 the mode of the address is VOIDmode or an integer mode.
8040 Update call to aarch64_classify_symbol.
8041 (aarch64_classify_symbolic_expression): Update call to
8042 aarch64_classify_symbol.
8043 (aarch64_const_vec_all_in_range_p): New function.
8044 (aarch64_print_vector_float_operand): Likewise.
8045 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
8046 "vN" for FP registers with SVE modes. Handle (const ...) vectors
8047 and the FP immediates 1.0 and 0.5.
8048 (aarch64_print_address_internal): Handle SVE addresses.
8049 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
8050 (aarch64_regno_regclass): Handle predicate registers.
8051 (aarch64_secondary_reload): Handle big-endian reloads of SVE
8053 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
8054 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
8055 (aarch64_convert_sve_vector_bits): New function.
8056 (aarch64_override_options): Use it to handle -msve-vector-bits=.
8057 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8059 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
8060 Handle SVE vector and predicate modes. Accept VL-based constants
8061 that need only one temporary register, and VL offsets that require
8062 no temporary registers.
8063 (aarch64_conditional_register_usage): Mark the predicate registers
8064 as fixed if SVE isn't available.
8065 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
8066 Return true for SVE vector and predicate modes.
8067 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
8068 rather than an unsigned int. Handle SVE modes.
8069 (aarch64_preferred_simd_mode): Update call accordingly. Handle
8071 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
8073 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8074 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8075 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8076 (aarch64_sve_float_mul_immediate_p): New functions.
8077 (aarch64_sve_valid_immediate): New function.
8078 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
8079 Explicitly reject structure modes. Check for INDEX constants.
8080 Handle PTRUE and PFALSE constants.
8081 (aarch64_check_zero_based_sve_index_immediate): New function.
8082 (aarch64_simd_imm_zero_p): Delete.
8083 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
8084 vector modes. Accept constants in the range of CNT[BHWD].
8085 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
8086 ask for an Advanced SIMD mode.
8087 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
8088 (aarch64_simd_vector_alignment): Handle SVE predicates.
8089 (aarch64_vectorize_preferred_vector_alignment): New function.
8090 (aarch64_simd_vector_alignment_reachable): Use it instead of
8092 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
8093 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
8095 (MAX_VECT_LEN): Delete.
8096 (expand_vec_perm_d): Add a vec_flags field.
8097 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
8098 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
8099 (aarch64_evpc_ext): Don't apply a big-endian lane correction
8101 (aarch64_evpc_rev): Rename to...
8102 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
8103 (aarch64_evpc_rev_global): New function.
8104 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
8105 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
8107 (aarch64_evpc_sve_tbl): New function.
8108 (aarch64_expand_vec_perm_const_1): Update after rename of
8109 aarch64_evpc_rev. Handle SVE permutes too, trying
8110 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
8111 than aarch64_evpc_tbl.
8112 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
8113 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
8114 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
8115 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
8116 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
8117 (aarch64_expand_sve_vcond): New functions.
8118 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
8119 of aarch64_vector_mode_p.
8120 (aarch64_dwarf_poly_indeterminate_value): New function.
8121 (aarch64_compute_pressure_classes): Likewise.
8122 (aarch64_can_change_mode_class): Likewise.
8123 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
8124 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
8125 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
8126 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
8127 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
8128 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
8129 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
8130 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
8132 (Dn, Dl, Dr): Accept const as well as const_vector.
8133 (Dz): Likewise. Compare against CONST0_RTX.
8134 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
8135 of "vector" where appropriate.
8136 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
8137 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
8138 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
8139 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
8140 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
8141 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
8142 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
8143 (v_int_equiv): Extend to SVE modes.
8144 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
8146 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
8147 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
8148 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
8149 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
8150 (SVE_COND_FP_CMP): New int iterators.
8151 (perm_hilo): Handle the new unpack unspecs.
8152 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
8154 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
8155 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
8156 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
8157 (aarch64_equality_operator, aarch64_constant_vector_operand)
8158 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
8159 (aarch64_sve_nonimmediate_operand): Likewise.
8160 (aarch64_sve_general_operand): Likewise.
8161 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
8162 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
8163 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
8164 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
8165 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
8166 (aarch64_sve_float_arith_immediate): Likewise.
8167 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
8168 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
8169 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
8170 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
8171 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
8172 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
8173 (aarch64_sve_float_arith_operand): Likewise.
8174 (aarch64_sve_float_arith_with_sub_operand): Likewise.
8175 (aarch64_sve_float_mul_operand): Likewise.
8176 (aarch64_sve_vec_perm_operand): Likewise.
8177 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
8178 (aarch64_mov_operand): Accept const_poly_int and const_vector.
8179 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
8180 as well as const_vector.
8181 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
8182 in file. Use CONST0_RTX and CONSTM1_RTX.
8183 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
8184 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
8185 Use aarch64_simd_imm_zero.
8186 * config/aarch64/aarch64-sve.md: New file.
8187 * config/aarch64/aarch64.md: Include it.
8188 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
8189 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
8190 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
8191 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
8192 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
8193 (sve): New attribute.
8194 (enabled): Disable instructions with the sve attribute unless
8196 (movqi, movhi): Pass CONST_POLY_INT operaneds through
8197 aarch64_expand_mov_immediate.
8198 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
8199 CNT[BHSD] immediates.
8200 (movti): Split CONST_POLY_INT moves into two halves.
8201 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
8202 Split additions that need a temporary here if the destination
8203 is the stack pointer.
8204 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
8205 (*add<mode>3_poly_1): New instruction.
8206 (set_clobber_cc): New expander.
8208 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8210 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
8211 parameter and use it instead of GET_MODE_SIZE (innermode). Use
8212 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
8213 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
8214 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
8215 Change innermode from fixed_mode_size to machine_mode.
8216 (simplify_subreg): Update call accordingly. Handle a constant-sized
8217 subreg of a variable-length CONST_VECTOR.
8219 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8220 Alan Hayward <alan.hayward@arm.com>
8221 David Sherwood <david.sherwood@arm.com>
8223 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
8224 (add_offset_to_base): New function, split out from...
8225 (create_mem_ref): ...here. When handling a scale other than 1,
8226 check first whether the address is valid without the offset.
8227 Add it into the base if so, leaving the index and scale as-is.
8229 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8232 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
8233 fold_for_warn before checking if arg2 is INTEGER_CST.
8235 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
8237 * config/rs6000/predicates.md (load_multiple_operation): Delete.
8238 (store_multiple_operation): Delete.
8239 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
8240 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
8241 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
8242 guarded by TARGET_STRING.
8243 (rs6000_output_load_multiple): Delete.
8244 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
8245 OPTION_MASK_STRING / TARGET_STRING handling.
8246 (print_operand) <'N', 'O'>: Add comment that these are unused now.
8247 (const rs6000_opt_masks) <"string">: Change mask to 0.
8248 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
8249 (MASK_STRING): Delete.
8250 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
8252 (load_multiple): Delete.
8259 (store_multiple): Delete.
8266 (movmemsi_8reg): Delete.
8267 (corresponding unnamed define_insn): Delete.
8268 (movmemsi_6reg): Delete.
8269 (corresponding unnamed define_insn): Delete.
8270 (movmemsi_4reg): Delete.
8271 (corresponding unnamed define_insn): Delete.
8272 (movmemsi_2reg): Delete.
8273 (corresponding unnamed define_insn): Delete.
8274 (movmemsi_1reg): Delete.
8275 (corresponding unnamed define_insn): Delete.
8276 * config/rs6000/rs6000.opt (mno-string): New.
8277 (mstring): Replace by deprecation warning stub.
8278 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
8280 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8282 * regrename.c (regrename_do_replace): If replacing the same
8283 reg multiple times, try to reuse last created gen_raw_REG.
8286 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
8287 main to workaround a bug in GDB.
8289 2018-01-12 Tom de Vries <tom@codesourcery.com>
8292 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
8294 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
8296 PR rtl-optimization/80481
8297 * ira-color.c (get_cap_member): New function.
8298 (allocnos_conflict_by_live_ranges_p): Use it.
8299 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
8300 (setup_slot_coalesced_allocno_live_ranges): Ditto.
8302 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
8305 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
8306 (*saddl_se_1): Ditto.
8308 (*ssubl_se_1): Ditto.
8310 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8312 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
8313 rather than wi::to_widest for DR_INITs.
8314 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
8315 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
8316 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
8318 (vect_analyze_group_access_1): Note that here.
8320 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8322 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
8323 polynomial type sizes.
8325 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8327 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
8328 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
8329 (gimple_add_tmp_var): Likewise.
8331 2018-01-12 Martin Liska <mliska@suse.cz>
8333 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
8334 (gimple_alloc_sizes): Likewise.
8335 (dump_gimple_statistics): Use PRIu64 in printf format.
8336 * gimple.h: Change uint64_t to int.
8338 2018-01-12 Martin Liska <mliska@suse.cz>
8340 * tree-core.h: Use uint64_t instead of int.
8341 * tree.c (tree_node_counts): Likewise.
8342 (tree_node_sizes): Likewise.
8343 (dump_tree_statistics): Use PRIu64 in printf format.
8345 2018-01-12 Martin Liska <mliska@suse.cz>
8347 * Makefile.in: As qsort_chk is implemented in vec.c, add
8348 vec.o to linkage of gencfn-macros.
8349 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
8350 passing the info to record_node_allocation_statistics.
8351 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
8353 * ggc-common.c (struct ggc_usage): Add operator== and use
8354 it in operator< and compare function.
8355 * mem-stats.h (struct mem_usage): Likewise.
8356 * vec.c (struct vec_usage): Remove operator< and compare
8357 function. Can be simply inherited.
8359 2018-01-12 Martin Jambor <mjambor@suse.cz>
8362 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
8363 * tree-ssa-math-opts.c: Include domwalk.h.
8364 (convert_mult_to_fma_1): New function.
8365 (fma_transformation_info): New type.
8366 (fma_deferring_state): Likewise.
8367 (cancel_fma_deferring): New function.
8368 (result_of_phi): Likewise.
8369 (last_fma_candidate_feeds_initial_phi): Likewise.
8370 (convert_mult_to_fma): Added deferring logic, split actual
8371 transformation to convert_mult_to_fma_1.
8372 (math_opts_dom_walker): New type.
8373 (math_opts_dom_walker::after_dom_children): New method, body moved
8374 here from pass_optimize_widening_mul::execute, added deferring logic
8376 (pass_optimize_widening_mul::execute): Moved most of code to
8377 math_opts_dom_walker::after_dom_children.
8378 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
8379 * config/i386/i386.c (ix86_option_override_internal): Added
8380 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
8382 2018-01-12 Richard Biener <rguenther@suse.de>
8385 * dwarf2out.c (gen_variable_die): Do not reset old_die for
8386 inline instance vars.
8388 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
8391 * config/rx/rx.c (rx_is_restricted_memory_address):
8394 2018-01-12 Richard Biener <rguenther@suse.de>
8396 PR tree-optimization/80846
8397 * target.def (split_reduction): New target hook.
8398 * targhooks.c (default_split_reduction): New function.
8399 * targhooks.h (default_split_reduction): Declare.
8400 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
8401 target requests first reduce vectors by combining low and high
8403 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
8404 (get_vectype_for_scalar_type_and_size): Export.
8405 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
8406 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
8407 * doc/tm.texi: Regenerate.
8408 * config/i386/i386.c (ix86_split_reduction): Implement
8409 TARGET_VECTORIZE_SPLIT_REDUCTION.
8411 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8414 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
8415 in PIC mode except for TARGET_VXWORKS_RTP.
8416 * config/sparc/sparc.c: Include cfgrtl.h.
8417 (TARGET_INIT_PIC_REG): Define.
8418 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
8419 (sparc_pic_register_p): New predicate.
8420 (sparc_legitimate_address_p): Use it.
8421 (sparc_legitimize_pic_address): Likewise.
8422 (sparc_delegitimize_address): Likewise.
8423 (sparc_mode_dependent_address_p): Likewise.
8424 (gen_load_pcrel_sym): Remove 4th parameter.
8425 (load_got_register): Adjust call to above. Remove obsolete stuff.
8426 (sparc_expand_prologue): Do not call load_got_register here.
8427 (sparc_flat_expand_prologue): Likewise.
8428 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
8429 (sparc_use_pseudo_pic_reg): New function.
8430 (sparc_init_pic_reg): Likewise.
8431 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8432 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
8434 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
8436 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
8437 Add item for branch_cost.
8439 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8441 PR rtl-optimization/83565
8442 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
8443 not extend the result to a larger mode for rotate operations.
8444 (num_sign_bit_copies1): Likewise.
8446 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8449 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
8451 Use values-Xc.o for -pedantic.
8452 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
8454 2018-01-12 Martin Liska <mliska@suse.cz>
8457 * ipa-devirt.c (final_warning_record::grow_type_warnings):
8459 (possible_polymorphic_call_targets): Use it.
8460 (ipa_devirt): Likewise.
8462 2018-01-12 Martin Liska <mliska@suse.cz>
8464 * profile-count.h (enum profile_quality): Use 0 as invalid
8465 enum value of profile_quality.
8467 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
8469 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
8470 -mext-string options.
8472 2018-01-12 Richard Biener <rguenther@suse.de>
8474 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
8475 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
8476 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
8478 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
8480 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
8482 * configure.ac (--with-long-double-format): Add support for the
8483 configuration option to change the default long double format on
8485 * config.gcc (powerpc*-linux*-*): Likewise.
8486 * configure: Regenerate.
8487 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
8488 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
8489 used without modification.
8491 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8493 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
8494 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
8495 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
8496 MISC_BUILTIN_SPEC_BARRIER.
8497 (rs6000_init_builtins): Likewise.
8498 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
8500 (speculation_barrier): New define_insn.
8501 * doc/extend.texi: Document __builtin_speculation_barrier.
8503 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8506 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
8507 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
8508 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
8510 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
8511 integral modes instead of "ss" and "sd".
8512 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
8513 vectors with 32-bit and 64-bit elements.
8514 (vecdupssescalarmodesuffix): New mode attribute.
8515 (vec_dup<mode>): Use it.
8517 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
8520 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
8521 frame if argument is passed on stack.
8523 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8526 * ree.c (combine_reaching_defs): Optimize also
8527 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
8528 reg2=any_extend(exp); reg1=reg2;, formatting fix.
8530 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8533 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
8535 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8538 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
8539 after they are computed.
8541 2018-01-11 Bin Cheng <bin.cheng@arm.com>
8543 PR tree-optimization/83695
8544 * gimple-loop-linterchange.cc
8545 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
8546 reset cached scev information after interchange.
8547 (pass_linterchange::execute): Remove call to scev_reset_htab.
8549 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8551 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
8552 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
8553 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
8554 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
8555 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
8556 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
8557 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
8558 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
8559 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
8560 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
8561 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
8562 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
8563 (V_lane_reg): Likewise.
8564 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
8566 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
8567 (vfmal_lane_low<mode>_intrinsic,
8568 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
8569 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
8570 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
8571 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
8572 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
8573 vfmsl_lane_high<mode>_intrinsic): New define_insns.
8575 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8577 * config/arm/arm-cpus.in (fp16fml): New feature.
8578 (ALL_SIMD): Add fp16fml.
8579 (armv8.2-a): Add fp16fml as an option.
8580 (armv8.3-a): Likewise.
8581 (armv8.4-a): Add fp16fml as part of fp16.
8582 * config/arm/arm.h (TARGET_FP16FML): Define.
8583 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
8585 * config/arm/arm-modes.def (V2HF): Define.
8586 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
8587 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
8588 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
8589 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
8590 vfmsl_low, vfmsl_high): New set of builtins.
8591 * config/arm/iterators.md (PLUSMINUS): New code iterator.
8592 (vfml_op): New code attribute.
8593 (VFMLHALVES): New int iterator.
8594 (VFML, VFMLSEL): New mode attributes.
8595 (V_reg): Define mapping for V2HF.
8596 (V_hi, V_lo): New mode attributes.
8597 (VF_constraint): Likewise.
8598 (vfml_half, vfml_half_selector): New int attributes.
8599 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
8601 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
8602 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
8604 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
8605 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
8606 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
8607 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
8609 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
8610 Document new effective target and option set.
8612 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8614 * config/arm/arm-cpus.in (armv8_4): New feature.
8615 (ARMv8_4a): New fgroup.
8616 (armv8.4-a): New arch.
8617 * config/arm/arm-tables.opt: Regenerate.
8618 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
8619 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
8620 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
8621 Add matching rules for -march=armv8.4-a and extensions.
8622 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
8624 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
8627 * config/rx/rx.md (BW): New mode attribute.
8628 (sync_lock_test_and_setsi): Add mode suffix to insn output.
8630 2018-01-11 Richard Biener <rguenther@suse.de>
8632 PR tree-optimization/83435
8633 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
8634 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
8635 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
8637 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8638 Alan Hayward <alan.hayward@arm.com>
8639 David Sherwood <david.sherwood@arm.com>
8641 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
8643 (aarch64_classify_address): Initialize it. Track polynomial offsets.
8644 (aarch64_print_address_internal): Use it to check for a zero offset.
8646 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8647 Alan Hayward <alan.hayward@arm.com>
8648 David Sherwood <david.sherwood@arm.com>
8650 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
8651 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
8652 Return a poly_int64 rather than a HOST_WIDE_INT.
8653 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
8654 rather than a HOST_WIDE_INT.
8655 * config/aarch64/aarch64.h (aarch64_frame): Protect with
8656 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
8657 hard_fp_offset, frame_size, initial_adjust, callee_offset and
8658 final_offset from HOST_WIDE_INT to poly_int64.
8659 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
8660 to_constant when getting the number of units in an Advanced SIMD
8662 (aarch64_builtin_vectorized_function): Check for a constant number
8664 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
8666 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
8667 attribute instead of GET_MODE_NUNITS.
8668 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
8669 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
8670 GET_MODE_SIZE for fixed-size registers.
8671 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
8672 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
8673 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
8674 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
8675 (aarch64_print_operand, aarch64_print_address_internal)
8676 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
8677 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
8678 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
8679 Handle polynomial GET_MODE_SIZE.
8680 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
8681 wider than SImode without modification.
8682 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
8683 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
8684 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
8685 passing and returning SVE modes.
8686 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
8687 rather than GEN_INT.
8688 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
8689 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
8690 (aarch64_allocate_and_probe_stack_space): Likewise.
8691 (aarch64_layout_frame): Cope with polynomial offsets.
8692 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
8693 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
8695 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
8696 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
8697 poly_int64 rather than a HOST_WIDE_INT.
8698 (aarch64_get_separate_components, aarch64_process_components)
8699 (aarch64_expand_prologue, aarch64_expand_epilogue)
8700 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
8701 (aarch64_anchor_offset): New function, split out from...
8702 (aarch64_legitimize_address): ...here.
8703 (aarch64_builtin_vectorization_cost): Handle polynomial
8704 TYPE_VECTOR_SUBPARTS.
8705 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
8707 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
8708 number of elements from the PARALLEL rather than the mode.
8709 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
8710 rather than GET_MODE_BITSIZE.
8711 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
8712 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
8713 (aarch64_expand_vec_perm_const_1): Handle polynomial
8714 d->perm.length () and d->perm elements.
8715 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
8716 Apply to_constant to d->perm elements.
8717 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
8718 polynomial CONST_VECTOR_NUNITS.
8719 (aarch64_move_pointer): Take amount as a poly_int64 rather
8721 (aarch64_progress_pointer): Avoid temporary variable.
8722 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
8723 the mode attribute instead of GET_MODE.
8725 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8726 Alan Hayward <alan.hayward@arm.com>
8727 David Sherwood <david.sherwood@arm.com>
8729 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
8730 x exists before using it.
8731 (aarch64_add_constant_internal): Rename to...
8732 (aarch64_add_offset_1): ...this. Replace regnum with separate
8733 src and dest rtxes. Handle the case in which they're different,
8734 including when the offset is zero. Replace scratchreg with an rtx.
8735 Use 2 additions if there is no spare register into which we can
8736 move a 16-bit constant.
8737 (aarch64_add_constant): Delete.
8738 (aarch64_add_offset): Replace reg with separate src and dest
8739 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
8740 Use aarch64_add_offset_1.
8741 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
8742 an rtx rather than an int. Take the delta as a poly_int64
8743 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
8744 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
8745 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
8746 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
8747 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
8749 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
8750 aarch64_add_constant.
8752 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8754 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
8755 Use scalar_float_mode.
8757 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8759 * config/aarch64/aarch64-simd.md
8760 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
8761 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
8762 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
8763 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
8764 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
8765 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
8766 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
8767 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
8768 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
8769 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
8771 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8774 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
8775 targ_options->x_arm_arch_string is non NULL.
8777 2018-01-11 Tamar Christina <tamar.christina@arm.com>
8779 * config/aarch64/aarch64.h
8780 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
8782 2018-01-11 Sudakshina Das <sudi.das@arm.com>
8785 * expmed.c (emit_store_flag_force): Swap if const op0
8786 and change VOIDmode to mode of op0.
8788 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8790 PR rtl-optimization/83761
8791 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
8792 than bytes to mode_for_size.
8794 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8797 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
8798 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
8801 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8804 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
8805 when in layout mode.
8806 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
8807 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
8810 2018-01-10 Michael Collison <michael.collison@arm.com>
8812 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
8813 * config/aarch64/aarch64-option-extension.def: Add
8814 AARCH64_OPT_EXTENSION of 'fp16fml'.
8815 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8816 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
8817 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
8818 * config/aarch64/constraints.md (Ui7): New constraint.
8819 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
8820 (VFMLA_SEL_W): Ditto.
8823 (VFMLA16_LOW): New int iterator.
8824 (VFMLA16_HIGH): Ditto.
8825 (UNSPEC_FMLAL): New unspec.
8826 (UNSPEC_FMLSL): Ditto.
8827 (UNSPEC_FMLAL2): Ditto.
8828 (UNSPEC_FMLSL2): Ditto.
8829 (f16mac): New code attribute.
8830 * config/aarch64/aarch64-simd-builtins.def
8831 (aarch64_fmlal_lowv2sf): Ditto.
8832 (aarch64_fmlsl_lowv2sf): Ditto.
8833 (aarch64_fmlalq_lowv4sf): Ditto.
8834 (aarch64_fmlslq_lowv4sf): Ditto.
8835 (aarch64_fmlal_highv2sf): Ditto.
8836 (aarch64_fmlsl_highv2sf): Ditto.
8837 (aarch64_fmlalq_highv4sf): Ditto.
8838 (aarch64_fmlslq_highv4sf): Ditto.
8839 (aarch64_fmlal_lane_lowv2sf): Ditto.
8840 (aarch64_fmlsl_lane_lowv2sf): Ditto.
8841 (aarch64_fmlal_laneq_lowv2sf): Ditto.
8842 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
8843 (aarch64_fmlalq_lane_lowv4sf): Ditto.
8844 (aarch64_fmlsl_lane_lowv4sf): Ditto.
8845 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
8846 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
8847 (aarch64_fmlal_lane_highv2sf): Ditto.
8848 (aarch64_fmlsl_lane_highv2sf): Ditto.
8849 (aarch64_fmlal_laneq_highv2sf): Ditto.
8850 (aarch64_fmlsl_laneq_highv2sf): Ditto.
8851 (aarch64_fmlalq_lane_highv4sf): Ditto.
8852 (aarch64_fmlsl_lane_highv4sf): Ditto.
8853 (aarch64_fmlalq_laneq_highv4sf): Ditto.
8854 (aarch64_fmlsl_laneq_highv4sf): Ditto.
8855 * config/aarch64/aarch64-simd.md:
8856 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
8857 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8858 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
8859 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8860 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
8861 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
8862 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
8863 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
8864 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
8865 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
8866 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
8867 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
8868 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
8869 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
8870 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
8871 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
8872 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
8873 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
8874 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
8875 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
8876 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
8877 (vfmlsl_low_u32): Ditto.
8878 (vfmlalq_low_u32): Ditto.
8879 (vfmlslq_low_u32): Ditto.
8880 (vfmlal_high_u32): Ditto.
8881 (vfmlsl_high_u32): Ditto.
8882 (vfmlalq_high_u32): Ditto.
8883 (vfmlslq_high_u32): Ditto.
8884 (vfmlal_lane_low_u32): Ditto.
8885 (vfmlsl_lane_low_u32): Ditto.
8886 (vfmlal_laneq_low_u32): Ditto.
8887 (vfmlsl_laneq_low_u32): Ditto.
8888 (vfmlalq_lane_low_u32): Ditto.
8889 (vfmlslq_lane_low_u32): Ditto.
8890 (vfmlalq_laneq_low_u32): Ditto.
8891 (vfmlslq_laneq_low_u32): Ditto.
8892 (vfmlal_lane_high_u32): Ditto.
8893 (vfmlsl_lane_high_u32): Ditto.
8894 (vfmlal_laneq_high_u32): Ditto.
8895 (vfmlsl_laneq_high_u32): Ditto.
8896 (vfmlalq_lane_high_u32): Ditto.
8897 (vfmlslq_lane_high_u32): Ditto.
8898 (vfmlalq_laneq_high_u32): Ditto.
8899 (vfmlslq_laneq_high_u32): Ditto.
8900 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
8901 (AARCH64_FL_FOR_ARCH8_4): New.
8902 (AARCH64_ISA_F16FML): New ISA flag.
8903 (TARGET_F16FML): New feature flag for fp16fml.
8904 (doc/invoke.texi): Document new fp16fml option.
8906 2018-01-10 Michael Collison <michael.collison@arm.com>
8908 * config/aarch64/aarch64-builtins.c:
8909 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
8910 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8911 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
8912 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
8913 (AARCH64_ISA_SHA3): New ISA flag.
8914 (TARGET_SHA3): New feature flag for sha3.
8915 * config/aarch64/iterators.md (sha512_op): New int attribute.
8916 (CRYPTO_SHA512): New int iterator.
8917 (UNSPEC_SHA512H): New unspec.
8918 (UNSPEC_SHA512H2): Ditto.
8919 (UNSPEC_SHA512SU0): Ditto.
8920 (UNSPEC_SHA512SU1): Ditto.
8921 * config/aarch64/aarch64-simd-builtins.def
8922 (aarch64_crypto_sha512hqv2di): New builtin.
8923 (aarch64_crypto_sha512h2qv2di): Ditto.
8924 (aarch64_crypto_sha512su0qv2di): Ditto.
8925 (aarch64_crypto_sha512su1qv2di): Ditto.
8926 (aarch64_eor3qv8hi): Ditto.
8927 (aarch64_rax1qv2di): Ditto.
8928 (aarch64_xarqv2di): Ditto.
8929 (aarch64_bcaxqv8hi): Ditto.
8930 * config/aarch64/aarch64-simd.md:
8931 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
8932 (aarch64_crypto_sha512su0qv2di): Ditto.
8933 (aarch64_crypto_sha512su1qv2di): Ditto.
8934 (aarch64_eor3qv8hi): Ditto.
8935 (aarch64_rax1qv2di): Ditto.
8936 (aarch64_xarqv2di): Ditto.
8937 (aarch64_bcaxqv8hi): Ditto.
8938 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
8939 (vsha512h2q_u64): Ditto.
8940 (vsha512su0q_u64): Ditto.
8941 (vsha512su1q_u64): Ditto.
8942 (veor3q_u16): Ditto.
8943 (vrax1q_u64): Ditto.
8945 (vbcaxq_u16): Ditto.
8946 * config/arm/types.md (crypto_sha512): New type attribute.
8947 (crypto_sha3): Ditto.
8948 (doc/invoke.texi): Document new sha3 option.
8950 2018-01-10 Michael Collison <michael.collison@arm.com>
8952 * config/aarch64/aarch64-builtins.c:
8953 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
8954 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8955 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
8956 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
8957 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
8958 (AARCH64_ISA_SM4): New ISA flag.
8959 (TARGET_SM4): New feature flag for sm4.
8960 * config/aarch64/aarch64-simd-builtins.def
8961 (aarch64_sm3ss1qv4si): Ditto.
8962 (aarch64_sm3tt1aq4si): Ditto.
8963 (aarch64_sm3tt1bq4si): Ditto.
8964 (aarch64_sm3tt2aq4si): Ditto.
8965 (aarch64_sm3tt2bq4si): Ditto.
8966 (aarch64_sm3partw1qv4si): Ditto.
8967 (aarch64_sm3partw2qv4si): Ditto.
8968 (aarch64_sm4eqv4si): Ditto.
8969 (aarch64_sm4ekeyqv4si): Ditto.
8970 * config/aarch64/aarch64-simd.md:
8971 (aarch64_sm3ss1qv4si): Ditto.
8972 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
8973 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
8974 (aarch64_sm4eqv4si): Ditto.
8975 (aarch64_sm4ekeyqv4si): Ditto.
8976 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
8977 (sm3part_op): Ditto.
8978 (CRYPTO_SM3TT): Ditto.
8979 (CRYPTO_SM3PART): Ditto.
8980 (UNSPEC_SM3SS1): New unspec.
8981 (UNSPEC_SM3TT1A): Ditto.
8982 (UNSPEC_SM3TT1B): Ditto.
8983 (UNSPEC_SM3TT2A): Ditto.
8984 (UNSPEC_SM3TT2B): Ditto.
8985 (UNSPEC_SM3PARTW1): Ditto.
8986 (UNSPEC_SM3PARTW2): Ditto.
8987 (UNSPEC_SM4E): Ditto.
8988 (UNSPEC_SM4EKEY): Ditto.
8989 * config/aarch64/constraints.md (Ui2): New constraint.
8990 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
8991 * config/arm/types.md (crypto_sm3): New type attribute.
8992 (crypto_sm4): Ditto.
8993 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
8994 (vsm3tt1aq_u32): Ditto.
8995 (vsm3tt1bq_u32): Ditto.
8996 (vsm3tt2aq_u32): Ditto.
8997 (vsm3tt2bq_u32): Ditto.
8998 (vsm3partw1q_u32): Ditto.
8999 (vsm3partw2q_u32): Ditto.
9000 (vsm4eq_u32): Ditto.
9001 (vsm4ekeyq_u32): Ditto.
9002 (doc/invoke.texi): Document new sm4 option.
9004 2018-01-10 Michael Collison <michael.collison@arm.com>
9006 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
9007 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
9008 (AARCH64_FL_FOR_ARCH8_4): New.
9009 (AARCH64_FL_V8_4): New flag.
9010 (doc/invoke.texi): Document new armv8.4-a option.
9012 2018-01-10 Michael Collison <michael.collison@arm.com>
9014 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9015 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
9016 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
9017 * config/aarch64/aarch64-option-extension.def: Add
9018 AARCH64_OPT_EXTENSION of 'sha2'.
9019 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
9020 (crypto): Disable sha2 and aes if crypto disabled.
9021 (crypto): Enable aes and sha2 if enabled.
9022 (simd): Disable sha2 and aes if simd disabled.
9023 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
9025 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
9026 (TARGET_SHA2): New feature flag for sha2.
9027 (TARGET_AES): New feature flag for aes.
9028 * config/aarch64/aarch64-simd.md:
9029 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
9030 conditional on TARGET_AES.
9031 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
9032 (aarch64_crypto_sha1hsi): Make pattern conditional
9034 (aarch64_crypto_sha1hv4si): Ditto.
9035 (aarch64_be_crypto_sha1hv4si): Ditto.
9036 (aarch64_crypto_sha1su1v4si): Ditto.
9037 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
9038 (aarch64_crypto_sha1su0v4si): Ditto.
9039 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
9040 (aarch64_crypto_sha256su0v4si): Ditto.
9041 (aarch64_crypto_sha256su1v4si): Ditto.
9042 (doc/invoke.texi): Document new aes and sha2 options.
9044 2018-01-10 Martin Sebor <msebor@redhat.com>
9046 PR tree-optimization/83781
9047 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
9050 2018-01-11 Martin Sebor <msebor@gmail.com>
9051 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9053 PR tree-optimization/83501
9054 PR tree-optimization/81703
9056 * tree-ssa-strlen.c (get_string_cst): Rename...
9057 (get_string_len): ...to this. Handle global constants.
9058 (handle_char_store): Adjust.
9060 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
9061 Jim Wilson <jimw@sifive.com>
9063 * config/riscv/riscv-protos.h (riscv_output_return): New.
9064 * config/riscv/riscv.c (struct machine_function): New naked_p field.
9065 (riscv_attribute_table, riscv_output_return),
9066 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
9067 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
9068 (riscv_compute_frame_info): Only compute frame->mask if not a naked
9070 (riscv_expand_prologue): Add early return for naked function.
9071 (riscv_expand_epilogue): Likewise.
9072 (riscv_function_ok_for_sibcall): Return false for naked function.
9073 (riscv_set_current_function): New.
9074 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
9075 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
9076 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
9077 * doc/extend.texi (RISC-V Function Attributes): New.
9079 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
9081 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
9082 check for 128-bit long double before checking TCmode.
9083 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
9084 128-bit long doubles before checking TFmode or TCmode.
9085 (FLOAT128_IBM_P): Likewise.
9087 2018-01-10 Martin Sebor <msebor@redhat.com>
9089 PR tree-optimization/83671
9090 * builtins.c (c_strlen): Unconditionally return zero for the empty
9092 Use -Warray-bounds for warnings.
9093 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
9094 for non-constant array indices with COMPONENT_REF, arrays of
9095 arrays, and pointers to arrays.
9096 (gimple_fold_builtin_strlen): Determine and set length range for
9097 non-constant character arrays.
9099 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
9102 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
9105 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
9107 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
9109 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9112 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
9113 VECTOR_MEM_ALTIVEC_OR_VSX_P.
9114 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
9115 indexed_or_indirect_operand predicate.
9116 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
9117 (*vsx_le_perm_load_v8hi): Likewise.
9118 (*vsx_le_perm_load_v16qi): Likewise.
9119 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
9120 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
9121 (*vsx_le_perm_store_v8hi): Likewise.
9122 (*vsx_le_perm_store_v16qi): Likewise.
9123 (eight unnamed splitters): Likewise.
9125 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9127 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
9128 * config/rs6000/emmintrin.h: Likewise.
9129 * config/rs6000/mmintrin.h: Likewise.
9130 * config/rs6000/xmmintrin.h: Likewise.
9132 2018-01-10 David Malcolm <dmalcolm@redhat.com>
9135 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
9137 * tree.c (tree_nop_conversion): Return true for location wrapper
9139 (maybe_wrap_with_location): New function.
9140 (selftest::check_strip_nops): New function.
9141 (selftest::test_location_wrappers): New function.
9142 (selftest::tree_c_tests): Call it.
9143 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
9144 (maybe_wrap_with_location): New decl.
9145 (EXPR_LOCATION_WRAPPER_P): New macro.
9146 (location_wrapper_p): New inline function.
9147 (tree_strip_any_location_wrapper): New inline function.
9149 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
9152 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
9153 stack_realign_offset for the largest alignment of stack slot
9155 (ix86_find_max_used_stack_alignment): New function.
9156 (ix86_finalize_stack_frame_flags): Use it. Set
9157 max_used_stack_alignment if we don't realign stack.
9158 * config/i386/i386.h (machine_function): Add
9159 max_used_stack_alignment.
9161 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
9163 * config/arm/arm.opt (-mbranch-cost): New option.
9164 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
9167 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
9170 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
9171 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
9173 2018-01-10 Richard Biener <rguenther@suse.de>
9176 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
9177 early out so it also covers the case where we have a non-NULL
9180 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9182 PR tree-optimization/83753
9183 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
9184 for non-strided grouped accesses if the number of elements is 1.
9186 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9189 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
9190 * i386.h (TARGET_USE_GATHER): Define.
9191 * x86-tune.def (X86_TUNE_USE_GATHER): New.
9193 2018-01-10 Martin Liska <mliska@suse.cz>
9196 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
9197 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
9199 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
9200 CLEANUP_NO_PARTITIONING is not set.
9202 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9204 * doc/rtl.texi: Remove documentation of (const ...) wrappers
9205 for vectors, as a partial revert of r254296.
9206 * rtl.h (const_vec_p): Delete.
9207 (const_vec_duplicate_p): Don't test for vector CONSTs.
9208 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
9209 * expmed.c (make_tree): Likewise.
9212 * common.md (E, F): Use CONSTANT_P instead of checking for
9214 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9215 checking for CONST_VECTOR.
9217 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9220 * predict.c (force_edge_cold): Handle in more sane way edges
9223 2018-01-09 Carl Love <cel@us.ibm.com>
9225 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
9227 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
9228 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
9229 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
9230 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
9231 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
9232 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
9233 * config/rs6000/rs6000-protos.h: Add extern defition for
9234 rs6000_generate_float2_double_code.
9235 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
9237 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
9238 (float2_v2df): Add define_expand.
9240 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
9243 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
9244 op_mode in the force_to_mode call.
9246 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9248 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
9249 instead of checking each element individually.
9250 (aarch64_evpc_uzp): Likewise.
9251 (aarch64_evpc_zip): Likewise.
9252 (aarch64_evpc_ext): Likewise.
9253 (aarch64_evpc_rev): Likewise.
9254 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
9255 instead of checking each element individually. Return true without
9257 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
9258 whether all selected elements come from the same input, instead of
9259 checking each element individually. Remove calls to gen_rtx_REG,
9260 start_sequence and end_sequence and instead assert that no rtl is
9263 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9265 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
9266 order of HIGH and CONST checks.
9268 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9270 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
9271 if the destination isn't an SSA_NAME.
9273 2018-01-09 Richard Biener <rguenther@suse.de>
9275 PR tree-optimization/83668
9276 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
9278 (canonicalize_loop_form): ... here, renamed from ...
9279 (canonicalize_loop_closed_ssa_form): ... this and amended to
9280 swap successor edges for loop exit blocks to make us use
9281 the RPO order we need for initial schedule generation.
9283 2018-01-09 Joseph Myers <joseph@codesourcery.com>
9285 PR tree-optimization/64811
9286 * match.pd: When optimizing comparisons with Inf, avoid
9287 introducing or losing exceptions from comparisons with NaN.
9289 2018-01-09 Martin Liska <mliska@suse.cz>
9292 * asan.c (shadow_mem_size): Add gcc_assert.
9294 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
9296 Don't save registers in main().
9299 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
9300 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
9301 * config/avr/avr.c (avr_set_current_function): Don't error if
9302 naked, OS_task or OS_main are specified at the same time.
9303 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
9305 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
9307 * common/config/avr/avr-common.c (avr_option_optimization_table):
9308 Switch on -mmain-is-OS_task for optimizing compilations.
9310 2018-01-09 Richard Biener <rguenther@suse.de>
9312 PR tree-optimization/83572
9313 * graphite.c: Include cfganal.h.
9314 (graphite_transform_loops): Connect infinite loops to exit
9315 and remove fake edges at the end.
9317 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9319 * ipa-inline.c (edge_badness): Revert accidental checkin.
9321 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9324 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
9325 symbols; not inline clones.
9327 2018-01-09 Jakub Jelinek <jakub@redhat.com>
9330 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
9331 hard registers. Formatting fixes.
9333 PR preprocessor/83722
9334 * gcc.c (try_generate_repro): Pass
9335 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
9336 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
9339 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
9340 Kito Cheng <kito.cheng@gmail.com>
9342 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
9343 (riscv_leaf_function_p): Delete.
9344 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
9346 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9348 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
9350 (do_ifelse): New function.
9351 (do_isel): New function.
9352 (do_sub3): New function.
9353 (do_add3): New function.
9354 (do_load_mask_compare): New function.
9355 (do_overlap_load_compare): New function.
9356 (expand_compare_loop): New function.
9357 (expand_block_compare): Call expand_compare_loop() when appropriate.
9358 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
9360 (-mblock-compare-inline-loop-limit): New option.
9362 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9365 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
9366 Reverse order of second and third operands in first alternative.
9367 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
9368 of first and second elements in UNSPEC_VPERMR vector.
9369 (altivec_expand_vec_perm_le): Likewise.
9371 2018-01-08 Jeff Law <law@redhat.com>
9373 PR rtl-optimizatin/81308
9374 * tree-switch-conversion.c (cfg_altered): New file scoped static.
9375 (process_switch): If group_case_labels makes a change, then set
9377 (pass_convert_switch::execute): If a switch is converted, then
9378 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
9380 PR rtl-optimization/81308
9381 * recog.c (split_all_insns): Conditionally cleanup the CFG after
9384 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
9386 PR target/83663 - Revert r255946
9387 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
9388 generation for cases where splatting a value is not useful.
9389 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
9390 across a vec_duplicate and a paradoxical subreg forming a vector
9391 mode to a vec_concat.
9393 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9395 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
9396 -march=armv8.3-a variants.
9397 * config/arm/t-multilib: Likewise.
9398 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
9400 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9402 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
9404 (cceq_ior_compare_complement): Give it a name so I can use it, and
9405 change boolean_or_operator predicate to boolean_operator so it can
9406 be used to generate a crand.
9407 (eqne): New code iterator.
9408 (bd/bd_neg): New code_attrs.
9409 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
9410 a single define_insn.
9411 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
9412 decrement (bdnzt/bdnzf/bdzt/bdzf).
9413 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
9414 with the new names of the branch decrement patterns, and added the
9415 names of the branch decrement conditional patterns.
9417 2018-01-08 Richard Biener <rguenther@suse.de>
9419 PR tree-optimization/83563
9420 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
9423 2018-01-08 Richard Biener <rguenther@suse.de>
9426 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
9428 2018-01-08 Richard Biener <rguenther@suse.de>
9430 PR tree-optimization/83685
9431 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
9432 references to abnormals.
9434 2018-01-08 Richard Biener <rguenther@suse.de>
9437 * dwarf2out.c (output_indirect_strings): Handle empty
9438 skeleton_debug_str_hash.
9439 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
9441 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9443 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
9444 (emit_store_direct): Likewise.
9445 (arc_trampoline_adjust_address): Likewise.
9446 (arc_asm_trampoline_template): New function.
9447 (arc_initialize_trampoline): Use asm_trampoline_template.
9448 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
9449 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
9450 * config/arc/arc.md (flush_icache): Delete pattern.
9452 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9454 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
9455 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
9458 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9461 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
9462 by not USED_FOR_TARGET.
9463 (make_pass_resolve_sw_modes): Likewise.
9465 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9467 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
9470 2018-01-08 Richard Biener <rguenther@suse.de>
9473 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
9475 2018-01-08 Richard Biener <rguenther@suse.de>
9478 * match.pd ((t * 2) / 2) -> t): Add missing :c.
9480 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
9483 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
9484 basic blocks with a small number of successors.
9485 (convert_control_dep_chain_into_preds): Improve handling of
9487 (dump_predicates): Split apart into...
9488 (dump_pred_chain): ...here...
9489 (dump_pred_info): ...and here.
9490 (can_one_predicate_be_invalidated_p): Add debugging printfs.
9491 (can_chain_union_be_invalidated_p): Improve check for invalidation
9493 (uninit_uses_cannot_happen): Avoid unnecessary if
9494 convert_control_dep_chain_into_preds yielded nothing.
9496 2018-01-06 Martin Sebor <msebor@redhat.com>
9498 PR tree-optimization/83640
9499 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
9500 subtracting negative offset from size.
9501 (builtin_access::overlap): Adjust offset bounds of the access to fall
9502 within the size of the object if possible.
9504 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
9506 PR rtl-optimization/83699
9507 * expmed.c (extract_bit_field_1): Restrict the vector usage of
9508 extract_bit_field_as_subreg to cases in which the extracted
9509 value is also a vector.
9511 * lra-constraints.c (process_alt_operands): Test for the equivalence
9512 substitutions when detecting a possible reload cycle.
9514 2018-01-06 Jakub Jelinek <jakub@redhat.com>
9517 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
9518 by default if flag_selective_schedling{,2}. Formatting fixes.
9520 PR rtl-optimization/83682
9521 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
9522 if it has non-VECTOR_MODE element mode.
9523 (vec_duplicate_p): Likewise.
9526 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
9527 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
9529 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9532 * config/i386/i386-builtin.def
9533 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
9534 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
9535 Require also OPTION_MASK_ISA_AVX512F in addition to
9536 OPTION_MASK_ISA_GFNI.
9537 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
9538 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
9539 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
9540 to OPTION_MASK_ISA_GFNI.
9541 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
9542 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
9543 OPTION_MASK_ISA_AVX512BW.
9544 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
9545 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
9546 addition to OPTION_MASK_ISA_GFNI.
9547 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
9548 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
9549 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
9550 to OPTION_MASK_ISA_GFNI.
9551 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
9552 a requirement for all ISAs rather than any of them with a few
9554 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
9556 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
9557 bitmasks to be enabled with 3 exceptions, instead of requiring any
9558 enabled ISA with lots of exceptions.
9559 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
9560 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
9561 Change avx512bw in isa attribute to avx512f.
9562 * config/i386/sgxintrin.h: Add license boilerplate.
9563 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
9564 to __AVX512F__ and __AVX512VL to __AVX512VL__.
9565 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
9566 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
9568 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
9569 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
9570 temporarily sse2 rather than sse if not enabled already.
9573 * config/i386/sse.md (VI248_VLBW): Rename to ...
9574 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
9575 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
9576 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
9577 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
9578 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
9579 mode iterator instead of VI248_VLBW.
9581 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
9583 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
9584 (record_modified): Skip clobbers; add debug output.
9585 (param_change_prob): Use sreal frequencies.
9587 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9589 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
9590 punt for user-aligned variables.
9592 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9594 * tree-chrec.c (chrec_contains_symbols): Return true for
9597 2018-01-05 Sudakshina Das <sudi.das@arm.com>
9600 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
9601 of (x|y) == x for BICS pattern.
9603 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9605 PR tree-optimization/83605
9606 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
9607 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
9610 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
9612 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
9613 * config/epiphany/rtems.h: New file.
9615 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9616 Uros Bizjak <ubizjak@gmail.com>
9619 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
9620 QIreg_operand instead of register_operand predicate.
9621 * config/i386/i386.c (ix86_rop_should_change_byte_p,
9622 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
9623 comments instead of -fmitigate[-_]rop.
9625 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9628 * cgraphunit.c (symbol_table::compile): Switch to text_section
9629 before calling assembly_start debug hook.
9630 * run-rtl-passes.c (run_rtl_passes): Likewise.
9633 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9635 * tree-vrp.c (extract_range_from_binary_expr_1): Check
9636 range_int_cst_p rather than !symbolic_range_p before calling
9637 extract_range_from_multiplicative_op_1.
9639 2018-01-04 Jeff Law <law@redhat.com>
9641 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
9642 redundant test in assertion.
9644 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9646 * doc/rtl.texi: Document machine_mode wrapper classes.
9648 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9650 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
9653 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9655 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
9656 the VEC_PERM_EXPR fold to fail.
9658 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9661 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
9662 to switched_sections.
9664 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9667 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
9670 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
9673 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
9674 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
9676 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9679 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
9680 is BLKmode and bitpos not zero or mode change is needed.
9682 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9685 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
9688 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
9691 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
9692 instead of MULT rtx. Update all corresponding splitters.
9694 (*ssub<modesuffix>): Ditto.
9696 (*cmp_sadd_di): Update split patterns.
9697 (*cmp_sadd_si): Ditto.
9698 (*cmp_sadd_sidi): Ditto.
9699 (*cmp_ssub_di): Ditto.
9700 (*cmp_ssub_si): Ditto.
9701 (*cmp_ssub_sidi): Ditto.
9702 * config/alpha/predicates.md (const23_operand): New predicate.
9703 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
9704 Look for ASHIFT, not MULT inner operand.
9705 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
9707 2018-01-04 Martin Liska <mliska@suse.cz>
9709 PR gcov-profile/83669
9710 * gcov.c (output_intermediate_file): Add version to intermediate
9712 * doc/gcov.texi: Document new field 'version' in intermediate
9713 file format. Fix location of '-k' option of gcov command.
9715 2018-01-04 Martin Liska <mliska@suse.cz>
9718 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
9720 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9722 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
9724 2018-01-03 Martin Sebor <msebor@redhat.com>
9726 PR tree-optimization/83655
9727 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
9728 checking calls with invalid arguments.
9730 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9732 * tree-vect-stmts.c (vect_get_store_rhs): New function.
9733 (vectorizable_mask_load_store): Delete.
9734 (vectorizable_call): Return false for masked loads and stores.
9735 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
9736 instead of gimple_assign_rhs1.
9737 (vectorizable_load): Handle IFN_MASK_LOAD.
9738 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
9740 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9742 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
9744 (vectorizable_mask_load_store): ...here.
9745 (vectorizable_load): ...and here.
9747 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9749 * tree-vect-stmts.c (vect_build_all_ones_mask)
9750 (vect_build_zero_merge_argument): New functions, split out from...
9751 (vectorizable_load): ...here.
9753 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9755 * tree-vect-stmts.c (vect_check_store_rhs): New function,
9757 (vectorizable_mask_load_store): ...here.
9758 (vectorizable_store): ...and here.
9760 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9762 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
9764 (vectorizable_mask_load_store): ...here.
9766 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9768 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
9769 (vect_model_store_cost): Take a vec_load_store_type instead of a
9771 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
9772 (vect_model_store_cost): Take a vec_load_store_type instead of a
9774 (vectorizable_mask_load_store): Update accordingly.
9775 (vectorizable_store): Likewise.
9776 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
9778 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9780 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
9781 IFN_MASK_LOAD calls here rather than...
9782 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
9784 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9785 Alan Hayward <alan.hayward@arm.com>
9786 David Sherwood <david.sherwood@arm.com>
9788 * expmed.c (extract_bit_field_1): For vector extracts,
9789 fall back to extract_bit_field_as_subreg if vec_extract
9792 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9793 Alan Hayward <alan.hayward@arm.com>
9794 David Sherwood <david.sherwood@arm.com>
9796 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
9797 they are variable or constant sized.
9798 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
9799 slots for constant-sized data.
9801 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9802 Alan Hayward <alan.hayward@arm.com>
9803 David Sherwood <david.sherwood@arm.com>
9805 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
9806 handling COND_EXPRs with boolean comparisons, try to find a better
9807 basis for the mask type than the boolean itself.
9809 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9811 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
9812 is calculated and how it can be overridden.
9813 * genmodes.c (max_bitsize_mode_any_mode): New variable.
9814 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
9816 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
9819 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9820 Alan Hayward <alan.hayward@arm.com>
9821 David Sherwood <david.sherwood@arm.com>
9823 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
9824 Remove the mode argument.
9825 (aarch64_simd_valid_immediate): Remove the mode and inverse
9827 * config/aarch64/iterators.md (bitsize): New iterator.
9828 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
9829 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
9830 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
9831 aarch64_simd_valid_immediate.
9832 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
9833 (aarch64_reg_or_bic_imm): Likewise.
9834 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
9835 with an insn_type enum and msl with a modifier_type enum.
9836 Replace element_width with a scalar_mode. Change the shift
9837 to unsigned int. Add constructors for scalar_float_mode and
9838 scalar_int_mode elements.
9839 (aarch64_vect_float_const_representable_p): Delete.
9840 (aarch64_can_const_movi_rtx_p)
9841 (aarch64_simd_scalar_immediate_valid_for_move)
9842 (aarch64_simd_make_constant): Update call to
9843 aarch64_simd_valid_immediate.
9844 (aarch64_advsimd_valid_immediate_hs): New function.
9845 (aarch64_advsimd_valid_immediate): Likewise.
9846 (aarch64_simd_valid_immediate): Remove mode and inverse
9847 arguments. Rewrite to use the above. Use const_vec_duplicate_p
9848 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
9849 and aarch64_float_const_representable_p on the result.
9850 (aarch64_output_simd_mov_immediate): Remove mode argument.
9851 Update call to aarch64_simd_valid_immediate and use of
9852 simd_immediate_info.
9853 (aarch64_output_scalar_simd_mov_immediate): Update call
9856 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9857 Alan Hayward <alan.hayward@arm.com>
9858 David Sherwood <david.sherwood@arm.com>
9860 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
9861 (mode_nunits): Likewise CONST_MODE_NUNITS.
9862 * machmode.def (ADJUST_NUNITS): Document.
9863 * genmodes.c (mode_data::need_nunits_adj): New field.
9864 (blank_mode): Update accordingly.
9865 (adj_nunits): New variable.
9866 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
9868 (emit_mode_size_inline): Set need_bytesize_adj for all modes
9869 listed in adj_nunits.
9870 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
9871 listed in adj_nunits. Don't emit case statements for such modes.
9872 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
9873 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
9874 nothing if adj_nunits is nonnull.
9875 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
9876 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
9877 (emit_mode_fbit): Update use of print_maybe_const_decl.
9878 (emit_move_size): Likewise. Treat the array as non-const
9880 (emit_mode_adjustments): Handle adj_nunits.
9882 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9884 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
9885 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
9886 (VECTOR_MODES): Use it.
9887 (make_vector_modes): Take the prefix as an argument.
9889 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9890 Alan Hayward <alan.hayward@arm.com>
9891 David Sherwood <david.sherwood@arm.com>
9893 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
9894 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
9895 for MODE_VECTOR_BOOL.
9896 * machmode.def (VECTOR_BOOL_MODE): Document.
9897 * genmodes.c (VECTOR_BOOL_MODE): New macro.
9898 (make_vector_bool_mode): New function.
9899 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
9901 * lto-streamer-in.c (lto_input_mode_table): Likewise.
9902 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
9904 * stor-layout.c (int_mode_for_mode): Likewise.
9905 * tree.c (build_vector_type_for_mode): Likewise.
9906 * varasm.c (output_constant_pool_2): Likewise.
9907 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
9908 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
9909 for MODE_VECTOR_BOOL.
9910 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
9911 of mode class checks.
9912 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
9913 instead of a list of mode class checks.
9914 (expand_vector_scalar_condition): Likewise.
9915 (type_for_widest_vector_mode): Handle BImode as an inner mode.
9917 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9918 Alan Hayward <alan.hayward@arm.com>
9919 David Sherwood <david.sherwood@arm.com>
9921 * machmode.h (mode_size): Change from unsigned short to
9923 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
9924 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9925 or if measurement_type is not polynomial.
9926 (fixed_size_mode::includes_p): Check for constant-sized modes.
9927 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
9928 return a poly_uint16 rather than an unsigned short.
9929 (emit_mode_size): Change the type of mode_size from unsigned short
9930 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
9931 (emit_mode_adjustments): Cope with polynomial vector sizes.
9932 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9934 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9936 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
9937 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
9938 * caller-save.c (setup_save_areas): Likewise.
9939 (replace_reg_with_saved_mem): Likewise.
9940 * calls.c (emit_library_call_value_1): Likewise.
9941 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
9942 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
9943 (gen_lowpart_for_combine): Likewise.
9944 * convert.c (convert_to_integer_1): Likewise.
9945 * cse.c (equiv_constant, cse_insn): Likewise.
9946 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
9947 (cselib_subst_to_values): Likewise.
9948 * dce.c (word_dce_process_block): Likewise.
9949 * df-problems.c (df_word_lr_mark_ref): Likewise.
9950 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
9951 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
9952 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
9953 (rtl_for_decl_location): Likewise.
9954 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
9955 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
9956 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
9957 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
9958 (expand_expr_real_1): Likewise.
9959 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
9960 (pad_below): Likewise.
9961 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9962 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
9963 * ira.c (get_subreg_tracking_sizes): Likewise.
9964 * ira-build.c (ira_create_allocno_objects): Likewise.
9965 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
9966 (ira_sort_regnos_for_alter_reg): Likewise.
9967 * ira-costs.c (record_operand_costs): Likewise.
9968 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
9969 (resolve_simple_move): Likewise.
9970 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
9971 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
9972 (lra_constraints): Likewise.
9973 (CONST_POOL_OK_P): Reject variable-sized modes.
9974 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
9975 (add_pseudo_to_slot, lra_spill): Likewise.
9976 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9977 * optabs-query.c (get_best_extraction_insn): Likewise.
9978 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9979 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
9980 (expand_mult_highpart, valid_multiword_target_p): Likewise.
9981 * recog.c (offsettable_address_addr_space_p): Likewise.
9982 * regcprop.c (maybe_mode_change): Likewise.
9983 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
9984 * regrename.c (build_def_use): Likewise.
9985 * regstat.c (dump_reg_info): Likewise.
9986 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
9987 (find_reloads, find_reloads_subreg_address): Likewise.
9988 * reload1.c (eliminate_regs_1): Likewise.
9989 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
9990 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
9991 (simplify_binary_operation_1, simplify_subreg): Likewise.
9992 * targhooks.c (default_function_arg_padding): Likewise.
9993 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
9994 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
9995 (verify_gimple_assign_ternary): Likewise.
9996 * tree-inline.c (estimate_move_cost): Likewise.
9997 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9998 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
9999 (get_address_cost_ainc): Likewise.
10000 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
10001 (vect_supportable_dr_alignment): Likewise.
10002 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10003 (vectorizable_reduction): Likewise.
10004 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
10005 (vectorizable_operation, vectorizable_load): Likewise.
10006 * tree.c (build_same_sized_truth_vector_type): Likewise.
10007 * valtrack.c (cleanup_auto_inc_dec): Likewise.
10008 * var-tracking.c (emit_note_insn_var_location): Likewise.
10009 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
10010 (ADDR_VEC_ALIGN): Likewise.
10012 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10013 Alan Hayward <alan.hayward@arm.com>
10014 David Sherwood <david.sherwood@arm.com>
10016 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
10018 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10019 or if measurement_type is polynomial.
10020 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
10021 * combine.c (make_extraction): Likewise.
10022 * dse.c (find_shift_sequence): Likewise.
10023 * dwarf2out.c (mem_loc_descriptor): Likewise.
10024 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
10025 (extract_bit_field, extract_low_bits): Likewise.
10026 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
10027 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
10028 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
10029 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
10030 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10031 * reload.c (find_reloads): Likewise.
10032 * reload1.c (alter_reg): Likewise.
10033 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
10034 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
10035 * tree-if-conv.c (predicate_mem_writes): Likewise.
10036 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
10037 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
10038 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
10039 * valtrack.c (dead_debug_insert_temp): Likewise.
10040 * varasm.c (mergeable_constant_section): Likewise.
10041 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
10043 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10044 Alan Hayward <alan.hayward@arm.com>
10045 David Sherwood <david.sherwood@arm.com>
10047 * expr.c (expand_assignment): Cope with polynomial mode sizes
10048 when assigning to a CONCAT.
10050 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10051 Alan Hayward <alan.hayward@arm.com>
10052 David Sherwood <david.sherwood@arm.com>
10054 * machmode.h (mode_precision): Change from unsigned short to
10056 (mode_to_precision): Return a poly_uint16 rather than an unsigned
10058 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
10059 or if measurement_type is not polynomial.
10060 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
10061 in which the mode is already known to be a scalar_int_mode.
10062 * genmodes.c (emit_mode_precision): Change the type of mode_precision
10063 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
10065 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10066 for GET_MODE_PRECISION.
10067 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10068 for GET_MODE_PRECISION.
10069 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
10071 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
10072 (expand_field_assignment, make_extraction): Likewise.
10073 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
10074 (get_last_value): Likewise.
10075 * convert.c (convert_to_integer_1): Likewise.
10076 * cse.c (cse_insn): Likewise.
10077 * expr.c (expand_expr_real_1): Likewise.
10078 * lra-constraints.c (simplify_operand_subreg): Likewise.
10079 * optabs-query.c (can_atomic_load_p): Likewise.
10080 * optabs.c (expand_atomic_load): Likewise.
10081 (expand_atomic_store): Likewise.
10082 * ree.c (combine_reaching_defs): Likewise.
10083 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
10084 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
10085 * tree.h (type_has_mode_precision_p): Likewise.
10086 * ubsan.c (instrument_si_overflow): Likewise.
10088 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10089 Alan Hayward <alan.hayward@arm.com>
10090 David Sherwood <david.sherwood@arm.com>
10092 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
10093 polynomial numbers of units.
10094 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
10095 (valid_vector_subparts_p): New function.
10096 (build_vector_type): Remove temporary shim and take the number
10097 of units as a poly_uint64 rather than an int.
10098 (build_opaque_vector_type): Take the number of units as a
10099 poly_uint64 rather than an int.
10100 * tree.c (build_vector_from_ctor): Handle polynomial
10101 TYPE_VECTOR_SUBPARTS.
10102 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
10103 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
10104 (build_vector_from_val): If the number of units is variable,
10105 use build_vec_duplicate_cst for constant operands and
10106 VEC_DUPLICATE_EXPR otherwise.
10107 (make_vector_type): Remove temporary is_constant ().
10108 (build_vector_type, build_opaque_vector_type): Take the number of
10109 units as a poly_uint64 rather than an int.
10110 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
10112 * cfgexpand.c (expand_debug_expr): Likewise.
10113 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
10114 (store_constructor, expand_expr_real_1): Likewise.
10115 (const_scalar_mask_from_tree): Likewise.
10116 * fold-const-call.c (fold_const_reduction): Likewise.
10117 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
10118 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
10119 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
10120 (fold_relational_const): Likewise.
10121 (native_interpret_vector): Likewise. Change the size from an
10122 int to an unsigned int.
10123 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
10124 TYPE_VECTOR_SUBPARTS.
10125 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
10126 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
10127 duplicating a non-constant operand into a variable-length vector.
10128 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
10129 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
10130 * ipa-icf.c (sem_variable::equals): Likewise.
10131 * match.pd: Likewise.
10132 * omp-simd-clone.c (simd_clone_subparts): Likewise.
10133 * print-tree.c (print_node): Likewise.
10134 * stor-layout.c (layout_type): Likewise.
10135 * targhooks.c (default_builtin_vectorization_cost): Likewise.
10136 * tree-cfg.c (verify_gimple_comparison): Likewise.
10137 (verify_gimple_assign_binary): Likewise.
10138 (verify_gimple_assign_ternary): Likewise.
10139 (verify_gimple_assign_single): Likewise.
10140 * tree-pretty-print.c (dump_generic_node): Likewise.
10141 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10142 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
10143 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
10144 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
10145 (vect_shift_permute_load_chain): Likewise.
10146 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
10147 (expand_vector_condition, optimize_vector_constructor): Likewise.
10148 (lower_vec_perm, get_compute_type): Likewise.
10149 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10150 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
10151 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
10152 (vect_recog_mask_conversion_pattern): Likewise.
10153 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
10154 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
10155 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10156 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
10157 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
10158 (vectorizable_shift, vectorizable_operation, vectorizable_store)
10159 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
10160 (supportable_widening_operation): Likewise.
10161 (supportable_narrowing_operation): Likewise.
10162 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
10164 * varasm.c (output_constant): Likewise.
10166 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10167 Alan Hayward <alan.hayward@arm.com>
10168 David Sherwood <david.sherwood@arm.com>
10170 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
10171 so that both the length == 3 and length != 3 cases set up their
10172 own permute vectors. Add comments explaining why we know the
10173 number of elements is constant.
10174 (vect_permute_load_chain): Likewise.
10176 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10177 Alan Hayward <alan.hayward@arm.com>
10178 David Sherwood <david.sherwood@arm.com>
10180 * machmode.h (mode_nunits): Change from unsigned char to
10182 (ONLY_FIXED_SIZE_MODES): New macro.
10183 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
10184 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
10185 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
10187 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
10188 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
10189 or if measurement_type is not polynomial.
10190 * genmodes.c (ZERO_COEFFS): New macro.
10191 (emit_mode_nunits_inline): Make mode_nunits_inline return a
10193 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
10194 Use ZERO_COEFFS when emitting initializers.
10195 * data-streamer.h (bp_pack_poly_value): New function.
10196 (bp_unpack_poly_value): Likewise.
10197 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10198 for GET_MODE_NUNITS.
10199 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10200 for GET_MODE_NUNITS.
10201 * tree.c (make_vector_type): Remove temporary shim and make
10202 the real function take the number of units as a poly_uint64
10203 rather than an int.
10204 (build_vector_type_for_mode): Handle polynomial nunits.
10205 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
10206 * emit-rtl.c (const_vec_series_p_1): Likewise.
10207 (gen_rtx_CONST_VECTOR): Likewise.
10208 * fold-const.c (test_vec_duplicate_folding): Likewise.
10209 * genrecog.c (validate_pattern): Likewise.
10210 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
10211 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10212 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
10213 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
10214 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
10215 * rtlanal.c (subreg_get_info): Likewise.
10216 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10217 (vect_grouped_load_supported): Likewise.
10218 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
10219 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
10220 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
10221 (simplify_const_unary_operation, simplify_binary_operation_1)
10222 (simplify_const_binary_operation, simplify_ternary_operation)
10223 (test_vector_ops_duplicate, test_vector_ops): Likewise.
10224 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
10225 instead of CONST_VECTOR_NUNITS.
10226 * varasm.c (output_constant_pool_2): Likewise.
10227 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
10228 explicit-encoded elements in the XVEC for variable-length vectors.
10230 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10232 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
10234 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10235 Alan Hayward <alan.hayward@arm.com>
10236 David Sherwood <david.sherwood@arm.com>
10238 * coretypes.h (fixed_size_mode): Declare.
10239 (fixed_size_mode_pod): New typedef.
10240 * builtins.h (target_builtins::x_apply_args_mode)
10241 (target_builtins::x_apply_result_mode): Change type to
10242 fixed_size_mode_pod.
10243 * builtins.c (apply_args_size, apply_result_size, result_vector)
10244 (expand_builtin_apply_args_1, expand_builtin_apply)
10245 (expand_builtin_return): Update accordingly.
10247 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10249 * cse.c (hash_rtx_cb): Hash only the encoded elements.
10250 * cselib.c (cselib_hash_rtx): Likewise.
10251 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
10252 CONST_VECTOR encoding.
10254 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10255 Jeff Law <law@redhat.com>
10258 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
10259 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
10260 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
10261 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
10264 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
10265 explicitly probe *sp in a noreturn function if there were any callee
10266 register saves or frame pointer is needed.
10268 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10271 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
10272 BLKmode for ternary, binary or unary expressions.
10275 * var-tracking.c (delete_vta_debug_insn): New inline function.
10276 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
10277 insns from get_insns () to NULL instead of each bb separately.
10278 Use delete_vta_debug_insn. No longer static.
10279 (vt_debug_insns_local, variable_tracking_main_1): Adjust
10280 delete_vta_debug_insns callers.
10281 * rtl.h (delete_vta_debug_insns): Declare.
10282 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
10283 instead of variable_tracking_main.
10285 2018-01-03 Martin Sebor <msebor@redhat.com>
10287 PR tree-optimization/83603
10288 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
10289 arguments past the endof the argument list in functions declared
10290 without a prototype.
10291 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
10292 Avoid checking when arguments are null.
10294 2018-01-03 Martin Sebor <msebor@redhat.com>
10297 * doc/extend.texi (attribute const): Fix a typo.
10298 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
10299 issuing -Wsuggest-attribute for void functions.
10301 2018-01-03 Martin Sebor <msebor@redhat.com>
10303 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
10304 offset_int::from instead of wide_int::to_shwi.
10305 (maybe_diag_overlap): Remove assertion.
10306 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
10307 * gimple-ssa-sprintf.c (format_directive): Same.
10308 (parse_directive): Same.
10309 (sprintf_dom_walker::compute_format_length): Same.
10310 (try_substitute_return_value): Same.
10312 2018-01-03 Jeff Law <law@redhat.com>
10314 PR middle-end/83654
10315 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
10316 non-constant residual for zero at runtime and avoid probing in
10317 that case. Reorganize code for trailing problem to mirror handling
10320 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10322 PR tree-optimization/83501
10323 * tree-ssa-strlen.c (get_string_cst): New.
10324 (handle_char_store): Call get_string_cst.
10326 2018-01-03 Martin Liska <mliska@suse.cz>
10328 PR tree-optimization/83593
10329 * tree-ssa-strlen.c: Include tree-cfg.h.
10330 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
10331 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
10332 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
10334 (strlen_dom_walker::before_dom_children): Call
10335 gimple_purge_dead_eh_edges. Dump tranformation with details
10337 (strlen_dom_walker::before_dom_children): Update call by adding
10338 new argument cleanup_eh.
10339 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
10341 2018-01-03 Martin Liska <mliska@suse.cz>
10344 * cif-code.def (VARIADIC_THUNK): New enum value.
10345 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
10348 2018-01-03 Jan Beulich <jbeulich@suse.com>
10350 * sse.md (mov<mode>_internal): Tighten condition for when to use
10351 vmovdqu<ssescalarsize> for TI and OI modes.
10353 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10355 Update copyright years.
10357 2018-01-03 Martin Liska <mliska@suse.cz>
10360 * ipa-visibility.c (function_and_variable_visibility): Skip
10361 functions with noipa attribure.
10363 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10365 * gcc.c (process_command): Update copyright notice dates.
10366 * gcov-dump.c (print_version): Ditto.
10367 * gcov.c (print_version): Ditto.
10368 * gcov-tool.c (print_version): Ditto.
10369 * gengtype.c (create_file): Ditto.
10370 * doc/cpp.texi: Bump @copying's copyright year.
10371 * doc/cppinternals.texi: Ditto.
10372 * doc/gcc.texi: Ditto.
10373 * doc/gccint.texi: Ditto.
10374 * doc/gcov.texi: Ditto.
10375 * doc/install.texi: Ditto.
10376 * doc/invoke.texi: Ditto.
10378 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10380 * vector-builder.h (vector_builder::m_full_nelts): Change from
10381 unsigned int to poly_uint64.
10382 (vector_builder::full_nelts): Update prototype accordingly.
10383 (vector_builder::new_vector): Likewise.
10384 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
10385 (vector_builder::operator ==): Likewise.
10386 (vector_builder::finalize): Likewise.
10387 * int-vector-builder.h (int_vector_builder::int_vector_builder):
10388 Take the number of elements as a poly_uint64 rather than an
10390 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
10391 from unsigned int to poly_uint64.
10392 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
10393 (vec_perm_indices::new_vector): Likewise.
10394 (vec_perm_indices::length): Likewise.
10395 (vec_perm_indices::nelts_per_input): Likewise.
10396 (vec_perm_indices::input_nelts): Likewise.
10397 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
10398 number of elements per input as a poly_uint64 rather than an
10399 unsigned int. Use the original encoding for variable-length
10400 vectors, rather than clamping each individual element.
10401 For the second and subsequent elements in each pattern,
10402 clamp the step and base before clamping their sum.
10403 (vec_perm_indices::series_p): Handle polynomial element counts.
10404 (vec_perm_indices::all_in_range_p): Likewise.
10405 (vec_perm_indices_to_tree): Likewise.
10406 (vec_perm_indices_to_rtx): Likewise.
10407 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
10408 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
10409 (tree_vector_builder::new_binary_operation): Handle polynomial
10410 element counts. Return false if we need to know the number
10411 of elements at compile time.
10412 * fold-const.c (fold_vec_perm): Punt if the number of elements
10413 isn't known at compile time.
10415 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10417 * vec-perm-indices.h (vec_perm_builder): Change element type
10418 from HOST_WIDE_INT to poly_int64.
10419 (vec_perm_indices::element_type): Update accordingly.
10420 (vec_perm_indices::clamp): Handle polynomial element_types.
10421 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10422 (vec_perm_indices::all_in_range_p): Likewise.
10423 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
10425 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
10426 polynomial vec_perm_indices element types.
10427 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
10428 * fold-const.c (fold_vec_perm): Likewise.
10429 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
10430 * tree-vect-generic.c (lower_vec_perm): Likewise.
10431 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10432 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
10433 element type to HOST_WIDE_INT.
10435 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10436 Alan Hayward <alan.hayward@arm.com>
10437 David Sherwood <david.sherwood@arm.com>
10439 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
10440 rather than an int. Use plus_constant.
10441 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
10442 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
10444 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10445 Alan Hayward <alan.hayward@arm.com>
10446 David Sherwood <david.sherwood@arm.com>
10448 * calls.c (emit_call_1, expand_call): Change struct_value_size from
10449 a HOST_WIDE_INT to a poly_int64.
10451 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10452 Alan Hayward <alan.hayward@arm.com>
10453 David Sherwood <david.sherwood@arm.com>
10455 * calls.c (load_register_parameters): Cope with polynomial
10456 mode sizes. Require a constant size for BLKmode parameters
10457 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
10458 forces a parameter to be padded at the lsb end in order to
10459 fill a complete number of words, require the parameter size
10460 to be ordered wrt UNITS_PER_WORD.
10462 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10463 Alan Hayward <alan.hayward@arm.com>
10464 David Sherwood <david.sherwood@arm.com>
10466 * reload1.c (spill_stack_slot_width): Change element type
10467 from unsigned int to poly_uint64_pod.
10468 (alter_reg): Treat mode sizes as polynomial.
10470 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10471 Alan Hayward <alan.hayward@arm.com>
10472 David Sherwood <david.sherwood@arm.com>
10474 * reload.c (complex_word_subreg_p): New function.
10475 (reload_inner_reg_of_subreg, push_reload): Use it.
10477 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10478 Alan Hayward <alan.hayward@arm.com>
10479 David Sherwood <david.sherwood@arm.com>
10481 * lra-constraints.c (process_alt_operands): Reject matched
10482 operands whose sizes aren't ordered.
10483 (match_reload): Refer to this check here.
10485 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10486 Alan Hayward <alan.hayward@arm.com>
10487 David Sherwood <david.sherwood@arm.com>
10489 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
10490 that the mode size is in the set {1, 2, 4, 8, 16}.
10492 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10493 Alan Hayward <alan.hayward@arm.com>
10494 David Sherwood <david.sherwood@arm.com>
10496 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
10497 Use plus_constant instead of gen_rtx_PLUS.
10499 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10500 Alan Hayward <alan.hayward@arm.com>
10501 David Sherwood <david.sherwood@arm.com>
10503 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
10504 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
10505 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
10506 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
10507 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
10508 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
10509 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
10510 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
10511 * config/i386/i386.c (ix86_push_rounding): ...this new function.
10512 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
10514 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
10515 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
10516 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
10517 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
10518 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
10519 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
10520 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
10521 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
10522 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
10523 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
10525 * expr.c (emit_move_resolve_push): Treat the input and result
10526 of PUSH_ROUNDING as a poly_int64.
10527 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
10528 (emit_push_insn): Likewise.
10529 * lra-eliminations.c (mark_not_eliminable): Likewise.
10530 * recog.c (push_operand): Likewise.
10531 * reload1.c (elimination_effects): Likewise.
10532 * rtlanal.c (nonzero_bits1): Likewise.
10533 * calls.c (store_one_arg): Likewise. Require the padding to be
10534 known at compile time.
10536 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10537 Alan Hayward <alan.hayward@arm.com>
10538 David Sherwood <david.sherwood@arm.com>
10540 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
10541 Use plus_constant instead of gen_rtx_PLUS.
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 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
10548 rather than an int.
10550 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10551 Alan Hayward <alan.hayward@arm.com>
10552 David Sherwood <david.sherwood@arm.com>
10554 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
10555 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
10556 via stack temporaries. Treat the mode size as polynomial too.
10558 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10559 Alan Hayward <alan.hayward@arm.com>
10560 David Sherwood <david.sherwood@arm.com>
10562 * expr.c (expand_expr_real_2): When handling conversions involving
10563 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
10564 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
10565 as a poly_uint64 too.
10567 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10568 Alan Hayward <alan.hayward@arm.com>
10569 David Sherwood <david.sherwood@arm.com>
10571 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
10573 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10574 Alan Hayward <alan.hayward@arm.com>
10575 David Sherwood <david.sherwood@arm.com>
10577 * combine.c (can_change_dest_mode): Handle polynomial
10578 REGMODE_NATURAL_SIZE.
10579 * expmed.c (store_bit_field_1): Likewise.
10580 * expr.c (store_constructor): Likewise.
10581 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
10582 and polynomial REGMODE_NATURAL_SIZE.
10583 (gen_lowpart_common): Likewise.
10584 * reginfo.c (record_subregs_of_mode): Likewise.
10585 * rtlanal.c (read_modify_subreg_p): Likewise.
10587 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10588 Alan Hayward <alan.hayward@arm.com>
10589 David Sherwood <david.sherwood@arm.com>
10591 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
10592 numbers of elements.
10594 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10595 Alan Hayward <alan.hayward@arm.com>
10596 David Sherwood <david.sherwood@arm.com>
10598 * match.pd: Cope with polynomial numbers of vector elements.
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 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
10605 in a POINTER_PLUS_EXPR.
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 * omp-simd-clone.c (simd_clone_subparts): New function.
10612 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
10613 (ipa_simd_modify_function_body): Likewise.
10615 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10616 Alan Hayward <alan.hayward@arm.com>
10617 David Sherwood <david.sherwood@arm.com>
10619 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
10620 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
10621 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
10622 (expand_vector_condition, vector_element): Likewise.
10623 (subparts_gt): New function.
10624 (get_compute_type): Use subparts_gt.
10625 (count_type_subparts): Delete.
10626 (expand_vector_operations_1): Use subparts_gt instead of
10627 count_type_subparts.
10629 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10630 Alan Hayward <alan.hayward@arm.com>
10631 David Sherwood <david.sherwood@arm.com>
10633 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
10634 (vect_compile_time_alias): ...this new function. Do the calculation
10635 on poly_ints rather than trees.
10636 (vect_prune_runtime_alias_test_list): Update call accordingly.
10638 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10639 Alan Hayward <alan.hayward@arm.com>
10640 David Sherwood <david.sherwood@arm.com>
10642 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
10644 (vect_schedule_slp_instance): Likewise.
10646 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10647 Alan Hayward <alan.hayward@arm.com>
10648 David Sherwood <david.sherwood@arm.com>
10650 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
10651 constant and extern definitions for variable-length vectors.
10652 (vect_get_constant_vectors): Note that the number of units
10653 is known to be constant.
10655 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10656 Alan Hayward <alan.hayward@arm.com>
10657 David Sherwood <david.sherwood@arm.com>
10659 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
10660 of units as polynomial. Choose between WIDE and NARROW based
10663 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10664 Alan Hayward <alan.hayward@arm.com>
10665 David Sherwood <david.sherwood@arm.com>
10667 * tree-vect-stmts.c (simd_clone_subparts): New function.
10668 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
10670 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10671 Alan Hayward <alan.hayward@arm.com>
10672 David Sherwood <david.sherwood@arm.com>
10674 * tree-vect-stmts.c (vectorizable_call): Treat the number of
10675 vectors as polynomial. Use build_index_vector for
10676 IFN_GOMP_SIMD_LANE.
10678 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10679 Alan Hayward <alan.hayward@arm.com>
10680 David Sherwood <david.sherwood@arm.com>
10682 * tree-vect-stmts.c (get_load_store_type): Treat the number of
10683 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
10684 for variable-length vectors.
10685 (vectorizable_mask_load_store): Treat the number of units as
10686 polynomial, asserting that it is constant if the condition has
10687 already been enforced.
10688 (vectorizable_store, vectorizable_load): Likewise.
10690 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10691 Alan Hayward <alan.hayward@arm.com>
10692 David Sherwood <david.sherwood@arm.com>
10694 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
10695 of units as polynomial. Punt if we can't tell at compile time
10696 which vector contains the final result.
10698 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10699 Alan Hayward <alan.hayward@arm.com>
10700 David Sherwood <david.sherwood@arm.com>
10702 * tree-vect-loop.c (vectorizable_induction): Treat the number
10703 of units as polynomial. Punt on SLP inductions. Use an integer
10704 VEC_SERIES_EXPR for variable-length integer reductions. Use a
10705 cast of such a series for variable-length floating-point
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 * tree.h (build_index_vector): Declare.
10713 * tree.c (build_index_vector): New function.
10714 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
10715 of units as polynomial, forcibly converting it to a constant if
10716 vectorizable_reduction has already enforced the condition.
10717 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
10718 to create a {1,2,3,...} vector.
10719 (vectorizable_reduction): Treat the number of units as polynomial.
10720 Choose vectype_in based on the largest scalar element size rather
10721 than the smallest number of units. Enforce the restrictions
10724 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10725 Alan Hayward <alan.hayward@arm.com>
10726 David Sherwood <david.sherwood@arm.com>
10728 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
10729 number of units as polynomial.
10731 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10732 Alan Hayward <alan.hayward@arm.com>
10733 David Sherwood <david.sherwood@arm.com>
10735 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
10736 * target.def (autovectorize_vector_sizes): Return the vector sizes
10737 by pointer, using vector_sizes rather than a bitmask.
10738 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
10739 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
10740 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
10742 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
10743 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
10744 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
10745 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
10746 * omp-general.c (omp_max_vf): Likewise.
10747 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10748 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
10749 * tree-vect-loop.c (vect_analyze_loop): Likewise.
10750 * tree-vect-slp.c (vect_slp_bb): Likewise.
10751 * doc/tm.texi: Regenerate.
10752 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
10754 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
10755 the vector size as a poly_uint64 rather than an unsigned int.
10756 (current_vector_size): Change from an unsigned int to a poly_uint64.
10757 (get_vectype_for_scalar_type): Update accordingly.
10758 * tree.h (build_truth_vector_type): Take the size and number of
10759 units as a poly_uint64 rather than an unsigned int.
10760 (build_vector_type): Add a temporary overload that takes
10761 the number of units as a poly_uint64 rather than an unsigned int.
10762 * tree.c (make_vector_type): Likewise.
10763 (build_truth_vector_type): Take the number of units as a poly_uint64
10764 rather than an unsigned int.
10766 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10767 Alan Hayward <alan.hayward@arm.com>
10768 David Sherwood <david.sherwood@arm.com>
10770 * target.def (get_mask_mode): Take the number of units and length
10771 as poly_uint64s rather than unsigned ints.
10772 * targhooks.h (default_get_mask_mode): Update accordingly.
10773 * targhooks.c (default_get_mask_mode): Likewise.
10774 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
10775 * doc/tm.texi: Regenerate.
10777 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10778 Alan Hayward <alan.hayward@arm.com>
10779 David Sherwood <david.sherwood@arm.com>
10781 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
10782 * omp-general.c (omp_max_vf): Likewise.
10783 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
10784 (expand_omp_simd): Handle polynomial safelen.
10785 * omp-low.c (omplow_simd_context): Add a default constructor.
10786 (omplow_simd_context::max_vf): Change from int to poly_uint64.
10787 (lower_rec_simd_input_clauses): Update accordingly.
10788 (lower_rec_input_clauses): Likewise.
10790 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10791 Alan Hayward <alan.hayward@arm.com>
10792 David Sherwood <david.sherwood@arm.com>
10794 * tree-vectorizer.h (vect_nunits_for_cost): New function.
10795 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
10796 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
10797 (vect_analyze_slp_cost): Likewise.
10798 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
10799 (vect_model_load_cost): Likewise.
10801 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10802 Alan Hayward <alan.hayward@arm.com>
10803 David Sherwood <david.sherwood@arm.com>
10805 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
10806 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
10807 from an unsigned int * to a poly_uint64_pod *.
10808 (calculate_unrolling_factor): New function.
10809 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
10811 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10812 Alan Hayward <alan.hayward@arm.com>
10813 David Sherwood <david.sherwood@arm.com>
10815 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
10816 from an unsigned int to a poly_uint64.
10817 (_loop_vec_info::slp_unrolling_factor): Likewise.
10818 (_loop_vec_info::vectorization_factor): Change from an int
10820 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
10821 (vect_get_num_vectors): New function.
10822 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
10823 (vect_get_num_copies): Use vect_get_num_vectors.
10824 (vect_analyze_data_ref_dependences): Change max_vf from an int *
10825 to an unsigned int *.
10826 (vect_analyze_data_refs): Change min_vf from an int * to a
10828 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10829 than an unsigned HOST_WIDE_INT.
10830 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
10831 (vect_analyze_data_ref_dependence): Change max_vf from an int *
10832 to an unsigned int *.
10833 (vect_analyze_data_ref_dependences): Likewise.
10834 (vect_compute_data_ref_alignment): Handle polynomial vf.
10835 (vect_enhance_data_refs_alignment): Likewise.
10836 (vect_prune_runtime_alias_test_list): Likewise.
10837 (vect_shift_permute_load_chain): Likewise.
10838 (vect_supportable_dr_alignment): Likewise.
10839 (dependence_distance_ge_vf): Take the vectorization factor as a
10840 poly_uint64 rather than an unsigned HOST_WIDE_INT.
10841 (vect_analyze_data_refs): Change min_vf from an int * to a
10843 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
10844 vfm1 as a poly_uint64 rather than an int. Make the same change
10845 for the returned bound_scalar.
10846 (vect_gen_vector_loop_niters): Handle polynomial vf.
10847 (vect_do_peeling): Likewise. Update call to
10848 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
10849 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
10851 * tree-vect-loop.c (vect_determine_vectorization_factor)
10852 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
10853 (vect_get_known_peeling_cost): Likewise.
10854 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
10855 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
10856 (vect_transform_loop): Likewise. Use the lowest possible VF when
10857 updating the upper bounds of the loop.
10858 (vect_min_worthwhile_factor): Make static. Return an unsigned int
10859 rather than an int.
10860 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
10861 polynomial unroll factors.
10862 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
10863 (vect_make_slp_decision): Likewise.
10864 (vect_supported_load_permutation_p): Likewise, and polynomial
10866 (vect_analyze_slp_cost): Handle polynomial vf.
10867 (vect_slp_analyze_node_operations): Likewise.
10868 (vect_slp_analyze_bb_1): Likewise.
10869 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10870 than an unsigned HOST_WIDE_INT.
10871 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
10872 (vectorizable_load): Handle polynomial vf.
10873 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
10875 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
10877 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10878 Alan Hayward <alan.hayward@arm.com>
10879 David Sherwood <david.sherwood@arm.com>
10881 * match.pd: Handle bit operations involving three constants
10882 and try to fold one pair.
10884 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10886 * tree-vect-loop-manip.c: Include gimple-fold.h.
10887 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
10888 niters_maybe_zero parameters. Handle other cases besides a step of 1.
10889 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
10890 Add a path that uses a step of VF instead of 1, but disable it
10892 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
10893 and niters_no_overflow parameters. Update calls to
10894 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
10895 Create a new SSA name if the latter choses to use a ste other
10896 than zero, and return it via niters_vector_mult_vf_var.
10897 * tree-vect-loop.c (vect_transform_loop): Update calls to
10898 vect_do_peeling, vect_gen_vector_loop_niters and
10899 slpeel_make_loop_iterate_ntimes.
10900 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
10901 (vect_gen_vector_loop_niters): Update declarations after above changes.
10903 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
10905 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
10906 128-bit round to integer instructions.
10907 (ceil<mode>2): Likewise.
10908 (btrunc<mode>2): Likewise.
10909 (round<mode>2): Likewise.
10911 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10913 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
10914 unaligned VSX load/store on P8/P9.
10915 (expand_block_clear): Allow the use of unaligned VSX
10916 load/store on P8/P9.
10918 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10920 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
10922 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
10923 swap associated with both a load and a store.
10925 2018-01-02 Andrew Waterman <andrew@sifive.com>
10927 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
10928 * config/riscv/riscv.md (clear_cache): Use it.
10930 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
10932 * web.c: Remove out-of-date comment.
10934 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10936 * expr.c (fixup_args_size_notes): Check that any existing
10937 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
10938 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
10939 (emit_single_push_insn): ...here.
10941 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10943 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
10944 (const_vector_encoded_nelts): New function.
10945 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
10946 (const_vector_int_elt, const_vector_elt): Declare.
10947 * emit-rtl.c (const_vector_int_elt_1): New function.
10948 (const_vector_elt): Likewise.
10949 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
10950 of CONST_VECTOR_ELT.
10952 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10954 * expr.c: Include rtx-vector-builder.h.
10955 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
10956 directly on the tree encoding.
10957 (const_vector_from_tree): Likewise.
10958 * optabs.c: Include rtx-vector-builder.h.
10959 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
10960 sequence of "u" values.
10961 * vec-perm-indices.c: Include rtx-vector-builder.h.
10962 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
10963 directly on the vec_perm_indices encoding.
10965 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10967 * doc/rtl.texi (const_vector): Describe new encoding scheme.
10968 * Makefile.in (OBJS): Add rtx-vector-builder.o.
10969 * rtx-vector-builder.h: New file.
10970 * rtx-vector-builder.c: Likewise.
10971 * rtl.h (rtx_def::u2): Add a const_vector field.
10972 (CONST_VECTOR_NPATTERNS): New macro.
10973 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
10974 (CONST_VECTOR_DUPLICATE_P): Likewise.
10975 (CONST_VECTOR_STEPPED_P): Likewise.
10976 (CONST_VECTOR_ENCODED_ELT): Likewise.
10977 (const_vec_duplicate_p): Check for a duplicated vector encoding.
10978 (unwrap_const_vec_duplicate): Likewise.
10979 (const_vec_series_p): Check for a non-duplicated vector encoding.
10980 Say that the function only returns true for integer vectors.
10981 * emit-rtl.c: Include rtx-vector-builder.h.
10982 (gen_const_vec_duplicate_1): Delete.
10983 (gen_const_vector): Call gen_const_vec_duplicate instead of
10984 gen_const_vec_duplicate_1.
10985 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
10986 (gen_const_vec_duplicate): Use rtx_vector_builder.
10987 (gen_const_vec_series): Likewise.
10988 (gen_rtx_CONST_VECTOR): Likewise.
10989 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
10990 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10991 Build a new vector rather than modifying a CONST_VECTOR in-place.
10992 (handle_special_swappables): Update call accordingly.
10993 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
10994 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10995 Build a new vector rather than modifying a CONST_VECTOR in-place.
10996 (handle_special_swappables): Update call accordingly.
10998 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11000 * simplify-rtx.c (simplify_const_binary_operation): Use
11001 CONST_VECTOR_ELT instead of XVECEXP.
11003 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11005 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
11006 the selector elements to be different from the data elements
11007 if the selector is a VECTOR_CST.
11008 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
11009 ssizetype for the selector.
11011 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11013 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
11014 before testing each element individually.
11015 * tree-vect-generic.c (lower_vec_perm): Likewise.
11017 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11019 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
11020 * selftest-run-tests.c (selftest::run_tests): Call it.
11021 * vector-builder.h (vector_builder::operator ==): New function.
11022 (vector_builder::operator !=): Likewise.
11023 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
11024 (vec_perm_indices::all_from_input_p): New function.
11025 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11026 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
11027 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
11028 instead of reading the VECTOR_CST directly. Detect whether both
11029 vector inputs are the same before constructing the vec_perm_indices,
11030 and update the number of inputs argument accordingly. Use the
11031 utility functions added above. Only construct sel2 if we need to.
11033 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11035 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
11036 the broadcast of the low byte.
11037 (expand_mult_highpart): Use an explicit encoding for the permutes.
11038 * optabs-query.c (can_mult_highpart_p): Likewise.
11039 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
11040 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11041 (vectorizable_bswap): Likewise.
11042 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
11043 explicit encoding for the power-of-2 permutes.
11044 (vect_permute_store_chain): Likewise.
11045 (vect_grouped_load_supported): Likewise.
11046 (vect_permute_load_chain): Likewise.
11048 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11050 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
11051 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
11052 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
11053 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11054 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
11055 (vect_gen_perm_mask_any): Likewise.
11057 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11059 * int-vector-builder.h: New file.
11060 * vec-perm-indices.h: Include int-vector-builder.h.
11061 (vec_perm_indices): Redefine as an int_vector_builder.
11062 (auto_vec_perm_indices): Delete.
11063 (vec_perm_builder): Redefine as a stand-alone class.
11064 (vec_perm_indices::vec_perm_indices): New function.
11065 (vec_perm_indices::clamp): Likewise.
11066 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
11067 (vec_perm_indices::new_vector): New function.
11068 (vec_perm_indices::new_expanded_vector): Update for new
11069 vec_perm_indices class.
11070 (vec_perm_indices::rotate_inputs): New function.
11071 (vec_perm_indices::all_in_range_p): Operate directly on the
11072 encoded form, without computing elided elements.
11073 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
11074 encoding. Update for new vec_perm_indices class.
11075 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
11076 the given vec_perm_builder.
11077 (expand_vec_perm_var): Update vec_perm_builder constructor.
11078 (expand_mult_highpart): Use vec_perm_builder instead of
11079 auto_vec_perm_indices.
11080 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
11081 vec_perm_indices instead of auto_vec_perm_indices. Use a single
11082 or double series encoding as appropriate.
11083 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
11084 vec_perm_indices instead of auto_vec_perm_indices.
11085 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11086 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11087 (vect_permute_store_chain): Likewise.
11088 (vect_grouped_load_supported): Likewise.
11089 (vect_permute_load_chain): Likewise.
11090 (vect_shift_permute_load_chain): Likewise.
11091 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11092 (vect_transform_slp_perm_load): Likewise.
11093 (vect_schedule_slp_instance): Likewise.
11094 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11095 (vectorizable_mask_load_store): Likewise.
11096 (vectorizable_bswap): Likewise.
11097 (vectorizable_store): Likewise.
11098 (vectorizable_load): Likewise.
11099 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
11100 vec_perm_indices instead of auto_vec_perm_indices. Use
11101 tree_to_vec_perm_builder to read the vector from a tree.
11102 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
11103 vec_perm_builder instead of a vec_perm_indices.
11104 (have_whole_vector_shift): Use vec_perm_builder and
11105 vec_perm_indices instead of auto_vec_perm_indices. Leave the
11106 truncation to calc_vec_perm_mask_for_shift.
11107 (vect_create_epilog_for_reduction): Likewise.
11108 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
11109 from auto_vec_perm_indices to vec_perm_indices.
11110 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11111 instead of changing individual elements.
11112 (aarch64_vectorize_vec_perm_const): Use new_vector to install
11113 the vector in d.perm.
11114 * config/arm/arm.c (expand_vec_perm_d::perm): Change
11115 from auto_vec_perm_indices to vec_perm_indices.
11116 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11117 instead of changing individual elements.
11118 (arm_vectorize_vec_perm_const): Use new_vector to install
11119 the vector in d.perm.
11120 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
11121 Update vec_perm_builder constructor.
11122 (rs6000_expand_interleave): Likewise.
11123 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
11124 (rs6000_expand_interleave): Likewise.
11126 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11128 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
11129 to qimode could truncate the indices.
11130 * optabs.c (expand_vec_perm_var): Likewise.
11132 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11134 * Makefile.in (OBJS): Add vec-perm-indices.o.
11135 * vec-perm-indices.h: New file.
11136 * vec-perm-indices.c: Likewise.
11137 * target.h (vec_perm_indices): Replace with a forward class
11139 (auto_vec_perm_indices): Move to vec-perm-indices.h.
11140 * optabs.h: Include vec-perm-indices.h.
11141 (expand_vec_perm): Delete.
11142 (selector_fits_mode_p, expand_vec_perm_var): Declare.
11143 (expand_vec_perm_const): Declare.
11144 * target.def (vec_perm_const_ok): Replace with...
11145 (vec_perm_const): ...this new hook.
11146 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
11147 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
11148 * doc/tm.texi: Regenerate.
11149 * optabs.def (vec_perm_const): Delete.
11150 * doc/md.texi (vec_perm_const): Likewise.
11151 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
11152 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
11153 expand_vec_perm for constant permutation vectors. Assert that
11154 the mode of variable permutation vectors is the integer equivalent
11155 of the mode that is being permuted.
11156 * optabs-query.h (selector_fits_mode_p): Declare.
11157 * optabs-query.c: Include vec-perm-indices.h.
11158 (selector_fits_mode_p): New function.
11159 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
11160 is defined, instead of checking whether the vec_perm_const_optab
11161 exists. Use targetm.vectorize.vec_perm_const instead of
11162 targetm.vectorize.vec_perm_const_ok. Check whether the indices
11163 fit in the vector mode before using a variable permute.
11164 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
11165 vec_perm_indices instead of an rtx.
11166 (expand_vec_perm): Replace with...
11167 (expand_vec_perm_const): ...this new function. Take the selector
11168 as a vec_perm_indices rather than an rtx. Also take the mode of
11169 the selector. Update call to shift_amt_for_vec_perm_mask.
11170 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
11171 Use vec_perm_indices::new_expanded_vector to expand the original
11172 selector into bytes. Check whether the indices fit in the vector
11173 mode before using a variable permute.
11174 (expand_vec_perm_var): Make global.
11175 (expand_mult_highpart): Use expand_vec_perm_const.
11176 * fold-const.c: Includes vec-perm-indices.h.
11177 * tree-ssa-forwprop.c: Likewise.
11178 * tree-vect-data-refs.c: Likewise.
11179 * tree-vect-generic.c: Likewise.
11180 * tree-vect-loop.c: Likewise.
11181 * tree-vect-slp.c: Likewise.
11182 * tree-vect-stmts.c: Likewise.
11183 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
11185 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
11186 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
11187 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
11188 (aarch64_vectorize_vec_perm_const): ...this new function.
11189 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11190 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11191 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
11192 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
11193 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11194 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11195 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
11197 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
11198 check for NEON modes.
11199 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
11200 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
11201 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
11202 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
11204 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
11205 the old VEC_PERM_CONST conditions.
11206 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
11207 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
11208 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
11209 (ia64_vectorize_vec_perm_const_ok): Merge into...
11210 (ia64_vectorize_vec_perm_const): ...this new function.
11211 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
11212 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
11213 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
11214 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
11215 * config/mips/mips.c (mips_expand_vec_perm_const)
11216 (mips_vectorize_vec_perm_const_ok): Merge into...
11217 (mips_vectorize_vec_perm_const): ...this new function.
11218 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
11219 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
11220 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
11221 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
11222 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
11223 (rs6000_expand_vec_perm_const): Delete.
11224 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
11226 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11227 (altivec_expand_vec_perm_const_le): Take each operand individually.
11228 Operate on constant selectors rather than rtxes.
11229 (altivec_expand_vec_perm_const): Likewise. Update call to
11230 altivec_expand_vec_perm_const_le.
11231 (rs6000_expand_vec_perm_const): Delete.
11232 (rs6000_vectorize_vec_perm_const_ok): Delete.
11233 (rs6000_vectorize_vec_perm_const): New function.
11234 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11235 an element count and rtx array.
11236 (rs6000_expand_extract_even): Update call accordingly.
11237 (rs6000_expand_interleave): Likewise.
11238 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
11239 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
11240 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
11241 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
11242 (rs6000_expand_vec_perm_const): Delete.
11243 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11244 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11245 (altivec_expand_vec_perm_const_le): Take each operand individually.
11246 Operate on constant selectors rather than rtxes.
11247 (altivec_expand_vec_perm_const): Likewise. Update call to
11248 altivec_expand_vec_perm_const_le.
11249 (rs6000_expand_vec_perm_const): Delete.
11250 (rs6000_vectorize_vec_perm_const_ok): Delete.
11251 (rs6000_vectorize_vec_perm_const): New function. Remove stray
11252 reference to the SPE evmerge intructions.
11253 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11254 an element count and rtx array.
11255 (rs6000_expand_extract_even): Update call accordingly.
11256 (rs6000_expand_interleave): Likewise.
11257 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
11258 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
11260 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11262 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11264 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
11265 vector mode and that that mode matches the mode of the data
11267 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
11268 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
11269 directly using expand_vec_perm_1 when forcing selectors into
11271 (expand_vec_perm_var): New function, split out from expand_vec_perm.
11273 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11275 * optabs-query.h (can_vec_perm_p): Delete.
11276 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
11277 * optabs-query.c (can_vec_perm_p): Split into...
11278 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
11279 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
11280 particular selector is valid.
11281 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11282 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11283 (vect_grouped_load_supported): Likewise.
11284 (vect_shift_permute_load_chain): Likewise.
11285 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11286 (vect_transform_slp_perm_load): Likewise.
11287 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11288 (vectorizable_bswap): Likewise.
11289 (vect_gen_perm_mask_checked): Likewise.
11290 * fold-const.c (fold_ternary_loc): Likewise. Don't take
11291 implementations of variable permutation vectors into account
11292 when deciding which selector to use.
11293 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
11294 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
11295 with a false third argument.
11296 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
11297 to test whether the constant selector is valid and can_vec_perm_var_p
11298 to test whether a variable selector is valid.
11300 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11302 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
11303 * optabs-query.c (can_vec_perm_p): Likewise.
11304 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
11305 instead of vec_perm_indices.
11306 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
11307 (vect_gen_perm_mask_checked): Likewise,
11308 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
11309 (vect_gen_perm_mask_checked): Likewise,
11311 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11313 * optabs-query.h (qimode_for_vec_perm): Declare.
11314 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
11315 (qimode_for_vec_perm): ...this new function.
11316 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
11318 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11320 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
11321 does not have a conditional at the top.
11323 2018-01-02 Richard Biener <rguenther@suse.de>
11325 * ipa-inline.c (big_speedup_p): Fix expression.
11327 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11330 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
11333 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11337 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
11338 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
11339 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
11340 cond_taken_branch_cost 3->4.
11342 2018-01-01 Jakub Jelinek <jakub@redhat.com>
11344 PR tree-optimization/83581
11345 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
11346 TODO_cleanup_cfg if any changes have been made.
11348 PR middle-end/83608
11349 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
11350 convert_modes if target mode has the right side, but different mode
11353 PR middle-end/83609
11354 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
11355 last argument when extracting from CONCAT. If either from_real or
11356 from_imag is NULL, use expansion through memory. If result is not
11357 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
11358 the parts directly to inner mode, if even that fails, use expansion
11361 PR middle-end/83623
11362 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
11363 check for bswap in mode rather than HImode and use that in expand_unop
11366 Copyright (C) 2018 Free Software Foundation, Inc.
11368 Copying and distribution of this file, with or without modification,
11369 are permitted in any medium without royalty provided the copyright
11370 notice and this notice are preserved.